version bump to 0.10.0

Dereferencing registers in `files` module instructions functions

.envrc

@@ -0,0 +1 @@
+use flake

.git-blame-ignore-revs

@@ -0,0 +1,2 @@
+# Resolved all lints and formatted the codebase
+9444e62df9820d6bfd96dbd8849e177bc5cecc2e

.github/actions/setup-rust/action.yml

@@ -0,0 +1,52 @@
+name: 'Setup Rust'
+inputs:
+  rust-version:
+    required: true
+    type: string
+  targets:
+    required: true
+    type: string
+  components:
+    required: false
+    default:
+  cache-context:
+    required: true
+    type: string
+
+runs:
+  using: "composite"
+  steps:
+  - uses: dtolnay/rust-toolchain@master
+    id: toolchain
+    with:
+      toolchain: ${{ inputs.rust-version }}
+      targets: ${{ inputs.targets }}
+      components: ${{ inputs.components }}
+
+  - name: Install i686 dependencies
+    if: "contains(inputs.targets,'i686')"
+    shell: bash
+    run: |
+      sudo dpkg --add-architecture i386
+      sudo apt-get update
+      sudo apt-get install libssl-dev:i386 gcc-multilib clang -y
+      echo "CC=clang" >> $GITHUB_ENV
+      echo "PKG_CONFIG_SYSROOT_DIR=/" >> $GITHUB_ENV
+
+  - uses: actions/cache@v4
+    with:
+      path: |
+        ~/.cargo/bin/
+        ~/.cargo/registry/index/
+        ~/.cargo/registry/cache/
+        ~/.cargo/git/db/
+        target/
+      key: ${{ inputs.cache-context }}_${{ inputs.targets }}_rustc-${{ steps.toolchain.outputs.cachekey }}_cargo-${{ hashFiles('**/Cargo.lock') }}
+
+  # Remove build artifacts for the current crate, since it will be rebuilt every
+  # run anyway, but keep dependency artifacts to cache them.
+  # Must be placed after actions/cache so its post step runs first.
+  - uses: pyTooling/Actions/with-post-step@v4.1.0
+    with:
+      main: bash ./.github/actions/setup-rust/cleanup.sh
+      post: bash ./.github/actions/setup-rust/cleanup.sh

.github/actions/setup-rust/cleanup.sh

@@ -0,0 +1,13 @@
+#!/usr/bin/env bash
+set -e
+
+echo Cleanup workspace build artifacts and extra target output
+
+# clean just the direct members of the current workspace, use cargo metadata to generalize to all rust projects
+cargo clean -p `cargo metadata --no-deps --offline --format-version 1 | jq -r '[.workspace_members[]|split(" ")|.[0]]|join(" ")'`
+
+# remove directories in /target/ that are not named `debug` or `release`
+before=`du -s target | awk '{print $1}'`
+find ./target -maxdepth 1 -type d ! -name debug ! -name release ! -name target -exec rm -r {} \;
+after=`du -s target | awk '{print $1}'`
+echo Deleted $(($before - $after)) bytes from target directory

.github/workflows/ci.yml

@@ -0,0 +1,238 @@
+name: CI
+
+on:
+  push:
+    branches: 
+      - master
+      - rebis-dev
+    tags:
+      - "v**"
+  pull_request:
+  schedule:
+    - cron: '0 0 * * 3' # At 12:00 AM, only on Wednesday
+  workflow_dispatch:
+
+permissions:
+  checks: write
+
+jobs:
+  style:
+    runs-on: ubuntu-22.04
+    steps:
+      - uses: actions/checkout@v4
+      - name: Setup Rust
+        uses: ./.github/actions/setup-rust
+        with:
+          rust-version: stable
+          targets: x86_64-unknown-linux-gnu
+          components: clippy, rustfmt
+          cache-context: style
+
+      - name: Check formatting
+        run: cargo fmt --check
+      - name: Check clippy
+        run: cargo clippy --no-deps --all-targets
+        if: ${{ !cancelled() }}
+
+  read-msrv:
+    runs-on: ubuntu-22.04
+    outputs: 
+      msrv: ${{ steps.read-declared-msrv.outputs.msrv }}
+    steps:
+      - uses: actions/checkout@v4
+      - id: read-declared-msrv
+        name: Read msrv from Cargo.toml rust_version field
+        run: echo "msrv=$(cargo metadata  --no-deps --format-version  1 | jq -r '.packages[] | select(.name = "scryer-prolog") | ."rust_version"')" >>  "$GITHUB_OUTPUT"
+          
+
+  build-test:
+    runs-on: ${{ matrix.os }}
+    needs: [read-msrv]
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          # operating systems
+          - { os: windows-latest, rust-version: stable,  target: 'x86_64-pc-windows-msvc',   publish: true }
+          - { os: macos-latest,   rust-version: stable,  target: 'x86_64-apple-darwin',      publish: true }
+          # architectures
+          - { os: ubuntu-22.04,   rust-version: stable,  target: 'x86_64-unknown-linux-gnu', publish: true }
+          - { os: ubuntu-22.04,   rust-version: stable,  target: 'i686-unknown-linux-gnu',   publish: true }
+          # FIXME(issue #2138): run wasm tests, failing to run since https://github.com/mthom/scryer-prolog/pull/2137 removed wasm-pack
+          - { os: ubuntu-22.04,   rust-version: nightly, target: 'wasm32-unknown-unknown',   publish: true, args: '--no-default-features' , test-args: '--no-run --no-default-features', use_swap: true }
+          # Cargo.toml rust-version
+          - { os: ubuntu-22.04,   rust-version: "${{ needs.read-msrv.outputs.msrv }}" ,  target: 'x86_64-unknown-linux-gnu'}
+          - { os: ubuntu-22.04,   rust-version: beta,    target: 'x86_64-unknown-linux-gnu'}
+          - { os: ubuntu-22.04,   rust-version: nightly, target: 'x86_64-unknown-linux-gnu', miri: true, components: "miri"} 
+    defaults:
+      run:
+        shell: bash
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actionhippie/swap-space@v1
+        if: matrix.use_swap
+        with:
+          size: 10G
+      - name: Setup Rust
+        uses: ./.github/actions/setup-rust
+        with:
+          rust-version: ${{ matrix.rust-version }}
+          targets: ${{ matrix.target }}
+          cache-context: ${{ matrix.os }}
+          components: ${{ matrix.components }}
+
+      # Build and test.
+      - name: Build library
+        run: cargo build --all-targets --target ${{ matrix.target }} ${{ matrix.args }} --verbose
+      - name: Test
+        run: cargo test --target ${{ matrix.target }} ${{ matrix.test-args }} --all
+
+      - name: Check miri
+        if: matrix.miri
+        run: cargo miri test
+
+      # On stable rust builds, build a binary and publish as a github actions
+      # artifact. These binaries could be useful for testing the pipeline but
+      # are only retained by github for 90 days.
+      - name: Build release binary
+        if: matrix.publish
+        run: |
+          cargo rustc --target ${{ matrix.target }} ${{ matrix.args }} --verbose --bin scryer-prolog --release
+          echo "$PWD/target/release" >> $GITHUB_PATH
+
+      - name: Install cargo-deb for creating debian packages
+        if: ${{ matrix.publish && contains(matrix.target,  'linux') }}
+        run: cargo install cargo-deb --force --locked
+        
+      - name: Build release debian packages
+        if: ${{ matrix.publish && contains(matrix.target,  'linux') }}
+        run: cargo deb --target ${{ matrix.target }}
+        
+      - name: Publish release binary artifact
+        if: matrix.publish
+        uses: actions/upload-artifact@v4
+        with:
+          path: |
+            target/${{ matrix.target }}/release/scryer-prolog*
+            target/${{ matrix.target }}/debian/scryer-prolog*.deb
+          name: scryer-prolog_${{ matrix.os }}_${{ matrix.target }}
+
+  logtalk-test:
+    # if: false # uncomment to disable job
+    runs-on: ubuntu-22.04
+    needs: [build-test]
+    steps:
+      # Download prebuilt ubuntu binary from build-test job, setup logtalk
+      - uses: actions/download-artifact@v4
+        with:
+          name: scryer-prolog_ubuntu-22.04_x86_64-unknown-linux-gnu
+      - run: |
+          chmod +x release/scryer-prolog
+          echo "$PWD/release" >> "$GITHUB_PATH"
+      - name: Install Logtalk
+        uses: logtalk-actions/setup-logtalk@master
+        with:
+          logtalk-version: "3.70.0"
+          logtalk-tool-dependencies: false
+
+      # Run logtalk  tests.
+      - name: Run Logtalk's prolog compliance test suite
+        working-directory: ${{ env.LOGTALKUSER }}/tests/prolog/
+        run: |
+          pwd
+          scryerlgt -g '{ack(tester)},halt.'
+          logtalk_tester -p scryer -g "set_logtalk_flag(clean,off)" -w -t 360 \
+            -f xunit \
+            -s "$LOGTALKUSER/tests/prolog" \
+            || echo "::warning ::logtalk compliance suite failed"
+            # -u "https://github.com/LogtalkDotOrg/logtalk3/tree/$LOGTALK_GIT_HASH/tests/prolog/" \
+      - name: Publish Logtalk test logs
+        uses: actions/upload-artifact@v4
+        with:
+          name: logtalk-test-logs
+          path: '${{ env.LOGTALKUSER }}/tests/prolog/logtalk_tester_logs'
+      - name: Publish Logtalk test results artifact
+        uses: actions/upload-artifact@v4
+        with:
+          name: logtalk-test-results
+          path: '${{ env.LOGTALKUSER }}/tests/prolog/**/*.xml'
+      - name: Publish Logtalk test summary
+        uses: EnricoMi/publish-unit-test-result-action/composite@master
+        with:
+          check_name: Logtalk test summary
+          files: '${{ env.LOGTALKUSER }}/tests/prolog/**/*.xml'
+          fail_on: nothing
+          comment_mode: off
+
+  report:
+    runs-on: ubuntu-22.04
+    steps:
+      - uses: actions/checkout@v4
+      - name: Setup Rust
+        uses: ./.github/actions/setup-rust
+        with:
+          rust-version: stable
+          targets: x86_64-unknown-linux-gnu
+          cache-context: report
+      - name: Install CLI tools
+        run: |
+          cargo install cargo2junit --force
+          version=`yq -ptoml -oj -r '.package[] | select(.name == "iai-callgrind") | .version' Cargo.lock`
+          echo installing iai-callgrind "$version"
+          cargo install iai-callgrind-runner --force --version "$version"
+          sudo apt install valgrind -y
+
+      - name: Test and report
+        run: |
+          RUSTC_BOOTSTRAP=1 cargo test --all -- -Z unstable-options --format json --report-time | cargo2junit > cargo_test_results.xml
+      - name: Publish cargo test results artifact
+        uses: actions/upload-artifact@v4
+        with:
+          name: cargo-test-results
+          path: cargo_test_results.xml
+      - name: Publish cargo test summary
+        uses: EnricoMi/publish-unit-test-result-action/composite@master
+        with:
+          check_name: Cargo test summary
+          files: cargo_test_results.xml
+          fail_on: nothing
+          comment_mode: off
+
+      - run: cargo build --all-targets --release
+      - run: cargo test --bench setup --release
+      - run: cargo bench --bench run_iai -- --save-summary=json
+      - run: cargo bench --bench run_criterion
+      - run: cargo bench --bench run_criterion -- --profile-time 60
+
+      - name: Publish benchmark results
+        uses: actions/upload-artifact@v4
+        with:
+          name: benchmark-results
+          path: |
+            target/criterion/*
+            target/iai/*
+            target/benchmark_inference_counts.json
+
+  # Publish binaries when building for a tag
+  release:
+    runs-on: ubuntu-20.04
+    needs: [build-test]
+    if: startsWith(github.ref, 'refs/tags/v')
+    steps:
+      - uses: actions/download-artifact@v4
+      - name: Zip binaries for release
+        run: |
+          zip scryer-prolog_macos-latest.zip ./scryer-prolog_macos-latest_x86_64-apple-darwin/scryer-prolog
+          zip scryer-prolog_ubuntu-22.04_i686.zip ./scryer-prolog_ubuntu-22.04_i686-unknown-linux-gnu/release/scryer-prolog ./scryer-prolog_ubuntu-22.04_i686-unknown-linux-gnu/debian/scryer-prolog*.deb
+          zip scryer-prolog_ubuntu-22.04_x86_64.zip ./scryer-prolog_ubuntu-22.04_x86_64-unknown-linux-gnu/release/scryer-prolog ./scryer-prolog_ubuntu-22.04_x86_64-unknown-linux-gnu/debian/scryer-prolog*.deb
+          zip scryer-prolog_windows-latest.zip ./scryer-prolog_windows-latest_x86_64-pc-windows-msvc/scryer-prolog.exe
+          zip scryer-prolog_wasm32.zip ./scryer-prolog_ubuntu-22.04_wasm32-unknown-unknown/scryer-prolog.wasm
+      - name: Release
+        uses: softprops/action-gh-release@v2
+        with:
+          files: |
+            scryer-prolog_macos-latest.zip
+            scryer-prolog_ubuntu-22.04_i686.zip
+            scryer-prolog_ubuntu-22.04_x86_64.zip
+            scryer-prolog_windows-latest.zip
+            scryer-prolog_wasm32.zip

.github/workflows/docker-publish.yml

@@ -2,10 +2,10 @@ name: Docker Publish
 
 on:
   push:
-    tags: [ 'v*.*.*' ]
-
-env:
-  IMAGE_NAME: mjt128/scryer-prolog
+    branches: 
+      - 'master'
+    tags: 
+      - 'v*.*.*'
 
 jobs:
   build:
@@ -14,37 +14,46 @@ jobs:
 
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v3
+        uses: actions/checkout@v4
 
       # Workaround: https://github.com/docker/build-push-action/issues/461
       - name: Setup Docker buildx
-        uses: docker/setup-buildx-action@79abd3f86f79a9d68a23c75a09a9a85889262adf
+        # https://github.com/docker/setup-buildx-action
+        uses: docker/setup-buildx-action@v3
 
       # Login against Docker registry
-      # https://github.com/docker/login-action
       - name: Log into registry
-        uses: docker/login-action@28218f9b04b4f3f62068d7b6ce6ca5b26e35336c
+        # https://github.com/docker/login-action
+        uses: docker/login-action@v3
         with:
           username: ${{ secrets.DOCKERHUB_USERNAME }}
           password: ${{ secrets.DOCKERHUB_TOKEN }}
 
       # Extract Docker image tag from git tag. E.g. if git tag is "v0.19.1" then use 
-      # Docker image tag "0.19.1". Tag "latest" is automatically synced with newest
-      # version.
-      # https://github.com/docker/metadata-action
+      # Docker image tag "0.19.1". The "latest" tag reflects the most recent build on
+      # master.
       - name: Extract Docker metadata
         id: meta
-        uses: docker/metadata-action@98669ae865ea3cffbcbaa878cf57c20bbf1c6c38
+        # https://github.com/docker/metadata-action
+        uses: docker/metadata-action@v5
         with:
-          images: docker.io/${{ env.IMAGE_NAME }}
+          images: docker.io/${{ secrets.DOCKERHUB_USERNAME }}/scryer-prolog
           tags: |
             type=semver,pattern={{version}}
+            type=raw,value=latest,enable={{is_default_branch}}
+            # type=raw,value=latest,enable=${{ github.ref == format('refs/heads/{0}', 'master') }}
 
       # Build and push Docker image with Buildx
-      # https://github.com/docker/build-push-action
       - name: Build and push Docker image
         id: build-and-push
-        uses: docker/build-push-action@ad44023a93711e3deb337508980b4b5e9bcdc5dc
+        # https://github.com/docker/build-push-action
+        uses: docker/build-push-action@v6
+        # v4 adds SLSA Provenance attestation which is 
+        # - unsupported by AWS Lambda 
+        # - limited support by Google Cloud Run 
+        #   > If deploying a multi-architecture image, the manifest list must include linux/amd64.
+        # see https://github.com/docker/build-push-action/releases/tag/v4.0.0
+        # we might want to disable this if its a problem for someone
         with:
           context: .
           push: true

.github/workflows/test.yml

@@ -1,71 +0,0 @@
-name: Test
-on: [push, pull_request]
-
-jobs:
-  build:
-    runs-on: ${{ matrix.os }}
-    strategy:
-      matrix:
-        os: [ubuntu-20.04, macos-10.15]
-        rust-version: [stable, beta]
-    steps:
-      - name: Checkout sources
-        uses: actions/checkout@v2
-      - name: Install Rust
-        uses: actions-rs/toolchain@v1
-        with:
-          profile: minimal
-          toolchain: ${{ matrix.rust-version }}
-          override: true
-      - name: Build lib
-        uses: actions-rs/cargo@v1
-        with:
-          command: rustc
-          args: --verbose --lib -- -D warnings
-      - name: Build bin
-        uses: actions-rs/cargo@v1
-        with:
-          command: rustc
-          args: --verbose --bin scryer-prolog -- -D warnings
-      - name: Test
-        uses: actions-rs/cargo@v1
-        with:
-          command: test
-          args: --verbose --all
-      - name: Num tests
-        uses: actions-rs/cargo@v1
-        continue-on-error: true
-        with:
-          command: test
-          args: --verbose --all --no-default-features --features num
-  msrv:
-    runs-on: ${{ matrix.os }}
-    strategy:
-      matrix:
-        os: [ubuntu-20.04, macos-10.15]
-    steps:
-      - name: Checkout sources
-        uses: actions/checkout@v2
-      - name: Install cargo-msrv
-        uses: baptiste0928/cargo-install@v1.1.0
-        with:
-          crate: cargo-msrv
-      - name: Verify MSRV
-        run: cargo msrv --verify
-  windows:
-    runs-on: windows-latest
-    defaults:
-      run:
-        shell: msys2 {0}
-    steps:
-      - name: Setup MSYS2
-        uses: msys2/setup-msys2@v2
-        with:
-          update: true
-          install: >-
-            base-devel
-            mingw-w64-x86_64-rust
-      - name: Checkout sources
-        uses: actions/checkout@v3
-      - name: Test on Windows
-        run: cargo test --verbose --all

.gitignore

@@ -1,5 +1,5 @@
 src/static_atoms.rs
 target/
-
+.direnv/
 
 

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "scryer-prolog"
-version = "0.9.1"
+version = "0.10.0"
 authors = ["Mark Thom <markjordanthom@gmail.com>"]
 edition = "2021"
 description = "A modern Prolog implementation written mostly in Rust."
@@ -10,67 +10,159 @@ license = "BSD-3-Clause"
 keywords = ["prolog", "prolog-interpreter", "prolog-system"]
 categories = ["command-line-utilities"]
 build = "build/main.rs"
-rust-version = "1.61"
+# Remember to check CI
+rust-version = "1.85"
+
+[lib]
+crate-type = ["cdylib", "rlib"]
 
 [features]
-default = ["rug"]
+default = ["ffi", "repl", "hostname", "tls", "http", "crypto-full"]
+ffi = ["dep:libffi"]
+repl = ["dep:crossterm", "dep:ctrlc", "dep:rustyline"]
+hostname = ["dep:hostname"]
+tls = ["dep:native-tls"]
+http = ["dep:warp", "dep:reqwest"]
+crypto-full = []
+
+[lints.rust]
+unexpected_cfgs = { level = "deny", check-cfg = [
+    'cfg(rust_version, values("1.87.0"))',
+] }
+
 
 [build-dependencies]
-indexmap = "1.0.2"
-proc-macro2 = "1.0.36"
-quote = "1.0.15"
-strum = "0.23"
-strum_macros = "0.23"
-syn = { version = "1.0.88", features = ['full', 'visit', 'extra-traits'] }
-to-syn-value = "0.1.0"
-to-syn-value_derive = "0.1.0"
+indexmap = "2.3.0"
+proc-macro2 = "1.0.86"
+quote = "1.0.36"
+strum = "0.26"
+strum_macros = "0.26"
+syn = { version = "2.0.72", features = ['full', 'visit', 'extra-traits'] }
+to-syn-value = "0.1.1"
+to-syn-value_derive = "0.1.1"
+version_check = "0.9.5"
 walkdir = "2"
 
 [dependencies]
+arcu = { version = "0.1.2", features = ["thread_local_counter"] }
+base64 = "0.22.1"
+bit-set = "0.8.0"
+bitvec = "1"
+blake2 = "0.10.6"
+bytes = "1"
+chrono = "0.4.38"
 cpu-time = "1.0.0"
-crossterm = "0.20.0"
+crrl = "0.9.0"
+dashu = { version = "0.4.2", features = ["rand"] }
+derive_more = "0.99.18"
 dirs-next = "2.0.0"
-divrem = "0.1.0"
-fxhash = "0.2.1"
-git-version = "0.3.4"
-hostname = "0.3.1"
-indexmap = "1.0.2"
-lazy_static = "1.4.0"
-lexical = "5.2.2"
-libc = "0.2.62"
-modular-bitfield = "0.11.2"
-ctrlc = "3.2.2"
-ordered-float = "2.6.0"
-phf = { version = "0.9",  features = ["macros"] }
-ref_thread_local = "0.0.0"
-rug = { version = "1.15.0", optional = true }
-rustyline = "9.0.0"
-ring = "0.16.13"
-ripemd160 = "0.8.0"
-sha3 = "0.8.2"
-blake2 = "0.8.1"
-crrl ="0.2.0"
-native-tls = "0.2.4"
-chrono = "0.4.11"
-select = "0.4.3"
-roxmltree = "0.11.0"
-base64 = "0.12.3"
-smallvec = "1.8.0"
-sodiumoxide = "0.2.6"
-static_assertions = "1.1.0"
-ryu = "1.0.9"
-hyper = { version = "0.14", features = ["full"] }
-hyper-tls = "0.5.0"
-tokio = { version = "1", features = ["full"] }
+divrem = "1.0.0"
 futures = "0.3"
+fxhash = "0.2.1"
+git-version = "0.3.9"
+indexmap = "2.3.0"
+lazy_static = "1.5.0"
+lexical = "7.0.4"
+libc = "0.2.155"
+libloading = "0.8"
+scryer-modular-bitfield = "0.11.4"
+num-order = { version = "1.2.0" }
+ordered-float = "5.0.0"
+phf = { version = "0.11", features = ["macros"] }
+puruspe = "0.4.1"
+rand = "0.8.5"
+ring = { version = "0.17.8", features = ["wasm32_unknown_unknown_js"] }
+ripemd = "0.1.3"
+roxmltree = "0.20.0"
+ryu = "1.0.18"
+sha3 = "0.10.8"
+smallvec = "1.13.2"
+static_assertions = "1.1.0"
+
+scraper = { version = "0.23.1", default-features = false, features = [
+    "errors",
+] }
+ego-tree = "0.10.0"
+
+
+serde_json = "1.0.122"
+serde = "1.0.204"
+parking_lot = "0.12.4"
+
+[target.'cfg(not(target_arch = "wasm32"))'.dependencies]
+crossterm = { version = "0.28.1", optional = true }
+ctrlc = { version = "3.4.4", optional = true }
+hostname = { version = "0.4.0", optional = true }
+libffi = { version = "4.0.0", optional = true }
+native-tls = { version = "0.2.12", optional = true }
+# the version requirement of reqwest is kept low for compatibility with old deno versions
+# that pin reqwest to 0.11.20
+reqwest = { version = "0.11.0", optional = true }
+rustyline = { version = "14.0.0", optional = true }
+tokio = { version = "1.39.2", features = ["full"] }
+warp = { version = "0.3.7", features = ["tls"], optional = true }
+
+[target.'cfg(target_arch = "wasm32")'.dependencies]
+getrandom = { version = "0.2.15", features = ["js"] }
+tokio = { version = "1.39.2", features = [
+    "sync",
+    "macros",
+    "io-util",
+    "rt",
+    "time",
+] }
+
+[target.'cfg(all(target_arch = "wasm32", target_os = "unknown"))'.dependencies]
+console_error_panic_hook = "0.1"
+wasm-bindgen = "0.2.92"
+wasm-bindgen-futures = "0.4"
+serde-wasm-bindgen = "0.6"
+web-sys = { version = "0.3", features = [
+    "Document",
+    "Window",
+    "Element",
+    "Performance",
+] }
+js-sys = "0.3"
+ouroboros = "0.18"
 
 [dev-dependencies]
-assert_cmd = "1.0.3"
-predicates-core = "1.0.2"
-serial_test = "0.5.1"
+current_platform = "0.2.0"
+maplit = "1.0.2"
+serial_test = "3.1.1"
 
-[patch.crates-io]
-modular-bitfield = { git = "https://github.com/mthom/modular-bitfield" }
+[target.'cfg(not(all(target_arch = "wasm32", target_os = "unknown")))'.dev-dependencies]
+criterion = "0.5.1"
+iai-callgrind = "0.12.1"
+trycmd = "0.15.6"
+
+[target.'cfg(not(any(target_os = "windows", all(target_arch = "wasm32", target_os = "unknown"))))'.dev-dependencies]
+pprof = { version = "0.13.0", features = ["criterion", "flamegraph"] }
+
+[profile.bench]
+lto = true
+opt-level = 3
 
 [profile.release]
-debug = true
+lto = true
+opt-level = 3
+debug = 2
+
+[profile.wasm-dev]
+inherits = "dev"
+opt-level = 1
+lto = "off"
+
+[profile.wasm-release]
+inherits = "release"
+lto = "off"
+panic = "abort"
+codegen-units = 256
+
+[[bench]]
+name = "run_criterion"
+harness = false
+
+[[bench]]
+name = "run_iai"
+harness = false

Dockerfile

@@ -1,18 +1,19 @@
 # See https://github.com/LukeMathWalker/cargo-chef
-ARG RUST_VERSION=1.60-buster
-FROM rust:${RUST_VERSION} as planner
+ARG DEBIAN_RELEASE=bookworm
+ARG RUST_VERSION=1-${DEBIAN_RELEASE}
+FROM rust:${RUST_VERSION} AS planner
 WORKDIR /scryer-prolog
 RUN cargo install cargo-chef 
 COPY . .
 RUN cargo chef prepare --recipe-path recipe.json
 
-FROM rust:${RUST_VERSION} as cacher
+FROM rust:${RUST_VERSION} AS cacher
 WORKDIR /scryer-prolog
 RUN cargo install cargo-chef
 COPY --from=planner /scryer-prolog/recipe.json recipe.json
 RUN cargo chef cook --release --recipe-path recipe.json
 
-FROM rust:${RUST_VERSION} as builder
+FROM rust:${RUST_VERSION} AS builder
 WORKDIR /scryer-prolog
 COPY . .
 # Copy over the cached dependencies
@@ -20,7 +21,14 @@ COPY --from=cacher /scryer-prolog/target target
 COPY --from=cacher $CARGO_HOME $CARGO_HOME
 RUN cargo build --release --bin scryer-prolog
 
-FROM debian:stable-slim
+FROM debian:${DEBIAN_RELEASE}-slim
+RUN apt-get update \
+    && apt-get install -y --no-install-recommends openssl \
+    && apt-get clean -y \
+    && rm -rf /var/lib/apt/lists/*
 COPY --from=builder /scryer-prolog/target/release/scryer-prolog /usr/local/bin
 ENV RUST_BACKTRACE=1
+# Sanity check the binary: if it can't be executed (e.g. if there are missing libraries) 
+# then fail the build
+RUN scryer-prolog --version
 ENTRYPOINT ["/usr/local/bin/scryer-prolog"]

INDEX.dj

@@ -0,0 +1,80 @@
+# Scryer Prolog
+
+```
+?- append("Hello, ", X, "Hello, Scryer Prolog!").
+   X = "Scryer Prolog!".
+```
+
+![scryer](scryer.png){width=128 style=float:right;} [Scryer Prolog](https://github.com/mthom/scryer-prolog) is a free software ISO Prolog system intended to be an industrial
+strength production environment *and* a testbed for bleeding edge research in
+logic and constraint programming.
+
+Some of the Scryer Prolog features are:
+
+* ISO standard compliant
+* Integrated constraint programming libraries: [clp(B)](/clpb.html), [clp(Z)](/clpz.html).
+* [Definite Clause Grammars](/dcgs.html)
+* Coroutining support ([`dif/2`](/dif.html), [`freeze/2`](/freeze.html), ...)
+* [Tabling and SLG resolution](/tabling.html)
+* Compact string representation
+* Network libraries ([TCP sockets](/sockets.html), [HTTP server](/http/http_server.html), [HTTP client](/http/http_open.html), ...)
+* [Cryptographical predicates](/crypto.html)
+* [Foreign Function Interface](/ffi.html)
+* WebAssembly support
+* Usable as a library
+* WAM based engine, cross-platform made in Rust
+* _and more..._
+
+Try Scryer Prolog without any installation! Use [Scryer Playground](https://play.scryer.pl), which uses the WASM version of Scryer Prolog.
+
+## What is Prolog?
+
+Prolog is a logic programming language created by [Alain Colmerauer](https://en.wikipedia.org/wiki/Alain_Colmerauer) and [Robert Kowalski](https://en.wikipedia.org/wiki/Robert_Kowalski) in 1972.
+The idea behind Prolog is try to express a task in language similar to First Order Logic.
+Prolog systems include _unification_ and _non-determinism_ as key concepts upon which we build programs.
+
+A Prolog program is made up of predicates which define a relation between its arguments. A predicate
+is made from clauses. A clause can be either a fact or a rule. There's also a toplevel, which we
+can use to ask and reason about our task.
+
+It's still to this day one of the best examples and one of the most popular languages in the field
+of logic programming. That's because Prolog allows us to elegantly solve many tasks with short and
+general programs.
+
+If you want a more detailed description of Prolog, check [A Tour of Prolog](https://www.youtube.com/watch?v=8XUutFBbUrg).
+
+If you want to learn more about Prolog history, check the videos [l'Aventure Prolog](https://www.youtube.com/watch?v=74Ig_QKndvE) and [50 years of Prolog and beyond](https://prologyear.logicprogramming.org/videos/PrologDay_Session_1_talk.mp4).
+
+## Where can I learn Prolog?
+
+There are a lot of classical Prolog books. Those books can teach you the basics of Prolog. Some
+examples are: _The Art of Prolog (Shapiro)_, _Programming in Prolog (Clocksin, Mellish)_ and _The Craft
+of Prolog (O'Keefe)_. However, most of them are not updated to _modern_ Prolog.
+We recommend _[The Power of Prolog (Markus Triska)](https://www.metalevel.at/prolog)_ for modern Prolog. For reference about
+the builtin Prolog modules and libraries in Scryer, check the documentation site. It's this!
+
+## Downloads
+
+The latest version of Scryer Prolog is *0.9.4*. And it's already useful for lots of tasks.
+
+| Windows (64 bits) | [Download](https://scryerprologrelease.blob.core.windows.net/release-094/scryer-prolog_windows-latest_x86_64-pc-windows-msvc.zip) |
+| macOS (Intel) | [Download](https://scryerprologrelease.blob.core.windows.net/release-094/scryer-prolog_macos-11_x86_64-apple-darwin.zip) |
+| macOS (ARM) | [Download](https://scryerprologrelease.blob.core.windows.net/release-094/scryer-prolog-macos-arm.zip) |
+| Linux (Ubuntu 20.04, 64 bits) | [Download](https://scryerprologrelease.blob.core.windows.net/release-094/scryer-prolog_ubuntu-20.04_x86_64-unknown-linux-gnu.zip) |
+| Linux (Ubuntu 22.04, 64 bits) | [Download](https://scryerprologrelease.blob.core.windows.net/release-094/scryer-prolog_ubuntu-22.04_x86_64-unknown-linux-gnu.zip) |
+| Linux (Ubuntu 22.04, 32 bits) | [Download](https://scryerprologrelease.blob.core.windows.net/release-094/scryer-prolog_ubuntu-22.04_i686-unknown-linux-gnu.zip) |
+
+Scryer Prolog can also be compiled from source, instructions are on the [GitHub README](https://github.com/mthom/scryer-prolog). It runs on Linux, macOS and Windows. Other operating systems may work but they're not regularly tested.
+
+If you're in Linux, maybe your distribution already has an Scryer Prolog package.
+
+There's also a [Docker image](https://github.com/mthom/scryer-prolog#docker-install) available.
+
+## Support and discussions
+
+If Scryer Prolog crashes or yields unexpected errors, consider filing
+an [issue](https://github.com/mthom/scryer-prolog/issues).
+
+To get in touch with the Scryer Prolog community, participate in
+[discussions](https://github.com/mthom/scryer-prolog/discussions)
+or visit our #scryer IRC channel on [Libera](https://libera.chat)!

README.md

@@ -1,4 +1,3 @@
-
 # Scryer Prolog
 
 Scryer Prolog aims to become to ISO Prolog what GHC is to Haskell: an open
@@ -6,14 +5,20 @@ source industrial strength production environment that is also a
 testbed for bleeding edge research in logic and constraint
 programming, which is itself written in a high-level language.
 
+**Scryer Prolog passes all tests** of
+[syntactic conformity](https://www.complang.tuwien.ac.at/ulrich/iso-prolog/conformity_testing),
+[`variable_names/1`](https://www.complang.tuwien.ac.at/ulrich/iso-prolog/variable_names) and
+[`dif/2`](https://www.complang.tuwien.ac.at/ulrich/iso-prolog/dif).
+
+The homepage of the project is: [**https://www.scryer.pl**](https://www.scryer.pl)
+
 ![Scryer Logo: Cryer](logo/scryer.png)
 
 ## Phase 1
 
 Produce an implementation of the Warren Abstract Machine in Rust, done
 according to the progression of languages in [Warren's Abstract
-Machine: A Tutorial
-Reconstruction](http://wambook.sourceforge.net/wambook.pdf).
+Machine: A Tutorial Reconstruction](https://github.com/mthom/scryer-prolog/blob/master/wambook/wambook.pdf).
 
 Phase 1 has been completed in that Scryer Prolog implements in some form
 all of the WAM book, including lists, cuts, Debray allocation, first
@@ -44,7 +49,7 @@ Extend Scryer Prolog to include the following, among other features:
   - [x] Support for `attribute_goals/2` and `project_attributes/2`
   - [x] `call_residue_vars/2`
 - [x] `if_/3` and related predicates, following the developments of the
-      paper "Indexing `dif/2`".
+      paper "[Indexing `dif/2`](https://arxiv.org/abs/1607.01590)".
 - [x] All-solutions predicates (`findall/{3,4}`, `bagof/3`, `setof/3`, `forall/2`).
 - [x] Clause creation and destruction (`asserta/1`, `assertz/1`,
       `retract/1`, `abolish/1`) with logical update semantics.
@@ -52,9 +57,9 @@ Extend Scryer Prolog to include the following, among other features:
       `bb_put/2` (non-backtrackable) and `bb_b_put/2`
       (backtrackable).
 - [x] Delimited continuations based on reset/3, shift/1 (documented in
-      "Delimited Continuations for Prolog").
+      "[Delimited Continuations for Prolog](https://biblio.ugent.be/publication/5646080/file/5646081)").
 - [x] Tabling library based on delimited continuations
-      (documented in "Tabling as a Library with Delimited Control").
+      (documented in "[Tabling as a Library with Delimited Control](https://biblio.ugent.be/publication/6880648/file/6885145.pdf)").
 - [x] A _redone_ representation of strings as difference lists of
       characters, using a packed internal representation.
 - [x] clp(B) and clp(ℤ) as builtin libraries.
@@ -66,10 +71,10 @@ Extend Scryer Prolog to include the following, among other features:
   - [ ] Replacing choice points pivoting on inlined semi-deterministic predicates
         (`atom`, `var`, etc) with if/else ladders. (_in progress_)
   - [ ] Inlining all built-ins and system call instructions.
-  - [ ] Greatly reducing the number of instructions used to compile disjunctives.
-  - [ ] Storing short atoms to heap cells without writing them to the atom table.
+  - [x] Greatly reducing the number of instructions used to compile disjunctives.
+  - [x] Storing short atoms to heap cells without writing them to the atom table.
 - [ ] A compacting garbage collector satisfying the five properties of
-      "Precise Garbage Collection in Prolog." (_in progress_)
+      "[Precise Garbage Collection in Prolog](https://www.complang.tuwien.ac.at/ulrich/papers/PDF/2008-ciclops.pdf)." (_in progress_)
 - [ ] Mode declarations.
 
 ## Phase 3
@@ -88,12 +93,12 @@ nice to have in the future. They'd make a good project for anyone wanting
 to contribute code to Scryer Prolog.
 
 1. Implement the global analysis techniques described in Peter van
-Roy's thesis, "Can Logic Programming Execute as Fast as Imperative
-Programming?"
+Roy's thesis, "[Can Logic Programming Execute as Fast as Imperative
+Programming?](https://www.info.ucl.ac.be/~pvr/Peter.thesis/Peter.thesis.html)"
 
 2. Add unum representation and arithmetic, using either an existing
 unum implementation or an ad hoc one. Unums are described in
-Gustafson's book "The End of Error."
+Gustafson's book "[The End of Error](http://www.johngustafson.net/unums.html)."
 
 3. Add concurrent tables to manage shared references to atoms and
 strings.
@@ -102,10 +107,17 @@ strings.
 
 ## Installing Scryer Prolog
 
-### Native Install
+### Binaries
+
+Precompiled binaries for several platforms are available for download
+at:
+
+**https://github.com/mthom/scryer-prolog/releases/latest**
+
+### Native Compilation
 
 First, install the latest stable version of
-[Rust](https://www.rust-lang.org/en-US/install.html) using your
+[Rust](https://www.rust-lang.org/tools/install) using your
 preferred method. Scryer tends to use features from newer Rust
 releases, whereas Rust packages in Linux distributions, Macports,
 etc. tend to lag behind. [rustup](http://rustup.rs) will keep your
@@ -113,17 +125,51 @@ Rust updated to the latest stable release; any existing Rust
 distribution should be uninstalled from your system before rustup is
 used.
 
-Currently the only way to install the latest version of Scryer is to
-clone directly from this git repository, which can be done as follows:
+
+> [!NOTE]
+> The minimum rust toolchain version required can be found in the [Cargo.toml](https://github.com/mthom/scryer-prolog/blob/master/Cargo.toml#L13)
+under the `package.rust-version` key.
+> The accuracy of this value is validated in CI
+
+### From a local git checkout
 
 ```
 $> git clone https://github.com/mthom/scryer-prolog
 $> cd scryer-prolog
-$> cargo run [--release]
+$> cargo build --release
 ```
 
-The optional `--release` flag will perform various optimizations,
-producing a faster executable.
+The `--release` flag performs various optimizations, producing a
+faster executable.
+
+After compilation, the executable `scryer-prolog` is available in the
+directory `target/release` and can be invoked to run the system.
+
+### Via `cargo install`
+
+#### From git
+
+```
+cargo install --locked --git https://github.com/mthom/scryer-prolog.git
+```
+
+Afterwards the `scryer-prolog` binary will be in the `$HOME/.cargo/bin` directory which is usually added to your PATH 
+during the installation of the rust toolchain.
+
+#### From Crates.io [![Crates.io Version](https://img.shields.io/crates/v/scryer-prolog)](https://crates.io/crates/scryer-prolog) ![Crates.io MSRV](https://img.shields.io/crates/msrv/scryer-prolog)
+
+> [!NOTE]
+> The lates crates.io release can be significantly behind the version available in the git repository
+> The crates.io badge in this sections title is a link to the crates.io page.
+> The msrv badge in the section title referece to the minimum rust toolchain version required to compile the latest crates.io release
+
+`scryer-prolog` is also release on crates.io and can be installed with
+
+```
+cargo install --locked scryer-prolog
+```
+
+### Caveats for Windows
 
 On Windows, Scryer Prolog is easier to build inside a [MSYS2](https://www.msys2.org/)
 environment as some crates may require native C compilation. However,
@@ -136,10 +182,119 @@ light.exe scryer-prolog.wixobj
 ```
 It will generate a very basic MSI file which installs the main executable and a shortcut in the Start Menu. It can be installed with a double-click. To uninstall, go to the Control Panel and uninstall as usual.
 
-Scryer Prolog must be built with **Rust 1.57 and up**.
+### Building WebAssembly
+
+Scryer Prolog has basic WebAssembly support. You can follow `wasm-pack`'s [official instructions](https://rustwasm.github.io/docs/wasm-pack/quickstart.html) to install `wasm-pack` and build it in any way you like.
+
+However, none of the [default features](https://doc.rust-lang.org/cargo/reference/features.html#the-default-feature) are currently supported. The preferred way of disabling them is passing [extra options](https://rustwasm.github.io/wasm-pack/book/commands/build.html#extra-options) to `wasm-pack`.
+
+For example, if you want a minimal working package without using any bundler like `webpack`, you can do this:
+```
+wasm-pack build --target web -- --no-default-features
+```
+Then a `pkg` directory will be created, containing everything you need for a webapp. You can test whether the package is successfully built by creating an html file, adapted from `wasm-bindgen`'s [official example](https://rustwasm.github.io/wasm-bindgen/examples/without-a-bundler.html) like this:
+
+```html
+<!DOCTYPE html>
+<html>
+
+<head>
+    <meta charset="UTF-8" />
+    <title>Scryer Prolog - Sudoku Solver Example</title>
+    <script type="module">
+        import initScryer, { MachineBuilder } from "./pkg/scryer_prolog.js";
+
+        // Initialize Scryer Prolog with WASM
+        const wasm = await fetch("./pkg/scryer_prolog_bg.wasm");
+        const module = await WebAssembly.compile(await wasm.arrayBuffer());
+        await initScryer(module);
+
+        // Set up the Prolog machine
+        const machine = new MachineBuilder().build();
+
+        // Knowledge base: Sudoku rules and problem definition
+        const kb = `
+        :- use_module(library(format)).
+        :- use_module(library(clpz)).
+        :- use_module(library(lists)).
+
+        sudoku(Rows) :-
+          length(Rows, 9), maplist(same_length(Rows), Rows),
+          append(Rows, Vs), Vs ins 1..9,
+          maplist(all_distinct, Rows),
+          transpose(Rows, Columns),
+          maplist(all_distinct, Columns),
+          Rows = [A,B,C,D,E,F,G,H,I],
+          blocks(A, B, C),
+          blocks(D, E, F),
+          blocks(G, H, I).
+
+        blocks([], [], []).
+        blocks([A,B,C|T1], [D,E,F|T2], [G,H,I|T3]) :-
+          all_distinct([A,B,C,D,E,F,G,H,I]),
+          blocks(T1, T2, T3).
+
+        problem(1, [[_,_,_,_,_,_,_,_,_],
+                    [_,_,_,_,_,3,_,8,5],
+                    [_,_,1,_,2,_,_,_,_],
+                    [_,_,_,5,_,7,_,_,_],
+                    [_,_,4,_,_,_,1,_,_],
+                    [_,9,_,_,_,_,_,_,_],
+                    [5,_,_,_,_,_,_,7,3],
+                    [_,_,2,_,1,_,_,_,_],
+                    [_,_,_,_,4,_,_,_,9]]).
+      `;
+
+        machine.consultModuleString("user", kb);
+
+        // Run the query
+        const query = "problem(1, Rows), sudoku(Rows), maplist(portray_clause, Rows).";
+        const answers = machine.runQuery(query);
+
+        const formattedSolutions = [];
+
+        // Format the answers
+        for (const solution of answers) {
+            const rows = solution.bindings["Rows"].list;
+
+            const grid = rows.map(row =>
+                row.list.map(cell => cell.integer)
+            );
+
+            const formatted = grid.map(row => `[${row.join(", ")}]`).join("\n");
+            formattedSolutions.push(formatted);
+        }
+
+        // Output results
+        const solutionDiv = document.querySelector("#soduku-solution");
+        for (const solution of formattedSolutions) {
+            const newPre = document.createElement("pre");
+            newPre.textContent = solution;
+            solutionDiv.appendChild(newPre);
+        }
+    </script>
+</head>
+
+<body>
+    <p>Sudoku solver returns:</p>
+    <div id="soduku-solution">
+
+    </div>
+</body>
+
+</html>
+```
+
+Then you can serve it with your favorite http server like `python -m http.server` or `npx serve`, and access the page with your browser.
 
 ### Docker Install
 
+Pre-built [Docker images are available on Docker Hub](https://hub.docker.com/r/mjt128/scryer-prolog/tags).
+The `latest` tag reflects the state on `master`, which might be unstable. 
+There are also tags for Scryer releases 0.9.2 and up. 
+Note though, that the base images are not kept up to date at the moment, 
+so be wary of security vulnerabilities (see [#2646](https://github.com/mthom/scryer-prolog/issues/2646)).
+
 First, install [Docker](https://docs.docker.com/get-docker/) on Linux,
 Windows, or Mac.
 
@@ -228,6 +383,35 @@ To quit Scryer Prolog, use the standard predicate `halt/0`:
 ?- halt.
 ```
 
+### Starting Scryer Prolog
+
+Scryer Prolog can be started from the command line by specifying
+options, files and additional arguments. All components are optional:
+
+<pre>
+scryer-prolog [OPTIONS] [FILES] [-- ARGUMENTS]
+</pre>
+
+The supported options are:
+
+```
+   -h, --help             Display help message
+   -v, --version          Print version information and exit
+   -g, --goal GOAL        Run the query GOAL after consulting files
+   -f                     Fast startup. Do not load initialization file (~/.scryerrc)
+   --no-add-history       Prevent adding input to history file (~/.scryer_history)
+```
+
+All specified Prolog files are consulted.
+
+After Prolog files, application-specific arguments can be specified on
+the command line. These arguments can be accessed from within Prolog
+applications with the predicate&nbsp;`argv/1`, which yields the list
+of arguments represented as strings.
+
+Prolog files can also be turned into *shell&nbsp;scripts* as explained in
+https://github.com/mthom/scryer-prolog/issues/2170#issuecomment-1821713993.
+
 ### Dynamic operators
 
 Scryer supports dynamic operators. Using the built-in
@@ -278,14 +462,14 @@ innovations of Scryer Prolog. This means that terms which appear as
 lists of characters to Prolog programs are stored in packed
 UTF-8&nbsp;encoding by the engine.
 
-Without this innovation, storing a list of characters in memory
-would use one memory&nbsp;cell per character, one memory&nbsp;cell per
-list constructor, and one memory&nbsp;cell for each tail that occurs
-in the list. Since one memory&nbsp;cell takes 8&nbsp;bytes on 64-bit
-machines, the packed representation used by Scryer&nbsp;Prolog yields
-an up&nbsp;to **24-fold&nbsp;reduction** of memory usage, and
-corresponding reduction of memory&nbsp;accesses when creating and
-processing strings.
+Without this innovation, storing a list of characters in memory would
+use one WAM memory&nbsp;cell per character, one cell per list
+constructor, and one cell for each tail that occurs in the list. Since
+one cell takes 8&nbsp;bytes in the WAM as implemented by
+Scryer&nbsp;Prolog, the packed representation yields an up&nbsp;to
+**24-fold&nbsp;reduction** of memory usage, and corresponding
+reduction of memory&nbsp;accesses when creating and processing
+strings.
 
 Scryer Prolog's compact internal string representation makes it
 ideally suited for the use case Prolog was originally developed for:
@@ -345,7 +529,7 @@ is unified with a term that contains that variable as a proper
 subterm. For efficiency, the *occurs&nbsp;check* is omitted by default
 in Scryer&nbsp;Prolog and many other Prolog systems.
 
-In Scryer Prolog, performing unifications which succeed only if the
+In Scryer Prolog, unifications which succeed only if the
 *occurs&nbsp;check* is omitted yield *cyclic&nbsp;terms*, also called
 *rational&nbsp;trees*. For example:
 
@@ -539,7 +723,7 @@ The modules that ship with Scryer&nbsp;Prolog are also called
   Probabilistic predicates and random number generators.
 * [`http/http_open`](src/lib/http/http_open.pl) Open a stream to
   read answers from web&nbsp;servers. HTTPS is also supported.
-* [`http/http_server`](src/lib/http/http_server.pl) Runs a HTTP/1.1 and HTTP/2.0 web server. Uses [Hyper](https://hyper.rs) as a backend. Supports some query and form handling.
+* [`http/http_server`](src/lib/http/http_server.pl) Runs a HTTP/1.1 and HTTP/2.0 web server. Uses [Warp](https://github.com/seanmonstar/warp) as a backend. Supports some query and form handling.
 * [`sgml`](src/lib/sgml.pl)
   `load_html/3` and `load_xml/3` represent HTML and XML&nbsp;documents
   as Prolog&nbsp;terms for convenient and efficient reasoning. Use
@@ -577,9 +761,13 @@ The modules that ship with Scryer&nbsp;Prolog are also called
   public key signatures and signature verification with&nbsp;Ed25519,
   ECDH key&nbsp;exchange over Curve25519 (X25519), authenticated symmetric
   encryption with ChaCha20-Poly1305, and reasoning about elliptic curves.
+* [`process`](src/lib/process.pl)
+  Create and manage parallel processes.
 * [`uuid`](src/lib/uuid.pl) UUIDv4 generation and hex representation
 * [`tls`](src/lib/tls.pl)
   Predicates for negotiating TLS connections explicitly.
+* [`numerics/special_functions`](src/lib/numerics/special_functions.pl)
+  Predicates for erf, gamma, beta, and related special functions.
 * [`ugraphs`](src/lib/ugraphs.pl) Graph manipulation library
 * [`simplex`](src/lib/simplex.pl) Providing `assignment/2`,
   `transportation/4` and other predicates for solving linear
@@ -673,10 +861,54 @@ not need additional tools and formalisms for its application, and
 further, it encourages declarative reasoning that can in principle
 also be performed automatically.
 
+## Applications
+
+Scryer Prolog's strong commitment to the Prolog ISO standard makes it
+ideally suited for use in corporations and government&nbsp;agencies
+that are subject to strict regulations pertaining to interoperability,
+standards&nbsp;compliance and warranty.
+
+Successful existing applications of Scryer Prolog include:
+
+- [DocLog](https://github.com/aarroyoc/doclog) which generates
+  Scryer's own documentation and homepage
+- [Grants4Companies](https://arxiv.org/abs/2406.15293): reasoning
+  about business&nbsp;grants in the Austrian public&nbsp;administration
+- parts of the [precautionary](https://github.com/dcnorris/precautionary/tree/main/exec/prolog)
+  package for the analysis of dose-escalation trials in the
+  safety-critical and highly regulated domain of oncology
+  trial&nbsp;design, described in [*An Executable Specification of
+  Oncology Dose-Escalation Protocols with&nbsp;Prolog*](https://arxiv.org/abs/2402.08334)
+  and culminating in&nbsp;[**DEDUCTION**](https://codeberg.org/dcnorris/DEDUCTION)
+- semantic reasoning and queries in [AD4M](https://github.com/coasys/ad4m),
+  an agent-centric distributed application meta-ontology.
+
+Scryer Prolog is also very well suited for teaching and learning
+Prolog, and for testing syntactic conformance and hence portability of
+existing Prolog&nbsp;programs.
+
+## Scryer Prolog Meetups
+
+Scryer Prolog Meetups are an excellent opportunity to present and get
+to know the latest developments in Scryer&nbsp;Prolog and its
+applications, to exchange ideas about current&nbsp;plans and
+future&nbsp;directions, and to discuss projects and visions
+in&nbsp;person.
+
+- [Scryer Prolog Meetup 2023](https://hsd-pbsa.de/veranstaltung/scryer-prolog-meetup-2023/)
+  in Düsseldorf, Germany. Its [announcement](https://github.com/mthom/scryer-prolog/discussions/1813)
+  and [discussion](https://github.com/mthom/scryer-prolog/discussions/2160).
+- [Scryer Prolog Meetup 2024](https://www.digitalaustria.gv.at/wissenswertes/events/scryerprologmeetup2024)
+  in Vienna, Austria. Its [announcement and discussion](https://github.com/mthom/scryer-prolog/discussions/2377).
+- **Save the date:** The [Scryer Prolog Meetup 2025](https://hsd-pbsa.de/veranstaltung/scryer-prolog-meetup-2025/)
+  will take place on Nov.&nbsp;13th and 14th&nbsp;2025 in Düsseldorf, Germany.
+  Its [announcement](https://github.com/mthom/scryer-prolog/discussions/2948).
+
 ## Support and discussions
 
 If Scryer Prolog crashes or yields unexpected errors, consider filing
 an&nbsp;[issue](https://github.com/mthom/scryer-prolog/issues).
 
 To get in touch with the Scryer Prolog community, participate in
-[discussions](https://github.com/mthom/scryer-prolog/discussions)!
+[discussions](https://github.com/mthom/scryer-prolog/discussions)
+or visit our #scryer IRC channel on [Libera](https://libera.chat)!

benches/README.md

@@ -0,0 +1,95 @@
+# About benches
+
+The `benches` directory contains benchmarks that test scryer-prolog performance.
+
+Benchmarks are run via two harnesses:
+
+* `criterion` - criterion performs statistical analysis of benchmark runs and is
+  great for benchmarking locally.
+* `iai-callgrind` - this runs the benchmark with callgrind, which is able to
+  precisely track the number of instructions executed during the run. This is
+  especially helpful in a public CI runner context where neighboring VMs can
+  cause a very high wall time variance. While instructions executed is only
+  correlated with the desired metric (wall time), this is a good tradeoff for CI
+  where that metric is unreliable.
+
+Run benchmarks with the following commands:
+
+```
+cargo bench --bench run_criterion
+
+# run a particular criterion benchmark
+cargo bench --bench run_criterion -- <benchmark_name>
+
+# run in profiling mode which outputs flamegraphs. Set profile time in seconds:
+cargo bench --bench run_criterion -- --profile-time <time>
+
+# to run iai, you need valgrind installed and to install iai-callgrind-runner
+# at the same version as is in Cargo.toml:
+cargo install iai-callgrind-runner --version 0.7.3
+
+cargo bench --bench run_iai
+```
+
+For consistency, both runners -- `run_iai.rs` and `run_criterion.rs` -- import
+the same setup code from `setup.rs`.
+
+## Setup
+
+`setup.rs` contains the setup code to run benchmarks. `fn prolog_benches()` at
+the top of the file is where the benchmarks are defined.
+
+Benchmarks are organized around running queries against a prolog module file.
+Before a benchmark starts, `benchmark.setup()` is called which reads the module
+file and initializes a new `scryer_prolog::machine::Machine`.
+
+Each benchmark measurement is done by running a query against the machine. In
+the case of criterion each query is run many times, in the case of iai it's run
+once.
+
+## Adding benchmarks
+
+This design is meant to suppoort defining lots of benchmarks.
+
+To add a new benchmark:
+
+* Add a new file `benches/[module].pl` that contains setup prolog code. Import
+  libraries, define predicates, etc.
+* Add a new section in `setup.rs::prolog_benchmarks()` that refers to to the
+  file and write a query to be benchmarked.
+* If the query mutates the machine, then use `Strategy::Fresh` so the criterion
+  benchmark will recreate a new machine for each benchmark run, otherwise use
+  `Strategy::Reuse` which has lower overhead. (This is not used by the iai
+  benchmark because it only runs once anyway.)
+
+Some tips:
+
+* The goal of benchmarking is to know if a library or engine change improved
+  performance or not.
+* Once a benchmark is defined and named, avoid changing it's definition. If a
+  benchmark needs to change to be more useful, give the new definition a new
+  name instead. This will prevent charts from showing wild changes in
+  performance just because the definition changed (see previous).
+* Aim for queries to execute in less than 0.5s realtime. Longer runtimes make it
+  easier for humans to see big differences, but benchmarks either run 10x slower
+  (iai) or execute repeatedly to attain statistical significance (criterion) and
+  in both cases benchmarking queries that take longer than about 0.5s are
+  cumbersome to run.
+* Consider that the library runtime actually parses the text output of the top
+  level. So don't use custom outputs or it will fail to parse. Also keep the
+  output small so it doesn't just benchmark the ouput parsing code.
+* DO test the output of the benchmark run, we don't want to count broken
+  benchmarks.
+
+## CI
+
+Both benchmark harnesses are run in `.github/workflows/ci.yaml` in the `report`
+job, and the results are published as build artifacts.
+
+## Todo
+
+- [ ] Currently, the execution time to load a module is not benchmarked. It
+  would be nice to have at least one benchmark for loading a module (probably a
+  big one).
+- [ ] Write a new action that downloads the test and benchmark results
+  artifacts, plots them over time, and publishes a report to github pages.

benches/edges.pl

@@ -0,0 +1,130 @@
+:- use_module(library(clpb)).
+:- use_module(library(assoc)).
+:- use_module(library(lists)).
+:- use_module(library(pairs)).
+
+/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
+    Contiguous United States and DC as they appear in SGB:
+    http://www-cs-faculty.stanford.edu/~uno/sgb.html
+- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
+
+edge(al, fl).
+edge(al, ga).
+edge(al, ms).
+edge(al, tn).
+edge(ar, la).
+edge(ar, mo).
+edge(ar, ms).
+edge(ar, ok).
+edge(ar, tn).
+edge(ar, tx).
+edge(az, ca).
+edge(az, nm).
+edge(az, nv).
+edge(az, ut).
+edge(ca, nv).
+edge(ca, or).
+edge(co, ks).
+edge(co, ne).
+edge(co, nm).
+edge(co, ok).
+edge(co, ut).
+edge(co, wy).
+edge(ct, ma).
+edge(ct, ny).
+edge(ct, ri).
+edge(dc, md).
+edge(dc, va).
+edge(de, md).
+edge(de, nj).
+edge(de, pa).
+edge(fl, ga).
+edge(ga, nc).
+edge(ga, sc).
+edge(ga, tn).
+edge(ia, il).
+edge(ia, mn).
+edge(ia, mo).
+edge(ia, ne).
+edge(ia, sd).
+edge(ia, wi).
+edge(id, mt).
+edge(id, nv).
+edge(id, or).
+edge(id, ut).
+edge(id, wa).
+edge(id, wy).
+edge(il, in).
+edge(il, ky).
+edge(il, mo).
+edge(il, wi).
+edge(in, ky).
+edge(in, mi).
+edge(in, oh).
+edge(ks, mo).
+edge(ks, ne).
+edge(ks, ok).
+edge(ky, mo).
+edge(ky, oh).
+edge(ky, tn).
+edge(ky, va).
+edge(ky, wv).
+edge(la, ms).
+edge(la, tx).
+edge(ma, nh).
+edge(ma, ny).
+edge(ma, ri).
+edge(ma, vt).
+edge(md, pa).
+edge(md, va).
+edge(md, wv).
+edge(me, nh).
+edge(mi, oh).
+edge(mi, wi).
+edge(mn, nd).
+edge(mn, sd).
+edge(mn, wi).
+edge(mo, ne).
+edge(mo, ok).
+edge(mo, tn).
+edge(ms, tn).
+edge(mt, nd).
+edge(mt, sd).
+edge(mt, wy).
+edge(nc, sc).
+edge(nc, tn).
+edge(nc, va).
+edge(nd, sd).
+edge(ne, sd).
+edge(ne, wy).
+edge(nh, vt).
+edge(nj, ny).
+edge(nj, pa).
+edge(nm, ok).
+edge(nm, tx).
+edge(nv, or).
+edge(nv, ut).
+edge(ny, pa).
+edge(ny, vt).
+edge(oh, pa).
+edge(oh, wv).
+edge(ok, tx).
+edge(or, wa).
+edge(pa, wv).
+edge(sd, wy).
+edge(tn, va).
+edge(ut, wy).
+edge(va, wv).
+
+independent_set(G, *(NBs)) :-
+        findall(U-V, (edge(U, V),G@<U), Edges),
+        setof(U, V^(member(U-V, Edges);member(V-U, Edges)), Nodes),
+        pairs_keys_values(Pairs, Nodes, _),
+        list_to_assoc(Pairs, Assoc),
+        maplist(not_both(Assoc), Edges, NBs).
+
+not_both(Assoc, U-V, ~BU + ~BV) :-
+        get_assoc(U, Assoc, BU),
+        get_assoc(V, Assoc, BV).
+
+independent_set_count(G, Count) :- independent_set(G, Sat), sat_count(Sat, Count).

benches/numlist.pl

@@ -0,0 +1,2 @@
+:- use_module(library(between)).
+run_numlist(Upper, Head) :- numlist(1, Upper, L), L = [Head|_].

benches/run_criterion.rs

@@ -0,0 +1,46 @@
+#[cfg(not(all(target_arch = "wasm32", target_os = "unknown")))]
+use criterion::{criterion_group, criterion_main, BatchSize, Criterion};
+
+#[cfg(not(all(target_arch = "wasm32", target_os = "unknown")))]
+#[cfg(not(target_os = "windows"))]
+use pprof::criterion::{Output, PProfProfiler};
+
+#[cfg(not(all(target_arch = "wasm32", target_os = "unknown")))]
+mod setup;
+
+#[cfg(not(all(target_arch = "wasm32", target_os = "unknown")))]
+fn bench_criterion(c: &mut Criterion) {
+    for (&name, bench) in setup::prolog_benches().iter() {
+        match bench.strategy {
+            setup::Strategy::Fresh => c.bench_function(name, |b| {
+                b.iter_batched(|| bench.setup(), |mut r| r(), BatchSize::LargeInput)
+            }),
+            setup::Strategy::Reuse => c.bench_function(name, |b| b.iter(bench.setup())),
+        };
+    }
+}
+#[cfg(not(all(target_arch = "wasm32", target_os = "unknown")))]
+#[cfg(not(target_os = "windows"))]
+fn config() -> Criterion {
+    Criterion::default()
+        .sample_size(20)
+        .with_profiler(PProfProfiler::new(100, Output::Flamegraph(None)))
+}
+
+#[cfg(target_os = "windows")]
+fn config() -> Criterion {
+    Criterion::default().sample_size(20)
+}
+
+#[cfg(not(all(target_arch = "wasm32", target_os = "unknown")))]
+criterion_group!(
+    name = benches;
+    config = config();
+    targets = bench_criterion
+);
+
+#[cfg(not(all(target_arch = "wasm32", target_os = "unknown")))]
+criterion_main!(benches);
+
+#[cfg(all(target_arch = "wasm32", target_os = "unknown"))]
+fn main() {}

benches/run_iai.rs

@@ -0,0 +1,38 @@
+#[cfg(not(all(target_arch = "wasm32", target_os = "unknown")))]
+mod setup;
+
+#[cfg(not(all(target_arch = "wasm32", target_os = "unknown")))]
+mod iai {
+    use iai_callgrind::{library_benchmark, library_benchmark_group, main};
+
+    use scryer_prolog::LeafAnswer;
+
+    use super::setup;
+
+    #[library_benchmark]
+    #[bench::count_edges(setup::prolog_benches()["count_edges"].setup())]
+    #[bench::numlist(setup::prolog_benches()["numlist"].setup())]
+    #[bench::csv_codename(setup::prolog_benches()["csv_codename"].setup())]
+    fn bench(mut run: impl FnMut() -> Vec<LeafAnswer>) -> Vec<LeafAnswer> {
+        run()
+    }
+
+    library_benchmark_group!(
+        name = benches;
+        benchmarks = bench
+    );
+
+    main!(library_benchmark_groups = benches);
+
+    pub fn call_main() {
+        main()
+    }
+}
+
+#[cfg(not(all(target_arch = "wasm32", target_os = "unknown")))]
+fn main() {
+    iai::call_main();
+}
+
+#[cfg(all(target_arch = "wasm32", target_os = "unknown"))]
+fn main() {}

benches/setup.rs

@@ -0,0 +1,140 @@
+use std::{collections::BTreeMap, fs, path::Path};
+
+use maplit::btreemap;
+use scryer_prolog::{LeafAnswer, Machine, MachineBuilder, Term};
+
+pub fn prolog_benches() -> BTreeMap<&'static str, PrologBenchmark> {
+    [
+        (
+            "count_edges",                       // name of the benchmark
+            "benches/edges.pl", // name of the prolog module file to load. use the same file in multiple benchmarks
+            "independent_set_count(ky, Count).", // query to benchmark in the context of the loaded module. consider making the query adjustable to tune the run time to ~0.1s
+            Strategy::Reuse,
+            btreemap! { "Count" => Term::integer(2869176) },
+        ),
+        (
+            "numlist",
+            "benches/numlist.pl",
+            "run_numlist(1000000, Head).",
+            Strategy::Reuse,
+            btreemap! { "Head" => Term::integer(1) },
+        ),
+        (
+            "csv_codename",
+            "benches/csv.pl",
+            "get_codename(\"0020\",Name).",
+            Strategy::Reuse,
+            btreemap! { "Name" => Term::string("SPACE") },
+        ),
+    ]
+    .map(|b| {
+        (
+            b.0,
+            PrologBenchmark {
+                name: b.0,
+                filename: b.1,
+                query: b.2,
+                strategy: b.3,
+                bindings: b.4,
+            },
+        )
+    })
+    .into()
+}
+
+pub enum Strategy {
+    #[allow(dead_code)]
+    Fresh,
+    Reuse,
+}
+
+#[allow(dead_code)]
+pub struct PrologBenchmark {
+    pub name: &'static str,
+    pub filename: &'static str,
+    pub query: &'static str,
+    pub strategy: Strategy,
+    pub bindings: BTreeMap<&'static str, Term>,
+}
+
+impl PrologBenchmark {
+    pub fn make_machine(&self) -> Machine {
+        let program = fs::read_to_string(self.filename).unwrap();
+        let module_name = Path::new(self.filename)
+            .file_stem()
+            .and_then(|s| s.to_str())
+            .unwrap();
+        let mut machine = MachineBuilder::default().build();
+        machine.load_module_string(module_name, program);
+        machine
+    }
+
+    #[cfg(not(all(target_arch = "wasm32", target_os = "unknown")))]
+    pub fn setup(&self) -> impl FnMut() -> Vec<LeafAnswer> {
+        let mut machine = self.make_machine();
+        let query = self.query;
+        move || {
+            use criterion::black_box;
+            black_box(
+                machine
+                    .run_query(black_box(query))
+                    .collect::<Result<Vec<_>, _>>()
+                    .unwrap(),
+            )
+        }
+    }
+}
+
+#[cfg(test)]
+mod test {
+
+    #[test]
+    fn validate_benchmarks() {
+        use super::prolog_benches;
+        use scryer_prolog::LeafAnswer;
+        use std::{fmt::Write, fs};
+
+        struct BenchResult {
+            pub name: &'static str,
+            pub setup_inference_count: u64,
+            pub query_inference_count: u64,
+        }
+
+        let mut results: Vec<BenchResult> = vec![];
+
+        for (_, r) in prolog_benches() {
+            let mut machine = r.make_machine();
+            let setup_inference_count = machine.get_inference_count();
+
+            let result: Vec<_> = machine
+                .run_query(r.query)
+                .collect::<Result<_, _>>()
+                .unwrap();
+            let query_inference_count = machine.get_inference_count() - setup_inference_count;
+
+            let expected = [LeafAnswer::from_bindings(r.bindings.clone())];
+            assert_eq!(result, expected, "validating benchmark {}", r.name);
+
+            results.push(BenchResult {
+                name: r.name,
+                setup_inference_count,
+                query_inference_count,
+            })
+        }
+
+        let mut json: String = Default::default();
+        json.push('[');
+        for r in results {
+            json.push('\n');
+            write!(
+                json,
+                r#"{{"name":"{}","setup_inference_count":{},"query_inference_count":{}}},"#,
+                r.name, r.setup_inference_count, r.query_inference_count
+            )
+            .unwrap();
+        }
+        json.pop(); // trailing comma
+        json.push_str("\n]");
+        fs::write("target/benchmark_inference_counts.json", json).expect("Unable to write file");
+    }
+}

build/main.rs

@@ -5,43 +5,48 @@ use instructions_template::generate_instructions_rs;
 use static_string_indexing::index_static_strings;
 
 use std::env;
-use std::fs;
 use std::fs::File;
 use std::io::Write;
 use std::path::Path;
+use std::path::PathBuf;
+use std::path::MAIN_SEPARATOR_STR;
 use std::process::{Command, Stdio};
 
-fn find_prolog_files(libraries: &mut File, prefix: &str, current_dir: &Path) {
+fn find_prolog_files(path_prefix: &str, current_dir: &Path) -> Vec<(String, PathBuf)> {
+    let mut libraries = vec![];
+
     let entries = match current_dir.read_dir() {
         Ok(entries) => entries,
-        Err(_) => return,
+        Err(_) => return libraries,
     };
 
     for entry in entries.filter_map(Result::ok).map(|e| e.path()) {
         if entry.is_dir() {
             if let Some(file_name) = entry.file_name() {
-                let new_prefix = prefix.to_owned() + file_name.to_str().unwrap() + "/";
-                find_prolog_files(libraries, &new_prefix, &entry);
+                let file_name = file_name.to_str().unwrap();
+                let new_path_prefix = format!("{path_prefix}{file_name}/");
+                let new_libs = find_prolog_files(&new_path_prefix, &entry);
+                libraries.extend(new_libs);
             }
         } else if entry.is_file() {
             let ext = std::ffi::OsStr::new("pl");
             if entry.extension() == Some(ext) {
-                let contain = String::from_utf8(fs::read(&entry).unwrap()).unwrap();
                 let name = entry.file_stem().unwrap().to_str().unwrap();
+                let lib_name = format!("{path_prefix}{name}");
 
-                let line = format!(
-                    "        m.insert(\"{}\",\n{:?});\n",
-                    prefix.to_owned() + name,
-                    contain
-                );
-
-                libraries.write_all(line.as_bytes()).unwrap();
+                libraries.push((lib_name, entry));
             }
         }
     }
+
+    libraries
 }
 
 fn main() {
+    if version_check::is_min_version("1.87.0").unwrap_or(false) {
+        println!(r#"cargo:rustc-cfg=rust_version="1.87.0""#);
+    }
+
     let has_rustfmt = Command::new("rustfmt")
         .arg("--version")
         .stdin(Stdio::inherit())
@@ -55,19 +60,23 @@ fn main() {
     let out_dir = env::var("OUT_DIR").unwrap();
     let dest_path = Path::new(&out_dir).join("libraries.rs");
 
-    let mut libraries = File::create(&dest_path).unwrap();
-    let lib_path = Path::new("src/lib");
+    let mut libraries = File::create(dest_path).unwrap();
+    let lib_path = Path::new("src").join("lib");
 
-    libraries
-        .write_all(
-            b"ref_thread_local::ref_thread_local! {
-    pub(crate) static managed LIBRARIES: IndexMap<&'static str, &'static str> = {
-        let mut m = IndexMap::new();\n",
+    let constants = find_prolog_files("", &lib_path);
+
+    let out_dir = std::env::var("OUT_DIR").unwrap();
+
+    writeln!(libraries, "{{").unwrap();
+    for (name, lib_path) in constants {
+        let path = format!("{}{}", MAIN_SEPARATOR_STR, lib_path.display());
+        writeln!(
+            libraries,
+            "m.insert(\"{name}\", include_str!(concat!(env!(\"CARGO_MANIFEST_DIR\"), {path:?})));"
         )
         .unwrap();
-
-    find_prolog_files(&mut libraries, "", &lib_path);
-    libraries.write_all(b"\n        m\n    };\n}\n").unwrap();
+    }
+    writeln!(libraries, "}}").unwrap();
 
     let instructions_path = Path::new(&out_dir).join("instructions.rs");
     let mut instructions_file = File::create(&instructions_path).unwrap();

build/static_string_indexing.rs

@@ -1,7 +1,7 @@
 use proc_macro2::TokenStream;
-use syn::*;
 use syn::parse::*;
 use syn::visit::*;
+use syn::*;
 
 use indexmap::IndexSet;
 
@@ -11,7 +11,9 @@ struct StaticStrVisitor {
 
 impl StaticStrVisitor {
     fn new() -> Self {
-        Self { static_strs: IndexSet::new() }
+        Self {
+            static_strs: IndexSet::new(),
+        }
     }
 }
 
@@ -33,7 +35,7 @@ impl Parse for ReadHeapCellExprAndArms {
         arms.push(input.parse()?);
 
         while !input.is_empty() {
-            if let Ok(_) = input.parse::<Token![,]>() {}
+            let _ = input.parse::<Token![,]>();
             arms.push(input.parse()?);
         }
 
@@ -50,7 +52,7 @@ impl Parse for MacroFnArgs {
         }
 
         while !input.is_empty() {
-            if let Ok(_) = input.parse::<Token![,]>() {}
+            let _ = input.parse::<Token![,]>();
             args.push(input.parse()?);
         }
 
@@ -63,27 +65,37 @@ impl<'ast> Visit<'ast> for StaticStrVisitor {
         let Macro { path, .. } = m;
 
         if path.is_ident("atom") {
-            if let Some(Lit::Str(string)) = m.parse_body::<Lit>().ok() {
+            if let Ok(Lit::Str(string)) = m.parse_body::<Lit>() {
                 self.static_strs.insert(string.value());
             }
         } else if path.is_ident("read_heap_cell") || path.is_ident("match_untyped_arena_ptr") {
-            if let Some(m) = m.parse_body::<ReadHeapCellExprAndArms>().ok() {
+            if let Ok(m) = m.parse_body::<ReadHeapCellExprAndArms>() {
                 self.visit_expr(&m.expr);
 
                 for e in m.arms {
                     self.visit_arm(&e);
                 }
             }
-        } else {
-            if let Some(m) = m.parse_body::<MacroFnArgs>().ok() {
-                for e in m.args {
-                    self.visit_expr(&e);
-                }
+        } else if let Ok(m) = m.parse_body::<MacroFnArgs>() {
+            for e in m.args {
+                self.visit_expr(&e);
             }
         }
     }
 }
 
+const INLINED_ATOM_MAX_LEN: usize = 6;
+
+fn static_string_index(string: &str, index: usize) -> u64 {
+    if !string.is_empty() && string.len() <= INLINED_ATOM_MAX_LEN && !string.contains('\u{0}') {
+        let mut string_buf: [u8; 8] = [0u8; 8];
+        string_buf[..string.len()].copy_from_slice(string.as_bytes());
+        (u64::from_le_bytes(string_buf) << 1) | 1
+    } else {
+        (index << 1) as u64
+    }
+}
+
 pub fn index_static_strings(instruction_rs_path: &std::path::Path) -> TokenStream {
     use quote::*;
 
@@ -114,14 +126,19 @@ pub fn index_static_strings(instruction_rs_path: &std::path::Path) -> TokenStrea
         match file.read_to_string(&mut src) {
             Ok(_) => {}
             Err(e) => {
-                panic!("error reading file: {:?}", e);
+                panic!("error reading file: {e:?}");
             }
         }
 
         let syntax = match syn::parse_file(&src) {
             Ok(s) => s,
             Err(e) => {
-                panic!("parse error: {} in file {:?}", e, path);
+                println!("cargo::warning=parse error: {e} in file {path:?}");
+                syn::File {
+                    shebang: None,
+                    attrs: vec![],
+                    items: vec![],
+                }
             }
         };
         Ok(syntax)
@@ -145,33 +162,48 @@ pub fn index_static_strings(instruction_rs_path: &std::path::Path) -> TokenStrea
         visitor.visit_file(&syntax);
     }
 
-    match process_filepath(instruction_rs_path) {
-        Ok(syntax) => visitor.visit_file(&syntax),
-        Err(_) => {}
+    if let Ok(syntax) = process_filepath(instruction_rs_path) {
+        visitor.visit_file(&syntax)
     }
 
-    let indices = (0..visitor.static_strs.len()).map(|i| i << 3);
-    let indices_iter = indices.clone();
+    let mut static_str_keys = vec![];
+    let mut static_strs = vec![];
+    let mut static_str_indices = vec![];
 
-    let static_strs_len = visitor.static_strs.len();
-    let static_strs: &Vec<_> = &visitor.static_strs.into_iter().collect();
+    let indices: Vec<u64> = visitor
+        .static_strs
+        .iter()
+        .map(|string| {
+            let index = static_string_index(string, static_strs.len());
+
+            static_str_keys.push(string);
+
+            if index & 1 == 1 {
+                index
+            } else {
+                static_str_indices.push(index);
+                static_strs.push(string);
+                index
+            }
+        })
+        .collect();
+
+    let static_strs_len = static_strs.len();
 
     quote! {
-        use phf;
-
-        static STRINGS: [&'static str; #static_strs_len] = [
+        static STRINGS: [&str; #static_strs_len] = [
             #(
                 #static_strs,
             )*
         ];
 
-        #[macro_export]
         macro_rules! atom {
-            #((#static_strs) => { Atom { index: #indices_iter } };)*
+            #((#static_str_keys) => { Atom { index: #indices } };)*
+            ($name:literal) => {compile_error!(concat!("unknown static atom ", $name))};
         }
 
         pub static STATIC_ATOMS_MAP: phf::Map<&'static str, Atom> = phf::phf_map! {
-            #(#static_strs => { Atom { index: #indices } },)*
+            #(#static_strs => { Atom { index: #static_str_indices } },)*
         };
     }
 }

doclog.config.pl

@@ -0,0 +1,14 @@
+project_name("Scryer Prolog").
+readme_file("INDEX.dj").
+source_lib_folder("src/lib").
+websource("https://github.com/mthom/scryer-prolog/tree/master/src/lib").
+omit(["ops_and_meta_predicates.pl", "tabling"]).
+learn_pages_source_folder("learn").
+learn_pages_categories(["First steps", "Tutorials"]).
+learn_pages([
+		   page("Let's play Brisca", "Tutorials", "lets-play-brisca.dj"),
+		   page("Test page", "First steps", "test-page.dj")
+]).
+copy_file("logo/scryer.png", "scryer.png").
+copy_file("learn/Spanish_deck_Fournier.jpg", "learn/Spanish_deck_Fournier.jpg").
+copy_file("learn/brisca-interactive.png", "learn/brisca-interactive.png").

flake.lock

@@ -0,0 +1,82 @@
+{
+  "nodes": {
+    "flake-utils": {
+      "inputs": {
+        "systems": "systems"
+      },
+      "locked": {
+        "lastModified": 1731533236,
+        "narHash": "sha256-l0KFg5HjrsfsO/JpG+r7fRrqm12kzFHyUHqHCVpMMbI=",
+        "owner": "numtide",
+        "repo": "flake-utils",
+        "rev": "11707dc2f618dd54ca8739b309ec4fc024de578b",
+        "type": "github"
+      },
+      "original": {
+        "owner": "numtide",
+        "repo": "flake-utils",
+        "type": "github"
+      }
+    },
+    "nixpkgs": {
+      "locked": {
+        "lastModified": 1744868846,
+        "narHash": "sha256-5RJTdUHDmj12Qsv7XOhuospjAjATNiTMElplWnJE9Hs=",
+        "owner": "NixOS",
+        "repo": "nixpkgs",
+        "rev": "ebe4301cbd8f81c4f8d3244b3632338bbeb6d49c",
+        "type": "github"
+      },
+      "original": {
+        "owner": "NixOS",
+        "ref": "nixpkgs-unstable",
+        "repo": "nixpkgs",
+        "type": "github"
+      }
+    },
+    "root": {
+      "inputs": {
+        "flake-utils": "flake-utils",
+        "nixpkgs": "nixpkgs",
+        "rust-overlay": "rust-overlay"
+      }
+    },
+    "rust-overlay": {
+      "inputs": {
+        "nixpkgs": [
+          "nixpkgs"
+        ]
+      },
+      "locked": {
+        "lastModified": 1745289264,
+        "narHash": "sha256-7nt+UJ7qaIUe2J7BdnEEph9n2eKEwxUwKS/QIr091uA=",
+        "owner": "oxalica",
+        "repo": "rust-overlay",
+        "rev": "3b7171858c20d5293360042936058fb0c4cb93a9",
+        "type": "github"
+      },
+      "original": {
+        "owner": "oxalica",
+        "repo": "rust-overlay",
+        "type": "github"
+      }
+    },
+    "systems": {
+      "locked": {
+        "lastModified": 1681028828,
+        "narHash": "sha256-Vy1rq5AaRuLzOxct8nz4T6wlgyUR7zLU309k9mBC768=",
+        "owner": "nix-systems",
+        "repo": "default",
+        "rev": "da67096a3b9bf56a91d16901293e51ba5b49a27e",
+        "type": "github"
+      },
+      "original": {
+        "owner": "nix-systems",
+        "repo": "default",
+        "type": "github"
+      }
+    }
+  },
+  "root": "root",
+  "version": 7
+}

flake.nix

@@ -0,0 +1,85 @@
+{
+  description = "A modern Prolog implementation written mostly in Rust";
+
+  inputs = {
+    nixpkgs.url = "github:NixOS/nixpkgs/nixpkgs-unstable";
+    flake-utils.url = "github:numtide/flake-utils";
+    rust-overlay = {
+      url = "github:oxalica/rust-overlay";
+      inputs.nixpkgs.follows = "nixpkgs";
+    };
+  };
+
+  outputs = { nixpkgs, flake-utils, rust-overlay, ... }:
+    let 
+      meta = (builtins.fromTOML (builtins.readFile ./Cargo.toml)).package;
+      inherit (meta) name version;
+      overlays = [ 
+        (import rust-overlay)
+        (self: super: {
+          rustToolchainDev = super.rust-bin.stable.latest.default.override {
+            extensions = [ "rust-src" "rust-analyzer" ];
+          };
+          rustToolchainDevWasm = super.rust-bin.stable.latest.default.override {
+            extensions = [ "rust-src" "rust-analyzer" ];
+            targets = [ "wasm32-unknown-unknown" ];
+          };
+          rustToolchainNightly = super.rust-bin.selectLatestNightlyWith (toolchain:
+            toolchain.default.override {
+              extensions = [ "rust-src" "rust-analyzer" "miri" ];
+            }
+          );
+        })
+      ];
+    in flake-utils.lib.eachDefaultSystem(system:
+      let 
+        pkgs = import nixpkgs { inherit system overlays; };
+        nativeBuildInputs = with pkgs; [ pkg-config ];
+        buildInputs = with pkgs; [ openssl ] ++
+                                 lib.optionals pkgs.stdenv.isDarwin [
+                                   pkgs.darwin.apple_sdk.frameworks.SystemConfiguration
+                                 ];
+      in
+      {
+        devShells = {
+          default = pkgs.mkShell.override { stdenv = pkgs.clangMultiStdenv; } {
+            nativeBuildInputs = nativeBuildInputs;
+            buildInputs = buildInputs ++ (with pkgs; [
+              rustToolchainDev
+            ]);
+          };
+          wasm-js = pkgs.mkShell.override { stdenv = pkgs.clangMultiStdenv; } {
+            nativeBuildInputs = nativeBuildInputs;
+            buildInputs = buildInputs ++ (with pkgs; [
+              wasm-pack
+              rustToolchainDevWasm
+            ]);
+            TARGET_CC = "${pkgs.clangMultiStdenv.cc}/bin/clang";
+            hardeningDisable = [ "all" ];
+          };
+          # For use with Miri and stuff like it
+          nightly = pkgs.mkShell {
+            nativeBuildInputs = nativeBuildInputs;
+            buildInputs = buildInputs ++ (with pkgs; [
+              rustToolchainNightly
+            ]);
+          };
+        };
+
+        packages = rec {
+          default = scryer-prolog;
+          scryer-prolog = pkgs.rustPlatform.buildRustPackage {
+            pname = name;
+            inherit version;
+            src = ./.;
+            nativeBuildInputs = nativeBuildInputs;
+            buildInputs = buildInputs;
+            cargoLock = {
+              lockFile = ./Cargo.lock;
+            };
+            release = true;
+          };
+        };
+      }
+    );
+}

learn/lets-play-brisca.dj

@@ -0,0 +1,457 @@
+# Let's play Brisca
+
+In this article, we'll see how to use modern Prolog to model the game of [Brisca](https://en.wikipedia.org/wiki/Brisca). First things first, what is Brisca?
+
+## Rules of the game
+
+Brisca is a traditional card game from Spain, although similar games both in name and in rules are spread across the Mediterranean. It's also similar to Tute. It is played using the traditional Spanish deck, usually with 40 cards in 4 suites (oros, espadas, bastos, copas). Here you have the famous Castillian pattern by Heraclio Fournier, the one I play in my family with my grandparents.
+
+![Heraclio Fournier Spanish cards](Spanish_deck_Fournier.jpg)
+
+How do you play it? Basically, players play in rounds. Every round each player on the table (usually four), selects one of his own three cards to leave one in the middle, and they do that in order. There are some rules, which I will explain later, and one person wins the round, taking all the cards with him. Players take new cards from the stock. Some cards get you points, some others don't. After there are no more cards in player's hands, we count the points, the players with more points, wins!
+
+The rules for knowing which players takes the round are the following:
+- Every game there's a trump suite, which was selected as a random card from the stock at the same time when the players got their three initial cards.
+- The first player in a round (the winner of the previous round) can choose whatever card he wants.
+- The rest of the players, in order to win the round, they need to improve the card of the first player. They can put a higher card of the same suite as the first player.
+- Or they can put a card from the trump suite, which is always better than a card from a non-trump suite.
+- But between cards in the trump suite, there's still an order.
+- However players are not required to put cards that improve the game if they don't wish to do that, they can just lose the round to discard themselves a bad card.
+- The order is the following: As (ace, number 1) -> 3 -> Rey (king, number 12) -> Caballo (horse, but usually it's a knight, number 11) -> Sota (a page, number 10) -> 7 -> 6 -> 5 -> 4 -> 2.
+
+## Modeling cards
+
+First, we need to decide a representation of our cards. Coming from another languages we can think that a good representation might be a class or a struct, with two fields, one for the number and the other for the suite, but Prolog doesn't have objects. We can use a list with two elements. But lists are better when we're dealing with variable length data. We could also use a compound term. This is the right choice if our fields are fixed.
+
+A compound term is defined by an atom, followed by the data itself enclosed by parenthesis and separated by comma. Like this: `card(oros, 4)`. Yes, very similar to predicates. In fact the only difference is how we use them, because they're the same. If we pass a compund term in the first level of a query, or inside a call/N, Prolog will treat it as code instead of data. This is one the the examples of Prolog being a homoiconic language.
+
+We can go further, Prolog is very flexible and we can define custom operators easily if we want. Those are also compound terms, but with a different syntax. There's an operator already defined that is very useful for us: the dash. We can just join two pieces of data with a dash, and they'll be together in the same structure. This is usually called "pair".
+
+So, using pairs we can model cards like this: `oros-4`, `espadas-7`.
+
+We can code a predicate that defines valid cards:
+
+```
+card(Suite-Number) :-
+    member(Suite, [oros, espadas, bastos, copas]),
+    member(Number, [rey, caballo, sota, 7, 6, 5, 4, 3, 2, as]).
+```
+
+## Counting points
+
+We are going to implement a counting predicate. The suite doesn't matter, only the number. First, we define a `card_score/2` predicate which relates a card to a score and viceversa.
+
+```
+card_score(_-as, 11).
+card_score(_-3, 10).
+card_score(_-rey, 4).
+card_score(_-caballo, 3).
+card_score(_-sota, 2).
+card_score(_-X, 0) :- member(X, [7, 6, 5, 4, 2]).
+```
+
+Now, if we want to know how many points gives an specific card, we can ask:
+
+```
+?- card_score(oros-rey, X).
+   X = 4
+;  false.
+```
+
+(Remember, to load a file named brisca.pl in Scryer Prolog, we can run `scryer-prolog brisca.pl` in your favourite shell.)
+
+But we can also ask what cards can give you that value:
+
+```
+?- card(X), card_score(X, 11).
+   X = oros-as
+;  X = espadas-as
+;  X = bastos-as
+;  X = copas-as.
+```
+
+You have probably noticed that I included the card predicate in the toplevel (which kind of acts like a type restriction), not in the `card_score/2`. However, does it make sense for card_score to work on something that is not a also a card? Both approaches are valid if applied correctly. Some people prefer to set domain restrictions like this externally, as it allows you to write more concise code and it usually performs better. However, having a card predicate in `card_score/2` is more correct since it's not possible to use that predicate in something that is not a card. We can choose either option thanks to Prolog being a dynamic language.
+
+In my case, my final version is going to be a predicate which includes the card restriction and uses a wrapper predicate to have the short and concise code:
+
+```
+card_score(X, N) :-
+    card(X),
+    card_score_(X, N).
+card_score_(_-as, 11).
+card_score_(_-3, 10).
+card_score_(_-rey, 4).
+card_score_(_-caballo, 3).
+card_score_(_-sota, 2).
+card_score_(_-X, 0) :- member(X, [7, 6, 5, 4, 2]).
+```
+
+Now we just need to work with a list of cards instead of a single card. This is a place where we can introduce DCGs. DCGs are the short name for [Definite Clause Grammars](https://www.metalevel.at/prolog/dcg), which is a shorthand notation used to describe sequences. We can use them to parse, generate, complete and check sequences in the form of lists.
+
+We can define a base case. If there are no more elements in the sequence, the score should be zero. If there's a Card, we can calculate its score and sum it together with the rest:
+
+```
+cards_score_(0) --> [].
+cards_score_(X) -->
+    [Card],
+    { card_score(Card, X0), #X #= #X0 + #X1 },
+    cards_score_(X1).
+```
+
+In DCGs, to match an item of the sequence, we use brackets. We use braces to introduce normal Prolog code. Calling other DCGs (in this case, the same, as it's a recursive one), it's just calling it again. Notice in this code that we are doing the addition of X0 and X1 when we still don't know the value of X1. This would be an error in the traditional arithmethic system of Prolog, but it's valid with clpz. clpz allows us to have a more declarative arithmethic, at least with integers.
+
+Now we can try this code using `phrase/2` which is needed to jump to a DCG.
+
+```
+?- phrase(cards_score_(X), [oros-as, oros-rey, oros-7]).
+   X = 15
+;  false.
+```
+
+It seems to work. However there's a small problem. if we try to do the reverse, generating a sequence of cards that give you a specific amount of points, we'll find trouble. The program won't end. This is basically because be default we are not fair enumerating. This means that it will go deep into the recursion, trying to create an infinite length sequence. We need a way so that it does iterative deepening (starts with sequences of length 1, then length 2, ...). Luckily, if we pass our list first through the `length/2` predicate, it will do exactly that. This is usually done at the outside.
+
+```
+cards_score(Cards, Score) :-
+    phrase(cards_score_(Score), Cards).
+
+cards_score_(0) --> [].
+cards_score_(X) -->
+    [Card],
+    { card_score(Card, X0), #X #= #X0 + #X1 },
+    cards_score_(X1).
+```
+
+So now:
+
+```
+?- length(X, _), cards_score(X, 15).
+   X = [oros-rey,oros-as]
+;  X = [oros-rey,espadas-as]
+;  X = [oros-rey,bastos-as]
+;  X = [oros-rey,copas-as]
+;  X = [oros-as,oros-rey]
+;  X = [oros-as,espadas-rey]
+;  X = [oros-as,bastos-rey]
+; ...
+```
+
+```
+?- cards_score(X, Y).
+   X = [], Y = 0
+;  X = [oros-rey], Y = 4
+;  X = [oros-caballo], Y = 3
+;  X = [oros-sota], Y = 2
+;  X = [oros-7], Y = 0
+;  X = [oros-6], Y = 0
+;  X = [oros-5], Y = 0
+;  X = [oros-4], Y = 0
+;  X = [oros-3], Y = 10
+;  X = [oros-2], Y = 0
+; ...
+```
+
+## Winner of a round
+
+Now, let's try to model the winner of a round in the game. First, we can model the pseudo-numerical order of the cards in a suite. We can start with a simple idea, if we have a list with the right order, we can take an element and see if the other card is in the higher section of the rest of the list.
+
+```
+card_higher_n(N0, N1) :-
+    Order = [as, 3, rey, caballo, sota, 7, 6, 5, 4, 2],
+    append(Highers, [N0|_], Order),
+    member(N1, Highers).
+```
+
+This works well, and it can show all of the 'Y is higher than X' relationships:
+
+```
+?- card_higher_n(X, Y).
+   X = 3, Y = as
+;  X = rey, Y = as
+;  X = rey, Y = 3
+;  X = caballo, Y = as
+;  X = caballo, Y = 3
+;  ...
+```
+
+For some people this might be enough. However, since we're in Scryer Prolog, we can also use reificated predicates. Using this technique, we can show not just the 'Y is higher than X' but also, the complete set of solutions of 'Y is NOT higher than X'. In order to do that we need to add a third argument to our predicate, which will be true or false (the predicate is true for the given X and Y or it's false). And we need to change `member/2` to a reified version of it: `memberd_t/3`.
+
+```
+card_higher_n(N0, N1, T) :-
+    Order = [as, 3, rey, caballo, sota, 7, 6, 5, 4, 2],
+    append(Highers, [N0|_], Order),
+    memberd_t(N1, Highers, T).
+```
+
+Now, we get for every combination a T truth value and also, in the case of false queries, we get the constraints that need to be held to get that result:
+
+```
+?- card_higher_n(X, Y, T).
+   X = as, T = false
+;  X = 3, Y = as, T = true
+;  X = 3, T = false, dif:dif(as,Y)
+;  X = rey, Y = as, T = true
+;  X = rey, Y = 3, T = true
+;  X = rey, T = false, dif:dif(3,Y), dif:dif(as,Y)
+;  X = caballo, Y = as, T = true
+;  X = caballo, Y = 3, T = true
+;  X = caballo, Y = rey, T = true
+;  X = caballo, T = false, dif:dif(3,Y), dif:dif(as,Y), dif:dif(rey,Y)
+```
+
+For example, if X = 3 and Y = as, then Y is higher than X. But if Y is different from as, then it's false.
+
+Using reified predicates can be wise if we want to raise ourselves to a more declarative way of working with Prolog, but it's newer and not applicable form some problems. In this case, it's fine, so we'll use them.
+
+Now, to model the relatioships of whole cards, we need to take into account the suites too. We're going to compare the suites, if they're the same, we just apply the pseudonumerical order. If they're different we check if the suite of the supposedly higher card is the trump suite. We can continue to use reified predicates from `library(reif)` like `if_/3`, which takes a reified predicate as a condition and `=/3` which does reified unification. We also expose our T truth argument in this predicate:
+
+```
+card_higher(Trump, Card, Higher, T) :-
+    card(Card),
+    card(Higher),
+    Card = S0-N0,
+    Higher = S1-N1,
+    if_(S0 = S1,
+	card_higher_n(N0, N1, T),
+	=(S1, Trump, T)
+       ).
+```
+
+The code allows us to generate every card combination with every trump value and their truth value:
+
+```
+?- card_higher(Trump, X, Y, T).
+   X = oros-rey, Y = oros-rey, T = false
+;  X = oros-rey, Y = oros-caballo, T = false
+;  X = oros-rey, Y = oros-sota, T = false
+;  X = oros-rey, Y = oros-7, T = false
+;  X = oros-rey, Y = oros-6, T = false
+;  X = oros-rey, Y = oros-5, T = false
+;  X = oros-rey, Y = oros-4, T = false
+;  X = oros-rey, Y = oros-3, T = true
+;  X = oros-rey, Y = oros-2, T = false
+;  X = oros-rey, Y = oros-as, T = true
+;  Trump = espadas, X = oros-rey, Y = espadas-rey, T = true
+;  X = oros-rey, Y = espadas-rey, T = false, dif:dif(espadas,Trump)
+;  Trump = espadas, X = oros-rey, Y = espadas-caballo, T = true
+;  X = oros-rey, Y = espadas-caballo, T = false, dif:dif(espadas,Trump)
+;  Trump = espadas, X = oros-rey, Y = espadas-sota, T = true
+;  X = oros-rey, Y = espadas-sota, T = false, dif:dif(espadas,Trump)
+```
+
+For example a espadas-caballo is higher than oros-rey if Trump = espadas, otherwise, it isn't.
+
+Now, let's check for the winner. For every card we need to check if the current one is higher value than the current highest. It's very easy, we just need to to do a fold! Just like in functional programming, we have a fold _predicate_. Actually, it's a fold-left, but for this case we need the right order, so we will need to reverse the list first. The, for our foldl, we're going to use a lambda predicate from `library(lambda)`.
+
+This library, allows us to write unnamed predicates that are very useful in metapredicates like `foldl/4`. The syntax might be a bit confusing if you're coming from other languages as it's a bit unique. You can find the complete description of the syntax in the [documentation page](https://www.scryer.pl/lambda.html).
+
+The code looks like this:
+
+```
+round_winner(Cards, Trump, WinnerCard) :-
+    reverse(Cards, [FirstCard|RestCards]),
+    foldl(Trump+\X^Y^Z^if_(card_higher(Trump, X, Y), Y = Z, X = Z), RestCards, FirstCard, WinnerCard). 
+```
+
+```
+?- length(Cards, 4), round_winner(Cards, Trump, Winner).
+   Cards = [oros-rey,oros-rey,oros-rey,oros-rey], Winner = oros-rey
+;  Cards = [oros-caballo,oros-rey,oros-rey,oros-rey], Winner = oros-rey
+;  Cards = [oros-sota,oros-rey,oros-rey,oros-rey], Winner = oros-rey
+;  Cards = [oros-7,oros-rey,oros-rey,oros-rey], Winner = oros-rey
+;  Cards = [oros-6,oros-rey,oros-rey,oros-rey], Winner = oros-rey
+;  Cards = [oros-5,oros-rey,oros-rey,oros-rey], Winner = oros-rey
+;  Cards = [oros-4,oros-rey,oros-rey,oros-rey], Winner = oros-rey
+;  Cards = [oros-3,oros-rey,oros-rey,oros-rey], Winner = oros-3
+;  Cards = [oros-2,oros-rey,oros-rey,oros-rey], Winner = oros-rey
+;  Cards = [oros-as,oros-rey,oros-rey,oros-rey], Winner = oros-as
+;  Cards = [espadas-rey,oros-rey,oros-rey,oros-rey], Trump = oros, Winner = oros-rey
+;  Cards = [espadas-rey,oros-rey,oros-rey,oros-rey], Winner = espadas-rey, dif:dif(oros,Trump)
+```
+
+# A complete game
+
+Now we have the basic pieces of the game. But we don't have a complete game. A game starts with someone dealing the first three cards to each player, a trump suite is selected, then we start the rounds: players put one card each, we choose a winner, the winner takes the cards and players get a new card from the stock. This all seems very procedural, but we're using Prolog. What can we do about it?
+
+Let's ask ourselves what is a procedure. It's a sequence. And we have already seen a way to describe sequences in Prolog! The DCGs.
+
+The basic idea however is having _explicit_ states. The predicates that we're going to write will take an state (a view of the world at a certain point) and will give us the next state.
+
+Let's define the state first. In a game of Brisca we have players. Each player has the three cards he can choose to put (less if we're running out of cards) and the cards he has got from winning rounds. Aditionally we have a stock, a trump suite and the order to play, which is usually from the player who won the last round and going to the right.  We could store the data in a list, with different compound terms:
+
+```
+[players([player(Name, PlayableCards, WonCards), player(Name, PlayableCards, WonCards), ...]), stock(Cards), trump(Trump)]
+```
+
+In order to express states using DCGs, we can use [semicontext notation](https://www.metalevel.at/prolog/dcg#semicontext):
+
+```
+state(S), [S] --> [S].
+state(S0, S), [S] --> [S0].
+```
+
+So, for example our predicate to reset the game, and leave all the different cards in the stock would like this:
+
+```
+reset -->
+    state(_, [stock(Cards)]),
+    {
+	setof(Card, card(Card), Cards)
+    }.
+```
+
+To shuffle them:
+
+```
+shuffle_cards -->
+    state(S0, S),
+    {
+	select(stock(Cards), S0, S1),
+	shuffle(Cards, ShuffledCards),
+	S = [stock(ShuffledCards)|S1]
+    }.
+```
+
+Where `shuffle/2` is a predicate that implements the Fisher-Yates algorithm in Prolog:
+
+```
+shuffle([], []).
+shuffle(Xs0, [Y|Ys]) :-
+    length(Xs0, N),
+    random_integer(0, N, R),
+    nth0(R, Xs0, Y, Xs),
+    shuffle(Xs, Ys).
+```
+
+We can also code more specific state accessors, which could be useful to know which part of the state each predicate can modify:
+
+```
+players(P0, P), [S] -->
+    [S0],
+    { select(players(P0), S0, S1), S = [players(P)|S1] }.
+```
+
+The whole state management topic in DCGs could be the whole topic for another article.
+
+The final DCG code for an interactive version of Brisca could be as follows:
+
+```
+brisca :-
+    phrase(brisca, [_], [_]).
+
+brisca -->
+    reset,
+    shuffle_cards,
+    set_trump,
+    create_players([aarroyoc, xijinping, donalddtrump, vonderleyen]),
+    play_rounds,
+    show_scores.
+
+reset -->
+    state(_, [stock(Cards)]),
+    {
+	setof(Card, card(Card), Cards)
+    }.
+shuffle_cards -->
+    state(S0, S),
+    {
+	select(stock(Cards), S0, S1),
+	shuffle(Cards, ShuffledCards),
+	S = [stock(ShuffledCards)|S1]
+    }.
+set_trump -->
+    state(S0, S),
+    {
+	member(stock(Cards), S0),
+	length(Cards, N),
+	nth1(N, Cards, LastCard),
+	LastCard = Trump-_,
+	S = [trump(Trump)|S0]
+    }.
+
+create_players(Names) -->
+    state(S0, S),
+    {
+	same_length(Players, Names),
+	maplist(\N^X^(X=player(N, [], [])), Names, Players),
+	S = [players(Players)|S0]
+    },
+    deal_one_card_per_player,
+    deal_one_card_per_player,
+    deal_one_card_per_player.    
+
+deal_one_card_per_player -->
+    state(S0, S),
+    {
+	select(players(Players), S0, S1),
+	select(stock(Cards), S1, S2),
+	deal_one_card_per_player(Players, Players1, Cards, Cards1),
+	S = [players(Players1), stock(Cards1)|S2]
+    }.
+
+deal_one_card_per_player([], [], Cs, Cs).
+deal_one_card_per_player(Ps, Ps, [], []).
+deal_one_card_per_player([P|Ps], [P1|Ps1], [C|Cs], Cs1) :-
+    P = player(N, A0, B0),
+    P1 = player(N, [C|A0], B0),
+    deal_one_card_per_player(Ps, Ps1, Cs, Cs1).
+
+play_rounds -->
+    players(P, P),
+    { P = [player(_, X, _)|_], length(X, 0) }.    
+play_rounds -->
+    players(P, P),
+    { P = [player(_, X, _)|_], length(X, N), N > 0 },
+    play_round,
+    play_rounds.
+
+play_round -->
+    state(S),
+    players(P0, P2),
+    {
+	member(trump(Trump), S),
+	format("Brisca round~nTrump is: ~a~n~n", [Trump])
+    },
+    play_players(P0, Cards),
+    {
+	round_winner(Cards, Trump, WinnerCard),
+	maplist(remove_card, P0, Cards, P1),
+	nth0(N, Cards, WinnerCard),
+	nth0(N, P1, WinnerPlayer0),
+	append(PBefore, [WinnerPlayer0|PAfter], P1),
+	WinnerPlayer0 = player(WinnerName, C, W0),
+	format("Winner card is ~w from ~w~n", [WinnerCard, WinnerName]),
+	append(W0, Cards, W1),
+	append([player(WinnerName, C, W1)|PAfter], PBefore, P2)
+    },
+    deal_one_card_per_player.
+
+remove_card(P0, C, P) :-
+    P0 = player(N, C0, W),
+    select(C, C0, C1),
+    P = player(N, C1, W).
+
+play_players([], []) --> [].
+play_players([P|Ps], [C|Cs]) -->
+    {
+	P = player(Name, SelectableCards, _),
+	format("It's ~a's turn!~n", [Name]),
+	format("Selectable cards: ~w~n", [SelectableCards]),
+	read(C),
+	member(C, SelectableCards)
+    },
+    play_players(Ps, Cs).
+play_players(Ps, Cs) --> play_players(Ps, Cs).
+
+show_scores -->
+    players(P, P),
+    {
+	maplist(\X^(
+		    X=player(N,_,Z),
+		    cards_score(Z, Y),
+		    format("Score of ~w is ~d~n", [N, Y])), P)
+    }.
+```
+
+Now, we can play! Of course, if you substitute interactive choosing (via `read/1`) with another method, you could simulate games, try AI strategies, etc
+
+![Playing a game of Brisca](brisca-interactive.png)

learn/test-page.dj

@@ -0,0 +1,3 @@
+# Test page
+
+This is a page about Scryer Prolog

scryer-prolog.wxs

@@ -4,7 +4,6 @@
     <Package Description="An open source industrial strength production environment for ISO Prolog that is also a testbed for bleeding edge research in logic and constraint programming, which is itself written in a high-level language." Platform="x64" Keywords="prolog" Id="*" Compressed="yes" InstallScope="perMachine" InstallerVersion="300" Languages="1033" SummaryCodepage="1252" Manufacturer="Scryer Prolog contributors"/>
     <Property Id="APPHELPLINK" Value="https://github.com/mthom/scryer-prolog"/>
     <Media Id="1" Cabinet="scryer.cab" EmbedCab="yes" />
-    
     <Directory Id="TARGETDIR" Name="SourceDir">
       <Directory Id="ProgramFilesFolder" Name="PFiles">
         <Directory Id="INSTALLDIR" Name="Scryer Prolog">
@@ -14,18 +13,16 @@
         </Directory>
       </Directory>
       <Directory Id="ProgramMenuFolder">
-	<Component Id="ApplicationShortcut" Guid="8c9b14a3-e7b1-4d30-a892-61d7371dcae2">
+        <Component Id="ApplicationShortcut" Guid="8c9b14a3-e7b1-4d30-a892-61d7371dcae2">
           <Shortcut Id="ApplicationStarMenuShortcut" Name="Scryer Prolog" Description="Launch Scryer Prolog" Target="[#ScryerPrologEXE]" WorkingDirectory="INSTALLDIR"/>
-	  <RemoveFolder Id="ApplicationShortcut" On="uninstall"/>
-	  <RegistryValue Root="HKCU" Key="Software\Microsoft\ScryerProlog" Name="installed" Type="integer" Value="1" KeyPath="yes"/>
-	</Component>
+          <RemoveFolder Id="ApplicationShortcut" On="uninstall"/>
+          <RegistryValue Root="HKCU" Key="Software\Microsoft\ScryerProlog" Name="installed" Type="integer" Value="1" KeyPath="yes"/>
+        </Component>
       </Directory>
     </Directory>
-
     <Feature Id="Complete" Level="1" Display="expand" ConfigurableDirectory="INSTALLDIR">
       <ComponentRef Id="MainExecutable"/>
       <ComponentRef Id="ApplicationShortcut"/>
     </Feature>
   </Product>
 </Wix>
-  

src/allocator.rs

@@ -1,14 +1,10 @@
 use crate::parser::ast::*;
-use crate::temp_v;
 
-use crate::fixtures::*;
 use crate::forms::*;
 use crate::instructions::*;
-use crate::machine::machine_indices::*;
 use crate::targets::*;
 
 use std::cell::Cell;
-use std::rc::Rc;
 
 pub(crate) trait Allocator {
     fn new() -> Self;
@@ -17,91 +13,45 @@ pub(crate) trait Allocator {
         &mut self,
         lvl: Level,
         context: GenContext,
-        code: &mut Code,
-    );
+        code: &mut CodeDeque,
+    ) -> RegType;
 
     fn mark_non_var<'a, Target: CompilationTarget<'a>>(
         &mut self,
         lvl: Level,
         context: GenContext,
         cell: &'a Cell<RegType>,
-        code: &mut Code,
+        code: &mut CodeDeque,
     );
 
+    #[allow(clippy::too_many_arguments)]
     fn mark_reserved_var<'a, Target: CompilationTarget<'a>>(
         &mut self,
-        var_name: Rc<String>,
+        var_num: usize,
         lvl: Level,
-        cell: &'a Cell<VarReg>,
+        cell: &Cell<VarReg>,
         term_loc: GenContext,
-        code: &mut Code,
+        code: &mut CodeDeque,
         r: RegType,
         is_new_var: bool,
     );
 
+    fn mark_cut_var(&mut self, var_num: usize, chunk_num: usize) -> RegType;
+
     fn mark_var<'a, Target: CompilationTarget<'a>>(
         &mut self,
-        var_name: Rc<String>,
+        var_num: usize,
         lvl: Level,
-        cell: &'a Cell<VarReg>,
+        cell: &Cell<VarReg>,
         context: GenContext,
-        code: &mut Code,
+        code: &mut CodeDeque,
     );
 
     fn reset(&mut self);
-    fn reset_contents(&mut self) {}
     fn reset_arg(&mut self, arg_num: usize);
-    fn reset_at_head(&mut self, args: &Vec<Term>);
+    fn reset_at_head(&mut self, args: &[Term]);
+    fn reset_contents(&mut self);
 
     fn advance_arg(&mut self);
-
-    fn bindings(&self) -> &AllocVarDict;
-    fn bindings_mut(&mut self) -> &mut AllocVarDict;
-
-    fn take_bindings(self) -> AllocVarDict;
     fn max_reg_allocated(&self) -> usize;
-
-    fn drain_var_data<'a>(
-        &mut self,
-        vs: VariableFixtures<'a>,
-        num_of_chunks: usize,
-    ) -> VariableFixtures<'a> {
-        let mut perm_vs = VariableFixtures::new();
-
-        for (var, (var_status, cells)) in vs.into_iter() {
-            match var_status {
-                VarStatus::Temp(chunk_num, tvd) => {
-                    self.bindings_mut()
-                        .insert(var.clone(), VarData::Temp(chunk_num, 0, tvd));
-
-                    if chunk_num + 1 == num_of_chunks {
-                        perm_vs.insert_last_chunk_temp_var(var);
-                    }
-                }
-                VarStatus::Perm(_) => {
-                    self.bindings_mut().insert(var.clone(), VarData::Perm(0));
-                    perm_vs.insert(var, (var_status, cells));
-                }
-            };
-        }
-
-        perm_vs
-    }
-
-    fn get(&self, var: Rc<String>) -> RegType {
-        self.bindings()
-            .get(&var)
-            .map_or(temp_v!(0), |v| v.as_reg_type())
-    }
-
-    fn is_unbound(&self, var: Rc<String>) -> bool {
-        self.get(var).reg_num() == 0
-    }
-
-    fn record_register(&mut self, var: Rc<String>, r: RegType) {
-        match self.bindings_mut().get_mut(&var).unwrap() {
-            &mut VarData::Temp(_, ref mut s, _) => *s = r.reg_num(),
-            &mut VarData::Perm(ref mut s) => *s = r.reg_num(),
-        }
-    }
 }

src/arithmetic.rs

@@ -1,3 +1,5 @@
+#![allow(clippy::new_without_default)] // annotating structs annotated with #[bitfield] doesn't work
+
 use crate::allocator::*;
 use crate::arena::*;
 use crate::atom_table::*;
@@ -5,16 +7,19 @@ use crate::debray_allocator::*;
 use crate::forms::*;
 use crate::instructions::*;
 use crate::iterators::*;
+use crate::offset_table::*;
 use crate::targets::QueryInstruction;
 use crate::types::*;
 
 use crate::parser::ast::*;
-use crate::parser::rug::ops::PowAssign;
-use crate::parser::rug::{Assign, Integer, Rational};
+use crate::parser::dashu::{Integer, Rational};
 
 use crate::machine::machine_errors::*;
 
-use ordered_float::*;
+use dashu::base::Abs;
+use dashu::base::BitTest;
+use num_order::NumOrd;
+use ordered_float::{Float, OrderedFloat};
 
 use std::cell::Cell;
 use std::cmp::{max, min, Ordering};
@@ -22,7 +27,6 @@ use std::convert::TryFrom;
 use std::f64;
 use std::num::FpCategory;
 use std::ops::Div;
-use std::rc::Rc;
 use std::vec::Vec;
 
 #[derive(Debug, Copy, Clone, PartialEq, Eq)]
@@ -53,7 +57,7 @@ pub(crate) struct ArithInstructionIterator<'a> {
     state_stack: Vec<TermIterState<'a>>,
 }
 
-pub(crate) type ArithCont = (Code, Option<ArithmeticTerm>);
+pub(crate) type ArithCont = (CodeDeque, Option<ArithmeticTerm>);
 
 impl<'a> ArithInstructionIterator<'a> {
     fn push_subterm(&mut self, lvl: Level, term: &'a Term) {
@@ -67,19 +71,6 @@ impl<'a> ArithInstructionIterator<'a> {
             Term::Clause(cell, name, terms) => {
                 TermIterState::Clause(Level::Shallow, 0, cell, *name, terms)
             }
-            /* match ClauseType::from(*name, terms.len()) {
-                ct @ ClauseType::Named(..) => {
-                    Ok(TermIterState::Clause(Level::Shallow, 0, cell, ct, terms))
-                }
-                ct @ ClauseType::Inlined(InlinedClauseType::IsFloat(_)) => {
-                    // let ct = ClauseType::Named(1, atom!("float"), CodeIndex::default());
-                    Ok(TermIterState::Clause(Level::Shallow, 0, cell, ct, terms))
-                }
-                _ => Err(ArithmeticError::NonEvaluableFunctor(
-                    Literal::Atom(*name),
-                    terms.len(),
-                )),
-            }?,*/
             Term::Literal(cell, cons) => TermIterState::Literal(Level::Shallow, cell, cons),
             Term::Cons(..) | Term::PartialString(..) | Term::CompleteString(..) => {
                 return Err(ArithmeticError::NonEvaluableFunctor(
@@ -87,7 +78,7 @@ impl<'a> ArithInstructionIterator<'a> {
                     2,
                 ))
             }
-            Term::Var(cell, var) => TermIterState::Var(Level::Shallow, cell, var.clone()),
+            Term::Var(cell, var_ptr) => TermIterState::Var(Level::Shallow, cell, var_ptr.clone()),
         };
 
         Ok(ArithInstructionIterator {
@@ -98,9 +89,9 @@ impl<'a> ArithInstructionIterator<'a> {
 
 #[derive(Debug)]
 pub(crate) enum ArithTermRef<'a> {
-    Literal(&'a Literal),
+    Literal(Literal),
     Op(Atom, usize), // name, arity.
-    Var(Level, &'a Cell<VarReg>, Rc<String>),
+    Var(Level, &'a Cell<VarReg>, VarPtr),
 }
 
 impl<'a> Iterator for ArithInstructionIterator<'a> {
@@ -127,9 +118,9 @@ impl<'a> Iterator for ArithInstructionIterator<'a> {
                         self.push_subterm(lvl.child_level(), &subterms[child_num]);
                     }
                 }
-                TermIterState::Literal(_, _, c) => return Some(Ok(ArithTermRef::Literal(c))),
-                TermIterState::Var(lvl, cell, var) => {
-                    return Some(Ok(ArithTermRef::Var(lvl, cell, var.clone())));
+                TermIterState::Literal(_, _, c) => return Some(Ok(ArithTermRef::Literal(*c))),
+                TermIterState::Var(lvl, cell, var_ptr) => {
+                    return Some(Ok(ArithTermRef::Var(lvl, cell, var_ptr)));
                 }
                 _ => {
                     return Some(Err(ArithmeticError::NonEvaluableFunctor(
@@ -169,16 +160,18 @@ fn push_literal(interm: &mut Vec<ArithmeticTerm>, c: &Literal) -> Result<(), Ari
     match c {
         Literal::Fixnum(n) => interm.push(ArithmeticTerm::Number(Number::Fixnum(*n))),
         Literal::Integer(n) => interm.push(ArithmeticTerm::Number(Number::Integer(*n))),
-        Literal::Float(n) => interm.push(ArithmeticTerm::Number(Number::Float(*n.as_ptr()))),
+        &Literal::F64(_offset, n) => {
+            interm.push(ArithmeticTerm::Number(Number::Float(n)));
+        }
         Literal::Rational(n) => interm.push(ArithmeticTerm::Number(Number::Rational(*n))),
         Literal::Atom(name) if name == &atom!("e") => interm.push(ArithmeticTerm::Number(
-            Number::Float(OrderedFloat(std::f64::consts::E))
+            Number::Float(OrderedFloat(std::f64::consts::E)),
         )),
         Literal::Atom(name) if name == &atom!("pi") => interm.push(ArithmeticTerm::Number(
-            Number::Float(OrderedFloat(std::f64::consts::PI))
+            Number::Float(OrderedFloat(std::f64::consts::PI)),
         )),
         Literal::Atom(name) if name == &atom!("epsilon") => interm.push(ArithmeticTerm::Number(
-            Number::Float(OrderedFloat(std::f64::EPSILON))
+            Number::Float(OrderedFloat(f64::EPSILON)),
         )),
         _ => return Err(ArithmeticError::NonEvaluableFunctor(*c, 0)),
     }
@@ -219,6 +212,8 @@ impl<'a> ArithmeticEvaluator<'a> {
             atom!("round") => Ok(Instruction::Round(a1, t)),
             atom!("ceiling") => Ok(Instruction::Ceiling(a1, t)),
             atom!("floor") => Ok(Instruction::Floor(a1, t)),
+            atom!("float_integer_part") => Ok(Instruction::FloatIntegerPart(a1, t)),
+            atom!("float_fractional_part") => Ok(Instruction::FloatFractionalPart(a1, t)),
             atom!("sign") => Ok(Instruction::Sign(a1, t)),
             atom!("\\") => Ok(Instruction::BitwiseComplement(a1, t)),
             _ => Err(ArithmeticError::NonEvaluableFunctor(Literal::Atom(name), 1)),
@@ -278,7 +273,7 @@ impl<'a> ArithmeticEvaluator<'a> {
                 let ninterm = if a1.interm_or(0) == 0 {
                     self.incr_interm()
                 } else {
-                    self.interm.push(a1.clone());
+                    self.interm.push(a1);
                     a1.interm_or(0)
                 };
 
@@ -320,41 +315,39 @@ impl<'a> ArithmeticEvaluator<'a> {
         src: &'a Term,
         term_loc: GenContext,
         arg: usize,
-    ) -> Result<ArithCont, ArithmeticError>
-    {
-        let mut code = vec![];
-        let mut iter = src.iter()?;
+    ) -> Result<ArithCont, ArithmeticError> {
+        let mut code = CodeDeque::new();
 
-        while let Some(term_ref) = iter.next() {
+        for term_ref in src.iter()? {
             match term_ref? {
-                ArithTermRef::Literal(c) => push_literal(&mut self.interm, c)?,
+                ArithTermRef::Literal(c) => push_literal(&mut self.interm, &c)?,
                 ArithTermRef::Var(lvl, cell, name) => {
-                    let r = if lvl == Level::Shallow {
-                        self.marker.mark_non_callable(
-                            name.clone(),
-                            arg,
-                            term_loc,
-                            cell,
-                            &mut code,
-                        )
-                    } else if term_loc.is_last() || cell.get().norm().reg_num() == 0 {
-                        self.marker.mark_var::<QueryInstruction>(
-                            name.clone(),
-                            lvl,
-                            cell,
-                            term_loc,
-                            &mut code,
-                        );
+                    let var_num = name.to_var_num().unwrap();
 
-                        self.marker.get_binding(&name).unwrap()
+                    let r = if lvl == Level::Shallow {
+                        self.marker
+                            .mark_non_callable(var_num, arg, term_loc, cell, &mut code)
+                    } else if term_loc.is_last() || cell.get().norm().reg_num() == 0 {
+                        let r = self.marker.get_binding(var_num);
+
+                        if r.reg_num() == 0 {
+                            self.marker.mark_var::<QueryInstruction>(
+                                var_num, lvl, cell, term_loc, &mut code,
+                            );
+                            cell.get().norm()
+                        } else {
+                            self.marker.increment_running_count(var_num);
+                            r
+                        }
                     } else {
+                        self.marker.increment_running_count(var_num);
                         cell.get().norm()
                     };
 
                     self.interm.push(ArithmeticTerm::Reg(r));
                 }
                 ArithTermRef::Op(name, arity) => {
-                    code.push(self.instr_from_clause(name, arity)?);
+                    code.push_back(self.instr_from_clause(name, arity)?);
                 }
             }
         }
@@ -364,37 +357,43 @@ impl<'a> ArithmeticEvaluator<'a> {
 }
 
 // integer division rounding function -- 9.1.3.1.
-pub(crate) fn rnd_i<'a>(n: &'a Number, arena: &mut Arena) -> Number {
+pub(crate) fn rnd_i(n: &'_ Number, arena: &mut Arena) -> Result<Number, EvalError> {
     match n {
         &Number::Integer(i) => {
-            if let Some(n) = i.to_i64() {
-                fixnum!(Number, n, arena)
+            if let Ok(value) = Fixnum::build_with_checked(&*i) {
+                Ok(Number::Fixnum(value))
             } else {
-                *n
+                Ok(*n)
             }
         }
-        &Number::Fixnum(_) => *n,
+        Number::Fixnum(_) => Ok(*n),
         &Number::Float(f) => {
             let f = f.floor();
 
-            const I64_MIN_TO_F: OrderedFloat<f64> = OrderedFloat(i64::MIN as f64);
-            const I64_MAX_TO_F: OrderedFloat<f64> = OrderedFloat(i64::MAX as f64);
+            const FIXNUM_MIN_TO_F: OrderedFloat<f64> = OrderedFloat(Fixnum::MIN as f64);
+            const FIXNUM_MAX_TO_F: OrderedFloat<f64> = OrderedFloat(Fixnum::MAX as f64);
 
-            if I64_MIN_TO_F <= f && f <= I64_MAX_TO_F {
-                fixnum!(Number, f.into_inner() as i64, arena)
+            if (FIXNUM_MIN_TO_F..=FIXNUM_MAX_TO_F).contains(&f) {
+                Ok(Number::Fixnum(
+                    // Safety: We checked that the value is in range
+                    unsafe { Fixnum::build_with_unchecked(f.into_inner() as i64) },
+                ))
             } else {
-                Number::Integer(arena_alloc!(Integer::from_f64(f.into_inner()).unwrap(), arena))
+                Ok(Number::Integer(arena_alloc!(
+                    Integer::try_from(classify_float(f.0)?).unwrap_or_else(|_| {
+                        unreachable!();
+                    }),
+                    arena
+                )))
             }
         }
-        &Number::Rational(ref r) => {
-            let r_ref = r.fract_floor_ref();
-            let (mut fract, mut floor) = (Rational::new(), Integer::new());
-            (&mut fract, &mut floor).assign(r_ref);
+        Number::Rational(ref r) => {
+            let floor = r.floor();
 
-            if let Some(floor) = floor.to_i64() {
-                fixnum!(Number, floor, arena)
+            if let Ok(value) = Fixnum::build_with_checked(&floor) {
+                Ok(Number::Fixnum(value))
             } else {
-                Number::Integer(arena_alloc!(floor, arena))
+                Ok(Number::Integer(arena_alloc!(floor, arena)))
             }
         }
     }
@@ -411,9 +410,9 @@ impl From<Fixnum> for Integer {
 pub(crate) fn rnd_f(n: &Number) -> f64 {
     match n {
         &Number::Fixnum(n) => n.get_num() as f64,
-        &Number::Integer(ref n) => n.to_f64(),
+        Number::Integer(ref n) => n.to_f64().value(),
         &Number::Float(OrderedFloat(f)) => f,
-        &Number::Rational(ref r) => r.to_f64(),
+        Number::Rational(ref r) => r.to_f64().value(),
     }
 }
 
@@ -433,7 +432,7 @@ fn classify_float(f: f64) -> Result<f64, EvalError> {
             }
         }
         FpCategory::Nan => Err(EvalError::Undefined),
-        _ => Ok(f)
+        _ => Ok(f),
     }
 }
 
@@ -444,12 +443,12 @@ pub(crate) fn float_fn_to_f(n: i64) -> Result<f64, EvalError> {
 
 #[inline]
 pub(crate) fn float_i_to_f(n: &Integer) -> Result<f64, EvalError> {
-    classify_float(n.to_f64())
+    classify_float(n.to_f64().value())
 }
 
 #[inline]
 pub(crate) fn float_r_to_f(r: &Rational) -> Result<f64, EvalError> {
-    classify_float(r.to_f64())
+    classify_float(r.to_f64().value())
 }
 
 #[inline]
@@ -541,39 +540,33 @@ impl PartialEq for Number {
     fn eq(&self, rhs: &Self) -> bool {
         match (self, rhs) {
             (&Number::Fixnum(n1), &Number::Fixnum(n2)) => n1.eq(&n2),
-            (&Number::Fixnum(n1), &Number::Integer(ref n2)) => n1.get_num().eq(&**n2),
-            (&Number::Integer(ref n1), &Number::Fixnum(n2)) => (&**n1).eq(&n2.get_num()),
-            (&Number::Fixnum(n1), &Number::Rational(ref n2)) => n1.get_num().eq(&**n2),
-            (&Number::Rational(ref n1), &Number::Fixnum(n2)) => (&**n1).eq(&n2.get_num()),
+            (&Number::Fixnum(n1), Number::Integer(ref n2)) => n1.get_num().num_eq(&**n2),
+            (Number::Integer(ref n1), &Number::Fixnum(n2)) => n1.num_eq(&n2.get_num()),
+            (&Number::Fixnum(n1), Number::Rational(ref n2)) => {
+                Integer::from(n1.get_num()).num_eq(&**n2)
+            }
+            (Number::Rational(ref n1), &Number::Fixnum(n2)) => {
+                n1.num_eq(&Integer::from(n2.get_num()))
+            }
             (&Number::Fixnum(n1), &Number::Float(n2)) => OrderedFloat(n1.get_num() as f64).eq(&n2),
             (&Number::Float(n1), &Number::Fixnum(n2)) => n1.eq(&OrderedFloat(n2.get_num() as f64)),
-            (&Number::Integer(ref n1), &Number::Integer(ref n2)) => n1.eq(n2),
-            (&Number::Integer(ref n1), Number::Float(n2)) => OrderedFloat(n1.to_f64()).eq(n2),
-            (&Number::Float(n1), &Number::Integer(ref n2)) => n1.eq(&OrderedFloat(n2.to_f64())),
-            (&Number::Integer(ref n1), &Number::Rational(ref n2)) => {
-                #[cfg(feature = "num")]
-                {
-                    &Rational::from(&**n1) == &**n2
-                }
-                #[cfg(not(feature = "num"))]
-                {
-                    &**n1 == &**n2
-                }
+            (Number::Integer(ref n1), Number::Integer(ref n2)) => n1.eq(n2),
+            (Number::Integer(ref n1), Number::Float(n2)) => {
+                OrderedFloat(n1.to_f64().value()).eq(n2)
             }
-            (&Number::Rational(ref n1), &Number::Integer(ref n2)) => {
-                #[cfg(feature = "num")]
-                {
-                    &**n1 == &Rational::from(&**n2)
-                }
-                #[cfg(not(feature = "num"))]
-                {
-                    &**n1 == &**n2
-                }
+            (&Number::Float(n1), Number::Integer(ref n2)) => {
+                n1.eq(&OrderedFloat(n2.to_f64().value()))
+            }
+            (Number::Integer(ref n1), Number::Rational(ref n2)) => n1.num_eq(&**n2),
+            (Number::Rational(ref n1), Number::Integer(ref n2)) => n1.num_eq(&**n2),
+            (Number::Rational(ref n1), &Number::Float(n2)) => {
+                OrderedFloat(n1.to_f64().value()).eq(&n2)
+            }
+            (&Number::Float(n1), Number::Rational(ref n2)) => {
+                n1.eq(&OrderedFloat(n2.to_f64().value()))
             }
-            (&Number::Rational(ref n1), &Number::Float(n2)) => OrderedFloat(n1.to_f64()).eq(&n2),
-            (&Number::Float(n1), &Number::Rational(ref n2)) => n1.eq(&OrderedFloat(n2.to_f64())),
             (&Number::Float(f1), &Number::Float(f2)) => f1.eq(&f2),
-            (&Number::Rational(ref r1), &Number::Rational(ref r2)) => r1.eq(&r2),
+            (Number::Rational(ref r1), Number::Rational(ref r2)) => r1.eq(r2),
         }
     }
 }
@@ -593,8 +586,8 @@ impl PartialOrd<usize> for Number {
                     (n as usize).partial_cmp(rhs)
                 }
             }
-            Number::Integer(n) => (&**n).partial_cmp(rhs),
-            Number::Rational(r) => (&**r).partial_cmp(rhs),
+            Number::Integer(n) => Some((n).num_cmp(rhs)),
+            Number::Rational(r) => Some((r).num_cmp(&Integer::from(*rhs))),
             Number::Float(f) => f.partial_cmp(&OrderedFloat(*rhs as f64)),
         }
     }
@@ -613,8 +606,8 @@ impl PartialEq<usize> for Number {
                     (n as usize).eq(rhs)
                 }
             }
-            Number::Integer(n) => (&**n).eq(rhs),
-            Number::Rational(r) => (&**r).eq(rhs),
+            Number::Integer(n) => (n).num_eq(rhs),
+            Number::Rational(r) => (r).num_eq(&Integer::from(*rhs)),
             Number::Float(f) => f.eq(&OrderedFloat(*rhs as f64)),
         }
     }
@@ -630,50 +623,42 @@ impl Ord for Number {
     fn cmp(&self, rhs: &Number) -> Ordering {
         match (self, rhs) {
             (&Number::Fixnum(n1), &Number::Fixnum(n2)) => n1.get_num().cmp(&n2.get_num()),
-            (&Number::Fixnum(n1), Number::Integer(n2)) => Integer::from(n1.get_num()).cmp(&*n2),
-            (Number::Integer(n1), &Number::Fixnum(n2)) => (&**n1).cmp(&Integer::from(n2.get_num())),
-            (&Number::Fixnum(n1), Number::Rational(n2)) => Rational::from(n1.get_num()).cmp(&*n2),
+            (&Number::Fixnum(n1), Number::Integer(n2)) => Integer::from(n1.get_num()).cmp(n2),
+            (Number::Integer(n1), &Number::Fixnum(n2)) => (**n1).cmp(&Integer::from(n2.get_num())),
+            (&Number::Fixnum(n1), Number::Rational(n2)) => Rational::from(n1.get_num()).cmp(n2),
             (Number::Rational(n1), &Number::Fixnum(n2)) => {
-                (&**n1).cmp(&Rational::from(n2.get_num()))
+                (**n1).cmp(&Rational::from(n2.get_num()))
             }
             (&Number::Fixnum(n1), &Number::Float(n2)) => OrderedFloat(n1.get_num() as f64).cmp(&n2),
             (&Number::Float(n1), &Number::Fixnum(n2)) => n1.cmp(&OrderedFloat(n2.get_num() as f64)),
             (&Number::Integer(n1), &Number::Integer(n2)) => (*n1).cmp(&*n2),
-            (&Number::Integer(n1), Number::Float(n2)) => OrderedFloat(n1.to_f64()).cmp(n2),
-            (&Number::Float(n1), &Number::Integer(ref n2)) => n1.cmp(&OrderedFloat(n2.to_f64())),
+            (&Number::Integer(n1), Number::Float(n2)) => OrderedFloat(n1.to_f64().value()).cmp(n2),
+            (&Number::Float(n1), Number::Integer(ref n2)) => {
+                n1.cmp(&OrderedFloat(n2.to_f64().value()))
+            }
             (&Number::Integer(n1), &Number::Rational(n2)) => {
-                #[cfg(feature = "num")]
-                {
-                    Rational::from(&**n1).cmp(n2)
-                }
-                #[cfg(not(feature = "num"))]
-                {
-                    (&*n1).partial_cmp(&*n2).unwrap_or(Ordering::Less)
-                }
+                (*n1).num_partial_cmp(&*n2).unwrap_or(Ordering::Less)
             }
             (&Number::Rational(n1), &Number::Integer(n2)) => {
-                #[cfg(feature = "num")]
-                {
-                    (&**n1).cmp(&Rational::from(&**n2))
-                }
-                #[cfg(not(feature = "num"))]
-                {
-                    (&*n1).partial_cmp(&*n2).unwrap_or(Ordering::Less)
-                }
+                (*n1).num_partial_cmp(&*n2).unwrap_or(Ordering::Less)
+            }
+            (&Number::Rational(n1), &Number::Float(n2)) => {
+                OrderedFloat(n1.to_f64().value()).cmp(&n2)
+            }
+            (&Number::Float(n1), &Number::Rational(n2)) => {
+                n1.cmp(&OrderedFloat(n2.to_f64().value()))
             }
-            (&Number::Rational(n1), &Number::Float(n2)) => OrderedFloat(n1.to_f64()).cmp(&n2),
-            (&Number::Float(n1), &Number::Rational(n2)) => n1.cmp(&OrderedFloat(n2.to_f64())),
             (&Number::Float(f1), &Number::Float(f2)) => f1.cmp(&f2),
             (&Number::Rational(r1), &Number::Rational(r2)) => (*r1).cmp(&*r2),
         }
     }
 }
 
-impl TryFrom<HeapCellValue> for Number {
+impl TryFrom<(HeapCellValue, &'_ F64Table)> for Number {
     type Error = ();
 
     #[inline]
-    fn try_from(value: HeapCellValue) -> Result<Number, Self::Error> {
+    fn try_from((value, f64_tbl): (HeapCellValue, &F64Table)) -> Result<Number, Self::Error> {
         read_heap_cell!(value,
            (HeapCellValueTag::Cons, c) => {
                match_untyped_arena_ptr!(c,
@@ -688,10 +673,11 @@ impl TryFrom<HeapCellValue> for Number {
                   }
                )
            }
-           (HeapCellValueTag::F64, n) => {
-               Ok(Number::Float(*n))
+           (HeapCellValueTag::F64Offset, offset) => {
+               let n = f64_tbl.get_entry(offset);
+               Ok(Number::Float(n))
            }
-           (HeapCellValueTag::Fixnum, n) => {
+           (HeapCellValueTag::Fixnum | HeapCellValueTag::CutPoint, n) => {
                Ok(Number::Fixnum(n))
            }
            _ => {
@@ -703,20 +689,20 @@ impl TryFrom<HeapCellValue> for Number {
 
 // Computes n ^ power. Ignores the sign of power.
 pub(crate) fn binary_pow(mut n: Integer, power: &Integer) -> Integer {
-    let mut power = Integer::from(power.abs_ref());
+    let mut power = power.abs();
 
-    if power == 0 {
-        return Integer::from(1);
+    if power.is_zero() {
+        return Integer::ONE;
     }
 
-    let mut oddand = Integer::from(1);
+    let mut oddand = Integer::ONE;
 
-    while power > 1 {
-        if power.is_odd() {
+    while power.num_gt(&1) {
+        if power.bit(0) {
             oddand *= &n;
         }
 
-        n.pow_assign(2);
+        n = n.pow(2);
         power >>= 1;
     }
 

src/atom_table.rs

@@ -1,28 +1,146 @@
+#![allow(clippy::new_without_default)] // annotating structs annotated with #[bitfield] doesn't work
+
 use crate::parser::ast::MAX_ARITY;
 use crate::raw_block::*;
 use crate::types::*;
 
-use std::borrow::Borrow;
 use std::cmp::Ordering;
 use std::hash::{Hash, Hasher};
 use std::mem;
+use std::ops::Deref;
 use std::ptr;
-use std::slice;
 use std::str;
+use std::sync::Arc;
+use std::sync::Mutex;
+use std::sync::RwLock;
+use std::sync::Weak;
 
+use arcu::atomic::Arcu;
+use arcu::epoch_counters::GlobalEpochCounterPool;
+use arcu::rcu_ref::RcuRef;
+use arcu::Rcu;
 use indexmap::IndexSet;
 
-use modular_bitfield::prelude::*;
+use scryer_modular_bitfield::prelude::*;
+
+#[bitfield]
+#[repr(u64)]
+#[derive(Copy, Clone, Debug)]
+pub struct AtomCell {
+    name: B48,
+    arity: B8,
+    #[allow(unused)]
+    f: bool,
+    #[allow(unused)]
+    m: bool,
+    #[allow(unused)]
+    is_inlined: bool,
+    #[allow(unused)]
+    tag: B5,
+}
+
+const INLINED_ATOM_MAX_LEN: usize = 6;
+
+const_assert!(INLINED_ATOM_MAX_LEN < mem::size_of::<AtomCell>());
+const_assert!(mem::size_of::<AtomCell>() == 8);
+
+const_assert!(INLINED_ATOM_MAX_LEN < mem::size_of::<Atom>());
+const_assert!(mem::size_of::<Atom>() == 8);
+
+impl AtomCell {
+    #[inline]
+    fn new_static(index: u64) -> Self {
+        // upper 23 bits of index must be 0
+        debug_assert_eq!(index & !((1 << 49) - 1), 0);
+        AtomCell::new()
+            .with_name(index)
+            .with_arity(0u8)
+            .with_m(false)
+            .with_f(false)
+            .with_is_inlined(false)
+            .with_tag(HeapCellValueTag::Atom as u8)
+    }
+
+    #[inline]
+    fn new_inlined(string: &str, arity: u8) -> Self {
+        debug_assert!(string.len() <= INLINED_ATOM_MAX_LEN);
+
+        let mut string_buf: [u8; 8] = [0u8; 8];
+        string_buf[..string.len()].copy_from_slice(string.as_bytes());
+        let encoding = u64::from_le_bytes(string_buf);
+
+        AtomCell::new()
+            .with_name(encoding)
+            .with_arity(arity)
+            .with_m(false)
+            .with_f(false)
+            .with_is_inlined(true)
+            .with_tag(HeapCellValueTag::Atom as u8)
+    }
+
+    #[inline]
+    pub fn new_char_inlined(c: char) -> Self {
+        if c == '\u{0}' {
+            return Self::new_static(NULL_ATOM.flat_index());
+        }
+
+        let mut char_buf = [0u8; 8];
+        c.encode_utf8(&mut char_buf);
+
+        let encoding = u64::from_le_bytes(char_buf);
+
+        AtomCell::new()
+            .with_name(encoding)
+            .with_arity(0u8)
+            .with_m(false)
+            .with_f(false)
+            .with_is_inlined(true)
+            .with_tag(HeapCellValueTag::Atom as u8)
+    }
+
+    #[inline]
+    pub fn build_with(atom_index: u64, arity: u8) -> Self {
+        debug_assert!((arity as usize) <= MAX_ARITY);
+
+        AtomCell::new()
+            .with_name(atom_index >> 1)
+            .with_arity(arity)
+            .with_f(false)
+            .with_m(false)
+            .with_is_inlined(atom_index & 1 == 1)
+            .with_tag(HeapCellValueTag::Atom as u8)
+    }
+
+    #[inline]
+    pub fn get_name(self) -> Atom {
+        Atom {
+            index: (self.name() << 1) | self.is_inlined() as u64,
+        }
+    }
+
+    #[inline]
+    pub fn get_arity(self) -> usize {
+        self.arity() as usize
+    }
+
+    #[inline]
+    pub fn get_name_and_arity(self) -> (Atom, usize) {
+        (self.get_name(), self.get_arity())
+    }
+}
 
 #[derive(Copy, Clone, Debug, PartialEq, Eq)]
 pub struct Atom {
-    pub index: usize,
+    pub index: u64,
 }
 
-const_assert!(mem::size_of::<Atom>() == 8);
-
 include!(concat!(env!("OUT_DIR"), "/static_atoms.rs"));
 
+// populate these in STRINGS so they can be used from build_functor
+const _: Atom = atom!(".");
+const _: Atom = atom!("[]");
+const NULL_ATOM: Atom = atom!("\0");
+
 impl<'a> From<&'a Atom> for Atom {
     #[inline]
     fn from(atom: &'a Atom) -> Self {
@@ -30,49 +148,36 @@ impl<'a> From<&'a Atom> for Atom {
     }
 }
 
-impl From<bool> for Atom {
-    #[inline]
-    fn from(value: bool) -> Self {
-        if value { atom!("true") } else { atom!("false") }
+impl indexmap::Equivalent<Atom> for str {
+    fn equivalent(&self, key: &Atom) -> bool {
+        &*key.as_str() == self
     }
 }
 
-#[cfg(test)]
-use std::cell::RefCell;
+impl PartialEq<str> for Atom {
+    fn eq(&self, other: &str) -> bool {
+        self.as_str().deref() == other
+    }
+}
+
+impl PartialEq<&str> for Atom {
+    fn eq(&self, &other: &&str) -> bool {
+        self.as_str().deref() == other
+    }
+}
 
 const ATOM_TABLE_INIT_SIZE: usize = 1 << 16;
 const ATOM_TABLE_ALIGN: usize = 8;
 
-#[cfg(test)]
-thread_local! {
-    static ATOM_TABLE_BUF_BASE: RefCell<*const u8> = RefCell::new(ptr::null_mut());
+#[inline(always)]
+fn global_atom_table() -> &'static RwLock<Weak<AtomTable>> {
+    static GLOBAL_ATOM_TABLE: RwLock<Weak<AtomTable>> = RwLock::new(Weak::new());
+    &GLOBAL_ATOM_TABLE
 }
 
-#[cfg(not(test))]
-static mut ATOM_TABLE_BUF_BASE: *const u8 = ptr::null_mut();
-
-#[cfg(test)]
-fn set_atom_tbl_buf_base(ptr: *const u8) {
-    ATOM_TABLE_BUF_BASE.with(|atom_table_buf_base| {
-        *atom_table_buf_base.borrow_mut() = ptr;
-    });
-}
-
-#[cfg(test)]
-pub(crate) fn get_atom_tbl_buf_base() -> *const u8 {
-    ATOM_TABLE_BUF_BASE.with(|atom_table_buf_base| *atom_table_buf_base.borrow())
-}
-
-#[cfg(not(test))]
-fn set_atom_tbl_buf_base(ptr: *const u8) {
-    unsafe {
-        ATOM_TABLE_BUF_BASE = ptr;
-    }
-}
-
-#[cfg(not(test))]
-pub(crate) fn get_atom_tbl_buf_base() -> *const u8 {
-    unsafe { ATOM_TABLE_BUF_BASE }
+#[inline(always)]
+fn arc_atom_table() -> Option<Arc<AtomTable>> {
+    global_atom_table().read().unwrap().upgrade()
 }
 
 impl RawBlockTraits for AtomTable {
@@ -90,9 +195,17 @@ impl RawBlockTraits for AtomTable {
 #[bitfield]
 #[derive(Copy, Clone, Debug)]
 struct AtomHeader {
-    #[allow(unused)] m: bool,
+    #[allow(unused)]
+    m: bool,
     len: B50,
-    #[allow(unused)] padding: B13,
+    #[allow(unused)]
+    padding: B13,
+}
+
+#[repr(C)]
+pub struct AtomData {
+    header: AtomHeader,
+    data: str,
 }
 
 impl AtomHeader {
@@ -101,71 +214,129 @@ impl AtomHeader {
     }
 }
 
-impl Borrow<str> for Atom {
-    #[inline]
-    fn borrow(&self) -> &str {
-        self.as_str()
-    }
-}
-
 impl Hash for Atom {
     #[inline]
     fn hash<H: Hasher>(&self, hasher: &mut H) {
         self.as_str().hash(hasher)
-        // hasher.write_usize(self.index)
     }
 }
 
-#[macro_export]
-macro_rules! is_char {
-    ($s:expr) => {
-        !$s.is_empty() && $s.chars().nth(1).is_none()
-    };
+pub enum AtomString<'a> {
+    Static(&'a str),
+    Inlined([u8; 8]),
+    Dynamic(AtomTableRef<str>),
+}
+
+#[inline(always)]
+fn inlined_to_str(bytes: &[u8; 8]) -> &str {
+    let slice_len = bytes
+        .iter()
+        .position(|&b| b == 0u8)
+        .unwrap_or(INLINED_ATOM_MAX_LEN);
+
+    unsafe { str::from_utf8_unchecked(&bytes[..slice_len]) }
+}
+
+impl std::fmt::Debug for AtomString<'_> {
+    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
+        std::fmt::Debug::fmt(self.deref(), f)
+    }
+}
+
+impl std::fmt::Display for AtomString<'_> {
+    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
+        std::fmt::Display::fmt(self.deref(), f)
+    }
+}
+
+impl std::ops::Deref for AtomString<'_> {
+    type Target = str;
+    fn deref(&self) -> &Self::Target {
+        match self {
+            Self::Static(reference) => reference,
+            Self::Inlined(inlined) => inlined_to_str(inlined),
+            Self::Dynamic(guard) => guard.deref(),
+        }
+    }
+}
+
+#[cfg(feature = "repl")]
+impl rustyline::completion::Candidate for AtomString<'_> {
+    fn display(&self) -> &str {
+        self.deref()
+    }
+
+    fn replacement(&self) -> &str {
+        self.deref()
+    }
 }
 
 impl Atom {
     #[inline]
-    pub fn buf(self) -> *const u8 {
-        let ptr = self.as_ptr();
-
-        if ptr.is_null() {
-            return ptr::null();
-        }
-
-        (ptr as usize + mem::size_of::<AtomHeader>()) as *const u8
+    fn new_inlined(string: &str) -> Self {
+        AtomCell::new_inlined(string, 0).get_name()
     }
 
     #[inline(always)]
-    pub fn is_static(self) -> bool {
-        self.index < STRINGS.len() << 3
-    }
-
-    #[inline(always)]
-    pub fn as_ptr(self) -> *const u8 {
-        if self.is_static() {
-            ptr::null()
+    fn is_static(self) -> bool {
+        if self.is_inlined() {
+            true
         } else {
-            (get_atom_tbl_buf_base() as usize + self.index - (STRINGS.len() << 3)) as *const u8
+            (self.flat_index() as usize) < STRINGS.len()
+        }
+    }
+
+    #[inline]
+    pub(crate) fn flat_index(self) -> u64 {
+        self.index >> 1
+    }
+
+    #[inline(always)]
+    pub(crate) fn is_inlined(self) -> bool {
+        self.index & 1 == 1
+    }
+
+    #[inline(always)]
+    fn as_ptr(self) -> Option<AtomTableRef<AtomData>> {
+        if self.is_static() {
+            None
+        } else {
+            let atom_table =
+                arc_atom_table().expect("We should only have an Atom while there is an AtomTable");
+
+            AtomTableRef::try_map(atom_table.inner.read(), |buf| unsafe {
+                let ptr = buf
+                    .block
+                    .base
+                    .add(self.flat_index() as usize - STRINGS.len());
+                // TODO use std::ptr::from_raw_parts instead when feature ptr_metadata is stable rust-lang/rust#81513
+                let atom_data = &*(std::ptr::slice_from_raw_parts(ptr, 0) as *const AtomData);
+                let len = atom_data.header.len();
+                Some(&*(std::ptr::slice_from_raw_parts(ptr, len as usize) as *const AtomData))
+            })
         }
     }
 
     #[inline(always)]
-    pub fn from(index: usize) -> Self {
+    pub fn from(index: u64) -> Self {
         Self { index }
     }
 
     #[inline(always)]
     pub fn len(self) -> usize {
-        if self.is_static() {
-            STRINGS[self.index >> 3].len()
+        if let Some(s) = self.inlined_str() {
+            s.len()
+        } else if self.is_static() {
+            let index = self.flat_index();
+            STRINGS[index as usize].len()
         } else {
-            unsafe { ptr::read(self.as_ptr() as *const AtomHeader).len() as _ }
+            let len: u64 = self.as_ptr().unwrap().header.len();
+            len as usize
         }
     }
 
-    #[inline(always)]
-    pub fn flat_index(self) -> u64 {
-        (self.index >> 3) as u64
+    pub fn is_empty(self) -> bool {
+        self.len() == 0
     }
 
     pub fn as_char(self) -> Option<char> {
@@ -175,48 +346,53 @@ impl Atom {
         let c1 = it.next();
         let c2 = it.next();
 
-        if c2.is_none() { c1 } else { None }
-    }
-
-    #[inline]
-    pub fn chars(&self) -> str::Chars {
-        self.as_str().chars()
-    }
-
-    #[inline]
-    pub fn as_str(&self) -> &str {
-        unsafe {
-            let ptr = self.as_ptr();
-
-            if ptr.is_null() {
-                return STRINGS[self.index >> 3];
-            }
-
-            let header = ptr::read::<AtomHeader>(ptr as *const _);
-            let len = header.len() as usize;
-            let buf = (ptr as usize + mem::size_of::<AtomHeader>()) as *mut u8;
-
-            str::from_utf8_unchecked(slice::from_raw_parts(buf, len))
+        if c2.is_none() {
+            c1
+        } else {
+            None
         }
     }
 
-    pub fn defrock_brackets(&self, atom_tbl: &mut AtomTable) -> Self {
+    #[inline]
+    fn inlined_str<'a>(&self) -> Option<AtomString<'a>> {
+        if self.is_inlined() {
+            Some(AtomString::Inlined(self.flat_index().to_le_bytes()))
+        } else {
+            None
+        }
+    }
+
+    #[inline]
+    pub fn as_str(&self) -> AtomString<'static> {
+        if let Some(s) = self.inlined_str() {
+            s
+        } else if self.is_static() {
+            let index = self.flat_index() as usize;
+            AtomString::Static(STRINGS[index])
+        } else if let Some(ptr) = self.as_ptr() {
+            AtomString::Dynamic(AtomTableRef::map(ptr, |ptr| &ptr.data))
+        } else {
+            AtomString::Static(STRINGS[(self.index >> 1) as usize])
+        }
+    }
+
+    pub fn defrock_brackets(&self, atom_tbl: &AtomTable) -> Self {
         let s = self.as_str();
 
-        let s = if s.starts_with('(') && s.ends_with(')') {
+        let sub_str = if s.starts_with('(') && s.ends_with(')') {
             &s['('.len_utf8()..s.len() - ')'.len_utf8()]
         } else {
             return *self;
         };
 
-        atom_tbl.build_with(s)
+        AtomTable::build_with(atom_tbl, sub_str)
     }
 }
 
 unsafe fn write_to_ptr(string: &str, ptr: *mut u8) {
     ptr::write(ptr as *mut _, AtomHeader::build_with(string.len() as u64));
-    let str_ptr = (ptr as usize + mem::size_of::<AtomHeader>()) as *mut u8;
-    ptr::copy_nonoverlapping(string.as_ptr(), str_ptr as *mut u8, string.len());
+    let str_ptr = ptr.add(mem::size_of::<AtomHeader>());
+    ptr::copy_nonoverlapping(string.as_ptr(), str_ptr, string.len());
 }
 
 impl PartialOrd for Atom {
@@ -229,138 +405,142 @@ impl PartialOrd for Atom {
 impl Ord for Atom {
     #[inline]
     fn cmp(&self, other: &Atom) -> Ordering {
-        self.as_str().cmp(other.as_str())
+        self.as_str().cmp(&*other.as_str())
     }
 }
 
 #[derive(Debug)]
-pub struct AtomTable {
+pub struct InnerAtomTable {
     block: RawBlock<AtomTable>,
-    pub table: IndexSet<Atom>,
+    pub table: Arcu<IndexSet<Atom>, GlobalEpochCounterPool>,
 }
 
-impl Drop for AtomTable {
-    fn drop(&mut self) {
-        self.block.deallocate();
+#[derive(Debug)]
+pub struct AtomTable {
+    inner: Arcu<InnerAtomTable, GlobalEpochCounterPool>,
+    // this lock is taking during resizing
+    update: Mutex<()>,
+}
+
+pub type AtomTableRef<M> = arcu::rcu_ref::RcuRef<InnerAtomTable, M>;
+
+impl InnerAtomTable {
+    #[inline(always)]
+    fn lookup_str(self: &InnerAtomTable, string: &str) -> Option<Atom> {
+        STATIC_ATOMS_MAP
+            .get(string)
+            .cloned()
+            .or_else(|| self.table.read().get(string).cloned())
     }
 }
 
 impl AtomTable {
     #[inline]
-    pub fn new() -> Self {
-        let table = Self {
-            block: RawBlock::new(),
-            table: IndexSet::new(),
-        };
-
-        set_atom_tbl_buf_base(table.block.base);
-        table
+    pub fn new() -> Arc<Self> {
+        let upgraded = global_atom_table().read().unwrap().upgrade();
+        // don't inline upgraded, otherwise temporary will be dropped too late in case of None
+        if let Some(atom_table) = upgraded {
+            atom_table
+        } else {
+            let mut guard = global_atom_table().write().unwrap();
+            // try to upgrade again in case we lost the race on the write lock
+            if let Some(atom_table) = guard.upgrade() {
+                atom_table
+            } else {
+                let atom_table = Arc::new(Self {
+                    inner: Arcu::new(
+                        InnerAtomTable {
+                            block: RawBlock::new(),
+                            table: Arcu::new(IndexSet::new(), GlobalEpochCounterPool),
+                        },
+                        GlobalEpochCounterPool,
+                    ),
+                    update: Mutex::new(()),
+                });
+                *guard = Arc::downgrade(&atom_table);
+                atom_table
+            }
+        }
     }
 
-    #[inline]
-    pub fn buf(&self) -> *const u8 {
-        self.block.base as *const u8
+    pub fn active_table(&self) -> RcuRef<IndexSet<Atom>, IndexSet<Atom>> {
+        self.inner.read().table.read()
     }
 
-    #[inline]
-    pub fn top(&self) -> *const u8 {
-        self.block.top
-    }
-
-    #[inline(always)]
-    fn lookup_str(&self, string: &str) -> Option<Atom> {
-        STATIC_ATOMS_MAP.get(string).or_else(|| self.table.get(string)).cloned()
-    }
-
-    pub fn build_with(&mut self, string: &str) -> Atom {
-        if let Some(atom) = self.lookup_str(string) {
-            return atom;
+    pub fn build_with(atom_table: &AtomTable, string: &str) -> Atom {
+        if !string.is_empty() && string.len() <= INLINED_ATOM_MAX_LEN && !string.contains('\u{0}') {
+            return Atom::new_inlined(string);
         }
 
-        unsafe {
+        loop {
+            let mut block_epoch = atom_table.inner.read();
+            let mut table_epoch = block_epoch.table.read();
+
+            if let Some(atom) = block_epoch.lookup_str(string) {
+                return atom;
+            }
+
+            // take a lock to prevent concurrent updates
+            let update_guard = atom_table.update.lock().unwrap();
+
+            let is_same_allocation = RcuRef::same_epoch(&block_epoch, &atom_table.inner.read());
+            let is_same_atom_list = RcuRef::same_epoch(&table_epoch, &block_epoch.table.read());
+
+            if !(is_same_allocation && is_same_atom_list) {
+                // some other thread raced us between our lookup and
+                // us aquring the update lock,
+                // try again
+                continue;
+            }
+
             let size = mem::size_of::<AtomHeader>() + string.len();
-            let align_offset = 8 * mem::align_of::<AtomHeader>();
-            let size = (size & !(align_offset - 1)) + align_offset;
+            let size = size.next_multiple_of(AtomTable::align());
 
-            let len_ptr = {
-                let mut ptr;
-
-                loop {
-                    ptr = self.block.alloc(size);
+            unsafe {
+                let len_ptr = loop {
+                    let ptr = block_epoch.block.alloc(size);
 
                     if ptr.is_null() {
-                        self.block.grow();
-                        set_atom_tbl_buf_base(self.block.base);
+                        // garbage collection would go here
+                        let new_block = block_epoch.block.grow_new().unwrap();
+                        let new_table = Arcu::new(table_epoch.clone(), GlobalEpochCounterPool);
+                        let new_alloc = InnerAtomTable {
+                            block: new_block,
+                            table: new_table,
+                        };
+                        atom_table.inner.replace(new_alloc);
+                        block_epoch = atom_table.inner.read();
+                        table_epoch = block_epoch.table.read();
                     } else {
-                        break;
+                        break ptr;
                     }
-                }
+                };
 
-                ptr
-            };
+                let ptr_base = block_epoch.block.base.addr();
 
-            let ptr_base = self.block.base as usize;
+                write_to_ptr(string, len_ptr);
 
-            write_to_ptr(string, len_ptr);
+                let atom = AtomCell::new()
+                    .with_name((STRINGS.len() + len_ptr.addr() - ptr_base) as u64)
+                    .with_arity(0)
+                    .with_f(false)
+                    .with_m(false)
+                    .with_is_inlined(false)
+                    .with_tag(HeapCellValueTag::Atom as u8)
+                    .get_name();
 
-            let atom = Atom {
-                index: (STRINGS.len() << 3) + len_ptr as usize - ptr_base,
-            };
+                let mut table = table_epoch.clone();
+                table.insert(atom);
+                block_epoch.table.replace(table);
 
-            self.table.insert(atom);
+                // expicit drop to ensure we don't accidentally drop it early
+                drop(update_guard);
 
-            atom
+                return atom;
+            }
         }
     }
 }
 
-#[bitfield]
-#[repr(u64)]
-#[derive(Copy, Clone, Debug)]
-pub struct AtomCell {
-    name: B46,
-    arity: B10,
-    #[allow(unused)] f: bool,
-    #[allow(unused)] m: bool,
-    #[allow(unused)] tag: B6,
-}
-
-impl AtomCell {
-    #[inline]
-    pub fn build_with(name: u64, arity: u16, tag: HeapCellValueTag) -> Self {
-        if arity > 0 {
-            debug_assert!(arity as usize <= MAX_ARITY);
-
-            AtomCell::new()
-                .with_name(name)
-                .with_arity(arity)
-                .with_f(false)
-                .with_tag(tag as u8)
-        } else {
-            AtomCell::new()
-                .with_name(name)
-                .with_f(false)
-                .with_tag(tag as u8)
-        }
-    }
-
-    #[inline]
-    pub fn get_index(self) -> usize {
-        self.name() as usize
-    }
-
-    #[inline]
-    pub fn get_name(self) -> Atom {
-        Atom::from(self.get_index() << 3)
-    }
-
-    #[inline]
-    pub fn get_arity(self) -> usize {
-        self.arity() as usize
-    }
-
-    #[inline]
-    pub fn get_name_and_arity(self) -> (Atom, usize) {
-        (Atom::from(self.get_index() << 3), self.get_arity())
-    }
-}
+unsafe impl Send for AtomTable {}
+unsafe impl Sync for AtomTable {}

src/bin/scryer-prolog.rs

@@ -1,11 +1,7 @@
-fn main() {
-    use std::sync::atomic::Ordering;
-    use scryer_prolog::*;
+fn main() -> std::process::ExitCode {
+    #[cfg(target_arch = "wasm32")]
+    return std::process::ExitCode::SUCCESS;
 
-    ctrlc::set_handler(move || {
-        scryer_prolog::machine::INTERRUPT.store(true, Ordering::Relaxed);
-    }).unwrap();
-
-    let mut wam = machine::Machine::new();
-    wam.run_top_level();
+    #[cfg(not(target_arch = "wasm32"))]
+    return scryer_prolog::run_binary();
 }

src/debray_allocator.rs

@@ -1,42 +1,280 @@
+use crate::allocator::*;
+use crate::codegen::SubsumedBranchHits;
+use crate::forms::{GenContext, Level};
+use crate::instructions::*;
+use crate::machine::disjuncts::VarData;
+use crate::parser::ast::*;
+use crate::targets::*;
+use crate::variable_records::*;
+
+use bit_set::*;
+use bitvec::prelude::*;
+use fxhash::FxBuildHasher;
 use indexmap::IndexMap;
 
-use crate::allocator::*;
-use crate::fixtures::*;
-use crate::forms::Level;
-use crate::instructions::*;
-use crate::machine::machine_indices::*;
-use crate::parser::ast::*;
-use crate::targets::CompilationTarget;
-
-use crate::temp_v;
-
-use fxhash::FxBuildHasher;
-
 use std::cell::Cell;
-use std::collections::BTreeSet;
-use std::rc::Rc;
+use std::collections::VecDeque;
+use std::ops::{Deref, DerefMut};
+
+pub type BranchHits = IndexMap<usize, BitVec, FxBuildHasher>; // key: var_num, value: branch arm occurrences.
+
+#[derive(Debug, Default)]
+pub struct BranchOccurrences {
+    pub hits: BranchHits,
+    pub shallow_safety: BitSet<usize>, // unset means safe, set means unsafe (after the branch merge)
+    pub deep_safety: BitSet<usize>,
+    pub num_branches: usize,
+    pub current_branch: usize,
+    pub subsumed_hits: SubsumedBranchHits,
+}
+
+impl BranchOccurrences {
+    fn new(num_branches: usize) -> Self {
+        Self {
+            hits: BranchHits::with_hasher(FxBuildHasher::default()),
+            shallow_safety: BitSet::default(),
+            deep_safety: BitSet::default(),
+            num_branches,
+            current_branch: 0,
+            subsumed_hits: SubsumedBranchHits::with_hasher(FxBuildHasher::default()),
+        }
+    }
+
+    pub(crate) fn add_branch_occurrence(&mut self, var_num: usize) {
+        debug_assert!(self.current_branch < self.num_branches);
+        let num_branches = self.num_branches;
+
+        let entry = self
+            .hits
+            .entry(var_num)
+            .or_insert_with(|| BitVec::repeat(false, num_branches));
+
+        entry.set(self.current_branch, true);
+        self.subsumed_hits.insert(var_num);
+    }
+}
+
+#[derive(Debug)]
+pub(crate) struct BranchStack {
+    stack: Vec<BranchOccurrences>,
+}
+
+impl Deref for BranchStack {
+    type Target = Vec<BranchOccurrences>;
+
+    #[inline]
+    fn deref(&self) -> &Self::Target {
+        &self.stack
+    }
+}
+
+impl DerefMut for BranchStack {
+    #[inline]
+    fn deref_mut(&mut self) -> &mut Self::Target {
+        &mut self.stack
+    }
+}
+
+impl BranchStack {
+    fn branch_subsumes(&self, branch: &BranchDesignator, sub_branch: &BranchDesignator) -> bool {
+        if branch.branch_stack_num < sub_branch.branch_stack_num {
+            if branch.branch_stack_num == 0 {
+                true
+            } else {
+                let idx = branch.branch_stack_num - 1;
+                self[idx].current_branch == branch.branch_num
+            }
+        } else {
+            branch == sub_branch
+        }
+    }
+
+    fn safety_unneeded_in_branch(
+        &self,
+        safety: &VarSafetyStatus,
+        branch: &BranchDesignator,
+    ) -> bool {
+        match safety {
+            VarSafetyStatus::Needed => false,
+            VarSafetyStatus::LocallyUnneeded(planter_branch) => {
+                self.branch_subsumes(planter_branch, branch)
+            }
+            VarSafetyStatus::GloballyUnneeded => true,
+        }
+    }
+
+    pub(crate) fn add_branch_occurrence(&mut self, var_num: usize) {
+        if let Some(occurrences) = self.last_mut() {
+            occurrences.add_branch_occurrence(var_num);
+        }
+    }
+
+    pub(crate) fn add_branch_stack(&mut self, num_branches: usize) {
+        self.push(BranchOccurrences::new(num_branches));
+    }
+
+    pub(crate) fn current_branch_designator(&self) -> BranchDesignator {
+        let branch_stack_num = self.len();
+        let branch_num = self
+            .last()
+            .map(|occurrences| occurrences.current_branch)
+            .unwrap_or(0);
+
+        BranchDesignator {
+            branch_stack_num,
+            branch_num,
+        }
+    }
+
+    #[inline]
+    pub(crate) fn incr_current_branch(&mut self) {
+        let branch_occurrences = self.last_mut().unwrap();
+        branch_occurrences.current_branch += 1;
+    }
+
+    #[inline]
+    pub(crate) fn drain_branches(
+        &mut self,
+        depth: usize,
+    ) -> std::vec::Drain<'_, BranchOccurrences> {
+        let start_idx = self.len() - depth;
+        self.drain(start_idx..)
+    }
+}
 
 #[derive(Debug)]
 pub(crate) struct DebrayAllocator {
-    bindings: IndexMap<Rc<String>, VarData, FxBuildHasher>,
+    pub(crate) var_data: VarData, // var_data replaces bindings.
+    pub(crate) branch_stack: BranchStack,
+    pub(crate) in_tail_position: bool,
     arg_c: usize,
     temp_lb: usize,
+    perm_lb: usize,
     arity: usize, // 0 if not at head.
-    contents: IndexMap<usize, Rc<String>, FxBuildHasher>,
-    in_use: BTreeSet<usize>,
+    shallow_temp_mappings: IndexMap<usize, usize, FxBuildHasher>,
+    in_use: BitSet<usize>, // deep and non-var allocations
+    temp_free_list: Vec<usize>,
+    perm_free_list: VecDeque<(usize, usize)>, // chunk_num, var_num
 }
 
 impl DebrayAllocator {
-    fn is_curr_arg_distinct_from(&self, var: &String) -> bool {
-        match self.contents.get(&self.arg_c) {
-            Some(t_var) if **t_var != *var => true,
+    pub(crate) fn add_branch(&mut self) {
+        let branch_designator = self.branch_stack.current_branch_designator();
+        let subsumed_hits = {
+            let branch_occurrences = self.branch_stack.last_mut().unwrap();
+
+            std::mem::replace(
+                &mut branch_occurrences.subsumed_hits,
+                SubsumedBranchHits::with_hasher(FxBuildHasher::default()),
+            )
+        };
+
+        for var_num in subsumed_hits {
+            match &mut self.var_data.records[var_num].allocation {
+                VarAlloc::Perm(_, ref mut allocation) => {
+                    if let PermVarAllocation::Done {
+                        shallow_safety,
+                        deep_safety,
+                        ..
+                    } = allocation
+                    {
+                        if !self
+                            .branch_stack
+                            .safety_unneeded_in_branch(shallow_safety, &branch_designator)
+                        {
+                            let branch_occurrences = self.branch_stack.last_mut().unwrap();
+                            branch_occurrences.shallow_safety.insert(var_num);
+                        }
+
+                        if !self
+                            .branch_stack
+                            .safety_unneeded_in_branch(deep_safety, &branch_designator)
+                        {
+                            let branch_occurrences = self.branch_stack.last_mut().unwrap();
+                            branch_occurrences.deep_safety.insert(var_num);
+                        }
+                    }
+
+                    *allocation = PermVarAllocation::Pending;
+                }
+                _ => unreachable!(),
+            }
+        }
+    }
+
+    pub(crate) fn pop_branch(&mut self, depth: usize, subsumed_hits: SubsumedBranchHits) {
+        let removed_branches = self.branch_stack.drain_branches(depth);
+
+        let (deep_safety, shallow_safety) = removed_branches.into_iter().fold(
+            (BitSet::default(), BitSet::default()),
+            |(mut deep_safety, mut shallow_safety), branch_occurrences| {
+                deep_safety.union_with(&branch_occurrences.deep_safety);
+                shallow_safety.union_with(&branch_occurrences.shallow_safety);
+
+                (deep_safety, shallow_safety)
+            },
+        );
+
+        let branch_designator = self.branch_stack.current_branch_designator();
+
+        let (deep_safety, shallow_safety) = match self.branch_stack.last_mut() {
+            Some(latest_branch) => {
+                latest_branch.deep_safety.union_with(&deep_safety);
+                latest_branch.shallow_safety.union_with(&shallow_safety);
+
+                (&latest_branch.deep_safety, &latest_branch.shallow_safety)
+            }
+            None => (&deep_safety, &shallow_safety),
+        };
+
+        for var_num in subsumed_hits.iter().cloned() {
+            let running_count = self.var_data.records[var_num].running_count;
+            let num_occurrences = self.var_data.records[var_num].num_occurrences;
+
+            match &mut self.var_data.records[var_num].allocation {
+                VarAlloc::Perm(_, allocation) => {
+                    let shallow_safety = VarSafetyStatus::needed_if(
+                        shallow_safety.contains(var_num),
+                        branch_designator,
+                    );
+
+                    let deep_safety = VarSafetyStatus::needed_if(
+                        deep_safety.contains(var_num),
+                        branch_designator,
+                    );
+
+                    if running_count < num_occurrences {
+                        *allocation = PermVarAllocation::Done {
+                            shallow_safety,
+                            deep_safety,
+                        };
+                    }
+                }
+                _ => unreachable!(),
+            }
+        }
+
+        if !self.branch_stack.is_empty() {
+            for var_num in subsumed_hits {
+                self.branch_stack.add_branch_occurrence(var_num);
+            }
+        }
+    }
+
+    fn is_curr_arg_distinct_from(&self, var_num: usize) -> bool {
+        match self.shallow_temp_mappings.get(&self.arg_c).cloned() {
+            Some(t_var) => t_var != var_num,
             _ => false,
         }
     }
 
-    fn occurs_shallowly_in_head(&self, var: &String, r: usize) -> bool {
-        match self.bindings.get(var).unwrap() {
-            &VarData::Temp(_, _, ref tvd) => tvd.use_set.contains(&(GenContext::Head, r)),
+    fn occurs_shallowly_in_head(&self, var_num: usize, r: usize) -> bool {
+        match &self.var_data.records[var_num].allocation {
+            VarAlloc::Temp {
+                temp_var_data,
+                term_loc: GenContext::Head,
+                ..
+            } => temp_var_data.use_set.contains(&(GenContext::Head, r)),
             _ => false,
         }
     }
@@ -44,13 +282,13 @@ impl DebrayAllocator {
     #[inline]
     fn is_in_use(&self, r: usize) -> bool {
         let in_use_range = r <= self.arity && r >= self.arg_c;
-        in_use_range || self.in_use.contains(&r)
+        in_use_range || self.in_use.contains(r)
     }
 
-    fn alloc_with_cr(&self, var: &String) -> usize {
-        match self.bindings.get(var) {
-            Some(&VarData::Temp(_, _, ref tvd)) => {
-                for &(_, reg) in tvd.use_set.iter() {
+    fn alloc_with_cr(&self, var_num: usize) -> usize {
+        match &self.var_data.records[var_num].allocation {
+            VarAlloc::Temp { temp_var_data, .. } => {
+                for &(_, reg) in temp_var_data.use_set.iter() {
                     if !self.is_in_use(reg) {
                         return reg;
                     }
@@ -59,11 +297,9 @@ impl DebrayAllocator {
                 let mut result = 0;
 
                 for reg in self.temp_lb.. {
-                    if !self.is_in_use(reg) {
-                        if !tvd.no_use_set.contains(&reg) {
-                            result = reg;
-                            break;
-                        }
+                    if !self.is_in_use(reg) && !temp_var_data.no_use_set.contains(reg) {
+                        result = reg;
+                        break;
                     }
                 }
 
@@ -73,10 +309,10 @@ impl DebrayAllocator {
         }
     }
 
-    fn alloc_with_ca(&self, var: &String) -> usize {
-        match self.bindings.get(var) {
-            Some(&VarData::Temp(_, _, ref tvd)) => {
-                for &(_, reg) in tvd.use_set.iter() {
+    fn alloc_with_ca(&self, var_num: usize) -> usize {
+        match &self.var_data.records[var_num].allocation {
+            VarAlloc::Temp { temp_var_data, .. } => {
+                for &(_, reg) in temp_var_data.use_set.iter() {
                     if !self.is_in_use(reg) {
                         return reg;
                     }
@@ -85,13 +321,12 @@ impl DebrayAllocator {
                 let mut result = 0;
 
                 for reg in self.temp_lb.. {
-                    if !self.is_in_use(reg) {
-                        if !tvd.no_use_set.contains(&reg) {
-                            if !tvd.conflict_set.contains(&reg) {
-                                result = reg;
-                                break;
-                            }
-                        }
+                    if !self.is_in_use(reg)
+                        && !temp_var_data.no_use_set.contains(reg)
+                        && !temp_var_data.conflict_set.contains(reg)
+                    {
+                        result = reg;
+                        break;
                     }
                 }
 
@@ -101,22 +336,26 @@ impl DebrayAllocator {
         }
     }
 
-    fn alloc_in_last_goal_hint(&self, chunk_num: usize) -> Option<(Rc<String>, usize)> {
+    fn alloc_in_last_goal_hint(&self, chunk_num: usize) -> Option<(usize, usize)> {
         // we want to allocate a register to the k^{th} parameter, par_k.
         // par_k may not be a temporary variable.
         let k = self.arg_c;
 
-        match self.contents.get(&k) {
+        match self.shallow_temp_mappings.get(&k).cloned() {
             Some(t_var) => {
                 // suppose this branch fires. then t_var is a
                 // temp. var. belonging to the current chunk.
                 // consider its use set. T == par_k iff
                 // (GenContext::Last(_), k) is in t_var.use_set.
 
-                let tvd = self.bindings.get(t_var).unwrap();
-                if let &VarData::Temp(_, _, ref tvd) = tvd {
-                    if !tvd.use_set.contains(&(GenContext::Last(chunk_num), k)) {
-                        return Some((t_var.clone(), self.alloc_with_ca(t_var)));
+                if let VarAlloc::Temp { temp_var_data, .. } =
+                    &self.var_data.records[t_var].allocation
+                {
+                    if !temp_var_data
+                        .use_set
+                        .contains(&(GenContext::Last(chunk_num), k))
+                    {
+                        return Some((t_var, self.alloc_with_ca(t_var)));
                     }
                 }
 
@@ -129,51 +368,50 @@ impl DebrayAllocator {
     fn evacuate_arg<'a, Target: CompilationTarget<'a>>(
         &mut self,
         chunk_num: usize,
-        code: &mut Code,
+        code: &mut CodeDeque,
     ) {
-        match self.alloc_in_last_goal_hint(chunk_num) {
-            Some((var, r)) => {
-                let k = self.arg_c;
+        if let Some((var_num, r)) = self.alloc_in_last_goal_hint(chunk_num) {
+            let k = self.arg_c;
 
-                if r != k {
-                    let r = RegType::Temp(r);
+            if r != k {
+                let r = RegType::Temp(r);
 
-                    code.push(Target::move_to_register(r, k));
+                code.push_back(Target::move_to_register(r, k));
 
-                    self.contents.swap_remove(&k);
-                    self.contents.insert(r.reg_num(), var.clone());
+                self.shallow_temp_mappings.swap_remove(&k);
+                self.shallow_temp_mappings.insert(r.reg_num(), var_num);
 
-                    self.record_register(var, r);
-                    self.in_use.insert(r.reg_num());
-                }
+                self.var_data.records[var_num]
+                    .allocation
+                    .set_register(r.reg_num());
+                self.in_use.insert(r.reg_num());
             }
-            _ => {}
         };
     }
 
     fn alloc_reg_to_var<'a, Target: CompilationTarget<'a>>(
         &mut self,
-        var: &String,
+        var_num: usize,
         lvl: Level,
         term_loc: GenContext,
-        target: &mut Vec<Instruction>,
+        target: &mut CodeDeque,
     ) -> usize {
         match term_loc {
             GenContext::Head => {
                 if let Level::Shallow = lvl {
                     self.evacuate_arg::<Target>(0, target);
-                    self.alloc_with_cr(var)
+                    self.alloc_with_cr(var_num)
                 } else {
-                    self.alloc_with_ca(var)
+                    self.alloc_with_ca(var_num)
                 }
             }
-            GenContext::Mid(_) => self.alloc_with_ca(var),
+            GenContext::Mid(_) => self.alloc_with_ca(var_num),
             GenContext::Last(chunk_num) => {
                 if let Level::Shallow = lvl {
                     self.evacuate_arg::<Target>(chunk_num, target);
-                    self.alloc_with_cr(var)
+                    self.alloc_with_cr(var_num)
                 } else {
-                    self.alloc_with_ca(var)
+                    self.alloc_with_ca(var_num)
                 }
             }
         }
@@ -182,38 +420,262 @@ impl DebrayAllocator {
     fn alloc_reg_to_non_var(&mut self) -> usize {
         let mut final_index = 0;
 
+        while let Some(r) = self.temp_free_list.pop() {
+            if !self.is_in_use(r) {
+                self.in_use.insert(r);
+                return r;
+            }
+        }
+
         for index in self.temp_lb.. {
-            if !self.in_use.contains(&index) {
+            if !self.in_use.contains(index) {
                 final_index = index;
+                self.in_use.insert(final_index);
                 break;
             }
         }
 
-        self.in_use.insert(final_index);
         self.temp_lb = final_index + 1;
         final_index
     }
 
-    fn in_place(&self, var: &String, term_loc: GenContext, r: RegType, k: usize) -> bool {
+    fn in_place(&self, var_num: usize, term_loc: GenContext, r: RegType, k: usize) -> bool {
         match term_loc {
             GenContext::Head if !r.is_perm() => r.reg_num() == k,
-            _ => match self.bindings().get(var).unwrap() {
-                &VarData::Temp(_, o, _) if r.reg_num() == k => o == k,
+            _ => match &self.var_data.records[var_num].allocation {
+                &VarAlloc::Temp { temp_reg, .. } if r.reg_num() == k => temp_reg == k,
                 _ => false,
             },
         }
     }
+
+    fn alloc_perm_var(&mut self, var_num: usize, chunk_num: usize) -> usize {
+        let p = if let Some(p) = self.pop_free_perm(chunk_num) {
+            p
+        } else {
+            let p = self.perm_lb;
+            self.perm_lb += 1;
+
+            p
+        };
+
+        self.var_data.records[var_num].allocation = VarAlloc::Perm(p, PermVarAllocation::done());
+        p
+    }
+
+    pub(crate) fn add_reg_to_free_list(&mut self, r: RegType) {
+        if let RegType::Temp(r) = r {
+            self.in_use.remove(r);
+            self.temp_free_list.push(r);
+        }
+    }
+
+    pub fn reset_free_list(&mut self) {
+        self.temp_free_list.clear();
+    }
+
+    #[inline(always)]
+    pub fn get_binding(&self, var_num: usize) -> RegType {
+        self.var_data.records[var_num].allocation.as_reg_type()
+    }
+
+    pub fn num_perm_vars(&self) -> usize {
+        self.perm_lb - 1
+    }
+
+    pub fn increment_running_count(&mut self, var_num: usize) {
+        self.var_data.records[var_num].running_count += 1;
+    }
+
+    fn add_perm_to_free_list(&mut self, chunk_num: usize, var_num: usize) {
+        if let VarAlloc::Perm(..) = &self.var_data.records[var_num].allocation {
+            self.perm_free_list.push_back((chunk_num, var_num));
+        }
+    }
+
+    fn pop_free_perm(&mut self, chunk_num: usize) -> Option<usize> {
+        while let Some((perm_chunk_num, var_num)) = self.perm_free_list.front().cloned() {
+            if chunk_num > perm_chunk_num {
+                self.perm_free_list.pop_front();
+
+                match &mut self.var_data.records[var_num].allocation {
+                    VarAlloc::Perm(p, PermVarAllocation::Pending) if *p > 0 => {
+                        return Some(std::mem::replace(p, 0));
+                    }
+                    _ => {}
+                }
+            } else {
+                return None;
+            }
+        }
+
+        None
+    }
+
+    pub(crate) fn free_var(&mut self, chunk_num: usize, var_num: usize) {
+        if let VarAlloc::Perm(_, allocation) = &mut self.var_data.records[var_num].allocation {
+            *allocation = PermVarAllocation::Pending;
+            self.add_perm_to_free_list(chunk_num, var_num);
+        }
+    }
+
+    pub(crate) fn mark_safe_var_unconditionally(&mut self, var_num: usize) {
+        let branch_designator = self.branch_stack.current_branch_designator();
+
+        match &mut self.var_data.records[var_num].allocation {
+            VarAlloc::Perm(
+                _,
+                PermVarAllocation::Done {
+                    deep_safety,
+                    shallow_safety,
+                    ..
+                },
+            ) => {
+                *deep_safety = VarSafetyStatus::unneeded(branch_designator);
+                *shallow_safety = VarSafetyStatus::unneeded(branch_designator);
+            }
+            VarAlloc::Temp { safety, .. } => {
+                *safety = VarSafetyStatus::unneeded(branch_designator);
+            }
+            _ => {
+                unreachable!()
+            }
+        }
+    }
+
+    fn mark_safe_var(&mut self, var_num: usize, lvl: Level, term_loc: GenContext) {
+        let branch_designator = self.branch_stack.current_branch_designator();
+
+        match &mut self.var_data.records[var_num].allocation {
+            VarAlloc::Perm(
+                _,
+                PermVarAllocation::Done {
+                    deep_safety,
+                    shallow_safety,
+                    ..
+                },
+            ) => {
+                // GetVariable in head chunk is considered safe.
+                if lvl == Level::Deep {
+                    *deep_safety = VarSafetyStatus::unneeded(branch_designator);
+                    *shallow_safety = VarSafetyStatus::unneeded(branch_designator);
+                } else if term_loc == GenContext::Head {
+                    *shallow_safety = VarSafetyStatus::GloballyUnneeded;
+                } else if let Some(&temp_var_num) = self.shallow_temp_mappings.get(&self.arg_c) {
+                    match &mut self.var_data.records[temp_var_num].allocation {
+                        VarAlloc::Temp {
+                            ref mut to_perm_var_num,
+                            ..
+                        } => {
+                            *to_perm_var_num = Some(var_num);
+                        }
+                        _ => unreachable!(),
+                    }
+                }
+            }
+            VarAlloc::Temp { ref mut safety, .. } => {
+                *safety = VarSafetyStatus::GloballyUnneeded;
+            }
+            _ => {
+                unreachable!()
+            }
+        }
+    }
+
+    fn argument_to_value<'a, Target: CompilationTarget<'a>>(
+        &mut self,
+        var_num: usize,
+        r: RegType,
+        arg_c: usize,
+    ) -> Instruction {
+        let branch_designator = self.branch_stack.current_branch_designator();
+
+        match &mut self.var_data.records[var_num].allocation {
+            VarAlloc::Perm(
+                _,
+                PermVarAllocation::Done {
+                    ref mut shallow_safety,
+                    ..
+                },
+            ) => {
+                if !self.in_tail_position
+                    || self
+                        .branch_stack
+                        .safety_unneeded_in_branch(shallow_safety, &branch_designator)
+                {
+                    Target::argument_to_value(r, arg_c)
+                } else {
+                    *shallow_safety = VarSafetyStatus::unneeded(branch_designator);
+                    Target::unsafe_argument_to_value(r, arg_c)
+                }
+            }
+            VarAlloc::Temp { .. } => {
+                debug_assert!(matches!(r, RegType::Temp(_)));
+                Target::argument_to_value(r, arg_c)
+            }
+            _ => {
+                unreachable!()
+            }
+        }
+    }
+
+    fn subterm_to_value<'a, Target: CompilationTarget<'a>>(
+        &mut self,
+        var_num: usize,
+        r: RegType,
+    ) -> Instruction {
+        let branch_designator = self.branch_stack.current_branch_designator();
+
+        match &mut self.var_data.records[var_num].allocation {
+            VarAlloc::Perm(
+                _,
+                PermVarAllocation::Done {
+                    ref mut deep_safety,
+                    ..
+                },
+            ) => {
+                if self
+                    .branch_stack
+                    .safety_unneeded_in_branch(deep_safety, &branch_designator)
+                {
+                    Target::subterm_to_value(r)
+                } else {
+                    *deep_safety = VarSafetyStatus::unneeded(branch_designator);
+                    Target::unsafe_subterm_to_value(r)
+                }
+            }
+            VarAlloc::Temp { ref mut safety, .. } => {
+                if self
+                    .branch_stack
+                    .safety_unneeded_in_branch(safety, &branch_designator)
+                {
+                    Target::subterm_to_value(r)
+                } else {
+                    *safety = VarSafetyStatus::unneeded(branch_designator);
+                    Target::unsafe_subterm_to_value(r)
+                }
+            }
+            _ => {
+                unreachable!()
+            }
+        }
+    }
 }
 
 impl Allocator for DebrayAllocator {
     fn new() -> DebrayAllocator {
-        DebrayAllocator {
+        Self {
+            var_data: VarData::default(),
+            in_tail_position: false,
             arity: 0,
             arg_c: 1,
             temp_lb: 1,
-            bindings: IndexMap::with_hasher(FxBuildHasher::default()),
-            contents: IndexMap::with_hasher(FxBuildHasher::default()),
-            in_use: BTreeSet::new(),
+            perm_lb: 1,
+            shallow_temp_mappings: IndexMap::with_hasher(FxBuildHasher::default()),
+            in_use: BitSet::default(),
+            temp_free_list: vec![],
+            perm_free_list: VecDeque::new(),
+            branch_stack: BranchStack { stack: vec![] },
         }
     }
 
@@ -221,12 +683,12 @@ impl Allocator for DebrayAllocator {
         &mut self,
         lvl: Level,
         term_loc: GenContext,
-        code: &mut Code,
-    ) {
+        code: &mut CodeDeque,
+    ) -> RegType {
         let r = RegType::Temp(self.alloc_reg_to_non_var());
 
         match lvl {
-            Level::Deep => code.push(Target::subterm_to_variable(r)),
+            Level::Deep => code.push_back(Target::subterm_to_variable(r)),
             Level::Root | Level::Shallow => {
                 let k = self.arg_c;
 
@@ -236,9 +698,11 @@ impl Allocator for DebrayAllocator {
 
                 self.arg_c += 1;
 
-                code.push(Target::argument_to_variable(r, k));
+                code.push_back(Target::argument_to_variable(r, k));
             }
         };
+
+        r
     }
 
     fn mark_non_var<'a, Target: CompilationTarget<'a>>(
@@ -246,7 +710,7 @@ impl Allocator for DebrayAllocator {
         lvl: Level,
         term_loc: GenContext,
         cell: &'a Cell<RegType>,
-        code: &mut Code,
+        code: &mut CodeDeque,
     ) {
         let r = cell.get();
 
@@ -273,39 +737,51 @@ impl Allocator for DebrayAllocator {
 
     fn mark_var<'a, Target: CompilationTarget<'a>>(
         &mut self,
-        var: Rc<String>,
+        var_num: usize,
         lvl: Level,
-        cell: &'a Cell<VarReg>,
+        cell: &Cell<VarReg>,
         term_loc: GenContext,
-        code: &mut Code,
+        code: &mut CodeDeque,
     ) {
-        let (r, is_new_var) = match self.get(var.clone()) {
+        let (r, is_new_var) = match self.get_binding(var_num) {
             RegType::Temp(0) => {
-                // here, r is temporary *and* unassigned.
-                let o = self.alloc_reg_to_var::<Target>(&var, lvl, term_loc, code);
+                let o = self.alloc_reg_to_var::<Target>(var_num, lvl, term_loc, code);
                 cell.set(VarReg::Norm(RegType::Temp(o)));
-
                 (RegType::Temp(o), true)
             }
             RegType::Perm(0) => {
-                let pr = cell.get().norm();
-                self.record_register(var.clone(), pr);
+                let p = self.alloc_perm_var(var_num, term_loc.chunk_num());
+                cell.set(VarReg::Norm(RegType::Perm(p)));
+                (RegType::Perm(p), true)
+            }
+            r @ RegType::Perm(_) => {
+                let is_new_var = match &mut self.var_data.records[var_num].allocation {
+                    VarAlloc::Perm(_, allocation) => {
+                        if allocation.pending() {
+                            *allocation = PermVarAllocation::done();
+                            true
+                        } else {
+                            false
+                        }
+                    }
+                    _ => unreachable!(),
+                };
 
-                (pr, true)
+                (r, is_new_var)
             }
             r => (r, false),
         };
 
-        self.mark_reserved_var::<Target>(var, lvl, cell, term_loc, code, r, is_new_var);
+        self.mark_reserved_var::<Target>(var_num, lvl, cell, term_loc, code, r, is_new_var);
     }
 
     fn mark_reserved_var<'a, Target: CompilationTarget<'a>>(
         &mut self,
-        var: Rc<String>,
+        var_num: usize,
         lvl: Level,
-        cell: &'a Cell<VarReg>,
+        cell: &Cell<VarReg>,
         term_loc: GenContext,
-        code: &mut Code,
+        code: &mut CodeDeque,
         r: RegType,
         is_new_var: bool,
     ) {
@@ -313,84 +789,114 @@ impl Allocator for DebrayAllocator {
             Level::Root | Level::Shallow => {
                 let k = self.arg_c;
 
-                if self.is_curr_arg_distinct_from(&var) {
+                if self.is_curr_arg_distinct_from(var_num) {
                     self.evacuate_arg::<Target>(term_loc.chunk_num(), code);
                 }
 
-                self.arg_c += 1;
-
                 cell.set(VarReg::ArgAndNorm(r, k));
 
-                if !self.in_place(&var, term_loc, r, k) {
+                if !self.in_place(var_num, term_loc, r, k) {
                     if is_new_var {
-                        code.push(Target::argument_to_variable(r, k));
+                        self.mark_safe_var(var_num, lvl, term_loc);
+                        code.push_back(Target::argument_to_variable(r, k));
                     } else {
-                        code.push(Target::argument_to_value(r, k));
+                        code.push_back(self.argument_to_value::<Target>(var_num, r, k));
                     }
                 }
+
+                self.arg_c += 1;
             }
             Level::Deep if is_new_var => {
                 if let GenContext::Head = term_loc {
-                    if self.occurs_shallowly_in_head(&var, r.reg_num()) {
-                        code.push(Target::subterm_to_value(r));
+                    if self.occurs_shallowly_in_head(var_num, r.reg_num()) {
+                        code.push_back(self.subterm_to_value::<Target>(var_num, r));
                     } else {
-                        code.push(Target::subterm_to_variable(r));
+                        self.mark_safe_var(var_num, lvl, term_loc);
+                        code.push_back(Target::subterm_to_variable(r));
                     }
                 } else {
-                    code.push(Target::subterm_to_variable(r));
+                    self.mark_safe_var(var_num, lvl, term_loc);
+                    code.push_back(Target::subterm_to_variable(r));
                 }
             }
-            Level::Deep => code.push(Target::subterm_to_value(r)),
-        };
+            Level::Deep => code.push_back(self.subterm_to_value::<Target>(var_num, r)),
+        }
+
+        let o = r.reg_num();
 
         if !r.is_perm() {
-            let o = r.reg_num();
+            self.shallow_temp_mappings.insert(o, var_num);
+        } else if r.is_perm() && is_new_var {
+            self.branch_stack.add_branch_occurrence(var_num);
+        }
 
-            self.contents.insert(o, var.clone());
-            self.record_register(var.clone(), r);
-            self.in_use.insert(o);
+        let record = &mut self.var_data.records[var_num];
+
+        record.allocation.set_register(o);
+
+        if record.running_count < record.num_occurrences {
+            record.running_count += 1;
+        } else {
+            self.free_var(term_loc.chunk_num(), var_num);
+        }
+
+        self.in_use.insert(o);
+    }
+
+    fn mark_cut_var(&mut self, var_num: usize, chunk_num: usize) -> RegType {
+        match self.get_binding(var_num) {
+            RegType::Perm(0) => RegType::Perm(self.alloc_perm_var(var_num, chunk_num)),
+            RegType::Temp(0) => {
+                let t = self.alloc_reg_to_non_var();
+
+                match &mut self.var_data.records[var_num].allocation {
+                    VarAlloc::Temp {
+                        temp_reg, safety, ..
+                    } => {
+                        *temp_reg = t;
+                        *safety = VarSafetyStatus::GloballyUnneeded;
+                    }
+                    _ => unreachable!(),
+                };
+
+                RegType::Temp(t)
+            }
+            r => r,
         }
     }
 
     fn reset(&mut self) {
-        self.bindings.clear();
-        self.contents.clear();
+        self.perm_lb = 1;
+        self.shallow_temp_mappings.clear();
         self.in_use.clear();
+        self.temp_free_list.clear();
     }
 
     fn reset_contents(&mut self) {
-        self.contents.clear();
         self.in_use.clear();
+        self.shallow_temp_mappings.clear();
+        self.temp_free_list.clear();
     }
 
     fn advance_arg(&mut self) {
         self.arg_c += 1;
     }
 
-    fn bindings(&self) -> &AllocVarDict {
-        &self.bindings
-    }
-
-    fn bindings_mut(&mut self) -> &mut AllocVarDict {
-        &mut self.bindings
-    }
-
-    fn take_bindings(self) -> AllocVarDict {
-        self.bindings
-    }
-
-    fn reset_at_head(&mut self, args: &Vec<Term>) {
+    fn reset_at_head(&mut self, args: &[Term]) {
         self.reset_arg(args.len());
         self.arity = args.len();
 
         for (idx, arg) in args.iter().enumerate() {
-            if let &Term::Var(_, ref var) = arg {
-                let r = self.get(var.clone());
+            if let Term::Var(_, ref var) = arg {
+                let var_num = var.to_var_num().unwrap();
+                let r = self.get_binding(var_num);
 
                 if !r.is_perm() && r.reg_num() == 0 {
                     self.in_use.insert(idx + 1);
-                    self.contents.insert(idx + 1, var.clone());
-                    self.record_register(var.clone(), temp_v!(idx + 1));
+                    self.shallow_temp_mappings.insert(idx + 1, var_num);
+                    self.var_data.records[var_num]
+                        .allocation
+                        .set_register(idx + 1);
                 }
             }
         }

src/ffi.rs

@@ -0,0 +1,936 @@
+/* How does FFI work?
+
+Each WAM machine has a ForeignFunctionTable instance that contains a table of functions and structs.
+
+Structs are defined via foreign_struct/2. Basic types are defined by libffi, but struct types need to
+be manually defined to get an ffi_type. Additionally, to recover structs from return arguments, we store
+fields and atom_fields, as a way to lookup the content of the struct (fields) and the nested structs (atom_fields).
+
+Functions are defined via use_foreign_module/2. It opens a library and leaks the memory of the library,
+to prevent Rust freeing the memory. There's no way to recover that memory at the moment. We get a pointer for
+each function and we build a CIF for each one, with the input arguments and the return argument.
+
+Exec happens via '$foreign_call', we find the function, we try to cast the values that we have to the definition
+of the function, we reserve memory for them and we build an array of pointers. To get the return argument, we
+reserve enough memory for the return and we build the Scryer values from them.
+
+Structs are a bit tricky as they need to be aligned. For that, we reserve enough memory (libffi calculates that)
+and for each field: we add to the pointer until we're aligned to the next data type we're going to write, we write it,
+and finally we add the pointer the size of what we've written.
+*/
+
+use crate::arena::Arena;
+use crate::atom_table::Atom;
+use crate::forms::Number;
+use crate::machine::machine_errors::DomainErrorType;
+use crate::parser::ast::{Fixnum, MightNotFitInFixnum};
+
+use dashu::Integer;
+use libffi::middle::{Arg, Cif, CodePtr, Type};
+use libloading::{Library, Symbol};
+use ordered_float::OrderedFloat;
+use std::alloc::{self, Layout};
+use std::collections::HashMap;
+use std::error::Error;
+use std::ffi::{c_char, c_void, CStr, CString};
+use std::fmt::Debug;
+use std::marker::PhantomData;
+use std::mem::ManuallyDrop;
+use std::ops::Deref;
+use std::ptr::NonNull;
+
+pub struct FunctionDefinition {
+    pub name: Atom,
+    pub return_value: Atom,
+    pub args: Vec<Atom>,
+}
+
+#[derive(Debug)]
+pub struct FunctionImpl {
+    cif: Cif,
+    args: Vec<FfiType>,
+    code_ptr: CodePtr,
+    return_type: FfiType,
+}
+
+impl FunctionImpl {
+    unsafe fn call_void(&self, args: &[Arg], _: &mut Arena) -> Result<Value, FfiError> {
+        self.cif.call::<()>(self.code_ptr, args);
+        Ok(Value::Number(Number::Fixnum(Fixnum::build_with(0))))
+    }
+
+    unsafe fn call_int<T>(&self, args: &[Arg], arena: &mut Arena) -> Result<Value, FfiError>
+    where
+        Integer: From<T>,
+        T: Copy + TryInto<i64> + MightNotFitInFixnum,
+    {
+        let n = self.cif.call::<T>(self.code_ptr, args);
+        Ok(Value::Number(fixnum!(Number, n, arena)))
+    }
+
+    unsafe fn call_float<T>(&self, args: &[Arg], _: &mut Arena) -> Result<Value, FfiError>
+    where
+        T: Into<f64>,
+    {
+        let n = self.cif.call::<T>(self.code_ptr, args);
+        Ok(Value::Number(Number::Float(OrderedFloat(n.into()))))
+    }
+
+    unsafe fn call_ptr(&self, args: &[Arg], arena: &mut Arena) -> Result<Value, FfiError> {
+        let ptr = unsafe { self.cif.call::<*mut c_void>(self.code_ptr, args) };
+        Ok(Value::Number(fixnum!(Number, ptr as isize, arena)))
+    }
+
+    unsafe fn call_cstr(&self, args: &[Arg], _: &mut Arena) -> Result<Value, FfiError> {
+        let ptr = unsafe {
+            self.cif
+                .call::<Option<NonNull<c_char>>>(self.code_ptr, args)
+        };
+
+        if let Some(cstr) = ptr {
+            Ok(Value::CString(
+                unsafe { CStr::from_ptr(cstr.as_ptr()) }.to_owned(),
+            ))
+        } else {
+            Ok(Value::Number(Number::Fixnum(Fixnum::build_with(0))))
+        }
+    }
+
+    unsafe fn call_struct(
+        &self,
+        return_type_name: Atom,
+        args: &[Arg],
+        arena: &mut Arena,
+        structs_table: &HashMap<Atom, StructImpl>,
+    ) -> Result<Value, FfiError> {
+        let struct_type = structs_table
+            .get(&return_type_name)
+            .ok_or(FfiError::StructNotFound(return_type_name))?;
+        let ffi_type = unsafe { *struct_type.ffi_type.as_raw_ptr() };
+
+        let layout = Layout::from_size_align(ffi_type.size, ffi_type.alignment.into())
+            .map_err(|_| FfiError::LayoutError)?;
+
+        let alloc = FfiStruct::new(layout, FfiAllocator::Rust)?;
+
+        unsafe {
+            libffi::raw::ffi_call(
+                self.cif.as_raw_ptr(),
+                Some(*self.code_ptr.as_safe_fun()),
+                alloc.ptr.as_ptr(),
+                args.as_ptr() as *mut *mut c_void,
+            )
+        };
+
+        let struct_val =
+            struct_type.read(alloc.ptr.as_ptr(), return_type_name, structs_table, arena);
+
+        drop(alloc);
+
+        struct_val
+    }
+
+    fn call(
+        &self,
+        args: &[Arg],
+        arena: &mut Arena,
+        structs_table: &HashMap<Atom, StructImpl>,
+    ) -> Result<Value, FfiError> {
+        let call_fn: unsafe fn(&Self, &[Arg], &mut Arena) -> Result<Value, FfiError> =
+            match self.return_type {
+                FfiType::Void => FunctionImpl::call_void,
+                FfiType::U8 => FunctionImpl::call_int::<u8>,
+                FfiType::I8 | FfiType::Bool => FunctionImpl::call_int::<i8>,
+                FfiType::U16 => FunctionImpl::call_int::<u16>,
+                FfiType::I16 => FunctionImpl::call_int::<i16>,
+                FfiType::U32 => FunctionImpl::call_int::<u32>,
+                FfiType::I32 => FunctionImpl::call_int::<i32>,
+                FfiType::U64 => FunctionImpl::call_int::<u64>,
+                FfiType::I64 => FunctionImpl::call_int::<i64>,
+                FfiType::F32 => FunctionImpl::call_float::<f32>,
+                FfiType::F64 => FunctionImpl::call_float::<f64>,
+                FfiType::Ptr => FunctionImpl::call_ptr,
+                FfiType::CStr => FunctionImpl::call_cstr,
+                FfiType::Struct(name) => {
+                    return unsafe { self.call_struct(name, args, arena, structs_table) }
+                }
+            };
+        unsafe { call_fn(self, args, arena) }
+    }
+}
+
+#[derive(Debug, Default)]
+pub struct ForeignFunctionTable {
+    table: HashMap<Atom, FunctionImpl>,
+    structs: HashMap<Atom, StructImpl>,
+}
+
+#[derive(Clone, Debug)]
+struct StructImpl {
+    ffi_type: Type,
+    fields: Vec<FfiType>,
+}
+
+impl StructImpl {
+    fn layout(&self) -> Result<Layout, FfiError> {
+        let ffi_type = unsafe { *self.ffi_type.as_raw_ptr() };
+        Layout::from_size_align(ffi_type.size, ffi_type.alignment.into())
+            .map_err(|_| FfiError::LayoutError)
+    }
+
+    fn build(
+        &self,
+        structs_table: &HashMap<Atom, StructImpl>,
+        struct_args: &mut [Value],
+    ) -> Result<FfiStruct, FfiError> {
+        let args = ArgValue::build_args(struct_args, &self.fields, structs_table)?;
+
+        let alloc = FfiStruct::new(self.layout()?, FfiAllocator::Rust)?;
+
+        let Ok(mut current_layout) = Layout::from_size_align(0, 1) else {
+            return Err(FfiError::LayoutError);
+        };
+
+        unsafe fn write_primitive<T>(
+            ptr: NonNull<c_void>,
+            layout: &mut Layout,
+            val: T,
+        ) -> Result<(), FfiError> {
+            let (new_layout, offset) = layout
+                .extend(Layout::new::<T>())
+                .map_err(|_| FfiError::LayoutError)?;
+            *layout = new_layout;
+            ptr.byte_offset(offset as isize).cast::<T>().write(val);
+            Ok(())
+        }
+
+        for arg in args {
+            unsafe {
+                match arg {
+                    ArgValue::U8(i) => write_primitive(alloc.ptr, &mut current_layout, i)?,
+                    ArgValue::I8(i) => write_primitive(alloc.ptr, &mut current_layout, i)?,
+                    ArgValue::U16(i) => write_primitive(alloc.ptr, &mut current_layout, i)?,
+                    ArgValue::I16(i) => write_primitive(alloc.ptr, &mut current_layout, i)?,
+                    ArgValue::U32(i) => write_primitive(alloc.ptr, &mut current_layout, i)?,
+                    ArgValue::I32(i) => write_primitive(alloc.ptr, &mut current_layout, i)?,
+                    ArgValue::U64(i) => write_primitive(alloc.ptr, &mut current_layout, i)?,
+                    ArgValue::I64(i) => write_primitive(alloc.ptr, &mut current_layout, i)?,
+                    ArgValue::F32(f) => write_primitive(alloc.ptr, &mut current_layout, f)?,
+                    ArgValue::F64(f) => write_primitive(alloc.ptr, &mut current_layout, f)?,
+                    ArgValue::Ptr(p, _) => write_primitive(alloc.ptr, &mut current_layout, p)?,
+                    ArgValue::Struct(arg) => {
+                        let Ok((new_layout, offset)) = current_layout.extend(arg.layout) else {
+                            return Err(FfiError::LayoutError);
+                        };
+
+                        current_layout = new_layout;
+
+                        std::ptr::copy(
+                            arg.ptr.as_ptr(),
+                            alloc.ptr.byte_offset(offset as isize).as_ptr(),
+                            arg.layout.size(),
+                        );
+                    }
+                }
+            }
+        }
+
+        if alloc.layout != current_layout.pad_to_align() {
+            // sanity check
+            return Err(FfiError::LayoutError);
+        }
+
+        Ok(alloc)
+    }
+
+    fn read(
+        &self,
+        ptr: *mut c_void,
+        struct_name: Atom,
+        struct_table: &HashMap<Atom, StructImpl>,
+        arena: &mut Arena,
+    ) -> Result<Value, FfiError> {
+        unsafe {
+            let mut returns = Vec::new();
+
+            unsafe fn read_primitive<T>(
+                ptr: *mut c_void,
+                layout: &mut Layout,
+            ) -> Result<T, FfiError> {
+                let (new_layout, offset) = layout
+                    .extend(Layout::new::<T>())
+                    .map_err(|_| FfiError::LayoutError)?;
+                *layout = new_layout;
+                let n = std::ptr::read::<T>(ptr.byte_offset(offset as isize).cast());
+                Ok(n)
+            }
+
+            unsafe fn read_int<T>(
+                ptr: *mut c_void,
+                layout: &mut Layout,
+                arena: &mut Arena,
+            ) -> Result<Value, FfiError>
+            where
+                T: Copy + TryInto<i64> + MightNotFitInFixnum,
+                Integer: From<T>,
+            {
+                let n = read_primitive::<T>(ptr, layout)?;
+                Ok(Value::Number(fixnum!(Number, n, arena)))
+            }
+
+            unsafe fn read_float<T>(
+                ptr: *mut c_void,
+                layout: &mut Layout,
+            ) -> Result<Value, FfiError>
+            where
+                T: Into<f64>,
+            {
+                let n = read_primitive::<T>(ptr, layout)?;
+                Ok(Value::Number(Number::Float(OrderedFloat(n.into()))))
+            }
+
+            let mut layout = Layout::from_size_align(0, 1).map_err(|_| FfiError::LayoutError)?;
+
+            for field_type in &self.fields {
+                let val = match field_type {
+                    FfiType::U8 => read_int::<u8>(ptr, &mut layout, arena),
+                    FfiType::I8 | FfiType::Bool => read_int::<i8>(ptr, &mut layout, arena),
+                    FfiType::U16 => read_int::<u16>(ptr, &mut layout, arena),
+                    FfiType::I16 => read_int::<i16>(ptr, &mut layout, arena),
+                    FfiType::U32 => read_int::<u32>(ptr, &mut layout, arena),
+                    FfiType::I32 => read_int::<i32>(ptr, &mut layout, arena),
+                    FfiType::U64 => read_int::<u64>(ptr, &mut layout, arena),
+                    FfiType::I64 => read_int::<i64>(ptr, &mut layout, arena),
+                    FfiType::Ptr => {
+                        let ptr = read_primitive::<*mut c_void>(ptr, &mut layout)?;
+                        Ok(Value::Number(fixnum!(Number, ptr as isize, arena)))
+                    }
+                    FfiType::CStr => {
+                        let ptr = read_primitive::<*mut c_void>(ptr, &mut layout)?;
+                        Ok(Value::CString(CStr::from_ptr(ptr.cast()).to_owned()))
+                    }
+                    FfiType::F32 => read_float::<f32>(ptr, &mut layout),
+                    FfiType::F64 => read_float::<f64>(ptr, &mut layout),
+                    FfiType::Struct(substruct) => {
+                        let Some(substruct_type) = struct_table.get(substruct) else {
+                            return Err(FfiError::StructNotFound(*substruct));
+                        };
+
+                        let ffi_type = *substruct_type.ffi_type.as_raw_ptr();
+                        let field_layout =
+                            Layout::from_size_align(ffi_type.size, ffi_type.alignment as usize)
+                                .map_err(|_| FfiError::LayoutError)?;
+                        let (new_layout, offset) = layout
+                            .extend(field_layout)
+                            .map_err(|_| FfiError::LayoutError)?;
+                        layout = new_layout;
+                        let field_ptr = ptr.byte_offset(offset as isize);
+                        let struct_val =
+                            substruct_type.read(field_ptr, *substruct, struct_table, arena)?;
+                        Ok(struct_val)
+                    }
+                    FfiType::Void => return Err(FfiError::VoidArgumentType),
+                };
+                returns.push(val?);
+            }
+            Ok(Value::Struct(struct_name, returns))
+        }
+    }
+}
+
+struct PointerArgs<'a, 'val> {
+    memory: Vec<Arg>,
+    phantom: PhantomData<&'a mut ArgValue<'val>>,
+}
+
+impl<'args, 'val> PointerArgs<'args, 'val> {
+    fn new(args: &'args [ArgValue<'val>]) -> Self {
+        let args = args
+            .iter()
+            .map(|arg| match arg {
+                ArgValue::U8(a) => libffi::middle::arg(a),
+                ArgValue::I8(a) => libffi::middle::arg(a),
+                ArgValue::U16(a) => libffi::middle::arg(a),
+                ArgValue::I16(a) => libffi::middle::arg(a),
+                ArgValue::U32(a) => libffi::middle::arg(a),
+                ArgValue::I32(a) => libffi::middle::arg(a),
+                ArgValue::U64(a) => libffi::middle::arg(a),
+                ArgValue::I64(a) => libffi::middle::arg(a),
+                ArgValue::F32(a) => libffi::middle::arg(a),
+                ArgValue::F64(a) => libffi::middle::arg(a),
+                ArgValue::Ptr(ptr, _) => Arg::new(ptr),
+                ArgValue::Struct(s) => unsafe {
+                    std::mem::transmute::<*mut c_void, Arg>(s.ptr.as_ptr())
+                },
+            })
+            .collect();
+
+        PointerArgs {
+            memory: args,
+            phantom: PhantomData,
+        }
+    }
+}
+
+impl Deref for PointerArgs<'_, '_> {
+    type Target = [Arg];
+
+    fn deref(&self) -> &Self::Target {
+        &self.memory
+    }
+}
+
+#[derive(Debug, Clone, Copy)]
+enum FfiType {
+    Void,
+    Bool,
+    U8,
+    I8,
+    U16,
+    I16,
+    U32,
+    I32,
+    U64,
+    I64,
+    F32,
+    F64,
+    Ptr,
+    CStr,
+    Struct(Atom),
+}
+
+impl FfiType {
+    fn from_atom(atom: &Atom) -> Self {
+        match atom {
+            atom!("sint64") | atom!("i64") => Self::I64,
+            atom!("sint32") | atom!("i32") => Self::I32,
+            atom!("sint16") | atom!("i16") => Self::I16,
+            atom!("sint8") | atom!("i8") => Self::I8,
+            atom!("uint64") | atom!("u64") => Self::U64,
+            atom!("uint32") | atom!("u32") => Self::U32,
+            atom!("uint16") | atom!("u16") => Self::U16,
+            atom!("uint8") | atom!("u8") => Self::U8,
+            atom!("bool") => Self::Bool,
+            atom!("void") => Self::Void,
+            atom!("cstr") => Self::CStr,
+            atom!("ptr") => Self::Ptr,
+            atom!("f32") => Self::F32,
+            atom!("f64") => Self::F64,
+            struct_name => Self::Struct(*struct_name),
+        }
+    }
+
+    fn to_type(self, structs_table: &HashMap<Atom, StructImpl>) -> Result<Type, FfiError> {
+        Ok(match self {
+            Self::I64 => libffi::middle::Type::i64(),
+            Self::I32 => libffi::middle::Type::i32(),
+            Self::I16 => libffi::middle::Type::i16(),
+            Self::I8 => libffi::middle::Type::i8(),
+            Self::U64 => libffi::middle::Type::u64(),
+            Self::U32 => libffi::middle::Type::u32(),
+            Self::U16 => libffi::middle::Type::u16(),
+            Self::U8 => libffi::middle::Type::u8(),
+            Self::Bool => libffi::middle::Type::i8(),
+            Self::Void => libffi::middle::Type::void(),
+            Self::CStr => libffi::middle::Type::pointer(),
+            Self::Ptr => libffi::middle::Type::pointer(),
+            Self::F32 => libffi::middle::Type::f32(),
+            Self::F64 => libffi::middle::Type::f64(),
+            Self::Struct(struct_name) => structs_table
+                .get(&struct_name)
+                .ok_or(FfiError::StructNotFound(struct_name))?
+                .ffi_type
+                .clone(),
+        })
+    }
+}
+
+enum ArgValue<'a> {
+    U8(u8),
+    I8(i8),
+    U16(u16),
+    I16(i16),
+    U32(u32),
+    I32(i32),
+    U64(u64),
+    I64(i64),
+    F32(f32),
+    F64(f64),
+    Ptr(*mut c_void, PhantomData<&'a CString>),
+    Struct(FfiStruct),
+}
+
+impl<'val> ArgValue<'val> {
+    fn new(
+        val: &'val mut Value,
+        arg_type: &FfiType,
+        structs_table: &HashMap<Atom, StructImpl>,
+    ) -> Result<Self, FfiError> {
+        match arg_type {
+            FfiType::U8 => Ok(Self::U8(val.as_int()?)),
+            FfiType::I8 | FfiType::Bool => Ok(Self::I8(val.as_int()?)),
+            FfiType::U16 => Ok(Self::U16(val.as_int()?)),
+            FfiType::I16 => Ok(Self::I16(val.as_int()?)),
+            FfiType::U32 => Ok(Self::U32(val.as_int()?)),
+            FfiType::I32 => Ok(Self::I32(val.as_int()?)),
+            FfiType::U64 => Ok(Self::U64(val.as_int()?)),
+            FfiType::I64 => Ok(Self::I64(val.as_int()?)),
+            FfiType::F32 => Ok(Self::F32(val.as_float()? as f32)),
+            FfiType::F64 => Ok(Self::F64(val.as_float()?)),
+            FfiType::Ptr => Ok(Self::Ptr(val.as_ptr()?, PhantomData)),
+            FfiType::CStr => Ok(Self::Ptr(val.as_ptr()?, PhantomData)),
+            FfiType::Struct(arg_type_name) => {
+                let (val_type_name, args) = val.as_struct()?;
+
+                if *arg_type_name != val_type_name {
+                    return Err(FfiError::ValueCast(*arg_type_name, val_type_name));
+                }
+
+                let Some(struct_type) = structs_table.get(&val_type_name) else {
+                    return Err(FfiError::StructNotFound(*arg_type_name));
+                };
+
+                Ok(Self::Struct(struct_type.build(structs_table, args)?))
+            }
+            FfiType::Void => Err(FfiError::VoidArgumentType),
+        }
+    }
+
+    fn build_args(
+        args: &'val mut [Value],
+        types: &[FfiType],
+        structs_table: &HashMap<Atom, StructImpl>,
+    ) -> Result<Vec<Self>, FfiError> {
+        if types.len() != args.len() {
+            return Err(FfiError::ArgCountMismatch);
+        }
+
+        args.iter_mut()
+            .zip(types)
+            .map(|(arg, arg_type)| ArgValue::new(arg, arg_type, structs_table))
+            .collect::<Result<Vec<_>, _>>()
+    }
+}
+
+struct FfiStruct {
+    ptr: NonNull<c_void>,
+    layout: Layout,
+    allocator: FfiAllocator,
+}
+
+#[derive(Debug, Clone, Copy)]
+pub(crate) enum FfiAllocator {
+    Rust,
+    C,
+}
+
+impl TryFrom<Atom> for FfiAllocator {
+    type Error = ();
+
+    fn try_from(value: Atom) -> Result<Self, Self::Error> {
+        match value {
+            atom!("rust") => Ok(Self::Rust),
+            atom!("c") => Ok(Self::C),
+            _ => Err(()),
+        }
+    }
+}
+
+impl FfiAllocator {
+    /// # Safety
+    ///
+    /// - layout must not have a size of 0
+    unsafe fn alloc(self, layout: Layout) -> Result<NonNull<c_void>, FfiError> {
+        let ptr = match self {
+            FfiAllocator::Rust => unsafe { alloc::alloc(layout).cast() },
+            FfiAllocator::C => unsafe { libc::malloc(layout.size()) },
+        };
+
+        NonNull::new(ptr).ok_or(FfiError::AllocationFailed)
+    }
+
+    /// # Safety
+    ///
+    /// - ptr must point to an allocation currently allocated by this allocator
+    /// - layout must match the layout that was used to allocate the allocation pointed to by ptr
+    unsafe fn dealloc(self, layout: Layout, ptr: NonNull<c_void>) {
+        match self {
+            FfiAllocator::Rust => unsafe { alloc::dealloc(ptr.as_ptr().cast(), layout) },
+            FfiAllocator::C => unsafe { libc::free(ptr.as_ptr()) },
+        }
+    }
+}
+
+impl FfiStruct {
+    fn new(layout: Layout, allocator: FfiAllocator) -> Result<Self, FfiError> {
+        assert_ne!(layout.size(), 0);
+        Ok(FfiStruct {
+            ptr: unsafe { allocator.alloc(layout) }?,
+            layout,
+            allocator,
+        })
+    }
+}
+
+impl Drop for FfiStruct {
+    fn drop(&mut self) {
+        unsafe { self.allocator.dealloc(self.layout, self.ptr) };
+    }
+}
+
+impl ForeignFunctionTable {
+    pub fn merge(&mut self, other: ForeignFunctionTable) {
+        self.table.extend(other.table);
+    }
+
+    pub fn define_struct(&mut self, name: Atom, atom_fields: Vec<Atom>) -> Result<(), FfiError> {
+        let fields: Vec<_> = atom_fields.iter().map(FfiType::from_atom).collect();
+        let struct_type = libffi::middle::Type::structure(
+            fields
+                .iter()
+                .map(|field| field.to_type(&self.structs))
+                .collect::<Result<Vec<_>, _>>()?,
+        );
+
+        unsafe {
+            // ensure that size and alignment of struct_type are set properly
+            use libffi::low::{ffi_abi_FFI_DEFAULT_ABI, prep_cif};
+            prep_cif(
+                &mut Default::default(),
+                ffi_abi_FFI_DEFAULT_ABI,
+                1,
+                struct_type.as_raw_ptr(),
+                [struct_type.as_raw_ptr()].as_mut_ptr(),
+            )?;
+        };
+
+        self.structs.insert(
+            name,
+            StructImpl {
+                ffi_type: struct_type,
+                fields,
+            },
+        );
+        Ok(())
+    }
+
+    pub(crate) fn load_library(
+        &mut self,
+        library_name: &str,
+        functions: &Vec<FunctionDefinition>,
+    ) -> Result<(), Box<dyn Error>> {
+        let mut ff_table: ForeignFunctionTable = Default::default();
+        let library = unsafe { Library::new(library_name) }?;
+        for function in functions {
+            let symbol_name: CString = CString::new(&*function.name.as_str())?;
+            let code_ptr: Symbol<*mut c_void> =
+                unsafe { library.get(symbol_name.as_bytes_with_nul()) }?;
+            let args: Vec<_> = function.args.iter().map(FfiType::from_atom).collect();
+            let return_type = FfiType::from_atom(&function.return_value);
+
+            let cif = libffi::middle::Cif::new(
+                args.iter()
+                    .map(|arg| arg.to_type(&self.structs))
+                    .collect::<Result<Vec<_>, _>>()?,
+                return_type.to_type(&self.structs)?,
+            );
+
+            ff_table.table.insert(
+                function.name,
+                FunctionImpl {
+                    cif,
+                    args,
+                    code_ptr: CodePtr(unsafe { code_ptr.into_raw() }.as_raw_ptr()),
+                    return_type,
+                },
+            );
+        }
+        std::mem::forget(library);
+        self.merge(ff_table);
+        Ok(())
+    }
+
+    pub fn exec(
+        &mut self,
+        fn_name: Atom,
+        mut args: Vec<Value>,
+        arena: &mut Arena,
+    ) -> Result<Value, FfiError> {
+        let fn_impl = self
+            .table
+            .get(&fn_name)
+            .ok_or(FfiError::FunctionNotFound(fn_name))?;
+
+        let args = ArgValue::build_args(&mut args, &fn_impl.args, &self.structs)?;
+
+        let args = PointerArgs::new(&args);
+
+        fn_impl.call(&args, arena, &self.structs)
+    }
+
+    pub fn allocate(
+        &mut self,
+        allocator: FfiAllocator,
+        kind: Atom,
+        mut args: Value,
+        arena: &mut Arena,
+    ) -> Result<Value, FfiError> {
+        fn allocate_primitive<T: Copy>(
+            allocator: FfiAllocator,
+            initial_value: T,
+            arena: &mut Arena,
+        ) -> Result<Value, FfiError> {
+            const { assert!(std::mem::size_of::<T>() != 0) };
+            let ptr = unsafe { allocator.alloc(Layout::new::<T>()) }?;
+            unsafe { ptr.cast::<T>().write(initial_value) };
+            Ok(Value::Number(fixnum!(
+                Number,
+                ptr.as_ptr().expose_provenance(),
+                arena
+            )))
+        }
+
+        match FfiType::from_atom(&kind) {
+            FfiType::Void => Err(FfiError::VoidArgumentType),
+            FfiType::Bool => {
+                let val = args.as_int::<u8>()?;
+                let init = match val {
+                    0 => false,
+                    1 => true,
+                    _ => return Err(FfiError::ValueOutOfRange(DomainErrorType::ZeroOrOne, args)),
+                };
+                allocate_primitive::<bool>(allocator, init, arena)
+            }
+            FfiType::U8 => allocate_primitive::<u8>(allocator, args.as_int()?, arena),
+            FfiType::I8 => allocate_primitive::<i8>(allocator, args.as_int()?, arena),
+            FfiType::U16 => allocate_primitive::<u16>(allocator, args.as_int()?, arena),
+            FfiType::I16 => allocate_primitive::<i16>(allocator, args.as_int()?, arena),
+            FfiType::U32 => allocate_primitive::<u32>(allocator, args.as_int()?, arena),
+            FfiType::I32 => allocate_primitive::<i32>(allocator, args.as_int()?, arena),
+            FfiType::U64 => allocate_primitive::<u64>(allocator, args.as_int()?, arena),
+            FfiType::I64 => allocate_primitive::<i64>(allocator, args.as_int()?, arena),
+            FfiType::F32 => allocate_primitive::<f32>(allocator, args.as_float()? as f32, arena),
+            FfiType::F64 => allocate_primitive::<f64>(allocator, args.as_float()?, arena),
+            FfiType::Ptr => allocate_primitive::<*mut c_void>(allocator, args.as_ptr()?, arena),
+            FfiType::CStr => Err(FfiError::CStrFieldType),
+            FfiType::Struct(_) => {
+                let Some(struct_impl) = self.structs.get(&kind) else {
+                    return Err(FfiError::StructNotFound(kind));
+                };
+
+                let (_, args) = args.as_struct()?;
+
+                let ffi_struct = struct_impl.build(&self.structs, args)?;
+
+                let ptr = ManuallyDrop::new(ffi_struct).ptr;
+
+                Ok(Value::Number(fixnum!(
+                    Number,
+                    ptr.as_ptr().expose_provenance(),
+                    arena
+                )))
+            }
+        }
+    }
+
+    pub fn read_ptr(
+        &mut self,
+        kind: Atom,
+        mut ptr: Value,
+        arena: &mut Arena,
+    ) -> Result<Value, FfiError> {
+        unsafe fn read_int<T>(ptr: NonNull<c_void>, arena: &mut Arena) -> Value
+        where
+            T: Copy + TryInto<i64> + MightNotFitInFixnum,
+            Integer: From<T>,
+        {
+            let n = ptr.cast::<T>().read();
+            Value::Number(fixnum!(Number, n, arena))
+        }
+
+        let ptr = ptr.as_ptr()?;
+
+        let Some(ptr) = NonNull::new(ptr) else {
+            return Err(FfiError::NullPtr);
+        };
+
+        match FfiType::from_atom(&kind) {
+            FfiType::Void => Err(FfiError::VoidArgumentType),
+            FfiType::Bool | FfiType::U8 => Ok(unsafe { read_int::<u8>(ptr, arena) }),
+            FfiType::I8 => Ok(unsafe { read_int::<i8>(ptr, arena) }),
+            FfiType::U16 => Ok(unsafe { read_int::<u16>(ptr, arena) }),
+            FfiType::I16 => Ok(unsafe { read_int::<i16>(ptr, arena) }),
+            FfiType::U32 => Ok(unsafe { read_int::<u32>(ptr, arena) }),
+            FfiType::I32 => Ok(unsafe { read_int::<i32>(ptr, arena) }),
+            FfiType::U64 => Ok(unsafe { read_int::<u64>(ptr, arena) }),
+            FfiType::I64 => Ok(unsafe { read_int::<i64>(ptr, arena) }),
+            FfiType::F32 => Ok(Value::Number(Number::Float(
+                (unsafe { ptr.cast::<f32>().read() } as f64).into(),
+            ))),
+            FfiType::F64 => Ok(Value::Number(Number::Float(
+                unsafe { ptr.cast::<f64>().read() }.into(),
+            ))),
+            FfiType::Ptr => {
+                let addr = unsafe { ptr.cast::<*mut c_void>().read() }.expose_provenance();
+                Ok(Value::Number(fixnum!(Number, addr, arena)))
+            }
+            FfiType::CStr => Ok(Value::CString(
+                unsafe { CStr::from_ptr(ptr.as_ptr().cast()) }.to_owned(),
+            )),
+            FfiType::Struct(_) => {
+                let Some(struct_impl) = self.structs.get(&kind) else {
+                    return Err(FfiError::StructNotFound(kind));
+                };
+
+                struct_impl.read(ptr.as_ptr(), kind, &self.structs, arena)
+            }
+        }
+    }
+
+    pub fn deallocate(
+        &mut self,
+        allocator: FfiAllocator,
+        kind: Atom,
+        mut ptr: Value,
+    ) -> Result<(), FfiError> {
+        fn deallocate_primitive<T: Copy>(allocator: FfiAllocator, ptr: NonNull<c_void>) {
+            const { assert!(std::mem::size_of::<T>() != 0) };
+            unsafe { allocator.dealloc(Layout::new::<T>(), ptr) };
+        }
+
+        let ptr = ptr.as_ptr()?;
+
+        let Some(ptr) = NonNull::new(ptr) else {
+            return Err(FfiError::NullPtr);
+        };
+
+        match FfiType::from_atom(&kind) {
+            FfiType::Void => return Err(FfiError::VoidArgumentType),
+            FfiType::Bool => deallocate_primitive::<bool>(allocator, ptr),
+            FfiType::U8 => deallocate_primitive::<u8>(allocator, ptr),
+            FfiType::I8 => deallocate_primitive::<i8>(allocator, ptr),
+            FfiType::U16 => deallocate_primitive::<u16>(allocator, ptr),
+            FfiType::I16 => deallocate_primitive::<i16>(allocator, ptr),
+            FfiType::U32 => deallocate_primitive::<u32>(allocator, ptr),
+            FfiType::I32 => deallocate_primitive::<i32>(allocator, ptr),
+            FfiType::U64 => deallocate_primitive::<u64>(allocator, ptr),
+            FfiType::I64 => deallocate_primitive::<i64>(allocator, ptr),
+            FfiType::F32 => deallocate_primitive::<f32>(allocator, ptr),
+            FfiType::F64 => deallocate_primitive::<f64>(allocator, ptr),
+            FfiType::Ptr => deallocate_primitive::<*mut c_void>(allocator, ptr),
+            FfiType::CStr => return Err(FfiError::CStrFieldType),
+            FfiType::Struct(_) => {
+                let Some(struct_impl) = self.structs.get(&kind) else {
+                    return Err(FfiError::StructNotFound(kind));
+                };
+
+                let layout = struct_impl.layout()?;
+
+                drop(FfiStruct {
+                    ptr,
+                    layout,
+                    allocator,
+                })
+            }
+        }
+        Ok(())
+    }
+}
+
+#[derive(Clone, Debug)]
+pub enum Value {
+    Number(Number),
+    CString(CString),
+    Struct(Atom, Vec<Value>),
+}
+
+impl Value {
+    fn as_int<I>(&self) -> Result<I, FfiError>
+    where
+        Integer: TryInto<I>,
+        i64: TryInto<I>,
+    {
+        match self {
+            Value::Number(Number::Integer(ibig_ptr)) => {
+                let ibig: &Integer = ibig_ptr;
+                ibig.clone().try_into().map_err(|_| {
+                    FfiError::ValueOutOfRange(DomainErrorType::FixedSizedInt, self.clone())
+                })
+            }
+            Value::Number(Number::Fixnum(fixnum)) => fixnum.get_num().try_into().map_err(|_| {
+                FfiError::ValueOutOfRange(DomainErrorType::FixedSizedInt, self.clone())
+            }),
+            _ => Err(FfiError::ValueOutOfRange(
+                DomainErrorType::FixedSizedInt,
+                self.clone(),
+            )),
+        }
+    }
+
+    fn as_float(&self) -> Result<f64, FfiError> {
+        match self {
+            &Value::Number(Number::Float(OrderedFloat(f))) => Ok(f),
+            _ => Err(FfiError::ValueOutOfRange(
+                DomainErrorType::F64,
+                self.clone(),
+            )),
+        }
+    }
+
+    fn as_ptr(&mut self) -> Result<*mut c_void, FfiError> {
+        match self {
+            Value::CString(ref mut cstr) => Ok(cstr.as_ptr().cast_mut().cast()),
+            Value::Number(Number::Fixnum(fixnum)) => Ok(std::ptr::with_exposed_provenance_mut(
+                fixnum.get_num() as usize,
+            )),
+            _ => Err(FfiError::ValueOutOfRange(
+                DomainErrorType::PtrLike,
+                self.clone(),
+            )),
+        }
+    }
+
+    fn as_struct(&mut self) -> Result<(Atom, &mut [Self]), FfiError> {
+        match self {
+            Value::Struct(name, values) => Ok((*name, values)),
+            _ => Err(FfiError::ValueOutOfRange(
+                DomainErrorType::FfiStruct,
+                self.clone(),
+            )),
+        }
+    }
+}
+
+#[derive(Debug)]
+pub enum FfiError {
+    ValueCast(Atom, Atom),
+    ValueOutOfRange(DomainErrorType, Value),
+    VoidArgumentType,
+    FunctionNotFound(Atom),
+    StructNotFound(Atom),
+    ArgCountMismatch,
+    AllocationFailed,
+    // LayoutError should never occour
+    LayoutError,
+    UnsupportedTypedef,
+    UnsupportedAbi,
+    CStrFieldType,
+    NullPtr,
+}
+
+impl std::fmt::Display for FfiError {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        std::fmt::Debug::fmt(self, f)
+    }
+}
+
+impl Error for FfiError {}
+
+impl From<libffi::low::Error> for FfiError {
+    fn from(value: libffi::low::Error) -> Self {
+        match value {
+            libffi::low::Error::Typedef => FfiError::UnsupportedTypedef,
+            libffi::low::Error::Abi => FfiError::UnsupportedAbi,
+        }
+    }
+}

src/fixtures.rs

@@ -1,320 +0,0 @@
-use crate::parser::ast::*;
-
-use crate::forms::*;
-use crate::instructions::*;
-use crate::iterators::*;
-
-use indexmap::{IndexMap, IndexSet};
-
-use std::cell::Cell;
-use std::collections::BTreeSet;
-use std::mem::swap;
-use std::rc::Rc;
-use std::vec::Vec;
-
-// labeled with chunk numbers.
-#[derive(Debug)]
-pub(crate) enum VarStatus {
-    Perm(usize),
-    Temp(usize, TempVarData), // Perm(chunk_num) | Temp(chunk_num, _)
-}
-
-pub(crate) type OccurrenceSet = BTreeSet<(GenContext, usize)>;
-
-// Perm: 0 initially, a stack register once processed.
-// Temp: labeled with chunk_num and temp offset (unassigned if 0).
-#[derive(Debug)]
-pub(crate) enum VarData {
-    Perm(usize),
-    Temp(usize, usize, TempVarData),
-}
-
-impl VarData {
-    pub(crate) fn as_reg_type(&self) -> RegType {
-        match self {
-            &VarData::Temp(_, r, _) => RegType::Temp(r),
-            &VarData::Perm(r) => RegType::Perm(r),
-        }
-    }
-}
-
-#[derive(Debug)]
-pub(crate) struct TempVarData {
-    pub(crate) last_term_arity: usize,
-    pub(crate) use_set: OccurrenceSet,
-    pub(crate) no_use_set: BTreeSet<usize>,
-    pub(crate) conflict_set: BTreeSet<usize>,
-}
-
-impl TempVarData {
-    pub(crate) fn new(last_term_arity: usize) -> Self {
-        TempVarData {
-            last_term_arity: last_term_arity,
-            use_set: BTreeSet::new(),
-            no_use_set: BTreeSet::new(),
-            conflict_set: BTreeSet::new(),
-        }
-    }
-
-    pub(crate) fn uses_reg(&self, reg: usize) -> bool {
-        for &(_, nreg) in self.use_set.iter() {
-            if reg == nreg {
-                return true;
-            }
-        }
-
-        return false;
-    }
-
-    pub(crate) fn populate_conflict_set(&mut self) {
-        if self.last_term_arity > 0 {
-            let arity = self.last_term_arity;
-            let mut conflict_set: BTreeSet<usize> = (1..arity).collect();
-
-            for &(_, reg) in self.use_set.iter() {
-                conflict_set.remove(&reg);
-            }
-
-            self.conflict_set = conflict_set;
-        }
-    }
-}
-
-type VariableFixture<'a> = (VarStatus, Vec<&'a Cell<VarReg>>);
-
-#[derive(Debug)]
-pub(crate) struct VariableFixtures<'a> {
-    perm_vars: IndexMap<Rc<String>, VariableFixture<'a>>,
-    last_chunk_temp_vars: IndexSet<Rc<String>>,
-}
-
-impl<'a> VariableFixtures<'a> {
-    pub(crate) fn new() -> Self {
-        VariableFixtures {
-            perm_vars: IndexMap::new(),
-            last_chunk_temp_vars: IndexSet::new(),
-        }
-    }
-
-    pub(crate) fn insert(&mut self, var: Rc<String>, vs: VariableFixture<'a>) {
-        self.perm_vars.insert(var, vs);
-    }
-
-    pub(crate) fn insert_last_chunk_temp_var(&mut self, var: Rc<String>) {
-        self.last_chunk_temp_vars.insert(var);
-    }
-
-    // computes no_use and conflict sets for all temp vars.
-    pub(crate) fn populate_restricting_sets(&mut self) {
-        // three stages:
-        // 1. move the use sets of each variable to a local IndexMap, use_set
-        // (iterate mutably, swap mutable refs).
-        // 2. drain use_set. For each use set of U, add into the
-        // no-use sets of appropriate variables T =/= U.
-        // 3. Move the use sets back to their original locations in the fixture.
-        // Compute the conflict set of u.
-
-        // 1.
-        let mut use_sets: IndexMap<Rc<String>, OccurrenceSet> = IndexMap::new();
-
-        for (var, &mut (ref mut var_status, _)) in self.iter_mut() {
-            if let &mut VarStatus::Temp(_, ref mut var_data) = var_status {
-                let mut use_set = OccurrenceSet::new();
-
-                swap(&mut var_data.use_set, &mut use_set);
-                use_sets.insert((*var).clone(), use_set);
-            }
-        }
-
-        for (u, use_set) in use_sets.drain(..) {
-            // 2.
-            for &(term_loc, reg) in use_set.iter() {
-                if let GenContext::Last(cn_u) = term_loc {
-                    for (ref t, &mut (ref mut var_status, _)) in self.iter_mut() {
-                        if let &mut VarStatus::Temp(cn_t, ref mut t_data) = var_status {
-                            if cn_u == cn_t && *u != ***t {
-                                if !t_data.uses_reg(reg) {
-                                    t_data.no_use_set.insert(reg);
-                                }
-                            }
-                        }
-                    }
-                }
-            }
-
-            // 3.
-            match self.get_mut(u).unwrap() {
-                &mut (VarStatus::Temp(_, ref mut u_data), _) => {
-                    u_data.use_set = use_set;
-                    u_data.populate_conflict_set();
-                }
-                _ => {}
-            };
-        }
-    }
-
-    fn get_mut(&mut self, u: Rc<String>) -> Option<&mut VariableFixture<'a>> {
-        self.perm_vars.get_mut(&u)
-    }
-
-    fn iter_mut(&mut self) -> indexmap::map::IterMut<Rc<String>, VariableFixture<'a>> {
-        self.perm_vars.iter_mut()
-    }
-
-    fn record_temp_info(&mut self, tvd: &mut TempVarData, arg_c: usize, term_loc: GenContext) {
-        match term_loc {
-            GenContext::Head | GenContext::Last(_) => {
-                tvd.use_set.insert((term_loc, arg_c));
-            }
-            _ => {}
-        };
-    }
-
-    pub(crate) fn vars_above_threshold(&self, index: usize) -> usize {
-        let mut var_count = 0;
-
-        for &(ref var_status, _) in self.values() {
-            if let &VarStatus::Perm(i) = var_status {
-                if i > index {
-                    var_count += 1;
-                }
-            }
-        }
-
-        var_count
-    }
-
-    pub(crate) fn mark_vars_in_chunk<I>(&mut self, iter: I, lt_arity: usize, term_loc: GenContext)
-    where
-        I: Iterator<Item = TermRef<'a>>,
-    {
-        let chunk_num = term_loc.chunk_num();
-        let mut arg_c = 1;
-
-        for term_ref in iter {
-            if let &TermRef::Var(lvl, cell, ref var) = &term_ref {
-                let mut status = self.perm_vars.swap_remove(var).unwrap_or((
-                    VarStatus::Temp(chunk_num, TempVarData::new(lt_arity)),
-                    Vec::new(),
-                ));
-
-                status.1.push(cell);
-
-                match status.0 {
-                    VarStatus::Temp(cn, ref mut tvd) if cn == chunk_num => {
-                        if let Level::Shallow = lvl {
-                            self.record_temp_info(tvd, arg_c, term_loc);
-                        }
-                    }
-                    _ => status.0 = VarStatus::Perm(chunk_num),
-                };
-
-                self.perm_vars.insert(var.clone(), status);
-            }
-
-            if let Level::Shallow = term_ref.level() {
-                arg_c += 1;
-            }
-        }
-    }
-
-    pub(crate) fn into_iter(self) -> indexmap::map::IntoIter<Rc<String>, VariableFixture<'a>> {
-        self.perm_vars.into_iter()
-    }
-
-    fn values(&self) -> indexmap::map::Values<Rc<String>, VariableFixture<'a>> {
-        self.perm_vars.values()
-    }
-
-    pub(crate) fn size(&self) -> usize {
-        self.perm_vars.len()
-    }
-
-    pub(crate) fn set_perm_vals(&self, has_deep_cuts: bool) {
-        let mut values_vec: Vec<_> = self
-            .values()
-            .filter_map(|ref v| match &v.0 {
-                &VarStatus::Perm(i) => Some((i, &v.1)),
-                _ => None,
-            })
-            .collect();
-
-        values_vec.sort_by_key(|ref v| v.0);
-
-        let offset = has_deep_cuts as usize;
-
-        for (i, (_, cells)) in values_vec.into_iter().rev().enumerate() {
-            for cell in cells {
-                cell.set(VarReg::Norm(RegType::Perm(i + 1 + offset)));
-            }
-        }
-    }
-}
-
-#[derive(Debug)]
-pub(crate) struct UnsafeVarMarker {
-    pub(crate) unsafe_vars: IndexMap<RegType, usize>,
-    pub(crate) safe_vars: IndexSet<RegType>,
-}
-
-impl UnsafeVarMarker {
-    pub(crate) fn new() -> Self {
-        UnsafeVarMarker {
-            unsafe_vars: IndexMap::new(),
-            safe_vars: IndexSet::new(),
-        }
-    }
-
-    pub(crate) fn from_safe_vars(safe_vars: IndexSet<RegType>) -> Self {
-        UnsafeVarMarker {
-            unsafe_vars: IndexMap::new(),
-            safe_vars,
-        }
-    }
-
-    pub(crate) fn mark_safe_vars(&mut self, query_instr: &Instruction) -> bool {
-        match query_instr {
-            &Instruction::PutVariable(r @ RegType::Temp(_), _) |
-            &Instruction::SetVariable(r) => {
-                self.safe_vars.insert(r);
-                true
-            }
-            _ => false,
-        }
-    }
-
-    pub(crate) fn mark_phase(&mut self, query_instr: &Instruction, phase: usize) {
-        match query_instr {
-            &Instruction::PutValue(r @ RegType::Perm(_), _) |
-            &Instruction::SetValue(r) => {
-                let p = self.unsafe_vars.entry(r).or_insert(0);
-                *p = phase;
-            }
-            _ => {}
-        }
-    }
-
-    pub(crate) fn mark_unsafe_vars(&mut self, query_instr: &mut Instruction, phase: usize) {
-        match query_instr {
-            &mut Instruction::PutValue(RegType::Perm(i), arg) => {
-                if let Some(p) = self.unsafe_vars.swap_remove(&RegType::Perm(i)) {
-                    if p == phase {
-                        *query_instr = Instruction::PutUnsafeValue(i, arg);
-                        self.safe_vars.insert(RegType::Perm(i));
-                    } else {
-                        self.unsafe_vars.insert(RegType::Perm(i), p);
-                    }
-                }
-            }
-            &mut Instruction::SetValue(r) => {
-                if !self.safe_vars.contains(&r) {
-                    *query_instr = Instruction::SetLocalValue(r);
-
-                    self.safe_vars.insert(r);
-                    self.unsafe_vars.remove(&r);
-                }
-            }
-            _ => {}
-        }
-    }
-}

src/forms.rs

@@ -1,15 +1,17 @@
 use crate::arena::*;
 use crate::atom_table::*;
+use crate::functor_macro::*;
 use crate::instructions::*;
-use crate::machine::heap::*;
+use crate::machine::disjuncts::VarData;
 use crate::machine::loader::PredicateQueue;
 use crate::machine::machine_errors::*;
 use crate::machine::machine_indices::*;
 use crate::parser::ast::*;
+use crate::parser::dashu::{Integer, Rational};
 use crate::parser::parser::CompositeOpDesc;
-use crate::parser::rug::{Integer, Rational};
 use crate::types::*;
 
+use dashu::base::Signed;
 use fxhash::FxBuildHasher;
 
 use indexmap::{IndexMap, IndexSet};
@@ -19,28 +21,11 @@ use std::cell::Cell;
 use std::collections::VecDeque;
 use std::convert::TryFrom;
 use std::fmt;
-use std::ops::AddAssign;
+use std::ops::{AddAssign, Deref, DerefMut};
 use std::path::PathBuf;
-use std::rc::Rc;
-
-use crate::{is_infix, is_postfix};
 
 pub type PredicateKey = (Atom, usize); // name, arity.
 
-pub type Predicate = Vec<PredicateClause>;
-
-// vars of predicate, toplevel offset.  Vec<Term> is always a vector
-// of vars (we get their adjoining cells this way).
-pub type JumpStub = Vec<Term>;
-
-#[derive(Debug, Clone)]
-pub enum TopLevel {
-    Fact(Term), // Term, line_num, col_num
-    Predicate(Predicate),
-    Query(Vec<QueryTerm>),
-    Rule(Rule), // Rule, line_num, col_num
-}
-
 #[derive(Debug, Clone, Copy)]
 pub enum AppendOrPrepend {
     Append,
@@ -57,7 +42,13 @@ impl AppendOrPrepend {
     }
 }
 
-#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+#[derive(Debug, Clone, Copy)]
+pub enum VarComparison {
+    Indistinct,
+    Distinct,
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash)]
 pub enum Level {
     Deep,
     Root,
@@ -79,38 +70,166 @@ pub enum CallPolicy {
     Counted,
 }
 
-#[derive(Debug, Clone)]
+#[derive(Debug, Copy, Clone, PartialEq, Eq, PartialOrd, Ord, Hash)]
+pub enum GenContext {
+    Head,
+    Mid(usize),
+    Last(usize), // Mid & Last: chunk_num
+}
+
+impl GenContext {
+    #[inline]
+    pub fn chunk_num(&self) -> usize {
+        match self {
+            GenContext::Head => 0,
+            &GenContext::Mid(cn) | &GenContext::Last(cn) => cn,
+        }
+    }
+
+    #[inline]
+    pub fn is_last(self) -> bool {
+        matches!(self, GenContext::Last(_))
+    }
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash)]
+pub enum ChunkType {
+    Head,
+    Mid,
+    Last,
+}
+
+#[derive(Debug)]
+pub enum RootIterationPolicy {
+    Iterated,
+    NotIterated,
+}
+
+impl RootIterationPolicy {
+    #[inline(always)]
+    pub fn iterable(&self) -> bool {
+        matches!(self, RootIterationPolicy::Iterated)
+    }
+}
+
+impl ChunkType {
+    #[inline(always)]
+    pub fn to_gen_context(self, chunk_num: usize) -> GenContext {
+        match self {
+            ChunkType::Head => GenContext::Head,
+            ChunkType::Mid => GenContext::Mid(chunk_num),
+            ChunkType::Last => GenContext::Last(chunk_num),
+        }
+    }
+
+    #[inline(always)]
+    pub fn is_last(self) -> bool {
+        self == ChunkType::Last
+    }
+}
+
+#[derive(Debug)]
+pub enum ChunkedTerms {
+    Branch(Vec<VecDeque<ChunkedTerms>>),
+    Chunk { terms: VecDeque<QueryTerm> },
+}
+
+#[derive(Debug)]
+pub struct ChunkedTermVec {
+    pub chunk_vec: VecDeque<ChunkedTerms>,
+}
+
+impl Deref for ChunkedTermVec {
+    type Target = VecDeque<ChunkedTerms>;
+
+    #[inline(always)]
+    fn deref(&self) -> &Self::Target {
+        &self.chunk_vec
+    }
+}
+
+impl DerefMut for ChunkedTermVec {
+    #[inline(always)]
+    fn deref_mut(&mut self) -> &mut Self::Target {
+        &mut self.chunk_vec
+    }
+}
+
+impl ChunkedTermVec {
+    #[allow(clippy::new_without_default)]
+    #[inline]
+    pub fn new() -> Self {
+        Self {
+            chunk_vec: VecDeque::new(),
+        }
+    }
+
+    pub fn reserve_branch(&mut self, capacity: usize) {
+        self.chunk_vec
+            .push_back(ChunkedTerms::Branch(Vec::with_capacity(capacity)));
+    }
+
+    #[inline]
+    pub fn add_chunk(&mut self) {
+        let chunk = ChunkedTerms::Chunk {
+            terms: VecDeque::from(vec![]),
+        };
+        self.chunk_vec.push_back(chunk);
+    }
+
+    pub fn push_chunk_term(&mut self, term: QueryTerm) {
+        match self.chunk_vec.back_mut() {
+            Some(ChunkedTerms::Branch(_)) => {
+                let chunk = ChunkedTerms::Chunk {
+                    terms: VecDeque::from(vec![term]),
+                };
+
+                self.chunk_vec.push_back(chunk);
+            }
+            Some(ChunkedTerms::Chunk { terms, .. }) => {
+                terms.push_back(term);
+            }
+            None => {
+                let chunk = ChunkedTerms::Chunk {
+                    terms: VecDeque::from(vec![term]),
+                };
+
+                self.chunk_vec.push_back(chunk);
+            }
+        }
+    }
+}
+
+#[derive(Debug)]
 pub enum QueryTerm {
     // register, clause type, subterms, clause call policy.
     Clause(Cell<RegType>, ClauseType, Vec<Term>, CallPolicy),
-    BlockedCut, // a cut which is 'blocked by letters', like the P term in P -> Q.
-    UnblockedCut(Cell<VarReg>),
-    GetLevelAndUnify(Cell<VarReg>, Rc<String>),
-    Jump(JumpStub),
+    Fail,
+    LocalCut { var_num: usize, cut_prev: bool }, // var_num
+    GlobalCut(usize),                            // var_num
+    GetCutPoint { var_num: usize, prev_b: bool },
+    GetLevel(usize), // var_num
 }
 
 impl QueryTerm {
-    pub(crate) fn set_call_policy(&mut self, cp: CallPolicy) {
-        match self {
-            &mut QueryTerm::Clause(_, _, _, ref mut clause_cp) => *clause_cp = cp,
-            _ => {}
-        }
-    }
-
     pub(crate) fn arity(&self) -> usize {
         match self {
-            &QueryTerm::Clause(_, _, ref subterms, ..) => subterms.len(),
-            &QueryTerm::BlockedCut | &QueryTerm::UnblockedCut(..) => 0,
-            &QueryTerm::Jump(ref vars) => vars.len(),
-            &QueryTerm::GetLevelAndUnify(..) => 1,
+            QueryTerm::Clause(_, _, subterms, ..) => subterms.len(),
+            &QueryTerm::GetLevel(_) | &QueryTerm::GetCutPoint { .. } => 1,
+            _ => 0,
         }
     }
 }
 
-#[derive(Debug, Clone)]
+#[derive(Debug)]
+pub struct Fact {
+    pub(crate) head: Term,
+}
+
+#[derive(Debug)]
 pub struct Rule {
-    pub(crate) head: (Atom, Vec<Term>, QueryTerm),
-    pub(crate) clauses: Vec<QueryTerm>,
+    pub(crate) head: (Atom, Vec<Term>),
+    pub(crate) clauses: ChunkedTermVec,
 }
 
 #[derive(Clone, Debug, Hash)]
@@ -156,11 +275,10 @@ impl ClauseInfo for Term {
     fn name(&self) -> Option<Atom> {
         match self {
             Term::Clause(_, name, terms) => {
-
                 match name {
                     atom!(":-") => {
                         match terms.len() {
-                            1 => None,            // a declaration.
+                            1 => None, // a declaration.
                             2 => terms[0].name(),
                             _ => Some(*name),
                         }
@@ -175,7 +293,7 @@ impl ClauseInfo for Term {
 
     fn arity(&self) -> usize {
         match self {
-            Term::Clause(_, name, terms) => match name.as_str() {
+            Term::Clause(_, name, terms) => match &*name.as_str() {
                 ":-" => match terms.len() {
                     1 => 0,
                     2 => terms[0].arity(),
@@ -201,30 +319,30 @@ impl ClauseInfo for Rule {
 impl ClauseInfo for PredicateClause {
     fn name(&self) -> Option<Atom> {
         match self {
-            &PredicateClause::Fact(ref term, ..) => term.name(),
-            &PredicateClause::Rule(ref rule, ..) => rule.name(),
+            PredicateClause::Fact(ref term, ..) => term.head.name(),
+            PredicateClause::Rule(ref rule, ..) => rule.name(),
         }
     }
 
     fn arity(&self) -> usize {
         match self {
-            &PredicateClause::Fact(ref term, ..) => term.arity(),
-            &PredicateClause::Rule(ref rule, ..) => rule.arity(),
+            PredicateClause::Fact(ref term, ..) => term.head.arity(),
+            PredicateClause::Rule(ref rule, ..) => rule.arity(),
         }
     }
 }
 
-#[derive(Debug, Clone)]
+#[derive(Debug)]
 pub enum PredicateClause {
-    Fact(Term),
-    Rule(Rule),
+    Fact(Fact, VarData),
+    Rule(Rule, VarData),
 }
 
 impl PredicateClause {
     pub(crate) fn args(&self) -> Option<&[Term]> {
         match self {
-            PredicateClause::Fact(term, ..) => match term {
-                Term::Clause(_, _, args) => Some(&args),
+            PredicateClause::Fact(term, ..) => match &term.head {
+                Term::Clause(_, _, args) => Some(args),
                 _ => None,
             },
             PredicateClause::Rule(rule, ..) => {
@@ -253,9 +371,9 @@ pub enum ModuleSource {
 
 impl ModuleSource {
     pub(crate) fn as_functor_stub(&self) -> MachineStub {
-        match self {
+        match *self {
             ModuleSource::Library(name) => {
-                functor!(atom!("library"), [atom(name)])
+                functor!(atom!("library"), [atom_as_cell(name)])
             }
             ModuleSource::File(name) => {
                 functor!(name)
@@ -290,17 +408,6 @@ pub struct OpDecl {
     pub(crate) name: Atom,
 }
 
-#[inline(always)]
-pub(crate) fn fixity(spec: u32) -> Fixity {
-    match spec {
-        XFY | XFX | YFX => Fixity::In,
-        XF | YF => Fixity::Post,
-        FX | FY => Fixity::Pre,
-        _ => unreachable!(),
-    }
-}
-
-
 impl OpDecl {
     #[inline]
     pub(crate) fn new(op_desc: OpDesc, name: Atom) -> Self {
@@ -317,15 +424,12 @@ impl OpDecl {
     }
 
     pub(crate) fn insert_into_op_dir(&self, op_dir: &mut OpDir) -> Option<OpDesc> {
-        let key = (self.name, fixity(self.op_desc.get_spec() as u32));
+        let key = (self.name, self.op_desc.get_spec().fixity());
 
-        match op_dir.get_mut(&key) {
-            Some(cell) => {
-                let (old_prec, old_spec) = cell.get();
-                cell.set(self.op_desc.get_prec(), self.op_desc.get_spec());
-                return Some(OpDesc::build_with(old_prec, old_spec));
-            }
-            None => {}
+        if let Some(cell) = op_dir.get_mut(&key) {
+            let (old_prec, old_spec) = cell.get();
+            cell.set(self.op_desc.get_prec(), self.op_desc.get_spec());
+            return Some(OpDesc::build_with(old_prec, old_spec));
         }
 
         op_dir.insert(key, self.op_desc)
@@ -336,9 +440,9 @@ impl OpDecl {
         existing_desc: Option<CompositeOpDesc>,
         op_dir: &mut OpDir,
     ) -> Result<(), SessionError> {
-        let (spec, name) = (self.op_desc.get_spec(), self.name.clone());
+        let (spec, name) = (self.op_desc.get_spec(), self.name);
 
-        if is_infix!(spec as u32) {
+        if spec.is_infix() {
             if let Some(desc) = existing_desc {
                 if desc.post > 0 {
                     return Err(SessionError::OpIsInfixAndPostFix(name));
@@ -346,7 +450,7 @@ impl OpDecl {
             }
         }
 
-        if is_postfix!(spec as u32) {
+        if spec.is_postfix() {
             if let Some(desc) = existing_desc {
                 if desc.inf > 0 {
                     return Err(SessionError::OpIsInfixAndPostFix(name));
@@ -367,35 +471,28 @@ pub enum AtomOrString {
 
 impl AtomOrString {
     #[inline]
-    pub fn as_atom(&self, atom_tbl: &mut AtomTable) -> Atom {
+    pub fn as_atom(&self, atom_tbl: &AtomTable) -> Atom {
         match self {
-            &AtomOrString::Atom(atom) => {
-                atom
-            }
-            AtomOrString::String(string) => {
-                atom_tbl.build_with(&string)
-            }
+            &AtomOrString::Atom(atom) => atom,
+            AtomOrString::String(string) => AtomTable::build_with(atom_tbl, string),
         }
     }
 
     #[inline]
-    pub fn as_str(&self) -> &str {
+    pub fn as_str(&self) -> AtomString<'_> {
         match self {
-            AtomOrString::Atom(atom) if atom == &atom!("[]") => "",
+            AtomOrString::Atom(atom) if atom == &atom!("[]") => AtomString::Static(""),
             AtomOrString::Atom(atom) => atom.as_str(),
-            AtomOrString::String(string) => string.as_str(),
+            AtomOrString::String(string) => AtomString::Static(string.as_str()),
         }
     }
+}
 
-    #[inline]
-    pub fn to_string(self) -> String {
-        match self {
-            AtomOrString::Atom(atom) => {
-                atom.as_str().to_owned()
-            }
-            AtomOrString::String(string) => {
-                string
-            }
+impl From<AtomOrString> for String {
+    fn from(val: AtomOrString) -> Self {
+        match val {
+            AtomOrString::Atom(atom) => atom.as_str().to_owned(),
+            AtomOrString::String(string) => string,
         }
     }
 }
@@ -437,7 +534,7 @@ pub(crate) fn fetch_op_spec(name: Atom, arity: usize, op_dir: &OpDir) -> Option<
             }
         }),
         1 => {
-            if let Some(op_desc) = op_dir.get(&(name.clone(), Fixity::Pre)) {
+            if let Some(op_desc) = op_dir.get(&(name, Fixity::Pre)) {
                 if op_desc.get_prec() > 0 {
                     return Some(*op_desc);
                 }
@@ -451,9 +548,7 @@ pub(crate) fn fetch_op_spec(name: Atom, arity: usize, op_dir: &OpDir) -> Option<
                 }
             })
         }
-        0 => {
-            fetch_atom_op_spec(name, None, op_dir)
-        }
+        0 => fetch_atom_op_spec(name, None, op_dir),
         _ => None,
     }
 }
@@ -485,10 +580,7 @@ pub struct Module {
 
 // Module's and related types are defined in forms.
 impl Module {
-    pub(crate) fn new(
-        module_decl: ModuleDecl,
-        listing_src: ListingSource,
-    ) -> Self {
+    pub(crate) fn new(module_decl: ModuleDecl, listing_src: ListingSource) -> Self {
         Module {
             module_decl,
             code_dir: CodeDir::with_hasher(FxBuildHasher::default()),
@@ -510,7 +602,7 @@ impl Module {
             meta_predicates: MetaPredicateDir::with_hasher(FxBuildHasher::default()),
             extensible_predicates: ExtensiblePredicates::with_hasher(FxBuildHasher::default()),
             local_extensible_predicates: LocalExtensiblePredicates::with_hasher(
-                FxBuildHasher::default()
+                FxBuildHasher::default(),
             ),
             listing_src: ListingSource::DynamicallyGenerated,
         }
@@ -534,9 +626,9 @@ impl Default for Number {
 impl fmt::Display for Number {
     fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
         match self {
-            Number::Float(fl) => write!(f, "{}", fl),
-            Number::Integer(n) => write!(f, "{}", n),
-            Number::Rational(r) => write!(f, "{}", r),
+            Number::Float(fl) => write!(f, "{fl}"),
+            Number::Integer(n) => write!(f, "{n}"),
+            Number::Rational(r) => write!(f, "{r}"),
             Number::Fixnum(n) => write!(f, "{}", n.get_num()),
         }
     }
@@ -605,17 +697,18 @@ impl ArenaFrom<isize> for Number {
 impl ArenaFrom<u32> for Number {
     #[inline]
     fn arena_from(value: u32, _arena: &mut Arena) -> Number {
-        Number::Fixnum(Fixnum::build_with(value as i64))
+        Number::Fixnum(Fixnum::build_with(value))
     }
 }
 
 impl ArenaFrom<i32> for Number {
     #[inline]
     fn arena_from(value: i32, _arena: &mut Arena) -> Number {
-        Number::Fixnum(Fixnum::build_with(value as i64))
+        Number::Fixnum(Fixnum::build_with(value))
     }
 }
 
+/*
 impl ArenaFrom<Number> for Literal {
     #[inline]
     fn arena_from(value: Number, arena: &mut Arena) -> Literal {
@@ -627,6 +720,21 @@ impl ArenaFrom<Number> for Literal {
         }
     }
 }
+*/
+
+impl ArenaFrom<u64> for HeapCellValue {
+    #[inline]
+    fn arena_from(value: u64, arena: &mut Arena) -> HeapCellValue {
+        HeapCellValue::from(fixnum!(Literal, value as i64, arena))
+    }
+}
+
+impl ArenaFrom<usize> for HeapCellValue {
+    #[inline]
+    fn arena_from(value: usize, arena: &mut Arena) -> HeapCellValue {
+        HeapCellValue::arena_from(value as u64, arena)
+    }
+}
 
 impl ArenaFrom<Number> for HeapCellValue {
     #[inline]
@@ -634,7 +742,7 @@ impl ArenaFrom<Number> for HeapCellValue {
         match value {
             Number::Fixnum(n) => fixnum_as_cell!(n),
             Number::Integer(n) => typed_arena_ptr_as_cell!(n),
-            Number::Float(OrderedFloat(n)) => HeapCellValue::from(float_alloc!(n, arena)),
+            Number::Float(n) => HeapCellValue::from(arena.f64_tbl.build_with(n)),
             Number::Rational(n) => typed_arena_ptr_as_cell!(n),
         }
     }
@@ -643,11 +751,15 @@ impl ArenaFrom<Number> for HeapCellValue {
 impl Number {
     pub(crate) fn sign(&self) -> Number {
         match self {
-            &Number::Float(f) if f == 0.0 => Number::Float(OrderedFloat(0f64)),
-            &Number::Float(f) => Number::Float(OrderedFloat(f.signum())),
+            Number::Float(f) if *f == 0.0 => Number::Float(OrderedFloat(0f64)),
+            Number::Float(f) => Number::Float(OrderedFloat(f.signum())),
             _ => {
                 if self.is_positive() {
-                    Number::Fixnum(Fixnum::build_with(1))
+                    if self.is_zero() {
+                        Number::Fixnum(Fixnum::build_with(0))
+                    } else {
+                        Number::Fixnum(Fixnum::build_with(1))
+                    }
                 } else if self.is_negative() {
                     Number::Fixnum(Fixnum::build_with(-1))
                 } else {
@@ -660,46 +772,43 @@ impl Number {
     #[inline]
     pub(crate) fn is_positive(&self) -> bool {
         match self {
-            &Number::Fixnum(n) => n.get_num() > 0,
-            &Number::Integer(ref n) => &**n > &0,
-            &Number::Float(f) => f.is_sign_positive(),
-            &Number::Rational(ref r) => &**r > &0,
+            Number::Fixnum(n) => n.get_num() > 0,
+            Number::Integer(ref n) => n.is_positive(),
+            Number::Float(f) => f.is_sign_positive(),
+            Number::Rational(ref r) => r.is_positive(),
         }
     }
 
     #[inline]
     pub(crate) fn is_negative(&self) -> bool {
         match self {
-            &Number::Fixnum(n) => n.get_num() < 0,
-            &Number::Integer(ref n) => &**n < &0,
-            &Number::Float(OrderedFloat(f)) => f.is_sign_negative() && OrderedFloat(f) != -0f64,
-            &Number::Rational(ref r) => &**r < &0,
+            Number::Fixnum(n) => n.get_num() < 0,
+            Number::Integer(ref n) => n.is_negative(),
+            &Number::Float(OrderedFloat(f)) => f.is_sign_negative() && f != -0f64,
+            Number::Rational(ref r) => r.is_negative(),
         }
     }
 
     #[inline]
     pub(crate) fn is_zero(&self) -> bool {
         match self {
-            &Number::Fixnum(n) => n.get_num() == 0,
-            &Number::Integer(ref n) => &**n == &0,
-            &Number::Float(f) => f == OrderedFloat(0f64) || f == OrderedFloat(-0f64),
-            &Number::Rational(ref r) => &**r == &0,
+            Number::Fixnum(n) => n.get_num() == 0,
+            Number::Integer(ref n) => n.is_zero(),
+            &Number::Float(OrderedFloat(f)) => f == 0.0 || f == -0.0,
+            Number::Rational(ref r) => r.is_zero(),
         }
     }
 
     #[inline]
     pub(crate) fn is_integer(&self) -> bool {
-        match self {
-            Number::Fixnum(_) | Number::Integer(_) => true,
-            _ => false,
-        }
+        matches!(self, Number::Fixnum(_) | Number::Integer(_))
     }
 }
 
-#[derive(Debug, Clone)]
+#[derive(Debug, Copy, Clone)]
 pub(crate) enum OptArgIndexKey {
-    Literal(usize, usize, Literal, Vec<Literal>), // index, IndexingCode location, opt arg, alternatives
-    List(usize, usize),                           // index, IndexingCode location
+    Literal(usize, usize, Literal, Option<Literal>), // index, IndexingCode location, opt arg, alternatives
+    List(usize, usize),                              // index, IndexingCode location
     None,
     Structure(usize, usize, Atom, usize), // index, IndexingCode location, name, arity
 }
@@ -782,7 +891,7 @@ impl ClauseIndexInfo {
     }
 }
 
-#[derive(Clone, Copy, Debug)]
+#[derive(Clone, Copy, Debug, Default)]
 pub(crate) struct PredicateInfo {
     pub(crate) is_extensible: bool,
     pub(crate) is_discontiguous: bool,
@@ -791,19 +900,6 @@ pub(crate) struct PredicateInfo {
     pub(crate) has_clauses: bool,
 }
 
-impl Default for PredicateInfo {
-    #[inline]
-    fn default() -> Self {
-        PredicateInfo {
-            is_extensible: false,
-            is_discontiguous: false,
-            is_dynamic: false,
-            is_multifile: false,
-            has_clauses: false,
-        }
-    }
-}
-
 impl PredicateInfo {
     #[inline]
     pub(crate) fn compile_incrementally(&self) -> bool {
@@ -812,8 +908,11 @@ impl PredicateInfo {
     }
 
     #[inline]
-    pub(crate) fn must_retract_local_clauses(&self) -> bool {
-        self.is_extensible && self.has_clauses && !self.is_discontiguous
+    pub(crate) fn must_retract_local_clauses(&self, is_cross_module_clause: bool) -> bool {
+        self.is_extensible
+            && self.has_clauses
+            && !self.is_discontiguous
+            && !(self.is_multifile && is_cross_module_clause)
     }
 }
 
@@ -859,7 +958,7 @@ impl LocalPredicateSkeleton {
 
     #[inline]
     pub(crate) fn add_retracted_dynamic_clause_info(&mut self, clause_info: ClauseIndexInfo) {
-        debug_assert_eq!(self.is_dynamic, true);
+        debug_assert!(self.is_dynamic);
 
         if self.retracted_dynamic_clauses.is_none() {
             self.retracted_dynamic_clauses = Some(vec![]);
@@ -902,19 +1001,17 @@ impl PredicateSkeleton {
         &mut self,
         clause_clause_loc: usize,
     ) -> Option<usize> {
-        let search_result = self.core.clause_clause_locs
-            .make_contiguous()[0..self.core.clause_assert_margin]
-            .binary_search_by(|loc| clause_clause_loc.cmp(&loc));
+        let search_result = self.core.clause_clause_locs.make_contiguous()
+            [0..self.core.clause_assert_margin]
+            .binary_search_by(|loc| clause_clause_loc.cmp(loc));
 
         match search_result {
             Ok(loc) => Some(loc),
-            Err(_) => {
-                self.core.clause_clause_locs
-                    .make_contiguous()[self.core.clause_assert_margin..]
-                    .binary_search_by(|loc| loc.cmp(&clause_clause_loc))
-                    .map(|loc| loc + self.core.clause_assert_margin)
-                    .ok()
-            }
+            Err(_) => self.core.clause_clause_locs.make_contiguous()
+                [self.core.clause_assert_margin..]
+                .binary_search_by(|loc| loc.cmp(&clause_clause_loc))
+                .map(|loc| loc + self.core.clause_assert_margin)
+                .ok(),
         }
     }
 }

src/functor_macro.rs

@@ -0,0 +1,798 @@
+//! A macro to construct functor terms ready to be written to the WAM
+//! heap.
+
+use crate::atom_table::*;
+use crate::instructions::IndexingCodePtr;
+use crate::machine::heap::Heap;
+use crate::parser::ast::Fixnum;
+use crate::types::*;
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub(crate) enum FunctorElement {
+    AbsoluteCell(HeapCellValue),
+    Cell(HeapCellValue),
+    InnerFunctor(u64, Vec<FunctorElement>),
+    String(u64, String),
+}
+
+// helper macros
+macro_rules! count {
+    () => (0);
+    ( $x:tt $($xs:tt)* ) => (1 + count!($($xs)*));
+}
+
+// core macros
+
+/*
+ * functor! is more declarative now, with fewer effects and more
+ * work done at compile time using const functions. With these
+ * advantages come new quirks: expressions must generally be wrapped
+ * in round parentheses for rustc to parse them. See the tests module
+ * below for examples, especially those involving atom!
+ * subexpressions.
+*/
+
+macro_rules! functor {
+    ($name:expr) => ({
+        vec![FunctorElement::Cell(atom_as_cell!($name))]
+    });
+    ($name:expr, [$($dt:ident($($value:tt),*)),+]) => ({
+        build_functor!([$($dt($($value),*)),*],
+                       [FunctorElement::Cell(atom_as_cell!($name, count!($($dt) *)))],
+                       1,
+                       [])
+    });
+}
+
+macro_rules! inner_functor {
+    ($name:expr, $res_len:expr, [$($dt:ident($($value:tt),*)),+]) => ({
+        build_functor!([$($dt($($value),*)),*],
+                       [FunctorElement::Cell(atom_as_cell!($name, count!($($dt) *)))],
+                       1 + $res_len,
+                       [])
+    });
+}
+
+macro_rules! build_functor {
+    ([], [$($res:expr),*], $res_len:expr, [$($subfunctor:expr),*]) => ({
+        vec![$($res,)* $($subfunctor),*]
+    });
+    ([indexing_code_ptr($e:expr) $(, $dt:ident($($value:tt),*))*],
+     [$($res:expr),*],
+     $res_len:expr,
+     [$($subfunctor:expr),*]) => ({
+        let (inner_functor, cell_size) = indexing_code_ptr($e);
+        let referent = if cell_size == 1 {
+            heap_loc_as_cell!(1u64 + count!($($dt)*) + $res_len)
+        } else {
+            str_loc_as_cell!(1u64 + count!($($dt)*) + $res_len)
+        };
+
+        build_functor!([$($dt($($value),*)),*],
+                       [$($res, )* FunctorElement::Cell(referent)],
+                       1 + cell_size + $res_len,
+                       [$($subfunctor, )* FunctorElement::InnerFunctor(cell_size, inner_functor)])
+    });
+    ([fixnum($e:expr) $(, $dt:ident($($value:tt),*))*],
+     [$($res:expr),*],
+     $res_len:expr,
+     [$($subfunctor:expr),*]) => ({
+         build_functor!([$($dt($($value),*)),*],
+                        [$($res, )* FunctorElement::Cell(fixnum_as_cell!(/*FIXME this is not safe*/ unsafe{Fixnum::build_with_unchecked($e as i64)}))],
+                        1 + $res_len,
+                        [$($subfunctor),*])
+    });
+    ([cell($e:expr) $(, $dt:ident($($value:tt),*))*],
+     [$($res:expr),*],
+     $res_len:expr,
+     [$($subfunctor:expr),*]) => ({
+         build_functor!([$($dt($($value),*)),*],
+                        [$($res, )* FunctorElement::AbsoluteCell($e)],
+                        1 + $res_len,
+                        [$($subfunctor),*])
+    });
+    ([literal($e:expr) $(, $dt:ident($($value:tt),*))*],
+     [$($res:expr),*],
+     $res_len:expr,
+     [$($subfunctor:expr),*]) => ({
+         build_functor!([$($dt($($value),*)),*],
+                        [$($res, )* FunctorElement::AbsoluteCell(HeapCellValue::from($e))],
+                        1 + $res_len,
+                        [$($subfunctor),*])
+    });
+    ([number($n:expr, $arena:expr) $(, $dt:ident($($value:tt),*))*],
+     [$($res:expr),*],
+     $res_len:expr,
+     [$($subfunctor:expr),*]) => ({
+         let number_cell = HeapCellValue::arena_from($n, $arena);
+
+         build_functor!([$($dt($($value),*)),*],
+                        [$($res, )* FunctorElement::Cell(number_cell)],
+                        1 + $res_len,
+                        [$($subfunctor),*])
+    });
+    ([list([]) $(, $dt:ident($($value:tt),*))*],
+     [$($res:expr),*],
+     $res_len:expr,
+     [$($subfunctor:expr),*]) => ({
+        build_functor!([$($dt($($value),*)),*],
+                       [$($res, )* FunctorElement::Cell(empty_list_as_cell!())],
+                       1 + $res_len,
+                       [$($subfunctor),*])
+    });
+    ([list([$id:ident($($id_value:tt),*) $(, $in_dt:ident($($in_value:tt),*))*]) $(, $dt:ident($($value:tt),*))*],
+     [$($res:expr),*],
+     $res_len:expr,
+     [$($subfunctor:expr),*]) => ({
+        build_functor!([functor((atom!(".")), [$id($($id_value),*), list([$($in_dt($($in_value),*)),*])])
+                        $(, $dt($($value),*))*],
+                       [$($res),*],
+                       $res_len,
+                       [$($subfunctor),*])
+    });
+    ([string($s:expr) $(, $dt:ident($($value:tt),*))*], [$($res:expr),*], $res_len:expr, [$($subfunctor:expr),*]) => ({
+        #[allow(unused_parens)]
+        let string = $s;
+        let pstr_len = cell_index!(Heap::compute_pstr_size(&string)) as u64;
+        let result_len = 1 + count!($($dt)*) + $res_len;
+
+        build_functor!([$($dt($($value),*)),*],
+                       [$($res, )* FunctorElement::Cell(pstr_loc_as_cell!(heap_index!(result_len as usize) as u64))],
+                       1 + $res_len + pstr_len,
+                       [$($subfunctor, )* FunctorElement::String(pstr_len, string)])
+    });
+    ([atom_as_cell($n:expr) $(, $dt:ident($($value:tt),*))*], [$($res:expr),*], $res_len:expr, [$($subfunctor:expr),*]) => ({
+        build_functor!([$($dt($($value),*)),*],
+                       [$($res, )* FunctorElement::Cell(atom_as_cell!($n))],
+                       1 + $res_len,
+                       [$($subfunctor),*])
+    });
+    ([functor($stub:expr) $(, $dt:ident($($value:tt),*))*], [$($res:expr),*], $res_len:expr, [$($subfunctor:expr),*]) => ({
+        let result_len = 1u64 + count!($($dt)*) + $res_len;
+        let inner_functor_size = cell_index!(Heap::compute_functor_byte_size(&$stub)) as u64;
+
+        build_functor!([$($dt($($value),*)),*],
+                       [$($res, )* FunctorElement::Cell(str_loc_as_cell!(result_len))],
+                       1 + $res_len + inner_functor_size,
+                       [$($subfunctor, )*
+                        FunctorElement::InnerFunctor(inner_functor_size, $stub)])
+    });
+    ([$id:ident($n:expr) $(, $dt:ident($($value:tt),*))*], [$($res:expr),*], $res_len:expr, [$($subfunctor:expr),*]) => ({
+        build_functor!([$($dt($($value),*)),*],
+                       [$($res, )* FunctorElement::Cell($id!($n))],
+                       1 + $res_len,
+                       [$($subfunctor),*])
+    });
+    ([functor($name:expr, [$($in_dt:ident($($in_value:tt),*)),+]) $(, $dt:ident($($value:tt),*))*],
+     [$($res:expr),*],
+     $res_len:expr,
+     [$($subfunctor:expr),*]) => ({
+         let result_len = 1u64 + count!($($dt)*) + $res_len;
+         let inner_functor = inner_functor!($name, 0, [$($in_dt($($in_value),*)),*]);
+         let inner_functor_size = cell_index!(Heap::compute_functor_byte_size(&inner_functor)) as u64;
+
+         build_functor!([$($dt($($value),*)),*],
+                        [$($res, )* FunctorElement::Cell(str_loc_as_cell!(result_len))],
+                        1 + $res_len + inner_functor_size,
+                        [$($subfunctor, )*
+                         FunctorElement::InnerFunctor(inner_functor_size, inner_functor)])
+    });
+}
+
+pub(crate) fn indexing_code_ptr(code_ptr: IndexingCodePtr) -> (Vec<FunctorElement>, u64) {
+    match code_ptr {
+        IndexingCodePtr::DynamicExternal(o) => {
+            (functor!(atom!("dynamic_external"), [fixnum(o)]), 2)
+        }
+        IndexingCodePtr::External(o) => (functor!(atom!("external"), [fixnum(o)]), 2),
+        IndexingCodePtr::Internal(o) => (functor!(atom!("internal"), [fixnum(o)]), 2),
+        IndexingCodePtr::Fail => (vec![FunctorElement::Cell(atom_as_cell!(atom!("fail")))], 1),
+    }
+}
+
+pub(crate) fn variadic_functor(
+    name: Atom,
+    arity: usize,
+    iter: impl Iterator<Item = Vec<FunctorElement>>,
+) -> Vec<FunctorElement> {
+    let mut arg_vec = vec![
+        FunctorElement::Cell(atom_as_cell!(name, arity)),
+        FunctorElement::Cell(list_loc_as_cell!(2)),
+    ];
+
+    let key_value_pairs: Vec<_> = iter.collect();
+    let num_items = key_value_pairs.len();
+    let mut functor_offset = 0;
+
+    for (idx, kv_func) in key_value_pairs.iter().enumerate() {
+        let functor_size = cell_index!(Heap::compute_functor_byte_size(kv_func));
+
+        arg_vec.push(FunctorElement::Cell(str_loc_as_cell!(
+            2 + num_items * 2 + functor_offset
+        )));
+        arg_vec.push(FunctorElement::Cell(list_loc_as_cell!(4 + 2 * idx)));
+
+        functor_offset += functor_size;
+    }
+
+    arg_vec.pop();
+    arg_vec.push(FunctorElement::Cell(empty_list_as_cell!()));
+
+    arg_vec.extend(key_value_pairs.into_iter().map(|kv_func| {
+        let inner_functor_size = cell_index!(Heap::compute_functor_byte_size(&kv_func));
+        FunctorElement::InnerFunctor(inner_functor_size as u64, kv_func)
+    }));
+
+    arg_vec
+}
+
+#[cfg(test)]
+#[allow(unused_parens)]
+mod tests {
+    use super::*;
+    use indexmap::indexmap;
+    use std::string::String;
+    use FunctorElement::*;
+
+    #[test]
+    fn basic_terms() {
+        let functor = functor!(
+            atom!("first"),
+            [atom_as_cell((atom!("a"))), char_as_cell('c')]
+        );
+
+        assert_eq!(functor.len(), 3);
+
+        assert_eq!(functor[0], Cell(atom_as_cell!(atom!("first"), 2)));
+        assert_eq!(functor[1], Cell(atom_as_cell!(atom!("a"))));
+        assert_eq!(functor[2], Cell(char_as_cell!('c')));
+
+        let functor = functor!(
+            atom!("second"),
+            [
+                atom_as_cell((atom!("a"))),
+                functor((atom!("b")), [fixnum(1), fixnum(2)]),
+                char_as_cell('c')
+            ]
+        );
+
+        assert_eq!(functor.len(), 5);
+
+        assert_eq!(functor[0], Cell(atom_as_cell!(atom!("second"), 3)));
+        assert_eq!(functor[1], Cell(atom_as_cell!(atom!("a"))));
+        assert_eq!(functor[2], Cell(str_loc_as_cell!(4)));
+        assert_eq!(functor[3], Cell(char_as_cell!('c')));
+        assert_eq!(
+            functor[4],
+            InnerFunctor(3, functor!(atom!("b"), [fixnum(1), fixnum(2)]))
+        );
+
+        let functor = functor!(
+            atom!("third"),
+            [
+                atom_as_cell((atom!("a"))),
+                functor((atom!("b")), [fixnum(1), fixnum(2)]),
+                functor((atom!("c")), [fixnum(1), fixnum(2)]),
+                char_as_cell('c')
+            ]
+        );
+
+        assert_eq!(functor.len(), 7);
+
+        assert_eq!(functor[0], Cell(atom_as_cell!(atom!("third"), 4)));
+        assert_eq!(functor[1], Cell(atom_as_cell!(atom!("a"))));
+        assert_eq!(functor[2], Cell(str_loc_as_cell!(5)));
+        assert_eq!(functor[3], Cell(str_loc_as_cell!(8)));
+        assert_eq!(functor[4], Cell(char_as_cell!('c')));
+        assert_eq!(
+            functor[5],
+            InnerFunctor(3, functor!(atom!("b"), [fixnum(1), fixnum(2)]))
+        );
+        assert_eq!(
+            functor[6],
+            InnerFunctor(3, functor!(atom!("c"), [fixnum(1), fixnum(2)]))
+        );
+
+        let functor = functor!(
+            atom!("fourth"),
+            [
+                atom_as_cell((atom!("a"))),
+                functor((atom!("b")), [fixnum(1), fixnum(2)]),
+                functor((atom!("c")), [fixnum(1)]),
+                functor((atom!("d")), [fixnum(453), fixnum(2)]),
+                char_as_cell('c')
+            ]
+        );
+
+        assert_eq!(functor.len(), 9);
+
+        assert_eq!(functor[0], Cell(atom_as_cell!(atom!("fourth"), 5)));
+        assert_eq!(functor[1], Cell(atom_as_cell!(atom!("a"))));
+        assert_eq!(functor[2], Cell(str_loc_as_cell!(6)));
+        assert_eq!(functor[3], Cell(str_loc_as_cell!(9)));
+        assert_eq!(functor[4], Cell(str_loc_as_cell!(11)));
+        assert_eq!(functor[5], Cell(char_as_cell!('c')));
+        assert_eq!(
+            functor[6],
+            InnerFunctor(3, functor!(atom!("b"), [fixnum(1), fixnum(2)]))
+        );
+        assert_eq!(
+            functor[7],
+            InnerFunctor(2, functor!(atom!("c"), [fixnum(1)]))
+        );
+        assert_eq!(
+            functor[8],
+            InnerFunctor(3, functor!(atom!("d"), [fixnum(453), fixnum(2)]))
+        );
+    }
+
+    #[test]
+    fn basic_terms_in_heap() {
+        let functor = functor!(
+            atom!("first"),
+            [atom_as_cell((atom!("a"))), char_as_cell('b')]
+        );
+
+        assert_eq!(functor.len(), 3);
+
+        let mut heap = Heap::new();
+        let mut functor_writer = Heap::functor_writer(functor);
+        let loc = functor_writer(&mut heap).unwrap();
+
+        assert_eq!(loc, str_loc_as_cell!(0));
+
+        assert_eq!(heap[0], atom_as_cell!(atom!("first"), 2));
+        assert_eq!(heap[1], atom_as_cell!(atom!("a")));
+        assert_eq!(heap[2], char_as_cell!('b'));
+
+        heap.truncate(2);
+
+        let functor = functor!(
+            atom!("second"),
+            [
+                atom_as_cell((atom!("a"))),
+                functor((atom!("b")), [fixnum(1), fixnum(2)]),
+                functor((atom!("c")), [fixnum(1), fixnum(2)]),
+                char_as_cell('b')
+            ]
+        );
+
+        assert_eq!(functor.len(), 7);
+
+        let mut functor_writer = Heap::functor_writer(functor);
+        let loc = functor_writer(&mut heap).unwrap();
+
+        assert_eq!(loc, str_loc_as_cell!(2));
+
+        assert_eq!(heap[2], atom_as_cell!(atom!("second"), 4));
+        assert_eq!(heap[3], atom_as_cell!(atom!("a")));
+        assert_eq!(heap[4], str_loc_as_cell!(7));
+        assert_eq!(heap[5], str_loc_as_cell!(10));
+        assert_eq!(heap[6], char_as_cell!('b'));
+        assert_eq!(heap[7], atom_as_cell!(atom!("b"), 2));
+        assert_eq!(heap[8], fixnum_as_cell!(Fixnum::build_with(1)));
+        assert_eq!(heap[9], fixnum_as_cell!(Fixnum::build_with(2)));
+        assert_eq!(heap[10], atom_as_cell!(atom!("c"), 2));
+        assert_eq!(heap[11], fixnum_as_cell!(Fixnum::build_with(1)));
+        assert_eq!(heap[12], fixnum_as_cell!(Fixnum::build_with(2)));
+    }
+
+    #[test]
+    fn nested_functors() {
+        let functor = functor!(
+            atom!("first"),
+            [
+                atom_as_cell((atom!("a"))),
+                functor(
+                    (atom!("d")),
+                    [
+                        fixnum(1),
+                        functor(
+                            (atom!("b")),
+                            [atom_as_cell((atom!("c"))), char_as_cell('c')]
+                        )
+                    ]
+                ),
+                functor((atom!("e")), [fixnum(453), fixnum(2)]),
+                char_as_cell('b')
+            ]
+        );
+
+        assert_eq!(functor.len(), 7);
+
+        assert_eq!(functor[0], Cell(atom_as_cell!(atom!("first"), 4)));
+        assert_eq!(functor[1], Cell(atom_as_cell!(atom!("a"))));
+        assert_eq!(functor[2], Cell(str_loc_as_cell!(5)));
+        assert_eq!(functor[3], Cell(str_loc_as_cell!(11)));
+        assert_eq!(functor[4], Cell(char_as_cell!('b')));
+        assert_eq!(
+            functor[5],
+            InnerFunctor(
+                6,
+                vec![
+                    Cell(atom_as_cell!(atom!("d"), 2)),
+                    Cell(fixnum_as_cell!(Fixnum::build_with(1))),
+                    Cell(str_loc_as_cell!(3)),
+                    InnerFunctor(
+                        3,
+                        functor!(atom!("b"), [atom_as_cell((atom!("c"))), char_as_cell('c')])
+                    )
+                ]
+            )
+        );
+        assert_eq!(
+            functor[6],
+            InnerFunctor(3, functor!(atom!("e"), [fixnum(453), fixnum(2)]))
+        );
+    }
+
+    #[test]
+    fn nested_functors_in_heap() {
+        let functor = functor!(
+            atom!("first"),
+            [
+                atom_as_cell((atom!("a"))),
+                functor(
+                    (atom!("second")),
+                    [
+                        fixnum(1),
+                        functor(
+                            (atom!("third")),
+                            [atom_as_cell((atom!("b"))), char_as_cell('c')]
+                        )
+                    ]
+                ),
+                functor((atom!("fourth")), [fixnum(453), fixnum(2)]),
+                char_as_cell('b')
+            ]
+        );
+
+        let mut heap = Heap::new();
+        let mut functor_writer = Heap::functor_writer(functor);
+        let loc = functor_writer(&mut heap).unwrap();
+
+        assert_eq!(loc, str_loc_as_cell!(0));
+
+        assert_eq!(heap.cell_len(), 14);
+
+        assert_eq!(heap[0], atom_as_cell!(atom!("first"), 4));
+        assert_eq!(heap[1], atom_as_cell!(atom!("a")));
+        assert_eq!(heap[2], str_loc_as_cell!(5));
+        assert_eq!(heap[3], str_loc_as_cell!(11));
+        assert_eq!(heap[4], char_as_cell!('b'));
+        assert_eq!(heap[5], atom_as_cell!(atom!("second"), 2));
+        assert_eq!(heap[6], fixnum_as_cell!(Fixnum::build_with(1)));
+        assert_eq!(heap[7], str_loc_as_cell!(8));
+        assert_eq!(heap[8], atom_as_cell!(atom!("third"), 2));
+        assert_eq!(heap[9], atom_as_cell!(atom!("b")));
+        assert_eq!(heap[10], char_as_cell!('c'));
+        assert_eq!(heap[11], atom_as_cell!(atom!("fourth"), 2));
+        assert_eq!(heap[12], fixnum_as_cell!(Fixnum::build_with(453)));
+        assert_eq!(heap[13], fixnum_as_cell!(Fixnum::build_with(2)));
+    }
+
+    #[test]
+    fn functors_with_strings_in_heap() {
+        let functor = functor!(atom!("first"), [string((String::from("a string")))]);
+
+        assert_eq!(functor.len(), 3);
+
+        let mut heap = Heap::new();
+        let mut functor_writer = Heap::functor_writer(functor);
+        let loc = functor_writer(&mut heap).unwrap();
+
+        assert_eq!(loc, str_loc_as_cell!(0));
+        assert_eq!(heap.cell_len(), 5);
+
+        assert_eq!(heap[0], atom_as_cell!(atom!("first"), 1));
+        assert_eq!(heap[1], pstr_loc_as_cell!(heap_index!(2)));
+        assert_eq!(
+            heap.slice_to_str(heap_index!(2), "a string".len()),
+            "a string"
+        );
+        assert_eq!(heap[4], empty_list_as_cell!());
+
+        heap.truncate(0);
+
+        let functor = functor!(
+            atom!("second"),
+            [string((String::from("a stuttered\0 string")))]
+        );
+
+        let mut functor_writer = Heap::functor_writer(functor);
+        functor_writer(&mut heap).unwrap();
+
+        assert_eq!(heap.cell_len(), 10);
+
+        assert_eq!(heap[0], atom_as_cell!(atom!("second"), 1));
+        assert_eq!(heap[1], pstr_loc_as_cell!(heap_index!(2)));
+        assert_eq!(
+            heap.slice_to_str(heap_index!(2), "a stuttered".len()),
+            "a stuttered"
+        );
+        assert_eq!(heap[4], list_loc_as_cell!(5));
+        assert_eq!(heap[5], char_as_cell!('\u{0}'));
+        assert_eq!(heap[6], pstr_loc_as_cell!(heap_index!(7)));
+        assert_eq!(
+            heap.slice_to_str(heap_index!(7), " string".len()),
+            " string"
+        );
+        assert_eq!(heap[9], empty_list_as_cell!());
+    }
+
+    #[test]
+    fn functors_with_lists_in_heap() {
+        let functor = functor!(
+            atom!("first"),
+            [list([fixnum(1), atom_as_cell((atom!("a"))), fixnum(2)])]
+        );
+
+        assert_eq!(functor.len(), 3);
+
+        let mut heap = Heap::new();
+        let mut functor_writer = Heap::functor_writer(functor);
+
+        functor_writer(&mut heap).unwrap();
+
+        assert_eq!(heap.cell_len(), 11);
+
+        assert_eq!(heap[0], atom_as_cell!(atom!("first"), 1));
+        assert_eq!(heap[1], str_loc_as_cell!(2));
+        assert_eq!(heap[2], atom_as_cell!(atom!("."), 2));
+        assert_eq!(heap[3], fixnum_as_cell!(Fixnum::build_with(1)));
+        assert_eq!(heap[4], str_loc_as_cell!(5));
+        assert_eq!(heap[5], atom_as_cell!(atom!("."), 2));
+        assert_eq!(heap[6], atom_as_cell!(atom!("a")));
+        assert_eq!(heap[7], str_loc_as_cell!(8));
+        assert_eq!(heap[8], atom_as_cell!(atom!("."), 2));
+        assert_eq!(heap[9], fixnum_as_cell!(Fixnum::build_with(2)));
+        assert_eq!(heap[10], empty_list_as_cell!());
+    }
+
+    #[test]
+    fn inlined_atoms() {
+        let atom_table = AtomTable::new();
+        let inlined = AtomTable::build_with(&atom_table, "inline");
+
+        assert!(inlined.is_inlined());
+        assert_eq!(&*inlined.as_str(), "inline");
+
+        let non_inlined = AtomTable::build_with(&atom_table, "longer non-inlined atom");
+
+        assert!(!non_inlined.is_inlined());
+        assert_eq!(&*non_inlined.as_str(), "longer non-inlined atom");
+    }
+
+    #[test]
+    fn functors_with_indexing_code_ptr() {
+        let code_ptr = IndexingCodePtr::Internal(0);
+        let functor = functor!(
+            atom!("first"),
+            [
+                string((String::from("a string"))),
+                indexing_code_ptr(code_ptr)
+            ]
+        );
+
+        let mut heap = Heap::new();
+        let mut functor_writer = Heap::functor_writer(functor);
+
+        functor_writer(&mut heap).unwrap();
+
+        assert_eq!(heap.cell_len(), 8);
+
+        assert_eq!(heap[0], atom_as_cell!(atom!("first"), 2));
+        assert_eq!(heap[1], pstr_loc_as_cell!(heap_index!(3)));
+        assert_eq!(heap[2], str_loc_as_cell!(6));
+        assert_eq!(
+            heap.slice_to_str(heap_index!(3), "a string".len()),
+            "a string"
+        );
+        assert_eq!(heap[5], empty_list_as_cell!());
+        assert_eq!(heap[6], atom_as_cell!(atom!("internal"), 1));
+        assert_eq!(heap[7], fixnum_as_cell!(Fixnum::build_with(0)));
+
+        heap.truncate(0);
+
+        let functor = functor!(
+            atom!("second"),
+            [
+                string((String::from("a string"))),
+                functor(
+                    (atom!("third")),
+                    [
+                        atom_as_cell((atom!("a"))),
+                        string((String::from("another string"))),
+                        indexing_code_ptr(code_ptr)
+                    ]
+                )
+            ]
+        );
+
+        let mut functor_writer = Heap::functor_writer(functor);
+        functor_writer(&mut heap).unwrap();
+
+        assert_eq!(heap.cell_len(), 15);
+
+        assert_eq!(heap[0], atom_as_cell!(atom!("second"), 2));
+        assert_eq!(heap[1], pstr_loc_as_cell!(heap_index!(3)));
+        assert_eq!(heap[2], str_loc_as_cell!(6));
+        assert_eq!(
+            heap.slice_to_str(heap_index!(3), "a string".len()),
+            "a string"
+        );
+        assert_eq!(heap[5], empty_list_as_cell!());
+        assert_eq!(heap[6], atom_as_cell!(atom!("third"), 3));
+        assert_eq!(heap[7], atom_as_cell!(atom!("a")));
+        assert_eq!(heap[8], pstr_loc_as_cell!(heap_index!(10)));
+        assert_eq!(heap[9], str_loc_as_cell!(13));
+        assert_eq!(
+            heap.slice_to_str(heap_index!(10), "another string".len()),
+            "another string"
+        );
+        assert_eq!(heap[12], empty_list_as_cell!());
+        assert_eq!(heap[13], atom_as_cell!(atom!("internal"), 1));
+        assert_eq!(heap[14], fixnum_as_cell!(Fixnum::build_with(0)));
+
+        let functor = functor!(
+            atom!("fourth"),
+            [
+                string((String::from("a string"))),
+                functor(
+                    (atom!("a")),
+                    [
+                        functor(
+                            (atom!("fifth")),
+                            [
+                                fixnum(5),
+                                string((String::from("another string"))),
+                                indexing_code_ptr(code_ptr)
+                            ]
+                        ),
+                        string((String::from("and another")))
+                    ]
+                )
+            ]
+        );
+
+        heap.truncate(0);
+
+        let mut functor_writer = Heap::functor_writer(functor);
+        functor_writer(&mut heap).unwrap();
+
+        assert_eq!(heap.cell_len(), 21);
+
+        assert_eq!(heap[0], atom_as_cell!(atom!("fourth"), 2));
+        assert_eq!(heap[1], pstr_loc_as_cell!(heap_index!(3)));
+        assert_eq!(heap[2], str_loc_as_cell!(6));
+        assert_eq!(
+            heap.slice_to_str(heap_index!(3), "a string".len()),
+            "a string"
+        );
+        assert_eq!(heap[5], empty_list_as_cell!());
+        assert_eq!(heap[6], atom_as_cell!(atom!("a"), 2));
+        assert_eq!(heap[7], str_loc_as_cell!(9));
+        assert_eq!(heap[8], pstr_loc_as_cell!(heap_index!(18))); // <-- wrong!
+        assert_eq!(heap[9], atom_as_cell!(atom!("fifth"), 3));
+        assert_eq!(heap[10], fixnum_as_cell!(Fixnum::build_with(5)));
+        assert_eq!(heap[11], pstr_loc_as_cell!(heap_index!(13)));
+        assert_eq!(heap[12], str_loc_as_cell!(16));
+        assert_eq!(
+            heap.slice_to_str(heap_index!(13), "another string".len()),
+            "another string"
+        );
+        assert_eq!(heap[15], empty_list_as_cell!());
+        assert_eq!(heap[16], atom_as_cell!(atom!("internal"), 1));
+        assert_eq!(heap[17], fixnum_as_cell!(Fixnum::build_with(0)));
+        assert_eq!(
+            heap.slice_to_str(heap_index!(18), "and another".len()),
+            "and another"
+        );
+        assert_eq!(heap[20], empty_list_as_cell!());
+
+        let constants = indexmap![
+            atom_as_cell!(atom!("a")) => IndexingCodePtr::External(2),
+            atom_as_cell!(atom!("d")) => IndexingCodePtr::External(7),
+        ];
+
+        let functor = variadic_functor(
+            atom!("switch_on_constants"),
+            1,
+            constants
+                .iter()
+                .map(|(c, ptr)| functor!(atom!(":"), [cell((*c)), indexing_code_ptr((*ptr))])),
+        );
+
+        heap.truncate(0);
+
+        let mut functor_writer = Heap::functor_writer(functor);
+        functor_writer(&mut heap).unwrap();
+
+        assert_eq!(heap[0], atom_as_cell!(atom!("switch_on_constants"), 1));
+        assert_eq!(heap[1], list_loc_as_cell!(2));
+        assert_eq!(heap[2], str_loc_as_cell!(6));
+        assert_eq!(heap[3], list_loc_as_cell!(4));
+        assert_eq!(heap[4], str_loc_as_cell!(11));
+        assert_eq!(heap[5], empty_list_as_cell!());
+        assert_eq!(heap[6], atom_as_cell!(atom!(":"), 2));
+        assert_eq!(heap[7], atom_as_cell!(atom!("a")));
+        assert_eq!(heap[8], str_loc_as_cell!(9));
+        assert_eq!(heap[9], atom_as_cell!(atom!("external"), 1));
+        assert_eq!(heap[10], fixnum_as_cell!(Fixnum::build_with(2)));
+        assert_eq!(heap[11], atom_as_cell!(atom!(":"), 2));
+        assert_eq!(heap[12], atom_as_cell!(atom!("d")));
+        assert_eq!(heap[13], str_loc_as_cell!(14));
+        assert_eq!(heap[14], atom_as_cell!(atom!("external"), 1));
+        assert_eq!(heap[15], fixnum_as_cell!(Fixnum::build_with(7)));
+    }
+
+    #[test]
+    fn undefined_procedure_functor() {
+        // existence_error
+        let culprit = functor!(atom!("/"), [atom_as_cell((atom!("a"))), fixnum(1)]);
+
+        let stub = functor!(
+            atom!("existence_error"),
+            [
+                atom_as_cell((atom!("procedure"))),
+                functor((culprit.clone()))
+            ]
+        );
+
+        println!("{stub:?}");
+
+        // now the error form
+        let lineless_error_form = functor!(atom!("error"), [functor(stub), functor(culprit)]);
+
+        println!("{lineless_error_form:?}");
+
+        let mut heap = Heap::new();
+        let mut functor_writer = Heap::functor_writer(lineless_error_form);
+
+        functor_writer(&mut heap).unwrap();
+
+        assert_eq!(heap[0], atom_as_cell!(atom!("error"), 2));
+        assert_eq!(heap[1], str_loc_as_cell!(3));
+        assert_eq!(heap[2], str_loc_as_cell!(9));
+        assert_eq!(heap[3], atom_as_cell!(atom!("existence_error"), 2));
+        assert_eq!(heap[4], atom_as_cell!(atom!("procedure")));
+        assert_eq!(heap[5], str_loc_as_cell!(6)); // is str_loc_as_cell!(3)
+        assert_eq!(heap[6], atom_as_cell!(atom!("/"), 2));
+        assert_eq!(heap[7], atom_as_cell!(atom!("a")));
+        assert_eq!(heap[8], fixnum_as_cell!(Fixnum::build_with(1)));
+        assert_eq!(heap[9], atom_as_cell!(atom!("/"), 2));
+        assert_eq!(heap[10], atom_as_cell!(atom!("a")));
+        assert_eq!(heap[11], fixnum_as_cell!(Fixnum::build_with(1)));
+    }
+
+    #[test]
+    fn argless_functor() {
+        let name = functor!(atom!("[]"));
+
+        assert_eq!(name.len(), 1);
+
+        let mut heap = Heap::new();
+        let mut functor_writer = Heap::functor_writer(name);
+        let loc = functor_writer(&mut heap).unwrap();
+
+        assert_eq!(loc, heap_loc_as_cell!(0));
+    }
+
+    #[test]
+    fn predefined_subfunctors() {
+        let stub = functor!(atom!("sub"), [atom_as_cell((atom!("[]")))]);
+        let name = functor!(atom!("super"), [functor(stub)]);
+
+        let mut heap = Heap::new();
+        let mut functor_writer = Heap::functor_writer(name);
+
+        functor_writer(&mut heap).unwrap();
+
+        assert_eq!(heap.cell_len(), 4);
+
+        assert_eq!(heap[0], atom_as_cell!(atom!("super"), 1));
+        assert_eq!(heap[1], str_loc_as_cell!(2));
+        assert_eq!(heap[2], atom_as_cell!(atom!("sub"), 1));
+        assert_eq!(heap[3], empty_list_as_cell!());
+    }
+}

src/http.rs

@@ -1,25 +1,23 @@
-use std::sync::Arc;
-use std::convert::Infallible;
+use bytes::{buf::Reader, Bytes};
+use std::sync::{Arc, Condvar, Mutex};
 
-use hyper::{Response, Request, Body};
-use tokio::sync::Mutex;
-use tokio::sync::mpsc::{channel, Receiver, Sender};
+use warp::http;
 
 pub struct HttpListener {
-    pub incoming: Receiver<HttpRequest>
+    pub incoming: std::sync::mpsc::Receiver<HttpRequest>,
 }
 
-#[derive(Debug)]
 pub struct HttpRequest {
-    pub request: Request<Body>,
+    pub request_data: HttpRequestData,
     pub response: HttpResponse,
 }
 
-pub type HttpResponse = Sender<Response<Body>>;
+pub type HttpResponse = Arc<(Mutex<bool>, Mutex<Option<warp::reply::Response>>, Condvar)>;
 
-pub async fn serve_req(req: Request<Body>, tx: Arc<Mutex<Sender<HttpRequest>>>) -> Result<Response<Body>, Infallible> {
-    let (response_tx, mut rx) = channel(1);
-    let http_request = HttpRequest { request: req, response: response_tx };
-    tx.lock().await.send(http_request).await.unwrap();
-    Ok(rx.recv().await.unwrap())
+pub struct HttpRequestData {
+    pub method: http::Method,
+    pub headers: http::HeaderMap,
+    pub path: String,
+    pub query: String,
+    pub body: Reader<Bytes>,
 }

src/indexing.rs

@@ -3,6 +3,7 @@ use crate::parser::ast::*;
 
 use crate::forms::*;
 use crate::instructions::*;
+use crate::types::HeapCellValue;
 
 use fxhash::FxBuildHasher;
 use indexmap::IndexMap;
@@ -112,15 +113,15 @@ impl<'a> IndexingCodeMergingPtr<'a> {
 
         match constant_key {
             Some(OptArgIndexKey::Literal(_, _, constant, _)) => {
-                constants.insert(*constant, constant_ptr);
+                constants.insert(HeapCellValue::from(*constant), constant_ptr);
+            }
+            _ if constant_ptr.is_external() => {
+                // this must be a defunct clause, because it's been deleted
+                // from the skeleton.
+                debug_assert!(constant_key.is_none());
             }
             _ => {
-                if let IndexingCodePtr::DynamicExternal(_) = constant_ptr {
-                    // this must be a defunct clause, because it's been deleted
-                    // from the skeleton.
-                } else {
-                    unreachable!()
-                }
+                unreachable!()
             }
         }
 
@@ -144,23 +145,26 @@ impl<'a> IndexingCodeMergingPtr<'a> {
     fn add_static_indexed_choice_for_constant(
         &mut self,
         external: usize,
-        constant: Literal,
+        constant: HeapCellValue,
         index: usize,
     ) {
         let third_level_index = if self.append_or_prepend.is_append() {
             vec![
                 IndexedChoiceInstruction::Try(external),
-                IndexedChoiceInstruction::Trust(index)
-            ].into()
+                IndexedChoiceInstruction::Trust(index),
+            ]
+            .into()
         } else {
             vec![
                 IndexedChoiceInstruction::Try(index),
-                IndexedChoiceInstruction::Trust(external)
-            ].into()
+                IndexedChoiceInstruction::Trust(external),
+            ]
+            .into()
         };
 
         let indexing_code_len = self.indexing_code.len();
-        self.indexing_code.push(IndexingLine::IndexedChoice(third_level_index));
+        self.indexing_code
+            .push(IndexingLine::IndexedChoice(third_level_index));
 
         match &mut self.indexing_code[self.offset] {
             IndexingLine::Indexing(IndexingInstruction::SwitchOnConstant(ref mut constants)) => {
@@ -178,7 +182,7 @@ impl<'a> IndexingCodeMergingPtr<'a> {
     fn add_dynamic_indexed_choice_for_constant(
         &mut self,
         external: usize,
-        constant: Literal,
+        constant: HeapCellValue,
         index: usize,
     ) {
         let third_level_index = if self.append_or_prepend.is_append() {
@@ -188,7 +192,8 @@ impl<'a> IndexingCodeMergingPtr<'a> {
         };
 
         let indexing_code_len = self.indexing_code.len();
-        self.indexing_code.push(IndexingLine::DynamicIndexedChoice(third_level_index));
+        self.indexing_code
+            .push(IndexingLine::DynamicIndexedChoice(third_level_index));
 
         match &mut self.indexing_code[self.offset] {
             IndexingLine::Indexing(IndexingInstruction::SwitchOnConstant(ref mut constants)) => {
@@ -231,8 +236,8 @@ impl<'a> IndexingCodeMergingPtr<'a> {
 
     fn index_overlapping_constant(
         &mut self,
-        orig_constant: Literal,
-        overlapping_constant: Literal,
+        orig_constant: HeapCellValue,
+        overlapping_constant: HeapCellValue,
         index: usize,
     ) {
         loop {
@@ -275,10 +280,8 @@ impl<'a> IndexingCodeMergingPtr<'a> {
                             );
                         }
                         None | Some(IndexingCodePtr::Fail) => {
-                            constants.insert(
-                                overlapping_constant,
-                                IndexingCodePtr::External(index),
-                            );
+                            constants
+                                .insert(overlapping_constant, IndexingCodePtr::External(index));
                         }
                         Some(IndexingCodePtr::DynamicExternal(o)) => {
                             self.add_dynamic_indexed_choice_for_constant(
@@ -314,7 +317,7 @@ impl<'a> IndexingCodeMergingPtr<'a> {
         }
     }
 
-    fn index_constant(&mut self, constant: Literal, index: usize) {
+    fn index_constant(&mut self, constant: HeapCellValue, index: usize) {
         loop {
             let indexing_code_len = self.indexing_code.len();
 
@@ -345,16 +348,10 @@ impl<'a> IndexingCodeMergingPtr<'a> {
                 IndexingLine::Indexing(IndexingInstruction::SwitchOnConstant(constants)) => {
                     match constants.get(&constant).cloned() {
                         None | Some(IndexingCodePtr::Fail) if self.is_dynamic => {
-                            constants.insert(
-                                constant,
-                                IndexingCodePtr::DynamicExternal(index),
-                            );
+                            constants.insert(constant, IndexingCodePtr::DynamicExternal(index));
                         }
                         None | Some(IndexingCodePtr::Fail) => {
-                            constants.insert(
-                                constant,
-                                IndexingCodePtr::External(index),
-                            );
+                            constants.insert(constant, IndexingCodePtr::External(index));
                         }
                         Some(IndexingCodePtr::DynamicExternal(o)) => {
                             self.add_dynamic_indexed_choice_for_constant(o, constant, index);
@@ -432,17 +429,20 @@ impl<'a> IndexingCodeMergingPtr<'a> {
         let third_level_index = if self.append_or_prepend.is_append() {
             vec![
                 IndexedChoiceInstruction::Try(external),
-                IndexedChoiceInstruction::Trust(index)
-            ].into()
+                IndexedChoiceInstruction::Trust(index),
+            ]
+            .into()
         } else {
             vec![
                 IndexedChoiceInstruction::Try(index),
-                IndexedChoiceInstruction::Trust(external)
-            ].into()
+                IndexedChoiceInstruction::Trust(external),
+            ]
+            .into()
         };
 
         let indexing_code_len = self.indexing_code.len();
-        self.indexing_code.push(IndexingLine::IndexedChoice(third_level_index));
+        self.indexing_code
+            .push(IndexingLine::IndexedChoice(third_level_index));
 
         match &mut self.indexing_code[self.offset] {
             IndexingLine::Indexing(IndexingInstruction::SwitchOnStructure(ref mut structures)) => {
@@ -470,7 +470,8 @@ impl<'a> IndexingCodeMergingPtr<'a> {
         };
 
         let indexing_code_len = self.indexing_code.len();
-        self.indexing_code.push(IndexingLine::DynamicIndexedChoice(third_level_index));
+        self.indexing_code
+            .push(IndexingLine::DynamicIndexedChoice(third_level_index));
 
         match &mut self.indexing_code[self.offset] {
             IndexingLine::Indexing(IndexingInstruction::SwitchOnStructure(ref mut structures)) => {
@@ -584,13 +585,15 @@ impl<'a> IndexingCodeMergingPtr<'a> {
                         let third_level_index = if self.append_or_prepend.is_append() {
                             vec![
                                 IndexedChoiceInstruction::Try(o),
-                                IndexedChoiceInstruction::Trust(index)
-                            ].into()
+                                IndexedChoiceInstruction::Trust(index),
+                            ]
+                            .into()
                         } else {
                             vec![
                                 IndexedChoiceInstruction::Try(index),
-                                IndexedChoiceInstruction::Trust(o)
-                            ].into()
+                                IndexedChoiceInstruction::Trust(o),
+                            ]
+                            .into()
                         };
 
                         self.indexing_code
@@ -613,7 +616,7 @@ pub(crate) fn merge_clause_index(
     target_indexing_code: &mut Vec<IndexingLine>,
     skeleton: &mut [ClauseIndexInfo], // the clause to be merged is the last element in the skeleton.
     retracted_clauses: &Option<Vec<ClauseIndexInfo>>,
-    new_clause_loc: usize,            // the absolute location of the new clause in the code vector.
+    new_clause_loc: usize, // the absolute location of the new clause in the code vector.
     append_or_prepend: AppendOrPrepend,
 ) {
     let opt_arg_index_key = match append_or_prepend {
@@ -631,14 +634,13 @@ pub(crate) fn merge_clause_index(
     match &opt_arg_index_key {
         OptArgIndexKey::Literal(_, index_loc, constant, ref overlapping_constants) => {
             let offset = new_clause_loc - index_loc + 1;
-            merging_ptr.index_constant(*constant, offset);
+            merging_ptr.index_constant(HeapCellValue::from(*constant), offset);
 
-            for overlapping_constant in overlapping_constants {
+            if let Some(overlapping_constant) = overlapping_constants {
                 merging_ptr.offset = 0;
-
                 merging_ptr.index_overlapping_constant(
-                    *constant,
-                    *overlapping_constant,
+                    HeapCellValue::from(*constant),
+                    HeapCellValue::from(*overlapping_constant),
                     offset,
                 );
             }
@@ -666,8 +668,8 @@ pub(crate) fn merge_clause_index(
 
 pub(crate) fn remove_constant_indices(
     constant: Literal,
-    overlapping_constants: &[Literal],
-    indexing_code: &mut Vec<IndexingLine>,
+    overlapping_constants: Option<Literal>,
+    indexing_code: &mut [IndexingLine],
     offset: usize,
 ) {
     let mut index = 0;
@@ -695,7 +697,7 @@ pub(crate) fn remove_constant_indices(
 
     let mut constants_index = 0;
 
-    for constant in iter {
+    for constant in iter.map(|l| HeapCellValue::from(*l)) {
         loop {
             match &mut indexing_code[index] {
                 IndexingLine::Indexing(IndexingInstruction::SwitchOnConstant(
@@ -703,13 +705,11 @@ pub(crate) fn remove_constant_indices(
                 )) => {
                     constants_index = index;
 
-                    let constant = *constant;
-
                     match constants.get(&constant).cloned() {
                         Some(IndexingCodePtr::DynamicExternal(_))
                         | Some(IndexingCodePtr::External(_))
                         | Some(IndexingCodePtr::Fail) => {
-                            constants.remove(&constant);
+                            constants.swap_remove(&constant);
                             break;
                         }
                         Some(IndexingCodePtr::Internal(o)) => {
@@ -742,7 +742,7 @@ pub(crate) fn remove_constant_indices(
                                 IndexingLine::Indexing(IndexingInstruction::SwitchOnConstant(
                                     ref mut constants,
                                 )) => {
-                                    constants.insert(*constant, ext);
+                                    constants.insert(constant, ext);
                                 }
                                 _ => {
                                     unreachable!()
@@ -775,7 +775,7 @@ pub(crate) fn remove_constant_indices(
                                 IndexingLine::Indexing(IndexingInstruction::SwitchOnConstant(
                                     ref mut constants,
                                 )) => {
-                                    constants.insert(*constant, ext);
+                                    constants.insert(constant, ext);
                                 }
                                 _ => {
                                     unreachable!()
@@ -813,7 +813,7 @@ pub(crate) fn remove_constant_indices(
 pub(crate) fn remove_structure_index(
     name: Atom,
     arity: usize,
-    indexing_code: &mut Vec<IndexingLine>,
+    indexing_code: &mut [IndexingLine],
     offset: usize,
 ) {
     let mut index = 0;
@@ -845,10 +845,10 @@ pub(crate) fn remove_structure_index(
             IndexingLine::Indexing(IndexingInstruction::SwitchOnStructure(ref mut structures)) => {
                 structures_index = index;
 
-                match structures.get(&(name.clone(), arity)).cloned() {
+                match structures.get(&(name, arity)).cloned() {
                     Some(IndexingCodePtr::DynamicExternal(_))
                     | Some(IndexingCodePtr::External(_)) => {
-                        structures.remove(&(name.clone(), arity));
+                        structures.swap_remove(&(name, arity));
                         break;
                     }
                     Some(IndexingCodePtr::Internal(o)) => {
@@ -879,7 +879,7 @@ pub(crate) fn remove_structure_index(
                             IndexingLine::Indexing(IndexingInstruction::SwitchOnStructure(
                                 ref mut structures,
                             )) => {
-                                structures.insert((name.clone(), arity), ext);
+                                structures.insert((name, arity), ext);
                             }
                             _ => {
                                 unreachable!()
@@ -910,7 +910,7 @@ pub(crate) fn remove_structure_index(
                             IndexingLine::Indexing(IndexingInstruction::SwitchOnStructure(
                                 ref mut structures,
                             )) => {
-                                structures.insert((name.clone(), arity), ext);
+                                structures.insert((name, arity), ext);
                             }
                             _ => {
                                 unreachable!()
@@ -950,7 +950,7 @@ pub(crate) fn remove_structure_index(
     }
 }
 
-pub(crate) fn remove_list_index(indexing_code: &mut Vec<IndexingLine>, offset: usize) {
+pub(crate) fn remove_list_index(indexing_code: &mut [IndexingLine], offset: usize) {
     let mut index = 0;
 
     match &mut indexing_code[index] {
@@ -1030,12 +1030,12 @@ pub(crate) fn remove_list_index(indexing_code: &mut Vec<IndexingLine>, offset: u
 
 pub(crate) fn remove_index(
     opt_arg_index_key: &OptArgIndexKey,
-    indexing_code: &mut Vec<IndexingLine>,
+    indexing_code: &mut [IndexingLine],
     clause_loc: usize,
 ) {
     match opt_arg_index_key {
         OptArgIndexKey::Literal(_, _, constant, ref overlapping_constants) => {
-            remove_constant_indices(*constant, overlapping_constants, indexing_code, clause_loc);
+            remove_constant_indices(*constant, *overlapping_constants, indexing_code, clause_loc);
         }
         OptArgIndexKey::Structure(_, _, name, arity) => {
             remove_structure_index(*name, *arity, indexing_code, clause_loc);
@@ -1051,93 +1051,72 @@ pub(crate) fn remove_index(
 
 #[inline]
 fn cap_choice_seq(prelude: &mut [IndexedChoiceInstruction]) {
-    prelude.first_mut().map(|instr| {
+    if let Some(instr) = prelude.first_mut() {
         *instr = IndexedChoiceInstruction::Try(instr.offset());
-    });
+    }
 
     cap_choice_seq_with_trust(prelude);
 }
 
 #[inline]
 fn cap_choice_seq_with_trust(prelude: &mut [IndexedChoiceInstruction]) {
-    prelude.last_mut().map(|instr| {
-        if let IndexedChoiceInstruction::Retry(i) = instr {
-            *instr = IndexedChoiceInstruction::Trust(*i);
+    if let Some(instr) = prelude.last_mut() {
+        match instr {
+            IndexedChoiceInstruction::Retry(i) => {
+                *instr = IndexedChoiceInstruction::Trust(*i);
+            }
+            IndexedChoiceInstruction::DefaultRetry(i) => {
+                *instr = IndexedChoiceInstruction::DefaultTrust(*i);
+            }
+            _ => {}
         }
-    });
+    }
 }
 
 #[inline]
 fn uncap_choice_seq_with_trust(prelude: &mut [IndexedChoiceInstruction]) {
-    prelude.last_mut().map(|instr| {
-        if let IndexedChoiceInstruction::Trust(i) = instr {
-            *instr = IndexedChoiceInstruction::Retry(*i);
+    if let Some(instr) = prelude.last_mut() {
+        match instr {
+            IndexedChoiceInstruction::Trust(i) => {
+                *instr = IndexedChoiceInstruction::Retry(*i);
+            }
+            IndexedChoiceInstruction::DefaultTrust(i) => {
+                *instr = IndexedChoiceInstruction::DefaultRetry(*i);
+            }
+            _ => {}
         }
-    });
+    }
 }
 
 #[inline]
 fn uncap_choice_seq_with_try(prelude: &mut [IndexedChoiceInstruction]) {
-    prelude.first_mut().map(|instr| {
+    if let Some(instr) = prelude.first_mut() {
         if let IndexedChoiceInstruction::Try(i) = instr {
             *instr = IndexedChoiceInstruction::Retry(*i);
         }
-    });
+    }
 }
 
-pub(crate) fn constant_key_alternatives(
-    constant: Literal,
-    atom_tbl: &mut AtomTable,
-    // arena: &mut Arena,
-) -> Vec<Literal> {
-    let mut constants = vec![];
+pub(crate) fn constant_key_alternatives(constant: Literal) -> Option<Literal> {
+    let n = match &constant {
+        Literal::Rational(n) if n.denominator().is_one() => n.numerator(),
+        Literal::Integer(n) => n,
+        _ => return None,
+    };
 
-    match constant {
-        Literal::Atom(ref name) => {
-            if let Some(c) = name.as_char() {
-                constants.push(Literal::Char(c));
-            }
-        }
-        Literal::Char(c) => {
-            let atom = atom_tbl.build_with(&c.to_string());
-            constants.push(Literal::Atom(atom));
-        }
-        /*
-        Literal::Fixnum(ref n) => {
-            constants.push(Literal::Integer(arena_alloc!(n, arena))); //Rc::new(Integer::from(*n))));
-
-            /*
-            if *n >= 0 {
-                if let Ok(n) = usize::try_from(*n) {
-                    constants.push(Literal::Usize(n));
-                }
-            }
-            */
-        }
-        */
-        Literal::Integer(ref n) => {
-            if let Some(n) = n.to_isize() {
-                Fixnum::build_with_checked(n as i64).map(|n| {
-                    constants.push(Literal::Fixnum(n));
-                }).unwrap();
-            }
-        }
-        _ => {}
-    }
-
-    constants
+    Fixnum::build_with_checked(n).map(Literal::Fixnum).ok()
 }
 
 #[derive(Debug)]
 pub(crate) struct StaticCodeIndices {
-    constants: IndexMap<Literal, VecDeque<IndexedChoiceInstruction>, FxBuildHasher>,
+    constants: IndexMap<HeapCellValue, VecDeque<IndexedChoiceInstruction>, FxBuildHasher>,
     lists: VecDeque<IndexedChoiceInstruction>,
     structures: IndexMap<(Atom, usize), VecDeque<IndexedChoiceInstruction>, FxBuildHasher>,
 }
 
 #[derive(Debug)]
 pub(crate) struct DynamicCodeIndices {
-    constants: IndexMap<Literal, VecDeque<usize>, FxBuildHasher>,
+    constants: IndexMap<HeapCellValue, VecDeque<usize>, FxBuildHasher>,
     lists: VecDeque<usize>,
     structures: IndexMap<(Atom, usize), VecDeque<usize>, FxBuildHasher>,
 }
@@ -1147,11 +1126,19 @@ pub(crate) trait Indexer {
 
     fn new() -> Self;
 
-    fn constants(&mut self) -> &mut IndexMap<Literal, VecDeque<Self::ThirdLevelIndex>, FxBuildHasher>;
+    fn constants(
+        &mut self,
+    ) -> &mut IndexMap<HeapCellValue, VecDeque<Self::ThirdLevelIndex>, FxBuildHasher>;
     fn lists(&mut self) -> &mut VecDeque<Self::ThirdLevelIndex>;
-    fn structures(&mut self) -> &mut IndexMap<(Atom, usize), VecDeque<Self::ThirdLevelIndex>, FxBuildHasher>;
+    fn structures(
+        &mut self,
+    ) -> &mut IndexMap<(Atom, usize), VecDeque<Self::ThirdLevelIndex>, FxBuildHasher>;
 
-    fn compute_index(is_initial_index: bool, index: usize) -> Self::ThirdLevelIndex;
+    fn compute_index(
+        is_initial_index: bool,
+        index: usize,
+        non_counted_bt: bool,
+    ) -> Self::ThirdLevelIndex;
 
     fn second_level_index<IndexKey: Eq + Hash>(
         indices: IndexMap<IndexKey, VecDeque<Self::ThirdLevelIndex>, FxBuildHasher>,
@@ -1187,7 +1174,9 @@ impl Indexer for StaticCodeIndices {
     }
 
     #[inline]
-    fn constants(&mut self) -> &mut IndexMap<Literal, VecDeque<IndexedChoiceInstruction>, FxBuildHasher> {
+    fn constants(
+        &mut self,
+    ) -> &mut IndexMap<HeapCellValue, VecDeque<IndexedChoiceInstruction>, FxBuildHasher> {
         &mut self.constants
     }
 
@@ -1197,13 +1186,21 @@ impl Indexer for StaticCodeIndices {
     }
 
     #[inline]
-    fn structures(&mut self) -> &mut IndexMap<(Atom, usize), VecDeque<IndexedChoiceInstruction>, FxBuildHasher> {
+    fn structures(
+        &mut self,
+    ) -> &mut IndexMap<(Atom, usize), VecDeque<IndexedChoiceInstruction>, FxBuildHasher> {
         &mut self.structures
     }
 
-    fn compute_index(is_initial_index: bool, index: usize) -> IndexedChoiceInstruction {
+    fn compute_index(
+        is_initial_index: bool,
+        index: usize,
+        non_counted_bt: bool,
+    ) -> IndexedChoiceInstruction {
         if is_initial_index {
             IndexedChoiceInstruction::Try(index + 1)
+        } else if non_counted_bt {
+            IndexedChoiceInstruction::DefaultRetry(index + 1)
         } else {
             IndexedChoiceInstruction::Retry(index + 1)
         }
@@ -1220,10 +1217,8 @@ impl Indexer for StaticCodeIndices {
                 index_locs.insert(key, IndexingCodePtr::Internal(prelude.len() + 1));
                 cap_choice_seq_with_trust(code.make_contiguous());
                 prelude.push_back(IndexingLine::from(code));
-            } else {
-                code.front().map(|i| {
-                    index_locs.insert(key, IndexingCodePtr::External(i.offset()));
-                });
+            } else if let Some(i) = code.front() {
+                index_locs.insert(key, IndexingCodePtr::External(i.offset()));
             }
         }
 
@@ -1231,7 +1226,9 @@ impl Indexer for StaticCodeIndices {
     }
 
     fn switch_on<IndexKey: Eq + Hash>(
-        mut instr_fn: impl FnMut(IndexMap<IndexKey, IndexingCodePtr, FxBuildHasher>) -> IndexingInstruction,
+        mut instr_fn: impl FnMut(
+            IndexMap<IndexKey, IndexingCodePtr, FxBuildHasher>,
+        ) -> IndexingInstruction,
         index: &mut IndexMap<IndexKey, VecDeque<IndexedChoiceInstruction>, FxBuildHasher>,
         prelude: &mut VecDeque<IndexingLine>,
     ) -> IndexingCodePtr {
@@ -1258,7 +1255,7 @@ impl Indexer for StaticCodeIndices {
     ) -> IndexingCodePtr {
         if lists.len() > 1 {
             cap_choice_seq_with_trust(lists.make_contiguous());
-            let lists = mem::replace(lists, VecDeque::new());
+            let lists = std::mem::take(lists);
             prelude.push_back(IndexingLine::from(lists));
 
             IndexingCodePtr::Internal(1)
@@ -1303,7 +1300,7 @@ impl Indexer for DynamicCodeIndices {
     }
 
     #[inline]
-    fn constants(&mut self) -> &mut IndexMap<Literal, VecDeque<usize>, FxBuildHasher> {
+    fn constants(&mut self) -> &mut IndexMap<HeapCellValue, VecDeque<usize>, FxBuildHasher> {
         &mut self.constants
     }
 
@@ -1318,7 +1315,7 @@ impl Indexer for DynamicCodeIndices {
     }
 
     #[inline]
-    fn compute_index(_: bool, index: usize) -> usize {
+    fn compute_index(_: bool, index: usize, _: bool) -> usize {
         index + 1
     }
 
@@ -1331,11 +1328,11 @@ impl Indexer for DynamicCodeIndices {
         for (key, code) in indices.into_iter() {
             if code.len() > 1 {
                 index_locs.insert(key, IndexingCodePtr::Internal(prelude.len() + 1));
-                prelude.push_back(IndexingLine::DynamicIndexedChoice(code.into_iter().collect()));
-            } else {
-                code.front().map(|i| {
-                    index_locs.insert(key, IndexingCodePtr::DynamicExternal(*i));
-                });
+                prelude.push_back(IndexingLine::DynamicIndexedChoice(
+                    code.into_iter().collect(),
+                ));
+            } else if let Some(i) = code.front() {
+                index_locs.insert(key, IndexingCodePtr::DynamicExternal(*i));
             }
         }
 
@@ -1343,7 +1340,9 @@ impl Indexer for DynamicCodeIndices {
     }
 
     fn switch_on<IndexKey: Eq + Hash>(
-        mut instr_fn: impl FnMut(IndexMap<IndexKey, IndexingCodePtr, FxBuildHasher>) -> IndexingInstruction,
+        mut instr_fn: impl FnMut(
+            IndexMap<IndexKey, IndexingCodePtr, FxBuildHasher>,
+        ) -> IndexingInstruction,
         index: &mut IndexMap<IndexKey, VecDeque<usize>, FxBuildHasher>,
         prelude: &mut VecDeque<IndexingLine>,
     ) -> IndexingCodePtr {
@@ -1369,8 +1368,10 @@ impl Indexer for DynamicCodeIndices {
         prelude: &mut VecDeque<IndexingLine>,
     ) -> IndexingCodePtr {
         if lists.len() > 1 {
-            let lists = mem::replace(lists, VecDeque::new());
-            prelude.push_back(IndexingLine::DynamicIndexedChoice(lists.into_iter().collect()));
+            let lists = std::mem::take(lists);
+            prelude.push_back(IndexingLine::DynamicIndexedChoice(
+                lists.into_iter().collect(),
+            ));
             IndexingCodePtr::Internal(1)
         } else {
             lists
@@ -1400,61 +1401,62 @@ impl Indexer for DynamicCodeIndices {
 pub(crate) struct CodeOffsets<I: Indexer> {
     indices: I,
     optimal_index: usize,
+    non_counted_bt: bool,
 }
 
 impl<I: Indexer> CodeOffsets<I> {
-    pub(crate) fn new(indices: I, optimal_index: usize) -> Self {
+    pub(crate) fn new(indices: I, optimal_index: usize, non_counted_bt: bool) -> Self {
         CodeOffsets {
             indices,
             optimal_index,
+            non_counted_bt,
         }
     }
 
     fn index_list(&mut self, index: usize) {
         let is_initial_index = self.indices.lists().is_empty();
-        let index = I::compute_index(is_initial_index, index);
+        let index = I::compute_index(is_initial_index, index, self.non_counted_bt);
         self.indices.lists().push_back(index);
     }
 
-    fn index_constant(
-        &mut self,
-        atom_tbl: &mut AtomTable,
-        constant: Literal,
-        index: usize,
-    ) -> Vec<Literal> {
-        let overlapping_constants = constant_key_alternatives(constant, atom_tbl);
-        let code = self.indices.constants().entry(constant).or_insert(VecDeque::new());
+    fn index_constant(&mut self, constant: Literal, index: usize) -> Option<Literal> {
+        let overlapping_constant_opt = constant_key_alternatives(constant);
+        let code = self
+            .indices
+            .constants()
+            .entry(HeapCellValue::from(constant))
+            .or_default();
 
         let is_initial_index = code.is_empty();
-        code.push_back(I::compute_index(is_initial_index, index));
+        code.push_back(I::compute_index(
+            is_initial_index,
+            index,
+            self.non_counted_bt,
+        ));
 
-        for constant in &overlapping_constants {
-            let code = self
-                .indices
-                .constants()
-                .entry(*constant)
-                .or_insert(VecDeque::new());
+        if let Some(constant) = overlapping_constant_opt.map(HeapCellValue::from) {
+            let code = self.indices.constants().entry(constant).or_default();
 
             let is_initial_index = code.is_empty();
-            let index = I::compute_index(is_initial_index, index);
+            let index = I::compute_index(is_initial_index, index, self.non_counted_bt);
 
             code.push_back(index);
         }
 
-        overlapping_constants
+        overlapping_constant_opt
     }
 
     fn index_structure(&mut self, name: Atom, arity: usize, index: usize) -> usize {
-        let code = self
-            .indices
-            .structures()
-            .entry((name.clone(), arity))
-            .or_insert(VecDeque::new());
+        let code = self.indices.structures().entry((name, arity)).or_default();
 
         let code_len = code.len();
         let is_initial_index = code.is_empty();
 
-        code.push_back(I::compute_index(is_initial_index, index));
+        code.push_back(I::compute_index(
+            is_initial_index,
+            index,
+            self.non_counted_bt,
+        ));
         code_len
     }
 
@@ -1463,25 +1465,24 @@ impl<I: Indexer> CodeOffsets<I> {
         optimal_arg: &Term,
         index: usize,
         clause_index_info: &mut ClauseIndexInfo,
-        atom_tbl: &mut AtomTable,
     ) {
         match optimal_arg {
             &Term::Clause(_, atom!("."), ref terms) if terms.len() == 2 => {
                 clause_index_info.opt_arg_index_key = OptArgIndexKey::List(self.optimal_index, 0);
                 self.index_list(index);
             }
-            &Term::Cons(..) | &Term::Literal(_, Literal::String(_)) | &Term::PartialString(..) => {
+            &Term::Cons(..) | &Term::PartialString(..) | &Term::CompleteString(..) => {
                 clause_index_info.opt_arg_index_key = OptArgIndexKey::List(self.optimal_index, 0);
                 self.index_list(index);
             }
             &Term::Clause(_, name, ref terms) => {
                 clause_index_info.opt_arg_index_key =
-                    OptArgIndexKey::Structure(self.optimal_index, 0, name.clone(), terms.len());
+                    OptArgIndexKey::Structure(self.optimal_index, 0, name, terms.len());
 
                 self.index_structure(name, terms.len(), index);
             }
             &Term::Literal(_, constant) => {
-                let overlapping_constants = self.index_constant(atom_tbl, constant, index);
+                let overlapping_constants = self.index_constant(constant, index);
 
                 clause_index_info.opt_arg_index_key =
                     OptArgIndexKey::Literal(self.optimal_index, 0, constant, overlapping_constants);
@@ -1528,20 +1529,14 @@ impl<I: Indexer> CodeOffsets<I> {
             &mut prelude,
         );
 
-        match &mut str_loc {
-            IndexingCodePtr::Internal(ref mut i) => {
-                *i += emitted_switch_on_constant as usize; // con_loc.is_internal() as usize;
-            }
-            _ => {}
-        };
+        if let IndexingCodePtr::Internal(ref mut i) = &mut str_loc {
+            *i += emitted_switch_on_constant as usize; // con_loc.is_internal() as usize;
+        }
 
-        match &mut lst_loc {
-            IndexingCodePtr::Internal(ref mut i) => {
-                *i += emitted_switch_on_constant as usize; // con_loc.is_internal() as usize;
-                *i += emitted_switch_on_structure as usize; // str_loc.is_internal() as usize;
-            }
-            _ => {}
-        };
+        if let IndexingCodePtr::Internal(ref mut i) = &mut lst_loc {
+            *i += emitted_switch_on_constant as usize; // con_loc.is_internal() as usize;
+            *i += emitted_switch_on_structure as usize; // str_loc.is_internal() as usize;
+        }
 
         let var_offset = 1 + skip_stub_try_me_else as usize;
 

src/iterators.rs

@@ -5,47 +5,34 @@ use crate::parser::ast::*;
 
 use std::cell::Cell;
 use std::collections::VecDeque;
-use std::fmt;
 use std::iter::*;
 use std::rc::Rc;
 use std::vec::Vec;
 
+#[allow(clippy::borrowed_box)]
 #[derive(Debug, Clone)]
 pub(crate) enum TermRef<'a> {
     AnonVar(Level),
     Cons(Level, &'a Cell<RegType>, &'a Term, &'a Term),
     Literal(Level, &'a Cell<RegType>, &'a Literal),
     Clause(Level, &'a Cell<RegType>, Atom, &'a Vec<Term>),
-    PartialString(Level, &'a Cell<RegType>, &'a String, &'a Box<Term>),
-    CompleteString(Level, &'a Cell<RegType>, Atom),
-    Var(Level, &'a Cell<VarReg>, Rc<String>),
-}
-
-impl<'a> TermRef<'a> {
-    pub(crate) fn level(self) -> Level {
-        match self {
-            TermRef::AnonVar(lvl)
-            | TermRef::Cons(lvl, ..)
-            | TermRef::Literal(lvl, ..)
-            | TermRef::Var(lvl, ..)
-            | TermRef::Clause(lvl, ..)
-            | TermRef::CompleteString(lvl, ..)
-            | TermRef::PartialString(lvl, ..) => lvl,
-        }
-    }
+    PartialString(Level, &'a Cell<RegType>, Rc<String>, &'a Box<Term>),
+    CompleteString(Level, &'a Cell<RegType>, Rc<String>),
+    Var(Level, &'a Cell<VarReg>, VarPtr),
 }
 
+#[allow(clippy::borrowed_box)]
 #[derive(Debug)]
 pub(crate) enum TermIterState<'a> {
     AnonVar(Level),
-    Literal(Level, &'a Cell<RegType>, &'a Literal),
     Clause(Level, usize, &'a Cell<RegType>, Atom, &'a Vec<Term>),
+    Literal(Level, &'a Cell<RegType>, &'a Literal),
     InitialCons(Level, &'a Cell<RegType>, &'a Term, &'a Term),
     FinalCons(Level, &'a Cell<RegType>, &'a Term, &'a Term),
-    InitialPartialString(Level, &'a Cell<RegType>, &'a String, &'a Box<Term>),
-    FinalPartialString(Level, &'a Cell<RegType>, &'a String, &'a Box<Term>),
-    CompleteString(Level, &'a Cell<RegType>, Atom),
-    Var(Level, &'a Cell<VarReg>, Rc<String>),
+    InitialPartialString(Level, &'a Cell<RegType>, Rc<String>, &'a Box<Term>),
+    FinalPartialString(Level, &'a Cell<RegType>, Rc<String>, &'a Box<Term>),
+    CompleteString(Level, &'a Cell<RegType>, Rc<String>),
+    Var(Level, &'a Cell<VarReg>, VarPtr),
 }
 
 impl<'a> TermIterState<'a> {
@@ -60,12 +47,12 @@ impl<'a> TermIterState<'a> {
             }
             Term::Literal(cell, constant) => TermIterState::Literal(lvl, cell, constant),
             Term::PartialString(cell, string_buf, tail) => {
-                TermIterState::InitialPartialString(lvl, cell, string_buf, tail)
+                TermIterState::InitialPartialString(lvl, cell, string_buf.clone(), tail)
             }
-            Term::CompleteString(cell, atom) => {
-                TermIterState::CompleteString(lvl, cell, *atom)
+            Term::CompleteString(cell, string) => {
+                TermIterState::CompleteString(lvl, cell, string.clone())
             }
-            Term::Var(cell, var) => TermIterState::Var(lvl, cell, var.clone()),
+            Term::Var(cell, var_ptr) => TermIterState::Var(lvl, cell, var_ptr.clone()),
         }
     }
 }
@@ -81,6 +68,7 @@ impl<'a> QueryIterator<'a> {
             .push(TermIterState::subterm_to_state(lvl, term));
     }
 
+    /*
     fn from_rule_head_clause(terms: &'a Vec<Term>) -> Self {
         let state_stack = terms
             .iter()
@@ -90,23 +78,21 @@ impl<'a> QueryIterator<'a> {
 
         QueryIterator { state_stack }
     }
+    */
 
     fn from_term(term: &'a Term) -> Self {
         let state = match term {
-            Term::AnonVar | Term::Cons(..) | Term::Literal(..) |
-            Term::PartialString(..) | Term::CompleteString(..) => {
+            Term::AnonVar
+            | Term::Cons(..)
+            | Term::Literal(..)
+            | Term::PartialString(..)
+            | Term::CompleteString(..) => {
                 return QueryIterator {
                     state_stack: vec![],
                 }
             }
-            Term::Clause(r, name, terms) => TermIterState::Clause(
-                Level::Root,
-                0,
-                r,
-                *name,
-                terms,
-            ),
-            Term::Var(cell, var) => TermIterState::Var(Level::Root, cell, var.clone()),
+            Term::Clause(r, name, terms) => TermIterState::Clause(Level::Root, 0, r, *name, terms),
+            Term::Var(cell, var_ptr) => TermIterState::Var(Level::Root, cell, var_ptr.clone()),
         };
 
         QueryIterator {
@@ -114,46 +100,27 @@ impl<'a> QueryIterator<'a> {
         }
     }
 
-    fn new(term: &'a QueryTerm) -> Self {
+    fn extend_state(&mut self, lvl: Level, term: &'a QueryTerm) {
         match term {
-            &QueryTerm::Clause(ref cell, ClauseType::CallN(_), ref terms, _) => {
-                let state = TermIterState::Clause(Level::Root, 1, cell, atom!("$call"), terms);
-                QueryIterator {
-                    state_stack: vec![state],
-                }
+            QueryTerm::Clause(ref cell, ClauseType::CallN(_), ref terms, _) => {
+                self.state_stack
+                    .push(TermIterState::Clause(lvl, 1, cell, atom!("$call"), terms));
             }
-            &QueryTerm::Clause(ref cell, ref ct, ref terms, _) => {
-                let state = TermIterState::Clause(Level::Root, 0, cell, ct.name(), terms);
-                QueryIterator {
-                    state_stack: vec![state],
-                }
+            QueryTerm::Clause(ref cell, ref ct, ref terms, _) => {
+                self.state_stack
+                    .push(TermIterState::Clause(lvl, 0, cell, ct.name(), terms));
             }
-            &QueryTerm::UnblockedCut(ref cell) => {
-                let state = TermIterState::Var(Level::Root, cell, Rc::new("!".to_string()));
-                QueryIterator {
-                    state_stack: vec![state],
-                }
-            }
-            &QueryTerm::GetLevelAndUnify(ref cell, ref var) => {
-                let state = TermIterState::Var(Level::Root, cell, var.clone());
-                QueryIterator {
-                    state_stack: vec![state],
-                }
-            }
-            &QueryTerm::Jump(ref vars) => {
-                let state_stack = vars
-                    .iter()
-                    .rev()
-                    .map(|t| TermIterState::subterm_to_state(Level::Shallow, t))
-                    .collect();
-
-                QueryIterator { state_stack }
-            }
-            &QueryTerm::BlockedCut => QueryIterator {
-                state_stack: vec![],
-            },
+            _ => {}
         }
     }
+
+    pub fn new(term: &'a QueryTerm) -> Self {
+        let mut iter = QueryIterator {
+            state_stack: vec![],
+        };
+        iter.extend_state(Level::Root, term);
+        iter
+    }
 }
 
 impl<'a> Iterator for QueryIterator<'a> {
@@ -191,20 +158,22 @@ impl<'a> Iterator for QueryIterator<'a> {
                     }
                 }
                 TermIterState::InitialCons(lvl, cell, head, tail) => {
-                    self.state_stack.push(TermIterState::FinalCons(lvl, cell, head, tail));
+                    self.state_stack
+                        .push(TermIterState::FinalCons(lvl, cell, head, tail));
 
                     self.push_subterm(lvl.child_level(), tail);
                     self.push_subterm(lvl.child_level(), head);
                 }
                 TermIterState::InitialPartialString(lvl, cell, string, tail) => {
-                    self.state_stack.push(TermIterState::FinalPartialString(lvl, cell, string, tail));
+                    self.state_stack
+                        .push(TermIterState::FinalPartialString(lvl, cell, string, tail));
                     self.push_subterm(lvl.child_level(), tail);
                 }
-                TermIterState::FinalPartialString(lvl, cell, atom, tail) => {
-                    return Some(TermRef::PartialString(lvl, cell, atom, tail));
+                TermIterState::FinalPartialString(lvl, cell, string, tail) => {
+                    return Some(TermRef::PartialString(lvl, cell, string, tail));
                 }
-                TermIterState::CompleteString(lvl, cell, atom) => {
-                    return Some(TermRef::CompleteString(lvl, cell, atom));
+                TermIterState::CompleteString(lvl, cell, string) => {
+                    return Some(TermRef::CompleteString(lvl, cell, string));
                 }
                 TermIterState::FinalCons(lvl, cell, head, tail) => {
                     return Some(TermRef::Cons(lvl, cell, head, tail));
@@ -212,8 +181,8 @@ impl<'a> Iterator for QueryIterator<'a> {
                 TermIterState::Literal(lvl, cell, constant) => {
                     return Some(TermRef::Literal(lvl, cell, constant));
                 }
-                TermIterState::Var(lvl, cell, var) => {
-                    return Some(TermRef::Var(lvl, cell, var));
+                TermIterState::Var(lvl, cell, var_ptr) => {
+                    return Some(TermRef::Var(lvl, cell, var_ptr));
                 }
             };
         }
@@ -225,7 +194,7 @@ impl<'a> Iterator for QueryIterator<'a> {
 #[derive(Debug)]
 pub(crate) struct FactIterator<'a> {
     state_queue: VecDeque<TermIterState<'a>>,
-    iterable_root: bool,
+    iterable_root: RootIterationPolicy,
 }
 
 impl<'a> FactIterator<'a> {
@@ -234,7 +203,7 @@ impl<'a> FactIterator<'a> {
             .push_back(TermIterState::subterm_to_state(lvl, term));
     }
 
-    pub(crate) fn from_rule_head_clause(terms: &'a Vec<Term>) -> Self {
+    pub(crate) fn from_rule_head_clause(terms: &'a [Term]) -> Self {
         let state_queue = terms
             .iter()
             .map(|bt| TermIterState::subterm_to_state(Level::Shallow, bt))
@@ -242,11 +211,11 @@ impl<'a> FactIterator<'a> {
 
         FactIterator {
             state_queue,
-            iterable_root: false,
+            iterable_root: RootIterationPolicy::NotIterated,
         }
     }
 
-    fn new(term: &'a Term, iterable_root: bool) -> Self {
+    fn new(term: &'a Term, iterable_root: RootIterationPolicy) -> Self {
         let states = match term {
             Term::AnonVar => {
                 vec![TermIterState::AnonVar(Level::Root)]
@@ -260,26 +229,26 @@ impl<'a> FactIterator<'a> {
                 head.as_ref(),
                 tail.as_ref(),
             )],
-            Term::PartialString(cell, string_buf, tail) => {
+            Term::PartialString(cell, string, tail) => {
                 vec![TermIterState::InitialPartialString(
                     Level::Root,
                     cell,
-                    string_buf,
+                    string.clone(),
                     tail,
                 )]
             }
-            Term::CompleteString(cell, atom) => {
+            Term::CompleteString(cell, string) => {
                 vec![TermIterState::CompleteString(
                     Level::Root,
                     cell,
-                    *atom,
+                    string.clone(),
                 )]
             }
             Term::Literal(cell, constant) => {
                 vec![TermIterState::Literal(Level::Root, cell, constant)]
             }
-            Term::Var(cell, var) => {
-                vec![TermIterState::Var(Level::Root, cell, var.clone())]
+            Term::Var(cell, var_ptr) => {
+                vec![TermIterState::Var(Level::Root, cell, var_ptr.clone())]
             }
         };
 
@@ -305,7 +274,7 @@ impl<'a> Iterator for FactIterator<'a> {
                     }
 
                     match lvl {
-                        Level::Root if !self.iterable_root => continue,
+                        Level::Root if !self.iterable_root.iterable() => continue,
                         _ => return Some(TermRef::Clause(lvl, cell, name, child_terms)),
                     };
                 }
@@ -325,8 +294,8 @@ impl<'a> Iterator for FactIterator<'a> {
                 TermIterState::Literal(lvl, cell, constant) => {
                     return Some(TermRef::Literal(lvl, cell, constant))
                 }
-                TermIterState::Var(lvl, cell, var) => {
-                    return Some(TermRef::Var(lvl, cell, var));
+                TermIterState::Var(lvl, cell, var_ptr) => {
+                    return Some(TermRef::Var(lvl, cell, var_ptr));
                 }
                 _ => {}
             }
@@ -336,197 +305,138 @@ impl<'a> Iterator for FactIterator<'a> {
     }
 }
 
-pub(crate) fn post_order_iter<'a>(term: &'a Term) -> QueryIterator<'a> {
+pub(crate) fn post_order_iter(term: &Term) -> QueryIterator<'_> {
     QueryIterator::from_term(term)
 }
 
-pub(crate) fn breadth_first_iter<'a>(term: &'a Term, iterable_root: bool) -> FactIterator<'a> {
+pub(crate) fn breadth_first_iter(
+    term: &Term,
+    iterable_root: RootIterationPolicy,
+) -> FactIterator<'_> {
     FactIterator::new(term, iterable_root)
 }
 
+#[derive(Debug, Copy, Clone)]
+enum ClauseIteratorState<'a> {
+    RemainingChunks(&'a VecDeque<ChunkedTerms>, usize),
+    RemainingBranches(&'a Vec<VecDeque<ChunkedTerms>>, usize),
+}
+
+#[derive(Debug, Clone)]
+pub(crate) enum ClauseItem<'a> {
+    FirstBranch(usize),
+    NextBranch,
+    BranchEnd(usize),
+    Chunk { terms: &'a VecDeque<QueryTerm> },
+}
+
 #[derive(Debug)]
-pub(crate) enum ChunkedTerm<'a> {
-    HeadClause(Atom, &'a Vec<Term>),
-    BodyTerm(&'a QueryTerm),
+pub(crate) struct ClauseIterator<'a> {
+    state_stack: Vec<ClauseIteratorState<'a>>,
+    remaining_chunks_on_stack: usize,
 }
 
-pub(crate) fn query_term_post_order_iter<'a>(query_term: &'a QueryTerm) -> QueryIterator<'a> {
-    QueryIterator::new(query_term)
-}
-
-impl<'a> ChunkedTerm<'a> {
-    pub(crate) fn post_order_iter(&self) -> QueryIterator<'a> {
-        match self {
-            &ChunkedTerm::BodyTerm(qt) => QueryIterator::new(qt),
-            &ChunkedTerm::HeadClause(_, terms) => QueryIterator::from_rule_head_clause(terms),
+fn state_from_chunked_terms(chunk_vec: &VecDeque<ChunkedTerms>) -> ClauseIteratorState<'_> {
+    if chunk_vec.len() == 1 {
+        if let Some(ChunkedTerms::Branch(ref branches)) = chunk_vec.front() {
+            return ClauseIteratorState::RemainingBranches(branches, 0);
         }
     }
+
+    ClauseIteratorState::RemainingChunks(chunk_vec, 0)
 }
 
-fn contains_cut_var<'a, Iter: Iterator<Item = &'a Term>>(terms: Iter) -> bool {
-    for term in terms {
-        if let &Term::Var(_, ref var) = term {
-            if var.as_str() == "!" {
-                return true;
+impl<'a> ClauseIterator<'a> {
+    pub fn new(clauses: &'a ChunkedTermVec) -> Self {
+        match state_from_chunked_terms(&clauses.chunk_vec) {
+            state @ ClauseIteratorState::RemainingBranches(..) => Self {
+                state_stack: vec![state],
+                remaining_chunks_on_stack: 0,
+            },
+            state @ ClauseIteratorState::RemainingChunks(..) => Self {
+                state_stack: vec![state],
+                remaining_chunks_on_stack: 1,
+            },
+        }
+    }
+
+    #[inline(always)]
+    pub fn in_tail_position(&self) -> bool {
+        self.remaining_chunks_on_stack == 0
+    }
+
+    fn branch_end_depth(&mut self) -> usize {
+        let mut depth = 1;
+
+        while let Some(state) = self.state_stack.pop() {
+            match state {
+                ClauseIteratorState::RemainingBranches(terms, focus) if terms.len() == focus => {
+                    depth += 1;
+                }
+                _ => {
+                    self.state_stack.push(state);
+                    break;
+                }
             }
         }
-    }
 
-    false
-}
-
-pub(crate) struct ChunkedIterator<'a> {
-    pub(crate) chunk_num: usize,
-    iter: Box<dyn Iterator<Item = ChunkedTerm<'a>> + 'a>,
-    deep_cut_encountered: bool,
-    cut_var_in_head: bool,
-}
-
-impl<'a> fmt::Debug for ChunkedIterator<'a> {
-    fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result {
-        fmt.debug_struct("ChunkedIterator")
-            .field("chunk_num", &self.chunk_num)
-            // Hacky solution.
-            .field("iter", &"Box<dyn Iterator<Item = ChunkedTerm<'a>> + 'a>")
-            .field("deep_cut_encountered", &self.deep_cut_encountered)
-            .field("cut_var_in_head", &self.cut_var_in_head)
-            .finish()
+        depth
     }
 }
 
-type ChunkedIteratorItem<'a> = (usize, usize, Vec<ChunkedTerm<'a>>);
-type RuleBodyIteratorItem<'a> = (usize, usize, Vec<&'a QueryTerm>);
-
-impl<'a> ChunkedIterator<'a> {
-    pub(crate) fn rule_body_iter(self) -> Box<dyn Iterator<Item = RuleBodyIteratorItem<'a>> + 'a> {
-        Box::new(self.filter_map(|(cn, lt_arity, terms)| {
-            let filtered_terms: Vec<_> = terms
-                .into_iter()
-                .filter_map(|ct| match ct {
-                    ChunkedTerm::BodyTerm(qt) => Some(qt),
-                    _ => None,
-                })
-                .collect();
-
-            if filtered_terms.is_empty() {
-                None
-            } else {
-                Some((cn, lt_arity, filtered_terms))
-            }
-        }))
-    }
-
-    pub(crate) fn from_rule_body(p1: &'a QueryTerm, clauses: &'a Vec<QueryTerm>) -> Self {
-        let inner_iter = Box::new(once(ChunkedTerm::BodyTerm(p1)));
-        let iter = inner_iter.chain(clauses.iter().map(|t| ChunkedTerm::BodyTerm(t)));
-
-        ChunkedIterator {
-            chunk_num: 0,
-            iter: Box::new(iter),
-            deep_cut_encountered: false,
-            cut_var_in_head: false,
-        }
-    }
-
-    pub(crate) fn from_rule(rule: &'a Rule) -> Self {
-        let &Rule {
-            head: (ref name, ref args, ref p1),
-            ref clauses,
-        } = rule;
-
-        let iter = once(ChunkedTerm::HeadClause(name.clone(), args));
-        let inner_iter = Box::new(once(ChunkedTerm::BodyTerm(p1)));
-        let iter = iter.chain(inner_iter.chain(clauses.iter().map(|t| ChunkedTerm::BodyTerm(t))));
-
-        ChunkedIterator {
-            chunk_num: 0,
-            iter: Box::new(iter),
-            deep_cut_encountered: false,
-            cut_var_in_head: false,
-        }
-    }
-
-    pub(crate) fn encountered_deep_cut(&self) -> bool {
-        self.deep_cut_encountered
-    }
-
-    fn take_chunk(&mut self, term: ChunkedTerm<'a>) -> (usize, usize, Vec<ChunkedTerm<'a>>) {
-        let mut arity = 0;
-        let mut item = Some(term);
-        let mut result = Vec::new();
-
-        while let Some(term) = item {
-            match term {
-                ChunkedTerm::HeadClause(_, terms) => {
-                    if contains_cut_var(terms.iter()) {
-                        self.cut_var_in_head = true;
-                    }
-
-                    result.push(term);
-                }
-                ChunkedTerm::BodyTerm(&QueryTerm::Jump(ref vars)) => {
-                    result.push(term);
-                    arity = vars.len();
-
-                    if contains_cut_var(vars.iter()) && !self.cut_var_in_head {
-                        self.deep_cut_encountered = true;
-                    }
-
-                    break;
-                }
-                ChunkedTerm::BodyTerm(&QueryTerm::BlockedCut) => {
-                    result.push(term);
-
-                    if self.chunk_num > 0 {
-                        self.deep_cut_encountered = true;
-                    }
-                }
-                ChunkedTerm::BodyTerm(&QueryTerm::GetLevelAndUnify(..)) => {
-                    self.deep_cut_encountered = true;
-
-                    result.push(term);
-                    arity = 1;
-                    break;
-                }
-                ChunkedTerm::BodyTerm(&QueryTerm::UnblockedCut(..)) => {
-                    self.deep_cut_encountered = true;
-                    result.push(term);
-                }
-                ChunkedTerm::BodyTerm(&QueryTerm::Clause(_, ClauseType::Inlined(_), ..)) => {
-                    result.push(term)
-                }
-                ChunkedTerm::BodyTerm(&QueryTerm::Clause(
-                    _,
-                    ClauseType::CallN(_),
-                    ref subterms,
-                    _,
-                )) => {
-                    result.push(term);
-                    arity = subterms.len() + 1;
-                    break;
-                }
-                ChunkedTerm::BodyTerm(qt) => {
-                    result.push(term);
-                    arity = qt.arity();
-                    break;
-                }
-            };
-
-            item = self.iter.next();
-        }
-
-        let chunk_num = self.chunk_num;
-        self.chunk_num += 1;
-
-        (chunk_num, arity, result)
-    }
-}
-
-impl<'a> Iterator for ChunkedIterator<'a> {
-    // the chunk number, last term arity, and vector of references.
-    type Item = ChunkedIteratorItem<'a>;
+impl<'a> Iterator for ClauseIterator<'a> {
+    type Item = ClauseItem<'a>;
 
     fn next(&mut self) -> Option<Self::Item> {
-        self.iter.next().map(|term| self.take_chunk(term))
+        while let Some(state) = self.state_stack.pop() {
+            match state {
+                ClauseIteratorState::RemainingChunks(chunks, focus) if focus < chunks.len() => {
+                    if focus + 1 < chunks.len() {
+                        self.state_stack
+                            .push(ClauseIteratorState::RemainingChunks(chunks, focus + 1));
+                    } else {
+                        self.remaining_chunks_on_stack -= 1;
+                    }
+
+                    match &chunks[focus] {
+                        ChunkedTerms::Branch(branches) => {
+                            self.state_stack
+                                .push(ClauseIteratorState::RemainingBranches(branches, 0));
+                        }
+                        ChunkedTerms::Chunk { ref terms } => {
+                            return Some(ClauseItem::Chunk { terms });
+                        }
+                    }
+                }
+                ClauseIteratorState::RemainingChunks(chunks, focus) => {
+                    debug_assert_eq!(chunks.len(), focus);
+                }
+                ClauseIteratorState::RemainingBranches(branches, focus)
+                    if focus < branches.len() =>
+                {
+                    self.state_stack
+                        .push(ClauseIteratorState::RemainingBranches(branches, focus + 1));
+                    let state = state_from_chunked_terms(&branches[focus]);
+
+                    if let ClauseIteratorState::RemainingChunks(..) = &state {
+                        self.remaining_chunks_on_stack += 1;
+                    }
+
+                    self.state_stack.push(state);
+
+                    return if focus == 0 {
+                        Some(ClauseItem::FirstBranch(branches.len()))
+                    } else {
+                        Some(ClauseItem::NextBranch)
+                    };
+                }
+                ClauseIteratorState::RemainingBranches(branches, focus) => {
+                    debug_assert_eq!(branches.len(), focus);
+                    return Some(ClauseItem::BranchEnd(self.branch_end_depth()));
+                }
+            }
+        }
+
+        None
     }
 }

src/lib.rs

@@ -1,36 +1,83 @@
+//! A free software ISO Prolog system.
 #![recursion_limit = "4112"]
+#![deny(missing_docs)]
 
 #[macro_use]
 extern crate static_assertions;
 
 #[macro_use]
-pub mod macros;
+pub(crate) mod macros;
 #[macro_use]
-pub mod atom_table;
+pub(crate) mod atom_table;
 #[macro_use]
-pub mod arena;
+pub(crate) mod arena;
+pub(crate) mod offset_table;
 #[macro_use]
-pub mod parser;
+pub(crate) mod parser;
+#[macro_use]
+pub(crate) mod functor_macro;
 mod allocator;
 mod arithmetic;
-pub mod codegen;
+pub(crate) mod codegen;
 mod debray_allocator;
-mod fixtures;
+#[cfg(feature = "ffi")]
+mod ffi;
 mod forms;
 mod heap_iter;
-pub mod heap_print;
+pub(crate) mod heap_print;
+#[cfg(feature = "http")]
 mod http;
 mod indexing;
+mod variable_records;
 #[macro_use]
-pub mod instructions {
+pub(crate) mod instructions {
     include!(concat!(env!("OUT_DIR"), "/instructions.rs"));
 }
 mod iterators;
-pub mod machine;
+pub(crate) mod machine;
 mod raw_block;
-pub mod read;
+pub(crate) mod read;
+#[cfg(feature = "repl")]
 mod repl_helper;
 mod targets;
-pub mod types;
+pub(crate) mod types;
 
-use instructions::instr;
+// Re-exports
+pub use machine::config::*;
+pub use machine::lib_machine::*;
+pub use machine::Machine;
+
+#[cfg(target_arch = "wasm32")]
+pub mod wasm;
+
+#[cfg(not(target_arch = "wasm32"))]
+/// The entry point for the Scryer Prolog CLI.
+pub fn run_binary() -> std::process::ExitCode {
+    use crate::atom_table::Atom;
+    use crate::machine::INTERRUPT;
+
+    #[cfg(feature = "repl")]
+    ctrlc::set_handler(move || {
+        INTERRUPT.store(true, std::sync::atomic::Ordering::Relaxed);
+    })
+    .unwrap();
+
+    #[cfg(target_arch = "wasm32")]
+    let runtime = tokio::runtime::Builder::new_current_thread()
+        .enable_all()
+        .build()
+        .unwrap();
+
+    #[cfg(not(target_arch = "wasm32"))]
+    let runtime = tokio::runtime::Builder::new_multi_thread()
+        .enable_all()
+        .build()
+        .unwrap();
+
+    runtime.block_on(async move {
+        let mut wam = MachineBuilder::default()
+            .with_streams(StreamConfig::stdio())
+            .build();
+        wam.run_module_predicate(atom!("$toplevel"), (atom!("$repl"), 0))
+    })
+}

src/lib/arithmetic.pl

@@ -1,4 +1,9 @@
-:- module(arithmetic, [expmod/4, lsb/2, msb/2, number_to_rational/2,
+/** Arithmetic predicates
+
+These predicates are additions to standard the arithmetic functions provided by `is/2`.
+*/
+
+:- module(arithmetic, [expmod/4, lcm/3, lsb/2, msb/2, number_to_rational/2,
                        number_to_rational/3, popcount/2,
                        rational_numerator_denominator/3]).
 
@@ -6,6 +11,10 @@
 :- use_module(library(error)).
 :- use_module(library(lists), [append/3, member/2]).
 
+
+%% expmod(+Base, +Expo, +Mod, -R).
+%
+% Modular exponentiation. Base, Expo and Mod must be integers.
 expmod(Base, Expo, Mod, R) :-
     (   member(N, [Base, Expo, Mod]), var(N) -> instantiation_error(expmod/4)
     ;   member(N, [Base, Expo, Mod]), \+ integer(N) ->
@@ -28,6 +37,25 @@ expmod_(Base0, Expo0, Mod, C, R) :-
     Base is (Base0 * Base0) mod Mod,
     expmod_(Base, Expo, Mod, C, R).
 
+%% lcm(+A, +B, -Lcm) is det.
+%
+% Calculates the Least common multiple for A and B: the smallest positive integer
+% that is divisible by both A and B.
+% 
+% A and B need to be integers.
+lcm(A, B, X) :-
+    builtins:must_be_number(A, lcm/2),
+    builtins:must_be_number(B, lcm/2),
+    (   \+ integer(A) -> type_error(integer, A, lcm/2)
+    ;   \+ integer(B) -> type_error(integer, B, lcm/2)
+    ;   (A = 0, B = 0) -> X = 0
+    ;   builtins:can_be_number(X, lcm/2),
+	X is abs(B) // gcd(A,B) * abs(A)
+    ).
+
+%% lsb(+X, -N).
+%
+% True iff N is the least significat bit of integer X
 lsb(X, N) :-
     builtins:must_be_number(X, lsb/2),
     (   \+ integer(X) -> type_error(integer, X, lsb/2)
@@ -37,6 +65,9 @@ lsb(X, N) :-
         msb_(X1, -1, N)
     ).
 
+%% msb(+X, -N).
+%
+% True iff N is the most significant bit of integer X
 msb(X, N) :-
     builtins:must_be_number(X, msb/2),
     (   \+ integer(X) -> type_error(integer, X, msb/2)
@@ -52,6 +83,9 @@ msb_(X, M, N) :-
     M1 is M + 1,
     msb_(X1, M1, N).
 
+%% number_to_rational(+Real, -Fraction).
+%
+% True iff given a number Real, Fraction is the same number represented as a fraction.
 number_to_rational(Real, Fraction) :-
     (   var(Real) -> instantiation_error(number_to_rational/2)
     ;   integer(Real) -> Fraction is Real rdiv 1
@@ -110,12 +144,20 @@ simplify_fraction(A0/B0, A/B) :-
     A is A0 // G,
     B is B0 // G.
 
+%% rational_numerator_denominator(+Fraction, -Numerator, -Denominator).
+%
+% True iff given a fraction Fraction, Numerator is the numerator of that fraction
+% and Denominator the denominator.
 rational_numerator_denominator(R, N, D) :-
     write_term_to_chars(R, [], Cs),
     append(Ns, [' ', r, d, i, v, ' '|Ds], Cs),
     number_chars(N, Ns),
     number_chars(D, Ds).
 
+%% popcount(+Number, -Bits1).
+%
+% True iff given an integer Number, Bits1 is the amount of 1 bits the binary representation
+% of that number has.
 popcount(X, N) :-
     must_be(integer, X),
     '$popcount'(X, N).

src/lib/assoc.pl

@@ -54,28 +54,27 @@
 
 :- use_module(library(lists)).
 
-/** <module> Binary associations
+/** Binary associations
 
 Assocs are Key-Value associations implemented as  a balanced binary tree
 (AVL tree).
 
-@see            library(pairs), library(rbtrees)
-@author         R.A.O'Keefe, L.Damas, V.S.Costa and Jan Wielemaker
+Authors: R.A.O'Keefe, L.Damas, V.S.Costa and Jan Wielemaker
 */
 
-:- meta_predicate map_assoc(1, ?).
-:- meta_predicate map_assoc(2, ?, ?).
+:- meta_predicate(map_assoc(1, ?)).
+:- meta_predicate(map_assoc(2, ?, ?)).
 
-%!  empty_assoc(?Assoc) is semidet.
+%% empty_assoc(?Assoc) is semidet.
 %
-%   Is true if Assoc is the empty association list.
+% Is true if Assoc is the empty association list.
 
 empty_assoc(t).
 
-%!  assoc_to_list(+Assoc, -Pairs) is det.
+%% assoc_to_list(+Assoc, -Pairs) is det.
 %
-%   Translate Assoc to a list Pairs of Key-Value pairs.  The keys
-%   in Pairs are sorted in ascending order.
+% Translate Assoc to a list Pairs of Key-Value pairs.  The keys
+% in Pairs are sorted in ascending order.
 
 assoc_to_list(Assoc, List) :-
     assoc_to_list(Assoc, List, []).
@@ -86,10 +85,10 @@ assoc_to_list(t(Key,Val,_,L,R), List, Rest) :-
 assoc_to_list(t, List, List).
 
 
-%!  assoc_to_keys(+Assoc, -Keys) is det.
+%% assoc_to_keys(+Assoc, -Keys) is det.
 %
-%   True if Keys is the list of keys   in Assoc. The keys are sorted
-%   in ascending order.
+% True if Keys is the list of keys   in Assoc. The keys are sorted
+% in ascending order.
 
 assoc_to_keys(Assoc, List) :-
     assoc_to_keys(Assoc, List, []).
@@ -100,11 +99,11 @@ assoc_to_keys(t(Key,_,_,L,R), List, Rest) :-
 assoc_to_keys(t, List, List).
 
 
-%!  assoc_to_values(+Assoc, -Values) is det.
+%% assoc_to_values(+Assoc, -Values) is det.
 %
-%   True if Values is the  list  of   values  in  Assoc.  Values are
-%   ordered in ascending  order  of  the   key  to  which  they were
-%   associated.  Values may contain duplicates.
+% True if Values is the  list  of   values  in  Assoc.  Values are
+% ordered in ascending  order  of  the   key  to  which  they were
+% associated.  Values may contain duplicates.
 
 assoc_to_values(Assoc, List) :-
     assoc_to_values(Assoc, List, []).
@@ -114,12 +113,12 @@ assoc_to_values(t(_,Value,_,L,R), List, Rest) :-
     assoc_to_values(R, More, Rest).
 assoc_to_values(t, List, List).
 
-%!  is_assoc(+Assoc) is semidet.
+%% is_assoc(+Assoc) is semidet.
 %
-%   True if Assoc is an association list. This predicate checks
-%   that the structure is valid, elements are in order, and tree
-%   is balanced to the extent guaranteed by AVL trees.  I.e.,
-%   branches of each subtree differ in depth by at most 1.
+% True if Assoc is an association list. This predicate checks
+% that the structure is valid, elements are in order, and tree
+% is balanced to the extent guaranteed by AVL trees.  I.e.,
+% branches of each subtree differ in depth by at most 1.
 
 is_assoc(Assoc) :-
     is_assoc(Assoc, _Min, _Max, _Depth).
@@ -151,12 +150,10 @@ balance(=,-).
 balance(<,<).
 balance(>,>).
 
-%!  gen_assoc(?Key, +Assoc, ?Value) is nondet.
+%% gen_assoc(?Key, +Assoc, ?Value) is nondet.
 %
-%   True if Key-Value is an association in Assoc. Enumerates keys in
-%   ascending order on backtracking.
-%
-%   @see get_assoc/3.
+% True if Key-Value is an association in Assoc. Enumerates keys in
+% ascending order on backtracking.
 
 gen_assoc(Key, Assoc, Value) :-
     (   ground(Key)
@@ -171,11 +168,11 @@ gen_assoc_(Key, t(_,_,_,_,R), Val) :-
     gen_assoc_(Key, R, Val).
 
 
-%!  get_assoc(+Key, +Assoc, -Value) is semidet.
+%% get_assoc(+Key, +Assoc, -Value) is semidet.
 %
-%   True if Key-Value is an association in Assoc.
+% True if Key-Value is an association in Assoc.
 %
-%   @error type_error(assoc, Assoc) if Assoc is not an association list.
+% Throws error: `type_error(assoc, Assoc)` if Assoc is not an association list.
 
 get_assoc(Key, Assoc, Val) :-
     must_be(assoc, Assoc),
@@ -201,9 +198,9 @@ get_assoc(>, Key, _, _, Tree, Val) :-
 % :- endif.
 
 
-%!  get_assoc(+Key, +Assoc0, ?Val0, ?Assoc, ?Val) is semidet.
+%% get_assoc(+Key, +Assoc0, ?Val0, ?Assoc, ?Val) is semidet.
 %
-%   True if Key-Val0 is in Assoc0 and Key-Val is in Assoc.
+% True if Key-Val0 is in Assoc0 and Key-Val is in Assoc.
 
 get_assoc(Key, t(K,V,B,L,R), Val, t(K,NV,B,NL,NR), NVal) :-
     compare(Rel, Key, K),
@@ -216,12 +213,12 @@ get_assoc(>, Key, V, L, R, Val, V, L, NR, NVal) :-
     get_assoc(Key, R, Val, NR, NVal).
 
 
-%!  list_to_assoc(+Pairs, -Assoc) is det.
+%% list_to_assoc(+Pairs, -Assoc) is det.
 %
-%   Create an association from a list Pairs of Key-Value pairs. List
-%   must not contain duplicate keys.
+% Create an association from a list Pairs of Key-Value pairs. List
+% must not contain duplicate keys.
 %
-%   @error domain_error(unique_key_pairs, List) if List contains duplicate keys
+% Throws error: `domain_error(unique_key_pairs, List)` if List contains duplicate keys
 
 list_to_assoc(List, Assoc) :-
     (  List = [] -> Assoc = t
@@ -246,13 +243,13 @@ list_to_assoc(N, List, More, Depth, t(K,V,Balance,L,R)) :-
     compare(B, RDepth, LDepth),
     balance(B, Balance).
 
-%!  ord_list_to_assoc(+Pairs, -Assoc) is det.
+%% ord_list_to_assoc(+Pairs, -Assoc) is det.
 %
-%   Assoc is created from an ordered list Pairs of Key-Value
-%   pairs. The pairs must occur in strictly ascending order of
-%   their keys.
+% Assoc is created from an ordered list Pairs of Key-Value
+% pairs. The pairs must occur in strictly ascending order of
+% their keys.
 %
-%   @error domain_error(key_ordered_pairs, List) if pairs are not ordered.
+% Throws error: `domain_error(key_ordered_pairs, List)` if pairs are not ordered.
 
 ord_list_to_assoc(Sorted, Assoc) :-
     (  Sorted = [] -> Assoc = t
@@ -263,9 +260,9 @@ ord_list_to_assoc(Sorted, Assoc) :-
            )
     ).
 
-%!  ord_pairs(+Pairs) is semidet
+%% ord_pairs(+Pairs) is semidet
 %
-%   True if Pairs is a list of Key-Val pairs strictly ordered by key.
+% True if Pairs is a list of Key-Val pairs strictly ordered by key.
 
 ord_pairs([K-_V|Rest]) :-
     ord_pairs(Rest, K).
@@ -274,9 +271,9 @@ ord_pairs([K-_V|Rest], K0) :-
     K0 @< K,
     ord_pairs(Rest, K).
 
-%!  map_assoc(:Pred, +Assoc) is semidet.
+%% map_assoc(:Pred, +Assoc) is semidet.
 %
-%   True if Pred(Value) is true for all values in Assoc.
+% True if Pred(Value) is true for all values in Assoc.
 
 map_assoc(Pred, T) :-
     map_assoc_(T, Pred).
@@ -287,10 +284,10 @@ map_assoc_(t(_,Val,_,L,R), Pred) :-
     call(Pred, Val),
     map_assoc_(R, Pred).
 
-%!  map_assoc(:Pred, +Assoc0, ?Assoc) is semidet.
+%% map_assoc(:Pred, +Assoc0, ?Assoc) is semidet.
 %
-%   Map corresponding values. True if Assoc is Assoc0 with Pred
-%   applied to all corresponding pairs of of values.
+% Map corresponding values. True if Assoc is Assoc0 with Pred
+% applied to all corresponding pairs of of values.
 
 map_assoc(Pred, T0, T) :-
     map_assoc_(T0, Pred, T).
@@ -302,9 +299,9 @@ map_assoc_(t(Key,Val,B,L0,R0), Pred, t(Key,Ans,B,L1,R1)) :-
     map_assoc_(R0, Pred, R1).
 
 
-%!  max_assoc(+Assoc, -Key, -Value) is semidet.
+%% max_assoc(+Assoc, -Key, -Value) is semidet.
 %
-%   True if Key-Value is in Assoc and Key is the largest key.
+% True if Key-Value is in Assoc and Key is the largest key.
 
 max_assoc(t(K,V,_,_,R), Key, Val) :-
     max_assoc(R, K, V, Key, Val).
@@ -314,9 +311,9 @@ max_assoc(t(K,V,_,_,R), _, _, Key, Val) :-
     max_assoc(R, K, V, Key, Val).
 
 
-%!  min_assoc(+Assoc, -Key, -Value) is semidet.
+%% min_assoc(+Assoc, -Key, -Value) is semidet.
 %
-%   True if Key-Value is in assoc and Key is the smallest key.
+% True if Key-Value is in assoc and Key is the smallest key.
 
 min_assoc(t(K,V,_,L,_), Key, Val) :-
     min_assoc(L, K, V, Key, Val).
@@ -326,10 +323,10 @@ min_assoc(t(K,V,_,L,_), _, _, Key, Val) :-
     min_assoc(L, K, V, Key, Val).
 
 
-%!  put_assoc(+Key, +Assoc0, +Value, -Assoc) is det.
+%% put_assoc(+Key, +Assoc0, +Value, -Assoc) is det.
 %
-%   Assoc is Assoc0, except that Key is associated with
-%   Value. This can be used to insert and change associations.
+% Assoc is Assoc0, except that Key is associated with
+% Value. This can be used to insert and change associations.
 
 put_assoc(Key, A0, Value, A) :-
     insert(A0, Key, Value, A, _).
@@ -361,11 +358,11 @@ table(<      , right   , -      , no        , no    ) :- !.
 table(>      , left    , -      , no        , no    ) :- !.
 table(>      , right   , -      , no        , yes   ) :- !.
 
-%!  del_min_assoc(+Assoc0, ?Key, ?Val, -Assoc) is semidet.
+%% del_min_assoc(+Assoc0, ?Key, ?Val, -Assoc) is semidet.
 %
-%   True if Key-Value  is  in  Assoc0   and  Key  is  the smallest key.
-%   Assoc is Assoc0 with Key-Value   removed. Warning: This will
-%   succeed with _no_ bindings for Key or Val if Assoc0 is empty.
+% True if Key-Value  is  in  Assoc0   and  Key  is  the smallest key.
+% Assoc is Assoc0 with Key-Value   removed. Warning: This will
+% succeed with _no_ bindings for Key or Val if Assoc0 is empty.
 
 del_min_assoc(Tree, Key, Val, NewTree) :-
     del_min_assoc(Tree, Key, Val, NewTree, _DepthChanged).
@@ -375,11 +372,11 @@ del_min_assoc(t(K,V,B,L,R), Key, Val, NewTree, Changed) :-
     del_min_assoc(L, Key, Val, NewL, LeftChanged),
     deladjust(LeftChanged, t(K,V,B,NewL,R), left, NewTree, Changed).
 
-%!  del_max_assoc(+Assoc0, ?Key, ?Val, -Assoc) is semidet.
+%% del_max_assoc(+Assoc0, ?Key, ?Val, -Assoc) is semidet.
 %
-%   True if Key-Value  is  in  Assoc0   and  Key  is  the greatest key.
-%   Assoc is Assoc0 with Key-Value   removed. Warning: This will
-%   succeed with _no_ bindings for Key or Val if Assoc0 is empty.
+% True if Key-Value  is  in  Assoc0   and  Key  is  the greatest key.
+% Assoc is Assoc0 with Key-Value   removed. Warning: This will
+% succeed with _no_ bindings for Key or Val if Assoc0 is empty.
 
 del_max_assoc(Tree, Key, Val, NewTree) :-
     del_max_assoc(Tree, Key, Val, NewTree, _DepthChanged).
@@ -389,10 +386,10 @@ del_max_assoc(t(K,V,B,L,R), Key, Val, NewTree, Changed) :-
     del_max_assoc(R, Key, Val, NewR, RightChanged),
     deladjust(RightChanged, t(K,V,B,L,NewR), right, NewTree, Changed).
 
-%!  del_assoc(+Key, +Assoc0, ?Value, -Assoc) is semidet.
+%% del_assoc(+Key, +Assoc0, ?Value, -Assoc) is semidet.
 %
-%   True if Key-Value is  in  Assoc0.   Assoc  is  Assoc0 with
-%   Key-Value removed.
+% True if Key-Value is  in  Assoc0.   Assoc  is  Assoc0 with
+% Key-Value removed.
 
 del_assoc(Key, A0, Value, A) :-
     delete(A0, Key, Value, A, _).

src/lib/atts.pl

@@ -1,8 +1,8 @@
 :- module(atts, [op(1199, fx, attribute),
-                 call_residue_vars/2,
                  term_attributed_variables/2]).
 
 :- use_module(library(dcgs)).
+:- use_module(library(error)).
 :- use_module(library(terms)).
 
 /* represent the list of attributes belonging to a variable,
@@ -19,77 +19,12 @@
     '$default_attr_list'(PGs, Module, AttrVar).
 '$default_attr_list'([], _, _) --> [].
 
-'$absent_attr'(V, Attr) :-
-    '$get_attr_list'(V, Ls),
-    '$absent_from_list'(Ls, Attr).
-
-'$absent_from_list'(X, Attr) :-
-    (  var(X) ->
-       true
-    ;  X = [L|Ls],
-       L \= Attr ->
-       '$absent_from_list'(Ls, Attr)
-    ).
-
-'$get_attr'(V, Attr) :-
-    '$get_attr_list'(V, Ls),
-    nonvar(Ls),
-    '$get_from_list'(Ls, V, Attr).
-
-'$get_from_list'([L|Ls], V, Attr) :-
-    nonvar(L),
-    (  L \= Attr ->
-       nonvar(Ls),
-       '$get_from_list'(Ls, V, Attr)
-    ;  L = Attr,
-       '$enqueue_attr_var'(V)
-    ).
-
-'$put_attr'(V, Attr) :-
-    '$get_attr_list'(V, Ls),
-    '$add_to_list'(Ls, V, Attr).
-
-'$add_to_list'(Ls, V, Attr) :-
-    (  var(Ls) ->
-       Ls = [Attr | _],
-       '$enqueue_attr_var'(V)
-    ;  Ls = [_ | Ls0],
-       '$add_to_list'(Ls0, V, Attr)
-    ).
-
-'$del_attr'(Ls0, _, _) :-
-    var(Ls0),
-    !.
-'$del_attr'(Ls0, V, Attr) :-
-    Ls0 = [Att | Ls1],
-    nonvar(Att),
-    (  Att \= Attr ->
-       '$del_attr_buried'(Ls0, Ls1, V, Attr)
-    ;  '$enqueue_attr_var'(V),
-       '$del_attr_head'(V),
-       '$del_attr'(Ls1, V, Attr)
-    ).
-
-'$del_attr_step'(Ls1, V, Attr) :-
-    (  nonvar(Ls1) ->
-       Ls1 = [_ | Ls2],
-       '$del_attr_buried'(Ls1, Ls2, V, Attr)
+'$absent_attr'(V, Module, Attr) :-
+    (  '$get_from_attr_list'(V, Module, Attr) ->
+       false
     ;  true
     ).
 
-%% assumptions: Ls0 is a list, Ls1 is its tail;
-%%              the head of Ls0 can be ignored.
-'$del_attr_buried'(Ls0, Ls1, V, Attr) :-
-    (  var(Ls1) -> true
-    ;  Ls1 = [Att | Ls2] ->
-       (  Att \= Attr ->
-          '$del_attr_buried'(Ls1, Ls2, V, Attr)
-       ;  '$enqueue_attr_var'(V),
-	      '$del_attr_non_head'(Ls0), %% set tail of Ls0 = tail of Ls1. can be undone by backtracking.
-	      '$del_attr_step'(Ls1, V, Attr)
-       )
-    ).
-
 '$copy_attr_list'(L, _Module, []) :- var(L), !.
 '$copy_attr_list'([Module0:Att|Atts], Module, CopiedAtts) :-
     (  Module0 == Module ->
@@ -145,38 +80,28 @@ put_attr(Name, Arity, Module) -->
     { functor(Attr, Name, Arity) },
     [(put_atts(V, +Attr) :-
           !,
-          functor(Attr, Head, Arity),
-	      functor(AttrForm, Head, Arity),
-	      '$get_attr_list'(V, Ls),
-	      atts:'$del_attr'(Ls, V, Module:AttrForm),
-	      atts:'$put_attr'(V, Module:Attr)),
-     (put_atts(V,  Attr) :-
+          '$put_to_attr_list'(V, Module, Attr)),
+     (put_atts(V, Attr) :-
           !,
-          functor(Attr, Head, Arity),
-	      functor(AttrForm, Head, Arity),
-	      '$get_attr_list'(V, Ls),
-	      atts:'$del_attr'(Ls, V, Module:AttrForm),
-	      atts:'$put_attr'(V, Module:Attr)),
+          '$put_to_attr_list'(V, Module, Attr)),
      (put_atts(V, -Attr) :-
           !,
-          functor(Attr, _, _),
-	      '$get_attr_list'(V, Ls),
-	      atts:'$del_attr'(Ls, V, Module:Attr))].
+          '$del_from_attr_list'(V, Module, Attr))].
 
 get_attr(Name, Arity, Module) -->
     { functor(Attr, Name, Arity) },
     [(get_atts(V, +Attr) :-
           !,
           functor(Attr, _, _),
-          atts:'$get_attr'(V, Module:Attr)),
+          atts:'$get_from_attr_list'(V, Module, Attr)),
      (get_atts(V,  Attr) :-
           !,
           functor(Attr, _, _),
-          atts:'$get_attr'(V, Module:Attr)),
+          atts:'$get_from_attr_list'(V, Module, Attr)),
      (get_atts(V, -Attr) :-
           !,
           functor(Attr, _, _),
-          atts:'$absent_attr'(V, Module:Attr))].
+          atts:'$absent_attr'(V, Module, Attr))].
 
 user:goal_expansion(Term, M:put_atts(Var, Attr)) :-
     nonvar(Term),
@@ -185,12 +110,5 @@ user:goal_expansion(Term, M:get_atts(Var, Attr)) :-
     nonvar(Term),
     Term = get_atts(Var, M, Attr).
 
-:- meta_predicate call_residue_vars(0, ?).
-
-call_residue_vars(Goal, Vars) :-
-    '$get_attr_var_queue_delim'(B),
-    call(Goal),
-    '$get_attr_var_queue_beyond'(B, Vars).
-
 term_attributed_variables(Term, Vars) :-
     '$term_attributed_variables'(Term, Vars).

src/lib/between.pl

@@ -1,3 +1,10 @@
+/** Predicates that generate integers
+
+These predicates can be used to reason about integers in a reduced domain that
+follow some property. `library(clpz)` provides another way of reasoning about
+integers that may also be interesting.
+*/
+
 :- module(between, [between/3, gen_int/1, gen_nat/1, numlist/2, numlist/3, repeat/1]).
 
 %% TODO: numlist/5.
@@ -5,6 +12,24 @@
 :- use_module(library(lists), [length/2]).
 :- use_module(library(error)).
 
+%% between(+Lower, +Upper, -X).
+%
+% Given Lower and Upper are both integer numbers, true iff X is an integer so that _Lower =< X =< Upper_.
+% Can be used both to check if X is between Lower and Upper or to generate an integer between
+% Lower and Upper.
+%
+% Examples:
+%
+% ```
+% ?- between(10, 20, 15).
+%    true.
+% ?- between(10, 20, 25).
+%    false.
+% ?- between(3, 5, X).
+%    X = 3
+% ;  X = 4
+% ;  X = 5.
+% ```
 between(Lower, Upper, X) :-
     must_be(integer, Lower),
     must_be(integer, Upper),
@@ -30,6 +55,9 @@ enumerate_nats(I0, N) :-
     I1 is I0 + 1,
     enumerate_nats(I1, N).
 
+%% gen_nat(?N)
+%
+% True iff N is a natural number.
 gen_nat(N) :-
     can_be(integer, N),
     (   var(N) -> enumerate_nats(0, N)
@@ -44,6 +72,9 @@ enumerate_ints(I0, N) :-
     I1 is I0 + 1,
     enumerate_ints(I1, N).
 
+%% gen_int(?N)
+%
+% True iff N is an integer.
 gen_int(N) :-
     can_be(integer, N),
     (   var(N) -> enumerate_ints(0, N)
@@ -55,9 +86,24 @@ repeat_integer(N) :-
 repeat_integer(N0) :-
     N0 > 0, N1 is N0 - 1, repeat_integer(N1).
 
+%% repeat(+N)
+%
+% Succeeds N times. This predicate is only included for compatibility and *should not be used*
+% because it lacks a declarative interpretation.
 repeat(N) :-
     must_be(integer, N), repeat_integer(N).
 
+%% numlist(?Upper, ?List)
+%
+% True iff List is the list of integers _[1, ..., Upper]_. Example:
+%
+% ```
+% ?- numlist(X, Y).
+%    X = 1, Y = [1],
+% ;  X = 2, Y = [1,2]
+% ;  X = 3, Y = [1,2,3]
+% ; ... .
+% ```
 numlist(Upper, List) :-
     (  integer(Upper) -> findall(X, between(1, Upper, X), List)
     ;  List = [_|_], length(List, Upper), findall(X, between(1, Upper, X), List)
@@ -106,5 +152,14 @@ gen_ints(L, U) :-
     ),
     L =< U.
 
+%% numlist(?Lower, ?Upper, ?List).
+%
+% True iff List is a list of the form _[Lower, ..., Upper]_.
+% Example:
+%
+% ```
+% ?- numlist(5, 10, X).
+%    X = [5,6,7,8,9,10].
+% ```
 numlist(Lower, Upper, List) :-
     gen_ints(Lower, Upper), findall(X, between(Lower, Upper, X), List).

src/lib/charsio.pl

@@ -1,9 +1,18 @@
+/** High-level predicates to work with chars and strings
+
+This module contains predicates that relates strings of chars
+to other representations, as well as high-level predicates to
+read and write chars.
+
+*/
+
 :- module(charsio, [char_type/2,
                     chars_utf8bytes/2,
                     get_single_char/1,
                     get_n_chars/3,
-                    read_line_to_chars/3,
+                    get_line_to_chars/3,
                     read_from_chars/2,
+                    read_term_from_chars/3,
                     write_term_to_chars/3,
                     chars_base64/3]).
 
@@ -11,6 +20,7 @@
 :- use_module(library(iso_ext)).
 :- use_module(library(error)).
 :- use_module(library(lists)).
+:- use_module(library(between)).
 :- use_module(library(iso_ext), [partial_string/1,partial_string/3]).
 
 fabricate_var_name(VarType, VarName, N) :-
@@ -65,18 +75,86 @@ extend_var_list_([V|Vs], N, VarList, NewVarList, VarType) :-
     ).
 
 
+%% char_type(?Char, ?Type).
+%
+% Type is one of the categories that Char fits in.
+% At least one of the arguments must be ground.
+% Possible categories are:
+%
+% - `alnum`
+% - `alpha`
+% - `alphabetic`
+% - `alphanumeric`
+% - `ascii`
+% - `ascii_graphic`
+% - `ascii_punctuation`
+% - `binary_digit`
+% - `control`
+% - `decimal_digit`
+% - `exponent`
+% - `graphic`
+% - `graphic_token`
+% - `hexadecimal_digit`
+% - `layout`
+% - `lower`
+% - `meta`
+% - `numeric`
+% - `octal_digit`
+% - `octet`
+% - `prolog`
+% - `sign`
+% - `solo`
+% - `symbolic_control`
+% - `symbolic_hexadecimal`
+% - `upper`
+% - `lower(Lower)`
+% - `upper(Upper)`
+% - `whitespace`
+%
+% An example:
+%
+% ```
+% ?- char_type(a, Type).
+%    Type = alnum
+% ;  Type = alpha
+% ;  Type = alphabetic
+% ;  Type = alphanumeric
+% ;  Type = ascii
+% ;  Type = ascii_graphic
+% ;  Type = hexadecimal_digit
+% ;  Type = lower
+% ;  Type = octet
+% ;  Type = prolog
+% ;  Type = symbolic_control
+% ;  Type = lower("a")
+% ;  Type = upper("A")
+% ;  false.
+% ```
+%
+% Note that uppercase and lowercase transformations use a string. This is because
+% some characters do not map 1:1 between lowercase and uppercase.
 char_type(Char, Type) :-
-        must_be(character, Char),
-        (   ground(Type) ->
-            (   ctype(Type) ->
-                '$char_type'(Char, Type)
-            ;   domain_error(char_type, Type, char_type/2)
-            )
-        ;   ctype(Type),
+        can_be(character, Char),
+        (   \+ ctype(Type) ->
+            domain_error(char_type, Type, char_type/2)
+        ;   true
+        ),
+        (   ground(Char) ->
+            ctype(Type),
             '$char_type'(Char, Type)
+        ;   ground(Type) ->
+            ccode(Code),
+            char_code(Char, Code),
+            '$char_type'(Char, Type)
+        ;   must_be(character, Char)
         ).
 
 
+% 0xD800 to 0xDFFF are surrogate code points used by UTF-16.
+
+ccode(Code) :- between(0, 0xD7FF, Code).
+ccode(Code) :- between(0xE000, 0x10FFFF, Code).
+
 ctype(alnum).
 ctype(alpha).
 ctype(alphabetic).
@@ -102,27 +180,68 @@ ctype(sign).
 ctype(solo).
 ctype(symbolic_control).
 ctype(symbolic_hexadecimal).
+ctype(lower(_)).
+ctype(upper(_)).
 ctype(upper).
 ctype(whitespace).
 
 
+%% get_single_char(-Char).
+%
+% Gets a single char from the current input stream.
 get_single_char(C) :-
     (  var(C) -> '$get_single_char'(C)
     ;  atom_length(C, 1) -> '$get_single_char'(C)
     ;  type_error(in_character, C, get_single_char/1)
     ).
 
-
+%% read_from_chars(+Chars, ?Term).
+%
+% Given a string made of chars which contains a representation of
+% a Prolog term, Term is the Prolog term represented. Example:
+%
+% ```
+% ?- read_from_chars("f(x,y).", X).
+%    X = f(x,y).
+% ```
 read_from_chars(Chars, Term) :-
     must_be(chars, Chars),
-    '$read_term_from_chars'(Chars, Term).
+    '$read_from_chars'(Chars, Term0),
+    Term = Term0.
 
+%% read_term_from_chars(+Chars, ?Term, +Options).
+%
+% Like `read_from_chars`, except the reader is configured according to
+% `Options` which are those of `read_term`.
+%
+% ```
+% ?- read_term_from_chars("f(X,y).", T, [variable_names(['X'=X])]).
+%    T = f(X,y).
+% ```
+read_term_from_chars(Chars, Term, Options) :-
+    must_be(chars, Chars),
+    builtins:parse_read_term_options(Options, [Singletons, VariableNames, Variables], read_term_from_chars/3),
+    '$read_term_from_chars'(Chars, Term0, Singletons, Variables, VariableNames),
+    Term = Term0.
 
+%% write_term_to_chars(+Term, +Options, -Chars).
+%
+% Given a Term which is a Prolog term and a set of options, Chars is
+% string representation of that term. Options available are:
+%
+%  * `ignore_ops(+Boolean)` if `true`, the generic term representation is used everywhere. In `false`
+%    (default), operators do not use that generic term representation.
+%  * `max_depth(+N)` if the term is nested deeper than N, print the reminder as ellipses.
+%    If N = 0 (default), there's no limit.
+%  * `numbervars(+Boolean)` if true, replaces `$VAR(N)` variables with letters, in order. Default is false.
+%  * `quoted(+Boolean)` if true, strings and atoms that need quotes to be valid Prolog syntax, are quoted. Default is false.
+%  * `variable_names(+List)` assign names to variables in term. List should be a list of terms of format `Name=Var`.
+%  * `double_quotes(+Boolean)` if true, strings are printed in double quotes rather than with list notation. Default is false.
 write_term_to_chars(_, Options, _) :-
     var(Options), instantiation_error(write_term_to_chars/3).
 write_term_to_chars(Term, Options, Chars) :-
     builtins:parse_write_options(Options,
-                                 [IgnoreOps, MaxDepth, NumberVars, Quoted, VNNames],
+                                 [DoubleQuotes, IgnoreOps, MaxDepth, NumberVars, Quoted, VNNames],
                                  write_term_to_chars/3),
     (  nonvar(Chars)  ->
        throw(error(uninstantiation_error(Chars), write_term_to_chars/3))
@@ -131,7 +250,7 @@ write_term_to_chars(Term, Options, Chars) :-
     ),
     term_variables(Term, Vars),
     extend_var_list(Vars, VNNames, NewVarNames, numbervars),
-    '$write_term_to_chars'(Chars, Term, IgnoreOps, NumberVars, Quoted, NewVarNames, MaxDepth).
+    '$write_term_to_chars'(Chars, Term, IgnoreOps, NumberVars, Quoted, NewVarNames, MaxDepth, DoubleQuotes).
 
 % Encodes Ch character to list of Bytes.
 char_utf8bytes(Ch, Bytes) :-
@@ -151,6 +270,17 @@ encode(Code, Prefix, Nb) -->
 % Maps characters and UTF-8 bytes.
 % If Cs is a variable, parses Bs as a list of UTF-8 bytes.
 % Otherwise, transform the list of characters Cs to UTF-8 bytes.
+
+%% chars_utf8bytes(?Chars, ?Bytes).
+%
+% Maps a string made of chars with a list of UTF-8 bytes. Some examples:
+%
+% ```
+% ?- chars_utf8bytes("Prolog", X).
+%   X = [80,114,111,108,111,103].
+% ?- chars_utf8bytes(X, [226, 136, 145]).
+%    X = "∑".
+% ```
 chars_utf8bytes(Cs, Bs) :-
   var(Cs), must_be(list, Bs) ->
     once(phrase(decode_utf8(Cs), Bs))
@@ -177,58 +307,66 @@ continuation(Code, Chars, Nb) --> [Byte],
 % each remaining continuation byte (if any) will raise 0xFFFD too
 continuation(_, ['\xFFFD\'|T], _) --> [_], decode_utf8(T).
 
-
-read_line_to_chars(Stream, Cs0, Cs) :-
+%% get_line_to_chars(+Stream, -Chars, +InitialChars).
+%
+% Reads chars from stream Stream until it finds a `\n` character.
+% InitialChars will be appended at the end of Chars
+get_line_to_chars(Stream, Cs0, Cs) :-
         '$get_n_chars'(Stream, 1, Char), % this also works for binary streams
         (   Char == [] -> Cs0 = Cs
         ;   Char = [C],
             Cs0 = [C|Rest],
             (   C == '\n' -> Rest = Cs
-            ;   read_line_to_chars(Stream, Rest, Cs)
+            ;   get_line_to_chars(Stream, Rest, Cs)
             )
         ).
 
-/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
-   Read N characters from Stream.
-
-   If N is a variable, read until EOF, unifying N with the number of
-   characters read.
-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
-
+%% get_n_chars(+Stream, ?N, -Chars).
+%
+% Read N chars from stream Stream. N can be an integer, in that case
+% only N chars are read, or a variable, unifying N with the number of chars
+% read until it found EOF.
 get_n_chars(Stream, N, Cs) :-
         can_be(integer, N),
         (   var(N) ->
-            read_to_eof(Stream, Cs),
+            get_to_eof(Stream, Cs),
             length(Cs, N)
         ;   N >= 0,
             '$get_n_chars'(Stream, N, Cs)
         ).
 
-read_to_eof(Stream, Cs) :-
-        '$get_n_chars'(Stream, 512, Cs0),
+get_n_chars_wrapper(Stream, N, Cs) :-
+        '$get_n_chars'(Stream, N, Cs).
+
+get_to_eof(Stream, Cs) :-
+        catch(get_n_chars_wrapper(Stream, 512, Cs0),
+              error(syntax_error(unexpected_end_of_file), _),
+              Cs0 = []),
         (   Cs0 == [] -> Cs = []
         ;   partial_string(Cs0, Cs, Rest),
-            read_to_eof(Stream, Rest)
+            get_to_eof(Stream, Rest)
         ).
 
-/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
-   Relation between a list of characters Cs and its Base64 encoding Bs,
-   also a list of characters.
-
-   At least one of the arguments must be instantiated.
-
-   Options are:
-
-      - padding(Boolean)
-        Whether to use padding: true (the default) or false.
-      - charset(C)
-        Either 'standard' (RFC 4648 §4, the default) or 'url' (RFC 4648 §5).
-
-   Example:
-
-      ?- chars_base64("hello", Bs, []).
-         Bs = "aGVsbG8=".
-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
+%% chars_base64(?Chars, ?Base64, +Options).
+%
+% Relation between a list of characters Cs and its Base64 encoding Bs,
+% also a list of characters.
+%
+% At least one of the arguments must be instantiated.
+%
+% Options are:
+%
+% - `padding(Boolean)`
+%   Whether to use padding: true (the default) or false.
+% - `charset(C)`
+%   Either 'standard' (RFC 4648 §4, the default) or 'url' (RFC 4648 §5).
+%
+% Example:
+%
+% ```
+% ?- chars_base64("hello", Bs, []).
+%    Bs = "aGVsbG8=".
+% ```
 
 chars_base64(Cs, Bs, Options) :-
         must_be(list, Options),

src/lib/clpb.pl

@@ -1,10 +1,29 @@
 /*  CLP(B): Constraint Logic Programming over Boolean Variables
 
-    Copyright (C): 2019 Markus Triska
-    All rights reserved.
-
+    Author:        Markus Triska
     E-mail:        triska@metalevel.at
-    WWW:           http://www.metalevel.at
+    WWW:           https://www.metalevel.at
+    Copyright (C): 2019-2025 Markus Triska
+
+    Permission is hereby granted, free of charge, to any person
+    obtaining a copy of this software and associated documentation
+    files (the "Software"), to deal in the Software without
+    restriction, including without limitation the rights to use, copy,
+    modify, merge, publish, distribute, sublicense, and/or sell copies
+    of the Software, and to permit persons to whom the Software is
+    furnished to do so, subject to the following conditions:
+
+    The above copyright notice and this permission notice shall be
+    included in all copies or substantial portions of the Software.
+
+    THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+    EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+    MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+    NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
+    HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+    WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+    OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+    DEALINGS IN THE SOFTWARE.
 
 */
 
@@ -17,8 +36,8 @@
 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
 
 :- module(clpb, [op(300, fy, ~),
-		         op(500, yfx, #),
-		         sat/1,
+                 op(500, yfx, #),
+                 sat/1,
                  taut/2,
                  labeling/1,
                  sat_count/2,
@@ -91,6 +110,46 @@ domain_error(Expectation, Term) :-
 type_error(Expectation, Term) :-
         type_error(Expectation, Term, unknown(Term)-1).
 
+/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
+Compatibility predicates.
+- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
+
+:- meta_predicate(include(1, ?, ?)).
+
+include(_, [], []).
+include(Goal, [L|Ls0], Ls) :-
+        (   call(Goal, L) ->
+            Ls = [L|Rest]
+        ;   Ls = Rest
+        ),
+        include(Goal, Ls0, Rest).
+
+:- meta_predicate(exclude(1, ?, ?)).
+
+exclude(_, [], []).
+exclude(Goal, [L|Ls0], Ls) :-
+        (   call(Goal, L) ->
+            Ls = Rest
+        ;   Ls = [L|Rest]
+        ),
+        exclude(Goal, Ls0, Rest).
+
+:- meta_predicate(partition(2,?,?,?,?)).
+
+partition(_, [], [], [], []).
+partition(Pred, [H|T], L, E, G) :-
+    call(Pred, H, Diff),
+    partition_(Diff, H, Pred, T, L, E, G).
+
+partition_(<, H, Pred, T, [H|Rest], E, G) :-
+    partition(Pred, T, Rest, E, G).
+partition_(=, H, Pred, T, L, [H|Rest], G) :-
+    partition(Pred, T, L, Rest, G).
+partition_(>, H, Pred, T, L, E, [H|Rest]) :-
+    partition(Pred, T, L, E, Rest).
+
+:- meta_predicate(partition(1,?,?,?)).
+
 partition(Pred, Ls0, As, Bs) :-
         include(Pred, Ls0, As),
         exclude(Pred, Ls0, Bs).
@@ -105,6 +164,262 @@ goal_expansion(del_attr(Var, Module), (var(Var) -> put_atts(Var, -Access);true))
         Access =.. [Module,_].
 
 
+/** Constraint Logic Programming over Boolean variables
+
+## Introduction
+
+This library provides CLP(B), Constraint Logic Programming over
+Boolean variables. It can be used to model and solve combinatorial
+problems such as verification, allocation and covering tasks.
+
+CLP(B) is an instance of the general CLP(_X_) scheme,
+extending logic programming with reasoning over specialised domains.
+
+The implementation is based on reduced and ordered Binary Decision
+Diagrams (BDDs).
+
+Benchmarks and usage examples of this library are available from:
+[*https://www.metalevel.at/clpb/*](https://www.metalevel.at/clpb/)
+
+## Boolean expressions
+
+A _Boolean expression_ is one of:
+
+| `0`                | false                                |
+| `1`                | true                                 |
+| _variable_         | unknown truth value                  |
+| _atom_             | universally quantified variable      |
+| `~` _Expr_           | logical NOT                          |
+| _Expr_ `+` _Expr_    | logical OR                           |
+| _Expr_ `*` _Expr_    | logical AND                          |
+| _Expr_ `#` _Expr_    | exclusive OR                         |
+| _Var_ `^` _Expr_     | existential quantification           |
+| _Expr_ `=:=` _Expr_  | equality                             |
+| _Expr_ `=\=` _Expr_  | disequality (same as #)              |
+| _Expr_ `=<` _Expr_   | less or equal (implication)          |
+| _Expr_ `>=` _Expr_   | greater or equal                     |
+| _Expr_ `<` _Expr_    | less than                            |
+| _Expr_ `>` _Expr_    | greater than                         |
+| `card(Is,Exprs)`     | cardinality constraint (_see below_) |
+| `+(Exprs)`         | n-fold disjunction (_see below_)     |
+| `*(Exprs)`         | n-fold conjunction (_see below_)     |
+
+where _Expr_ again denotes a Boolean expression.
+
+The Boolean expression `card(Is,Exprs)` is true iff the number of true
+expressions in the list `Exprs` is a member of the list `Is` of
+integers and integer ranges of the form `From-To`. For example, to
+state that precisely two of the three variables `X`, `Y` and `Z` are
+`true`, you can use `sat(card([2],[X,Y,Z]))`.
+
+`+(Exprs)` and `*(Exprs)` denote, respectively, the disjunction and
+conjunction of all elements in the list `Exprs` of Boolean
+expressions.
+
+Atoms denote parametric values that are universally quantified. All
+universal quantifiers appear implicitly in front of the entire
+expression. In residual goals, universally quantified variables always
+appear on the right-hand side of equations. Therefore, they can be
+used to express functional dependencies on input variables.
+
+## Interface predicates
+
+The most frequently used CLP(B) predicates are:
+
+    * `sat(+Expr)`
+      True iff the Boolean expression Expr is satisfiable.
+
+    * `taut(+Expr, -T)`
+      If Expr is a tautology with respect to the posted constraints, succeeds
+      with *T = 1*. If Expr cannot be satisfied, succeeds with *T = 0*.
+      Otherwise, it fails.
+
+    * `labeling(+Vs)`
+      Assigns truth values to the variables Vs such that all constraints
+      are satisfied.
+
+The unification of a CLP(B) variable _X_ with a term _T_ is equivalent
+to posting the constraint sat(X=:=T).
+
+## Examples
+
+Here is an example session with a few queries and their answers:
+
+```
+?- use_module(library(clpb)).
+   true.
+
+?- sat(X*Y).
+   X = 1, Y = 1.
+
+?- sat(X * ~X).
+   false.
+
+?- taut(X * ~X, T).
+   T = 0, clpb:sat(X=:=X).
+
+?- sat(X^Y^(X+Y)).
+   clpb:sat(X=:=X), clpb:sat(Y=:=Y).
+
+?- sat(X*Y + X*Z), labeling([X,Y,Z]).
+   X = 1, Y = 0, Z = 1
+;  X = 1, Y = 1, Z = 0
+;  X = 1, Y = 1, Z = 1.
+
+?- sat(X =< Y), sat(Y =< Z), taut(X =< Z, T).
+   T = 1, clpb:sat(X=:=X*Y), clpb:sat(Y=:=Y*Z).
+
+?- sat(1#X#a#b).
+   clpb:sat(X=:=a#b).
+```
+
+The pending residual goals constrain remaining variables to Boolean
+expressions and are declaratively equivalent to the original query.
+The last example illustrates that when applicable, remaining variables
+are expressed as functions of universally quantified variables.
+
+## Obtaining BDDs
+
+By default, CLP(B) residual goals appear in (approximately) algebraic
+normal form (ANF). This projection is often computationally expensive.
+We can assert `clpb:clpb_residuals(bdd)` to see the BDD representation
+of all constraints. This results in faster projection to residual
+goals, and is also useful for learning more about BDDs. For example:
+
+```
+?- asserta(clpb:clpb_residuals(bdd)).
+   true.
+
+?- sat(X#Y).
+node(3)- (v(X, 0)->node(2);node(1)),
+node(1)- (v(Y, 1)->true;false),
+node(2)- (v(Y, 1)->false;true).
+```
+
+Note that this representation cannot be pasted back on the toplevel,
+and its details are subject to change. Use copy_term/3 to obtain
+such answers as Prolog terms.
+
+The variable order of the BDD is determined by the order in which the
+variables first appear in constraints. To obtain different orders,
+we can for example use:
+
+```
+?- sat(+[1,Y,X]), sat(X#Y).
+node(3)- (v(Y, 0)->node(2);node(1)),
+node(1)- (v(X, 1)->true;false),
+node(2)- (v(X, 1)->false;true).
+```
+
+## Enabling monotonic CLP(B)
+
+In the default execution mode, CLP(B) constraints are _not_ monotonic.
+This means that _adding_ constraints can yield new solutions. For
+example:
+
+```
+?-          sat(X=:=1), X = 1+0.
+   false.
+
+?- X = 1+0, sat(X=:=1), X = 1+0.
+   X = 1+0.
+```
+
+This behaviour is highly problematic from a logical point of view, and
+it may render [*declarative
+debugging*](https://www.metalevel.at/prolog/debugging)
+techniques inapplicable.
+
+Assert `clpb:monotonic` to make CLP(B) *monotonic*. If this mode is
+enabled, then you must wrap CLP(B) variables with the functor
+`v/1`. For example:
+
+```
+?- asserta(clpb:monotonic).
+   true.
+
+?- sat(v(X)=:=1#1).
+   X = 0.
+```
+
+## Example: Pigeons
+
+In this example, we are attempting to place _I_ pigeons into _J_ holes
+in such a way that each hole contains at most one pigeon. One
+interesting property of this task is that it can be formulated using
+only _cardinality constraints_ (`card/2`). Another interesting aspect
+is that this task has no short resolution refutations in general.
+
+In the following, we use [*Prolog DCG
+notation*](https://www.metalevel.at/prolog/dcg) to describe a
+list `Cs` of CLP(B) constraints that must all be satisfied.
+
+```
+:- use_module(library(clpb)).
+:- use_module(library(clpz)).
+:- use_module(library(lists)).
+:- use_module(library(dcgs)).
+
+pigeon(I, J, Rows, Cs) :-
+        length(Rows, I), length(Row, J),
+        maplist(same_length(Row), Rows),
+        transpose(Rows, TRows),
+        phrase((all_cards(Rows,[1]),all_cards(TRows,[0,1])), Cs).
+
+all_cards([], _) --> [].
+all_cards([Ls|Lss], Cs) --> [card(Cs,Ls)], all_cards(Lss, Cs).
+```
+
+Example queries:
+
+```
+?- pigeon(9, 8, Rows, Cs), sat(*(Cs)).
+   false.
+
+?- pigeon(2, 3, Rows, Cs), sat(*(Cs)),
+   append(Rows, Vs), labeling(Vs),
+   maplist(portray_clause, Rows).
+[0,0,1].
+[0,1,0].
+etc.
+```
+
+## Example: Boolean circuit
+
+Consider a Boolean circuit that express the Boolean function =|XOR|=
+with 4 =|NAND|= gates. We can model such a circuit with CLP(B)
+constraints as follows:
+
+```
+:- use_module(library(clpb)).
+
+nand_gate(X, Y, Z) :- sat(Z =:= ~(X*Y)).
+
+xor(X, Y, Z) :-
+        nand_gate(X, Y, T1),
+        nand_gate(X, T1, T2),
+        nand_gate(Y, T1, T3),
+        nand_gate(T2, T3, Z).
+```
+
+Using universally quantified variables, we can show that the circuit
+does compute =|XOR|= as intended:
+
+```
+?- xor(x, y, Z).
+   clpb:sat(Z=:=x#y).
+```
+
+## Acknowledgments
+
+The interface predicates of this library follow the example of
+[*SICStus Prolog*](https://sicstus.sics.se).
+
+Use SICStus Prolog for higher performance in many cases.
+
+*/
+
+
 /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    Each CLP(B) variable belongs to exactly one BDD. Each CLP(B)
    variable gets an attribute (in module "clpb") of the form:
@@ -191,6 +506,10 @@ non_monotonic(X) :-
         ;   true
         ).
 
+:- meta_predicate(bdd_nodes(1, ?, ?)).
+:- meta_predicate(bdd_nodes_(1, ?, ?, ?)).
+:- meta_predicate(with_aux(1, ?)).
+
 /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    Rewriting to canonical expressions.
    Atoms are converted to variables with a special attribute.
@@ -831,8 +1150,9 @@ verify_attributes(Var, Other, Gs) :-
             (   integer(Other) ->
                 (   between(0, 1, Other) ->
                     root_get_formula_bdd(Root, Sat, BDD0),
+                    bdd_restriction(BDD0, I, Other, BDD),
                     root_put_formula_bdd(Root, Sat, BDD),
-                    Gs = [bdd_restriction(BDD0,I,Other,BDD),satisfiable_bdd(BDD)]
+                    Gs = [satisfiable_bdd(BDD)]
                 ;   no_truth_value(Other)
                 )
             ;   atom(Other) ->
@@ -932,7 +1252,7 @@ bdd_restriction_(Node, VI, Value, Res) -->
                   node_id(Node, ID) },
                 (   { I0 =:= VI } ->
                     (   { Value =:= 0 } -> { Res = Low }
-                    ;   { Value =:= 1 } -> { Res = High }
+                    ;   { Res = High }
                     )
                 ;   { I0 > VI } -> { Res = Node }
                 ;   state(G0), { get_assoc(ID, G0, Res) } -> []
@@ -1108,19 +1428,17 @@ indomain(1).
 %
 % Examples:
 %
-% ==
+% ```
 % ?- sat(A =< B), Vs = [A,B], sat_count(+[1|Vs], Count).
-% Vs = [A, B],
-% Count = 3,
-% sat(A=:=A*B).
+%    Vs = [A,B], Count = 3, clpb:sat(A=:=A*B).
 %
 % ?- length(Vs, 120),
 %    sat_count(+Vs, CountOr),
 %    sat_count(*(Vs), CountAnd).
-% Vs = [...],
-% CountOr = 1329227995784915872903807060280344575,
-% CountAnd = 1.
-% ==
+%    Vs = [...],
+%    CountOr = 1329227995784915872903807060280344575,
+%    CountAnd = 1.
+% ```
 
 
 
@@ -1233,11 +1551,15 @@ random_bindings(VNum, Node) -->
         { node_var_low_high(Node, Var, Low, High),
           bdd_count(Node, VNum, Total),
           bdd_count(Low, VNum, LCount) },
-        (   { maybe(LCount, Total) } ->
+        (   { weighted_maybe(LCount, Total) } ->
             [Var=0], random_bindings(VNum, Low)
         ;   [Var=1], random_bindings(VNum, High)
         ).
 
+weighted_maybe(K, N) :-
+        random_integer(0, N, X),
+        X < K.
+
 /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    Find solutions with maximum weight.
 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
@@ -1248,7 +1570,7 @@ random_bindings(VNum, Node) -->
 % linear objective function over Boolean variables Vs with integer
 % coefficients Weights. This predicate assigns 0 and 1 to the
 % variables in Vs such that all stated constraints are satisfied, and
-% Maximum is the maximum of sum(Weight_i*V_i) over all admissible
+% Maximum is the maximum of `sum(Weight_i*V_i)` over all admissible
 % assignments.  On backtracking, all admissible assignments that
 % attain the optimum are generated.
 %
@@ -1257,10 +1579,10 @@ random_bindings(VNum, Node) -->
 %
 % Example:
 %
-% ==
+% ```
 % ?- sat(A#B), weighted_maximum([1,2,1], [A,B,C], Maximum).
-% A = 0, B = 1, C = 1, Maximum = 3.
-% ==
+%    A = 0, B = 1, C = 1, Maximum = 3.
+% ```
 
 weighted_maximum(Ws, Vars, Max) :-
         must_be(list(integer), Ws),
@@ -1373,14 +1695,14 @@ skip_to_var_(Var, Weight, [Var0-Weight0|VWs0], VWs) -->
 
 attribute_goals(Var) -->
         { var_index_root(Var, _, Root) },
+        !,
         (   { root_get_formula_bdd(Root, Formula, BDD) } ->
             { del_bdd(Root) },
             (   { clpb_residuals(bdd) } ->
                 { bdd_nodes(BDD, Nodes),
                   phrase(nodes(Nodes), Ns) },
                 [clpb:'$clpb_bdd'(Ns)]
-            ;   { prepare_global_variables(BDD),
-                  phrase(sat_ands(Formula), Ands0),
+            ;   { phrase(sat_ands(Formula), Ands0),
                   ands_fusion(Ands0, Ands),
                   maplist(formula_anf, Ands, ANFs0),
                   sort(ANFs0, ANFs1),
@@ -1400,39 +1722,24 @@ attribute_goals(Var) -->
             booleans(RestVs)
         ;   boolean(Var)  % the variable may have occurred only in taut/2
         ).
+attribute_goals(Var) -->
+        { get_atts(Var, clpb_max(_)),
+          !,
+          put_atts(Var, -clpb_max(_)) }.
+attribute_goals(Var) -->
+        { get_atts(Var, clpb_bdd(BDD)),
+          ground(BDD),
+          put_atts(Var, -clpb_bdd(_)) }.
 
 del_clpb(Var) :-
         del_attr(Var, clpb),
-        del_attr(Var, clpb_hash).
-
-/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
-   To make residual projection work with recorded constraints, the
-   global counters must be adjusted so that new variables and nodes
-   also get new IDs. Also, clpb_next_id/2 is used to actually create
-   these counters, because creating them with b_setval/2 would make
-   them [] on backtracking, which is quite unfortunate in itself.
-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
+        del_attr(Var, clpb_hash),
+        del_attr(Var, clpb_atom).
 
 b_setval(K, T) :- bb_b_put(K, T).
 nb_setval(K, T) :- bb_put(K, T).
 b_getval(K, T) :- bb_get(K, T).
 
-prepare_global_variables(BDD) :-
-        clpb_next_id('$clpb_next_var', V0),
-        clpb_next_id('$clpb_next_node', N0),
-        bdd_nodes(BDD, Nodes),
-        foldl(max_variable_node, Nodes, V0-N0, MaxV0-MaxN0),
-        MaxV is MaxV0 + 1,
-        MaxN is MaxN0 + 1,
-        b_setval('$clpb_next_var', MaxV),
-        b_setval('$clpb_next_node', MaxN).
-
-max_variable_node(Node, V0-N0, V-N) :-
-        node_id(Node, N1),
-        node_varindex(Node, V1),
-        N is max(N0,N1),
-        V is max(V0,V1).
-
 /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    Fuse formulas that share the same variables into single conjunctions.
 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
@@ -1557,7 +1864,8 @@ booleans([B|Bs]) --> boolean(B), booleans(Bs).
 
 boolean(Var) -->
         { del_clpb(Var) },
-        (   { get_attr(Var, clpb_omit_boolean, true) } -> []
+        (   { get_attr(Var, clpb_omit_boolean, true) } ->
+            { put_atts(Var, -clpb_omit_boolean(_)) }
         ;   [clpb:sat(Var =:= Var)]
         ).
 
@@ -1659,49 +1967,3 @@ clpb_atom_var(Atom, Var) :-
             put_assoc(Atom, A0, Var, A),
             b_setval('$clpb_atoms', A)
         ).
-
-
-
-/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
-Compatibility predicates.
-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
-
-
-include(Goal, List, Is) :-
-        include_(List, Goal, Is).
-
-include_([], _, []).
-include_([X1|Xs1], P, Is) :-
-        (   call(P, X1)
-        ->  Is = [X1|Is1]
-        ;   Is = Is1
-        ),
-        include_(Xs1, P, Is1).
-
-
-exclude(Goal, List, Is) :-
-        exclude_(List, Goal, Is).
-
-exclude_([], _, []).
-exclude_([X1|Xs1], P, Is) :-
-        (   call(P, X1)
-        ->  Is = Is1
-        ;   Is = [X1|Is1]
-        ),
-        exclude_(Xs1, P, Is1).
-
-
-partition(Pred, List, Less, Equal, Greater) :-
-    partition_(List, Pred, Less, Equal, Greater).
-
-partition_([], _, [], [], []).
-partition_([H|T], Pred, L, E, G) :-
-    call(Pred, H, Diff),
-    partition_(Diff, H, Pred, T, L, E, G).
-
-partition_(<, H, Pred, T, [H|Rest], E, G) :-
-    partition_(T, Pred, Rest, E, G).
-partition_(=, H, Pred, T, L, [H|Rest], G) :-
-    partition_(T, Pred, L, Rest, G).
-partition_(>, H, Pred, T, L, E, [H|Rest]) :-
-    partition_(T, Pred, L, E, Rest).

src/lib/cont.pl

@@ -1,6 +1,6 @@
 :- module(cont, [reset/3, shift/1]).
 
-:- meta_predicate reset(0, ?, ?).
+:- meta_predicate(reset(0, ?, ?)).
 
 reset(Goal, Ball, Cont) :-
     call(Goal),

src/lib/csv.pl

@@ -1,54 +1,67 @@
-/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
-  Predicates for parsing CSV data
+/** Predicates for parsing CSV data
 
+## Read CSV files.
 
-  Read csv files
+Only two options with default values:
 
-  Only two options with default values :
-  - token_separator(',')
-  - with_header(true)
+- `token_separator(',')`
+- `with_header(true)`
 
-  Examples
+### Examples:
 
-  * parsing a csv string:
+Parsing a CSV string:
 
-  ?- use_module(library(csv)).
-  ?- use_module(library(dcgs)).
-  ?- phrase(parse_csv(Data), "col1,col2,col3,col4\none,2,,three").
-    Data = frame(["col1","col2","col3","col4"],[["one",2,[],"three"]]).
+```
+?- use_module(library(csv)).
+?- use_module(library(dcgs)).
+?- phrase(parse_csv(Data), "col1,col2,col3,col4\none,2,,three").
+   Data = frame(["col1","col2","col3","col4"],[["one",2,[],"three"]]).
+```
 
-  * with some options:
+With some options:
 
-  ?- phrase(parse_csv(Data, [with_header(false), token_separator(';')]), "one;2;;three").
-    Data = frame([],[["one",2,[],"three"]]).
+```
+?- phrase(parse_csv(Data, [with_header(false), token_separator(';')]), "one;2;;three").
+   Data = frame([],[["one",2,[],"three"]]).
+```
 
-  * parsing a csv file:
+Parsing a CSV file:
 
-  ?- use_module(library(csv)).
-  ?- use_module(library(pio)).
-  ?- phrase_from_file(parse_csv(frame(Header, Rows)), './test.csv').
+```
+?- use_module(library(csv)).
+?- use_module(library(pio)).
+?- phrase_from_file(parse_csv(frame(Header, Rows)), './test.csv').
+```
 
+## Write CSV files
 
-  Write csv files
+Four options with default values :
 
-  Four options with default values :
-  - line_separator('\n')
-  - token_separator(',')
-  - with_header(true)
-  - null_value(empty)
+- `line_separator('\n')`
+- `token_separator(',')`
+- `with_header(true)`
+- `null_value(empty)`
 
-  Examples
+### Examples
 
-  * writing a csv file:
+Writing a CSV file:
 
-  ?- use_module(library(csv)).
-  ?- write_csv('./test.csv', frame(["col1","col2","col3","col4"], [["one",2,[],"three"]])).
+```
+?- use_module(library(csv)).
+?- write_csv('./test.csv', frame(["col1","col2","col3","col4"], [["one",2,[],"three"]])).
+```
 
-  * with some options
+With some options
 
-  ?- use_module(library(csv)).
-  ?- write_csv('./test.csv', frame(["col1","col2","col3","col4"], [["one",2,[],"three"]]), [with_header(false), line_separator('\r\n'), token_separator(';'), null_value('\\N')]).
-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
+```
+?- use_module(library(csv)).
+?- write_csv('./test.csv', frame(
+      ["col1","col2","col3","col4"],
+      [["one",2,[],"three"]]
+    ),
+    [with_header(false), line_separator('\r\n'), token_separator(';'), null_value('\\N')]).
+```
+*/
 
 :- module(csv, [
   parse_csv//1,
@@ -208,7 +221,7 @@ row([X | Y], Opt) -->
   !,
   ( separator(Opt) ->
     row(Y, Opt)
-  ; end_token ->
+  ; end_token,
     { Y = [] }).
 
 

src/lib/dcgs.pl

@@ -1,10 +1,23 @@
+/** Support for Definite Clause Grammars.
+
+A Prolog definite clause grammar (DCG) describes a sequence. Operationally, DCGs
+can be used to parse, generate, complete and check sequences manifested as lists.
+
+Check [The Power of Prolog chapter on DCGs](https://www.metalevel.at/prolog/dcg)
+to learn more about them.
+*/
+
+
 :- module(dcgs,
           [op(1105, xfy, '|'),
            phrase/2,
            phrase/3,
+           phrase//2,
+           phrase//3,
            seq//1,
            seqq//1,
-           ... //0
+           ... //0,
+           (-->)/2
           ]).
 
 :- use_module(library(error)).
@@ -12,13 +25,56 @@
 :- use_module(library(lists), [append/3, member/2]).
 :- use_module(library(loader), [strip_module/3]).
 
-:- meta_predicate phrase(2, ?).
+:- meta_predicate(phrase(2, ?)).
 
-:- meta_predicate phrase(2, ?, ?).
+:- meta_predicate(phrase(2, ?, ?)).
+
+:- meta_predicate(phrase(3, ?, ?, ?)).
+
+:- meta_predicate(phrase(4, ?, ?, ?, ?)).
+
+:- meta_predicate(','(2, 2, ?, ?)).
+
+:- meta_predicate(;(2, 2, ?, ?)).
+
+%% phrase(+Body, ?Ls).
+%
+% True iff Body describes the list Ls. Body must be a DCG body.
+% It is equivalent to `phrase(Body, Ls, [])`.
+%
+% Examples:
+%
+% ```
+% as --> [].
+% as --> [a], as.
+%
+% ?- phrase(as, Ls).
+%    Ls = []
+% ;  Ls = "a"
+% ;  Ls = "aa"
+% ;  Ls = "aaa"
+% ;  ... .
+%
+% ?- phrase(as, "aaa").
+%    true.
+% ```
 
 phrase(GRBody, S0) :-
     phrase(GRBody, S0, []).
 
+%% phrase(+Body, ?Ls, ?Ls0).
+%
+% True iff Body describes part of the list Ls and the rest of Ls is Ls0.
+%
+% Example:
+%
+% ```
+% ?- phrase(seq(X), "aaa", Y).
+%    X = [], Y = "aaa"
+% ;  X = "a", Y = "aa"
+% ;  X = "aa", Y = "a"
+% ;  X = "aaa", Y = [].
+% ```
 phrase(GRBody, S0, S) :-
     strip_module(GRBody, M, GRBody1),
     (  var(GRBody) ->
@@ -30,12 +86,33 @@ phrase(GRBody, S0, S) :-
     ;  call(M:GRBody1, S0, S)
     ).
 
-
-module_call_qualified(M, Call, Call1) :-
-    (  nonvar(M) -> Call1 = M:Call
-    ;  Call = Call1
+phrase(GRBody, Arg, S0, S) :-
+    strip_module(GRBody, M, GRBody1),
+    (  var(GRBody) ->
+       instantiation_error(phrase/4)
+    ;  nonvar(GRBody1),
+       GRBody1 =.. GRBodys1,
+       append(GRBodys1, [Arg], GRBodys2),
+       GRBody2 =.. GRBodys2,
+       dcg_constr(GRBody2),
+       dcg_body(GRBody2, S0, S, GRBody3) ->
+       call(M:GRBody3)
+    ;  call(M:GRBody1, Arg, S0, S)
     ).
 
+phrase(GRBody, Arg1, Arg2, S0, S) :-
+    strip_module(GRBody, M, GRBody1),
+    (  var(GRBody) ->
+       instantiation_error(phrase/5)
+    ;  nonvar(GRBody1),
+       GRBody1 =.. GRBodys1,
+       append(GRBodys1, [Arg1,Arg2], GRBodys2),
+       GRBody2 =.. GRBodys2,
+       dcg_constr(GRBody2),
+       dcg_body(GRBody2, S0, S, GRBody3) ->
+       call(M:GRBody3)
+    ;  call(M:GRBody1, Arg1, Arg2, S0, S)
+    ).
 
 % The same version of the below two dcg_rule clauses, but with module scoping.
 dcg_rule(( M:NonTerminal, Terminals --> GRBody ), ( M:Head :- Body )) :-
@@ -63,7 +140,10 @@ dcg_rule(( NonTerminal --> GRBody ), ( Head :- Body )) :-
 dcg_non_terminal(NonTerminal, S0, S, Goal) :-
     NonTerminal =.. NonTerminalUniv,
     append(NonTerminalUniv, [S0, S], GoalUniv),
-    Goal =.. GoalUniv.
+    (  callable(NonTerminal) ->
+       Goal =.. GoalUniv
+    ;  Goal = NonTerminal % let call/N throw an error instead of throwing one here.
+    ).
 
 dcg_terminals(Terminals, S0, S, S0 = List) :-
     append(Terminals, S, List).
@@ -78,11 +158,12 @@ dcg_body(GRBody, S0, S, Body) :-
 dcg_body(NonTerminal, S0, S, Goal1) :-
     nonvar(NonTerminal),
     \+ dcg_constr(NonTerminal),
-    NonTerminal \= ( _ -> _ ),
-    NonTerminal \= ( \+ _ ),
     loader:strip_module(NonTerminal, M, NonTerminal0),
     dcg_non_terminal(NonTerminal0, S0, S, Goal0),
-    module_call_qualified(M, Goal0, Goal1).
+    (  functor(NonTerminal, (:), 2) ->
+       Goal1 = M:Goal0
+    ;  Goal1 = Goal0
+    ).
 
 % The following constructs in a grammar rule body
 % are defined in the corresponding subclauses.
@@ -94,9 +175,13 @@ dcg_constr(( _'|'_ )). % 7.14.6 - alternative
 dcg_constr({_}). % 7.14.7
 dcg_constr(call(_)). % 7.14.8
 dcg_constr(phrase(_)). % 7.14.9
+dcg_constr(phrase(_,_)). % extension of 7.14.9
+dcg_constr(phrase(_,_,_)). % extension of 7.14.9
 dcg_constr(!). % 7.14.10
-%% dcg_constr(\+ _). % 7.14.11 - not (existence implementation dep.)
-dcg_constr((_->_)). % 7.14.12 - if-then (existence implementation dep.)
+dcg_constr(\+ G_0) :- % 7.14.11 - not (existence implementation def.)
+    throw(error(representation_error(dcg_body), [culprit- (\+ G_0)])).
+dcg_constr((If->Then)) :- % 7.14.12 - if-then (existence implementation def.)
+    throw(error(representation_error(dcg_body), [culprit- (If->Then)])).
 
 % The principal functor of the first argument indicates
 % the construct to be expanded.
@@ -121,16 +206,24 @@ dcg_cbody(( GREither '|' GROr ), S0, S, ( Either ; Or )) :-
 dcg_cbody({Goal}, S0, S, ( Goal, S0 = S )).
 dcg_cbody(call(Cont), S0, S, call(Cont, S0, S)).
 dcg_cbody(phrase(Body), S0, S, phrase(Body, S0, S)).
+dcg_cbody(phrase(Body, Arg), S0, S, phrase(Body, Arg, S0, S)).
+dcg_cbody(phrase(Body, Arg1, Arg2), S0, S, phrase(Body, Arg1, Arg2, S0, S)).
 dcg_cbody(!, S0, S, ( !, S0 = S )).
-dcg_cbody(\+ GRBody, S0, S, ( \+ phrase(GRBody,S0,_), S0 = S )).
+% dcg_cbody(\+ GRBody, S0, S, ( \+ phrase(GRBody,S0,_), S0 = S )).
 dcg_cbody(( GRIf -> GRThen ), S0, S, ( If -> Then )) :-
     dcg_body(GRIf, S0, S1, If),
     dcg_body(GRThen, S1, S, Then).
 
+
+% When DCG expansion throws an exception – remove offending term and rethrow.
+user:term_expansion(throw_dcg_expansion_error(E), _) :-
+    throw(E).
 user:term_expansion(Term0, Term) :-
     nonvar(Term0),
-    dcg_rule(Term0, Term).
+    catch(dcg_rule(Term0, Term), E, Term = throw_dcg_expansion_error(E)).
 
+%% seq(Seq)//
+% 
 % Describes a sequence
 seq(Xs, Cs0,Cs) :-
    var(Xs),
@@ -141,10 +234,14 @@ seq(Xs, Cs0,Cs) :-
 seq([]) --> [].
 seq([E|Es]) --> [E], seq(Es).
 
+%% seqq(SeqOfSeqs)//
+%
 % Describes a sequence of sequences
 seqq([]) --> [].
 seqq([Es|Ess]) --> seq(Es), seqq(Ess).
 
+%% ...//
+%
 % Describes an arbitrary number of elements
 ...(Cs0,Cs) :-
    Cs0 == [],
@@ -152,8 +249,13 @@ seqq([Es|Ess]) --> seq(Es), seqq(Ess).
    Cs0 = Cs.
 ... --> [] | [_], ... .
 
+% defer instantiation errors until runtime. instantiations may be made
+% then.
+error_goal(error(instantiation_error, _Context), _).
 error_goal(error(E, must_be/2), error(E, must_be/2)).
 error_goal(error(E, (=..)/2), error(E, (=..)/2)).
+error_goal(error(representation_error(dcg_body), Context),
+           error(representation_error(dcg_body), Context)).
 error_goal(E, _) :- throw(E).
 
 user:goal_expansion(phrase(GRBody, S, S0), GRBody2) :-
@@ -163,6 +265,20 @@ user:goal_expansion(phrase(GRBody, S, S0), GRBody2) :-
           E,
           dcgs:error_goal(E, GRBody1)
          ),
-    module_call_qualified(M, GRBody1, GRBody2).
+    (  E = error(instantiation_error, _),
+       GRBody0 = [T|Ts] ->
+       GRBody2 = (error:must_be(list, [T|Ts]),
+                  lists:append([T|Ts], S0, S))
+    ;  GRBody = (_:_) ->
+       GRBody2 = M:GRBody1
+    ;  GRBody2 = GRBody1
+    ).
 
 user:goal_expansion(phrase(GRBody, S), phrase(GRBody, S, [])).
+
+
+% (-->)/2 behaves as if it didn't exist. We export (and define) it
+% only so that clauses for (-->)/2 cannot be asserted when
+% library(dcgs) is loaded.
+
+(_-->_) :- throw(error(existence_error(procedure,(-->)/2),(-->)/2)).

src/lib/debug.pl

@@ -1,4 +1,22 @@
-% Source: https://stackoverflow.com/a/30791637
+/** Declarative debugging.
+
+  This library provides three predicates with associated operators.
+  The operators can be placed in front of goals to debug Prolog
+  programs.
+
+  Of these predicates, the most frequently used is `(*)/1`, with
+  associated prefix operator `*` (star). Placing `*` in front of a
+  goal means to _generalize away_ the goal. `* Goal` acts as if `Goal`
+  did not appear at all in the source code. It is declaratively
+  equivalent to _commenting out_ the goal, and easier to write,
+  because `*` can also be placed in front of the last goal in a clause
+  without any additional changes.
+
+  Source: [https://stackoverflow.com/a/30791637](https://stackoverflow.com/a/30791637)
+
+*/
+
+
 
 :- module(debug, [
     op(900, fx, $),
@@ -11,16 +29,28 @@
 
 :- use_module(library(format), [portray_clause/1]).
 
-:- meta_predicate *(0).
-:- meta_predicate $(0).
-:- meta_predicate $-(0).
+:- meta_predicate(*(0)).
+:- meta_predicate($(0)).
+:- meta_predicate($-(0)).
+
+%% $-(Goal)
+%
+%  Portray exceptions thrown by Goal.
 
 $-(G_0) :-
    catch(G_0, Ex, ( portray_clause(exception:Ex:G_0), throw(Ex) ) ).
 
+%% $(Goal)
+%
+%  Provide a _trace_ for calls of Goal.
+
 $(G_0) :-
    portray_clause(call:G_0),
    $-G_0,
    portray_clause(exit:G_0).
 
+%% *(Goal)
+%
+%  Generalize away Goal.
+
 *(_).

src/lib/diag.pl

@@ -1,7 +1,160 @@
-:- module(diag, [wam_instructions/2]).
+:- module(diag, [wam_instructions/2, inlined_instructions/2]).
+
+/** Diagnostics library
+
+  The predicate `wam_instructions/2` _decompiles_ a predicate so that
+  we can inspect its Warren Abstract Machine (WAM) instructions.
+  In this way, we can verify and reason about compiled programs,
+  and detect opportunities for optimization.
+
+  For example, we have:
+
+```
+?- use_module(library(lists)).
+   true.
+?- use_module(library(diag)).
+   true.
+?- use_module(library(format)).
+   true.
+?- wam_instructions(append/3, Is),
+   maplist(portray_clause, Is).
+switch_on_term(1,external(1),external(2),external(6),fail).
+try_me_else(4).
+get_constant(level(shallow),[],x(1)).
+get_value(x(2),3).
+proceed.
+trust_me(0).
+get_list(level(shallow),x(1)).
+unify_variable(x(4)).
+unify_variable(x(1)).
+get_list(level(shallow),x(3)).
+unify_value(x(4)).
+unify_variable(x(3)).
+execute(append,3).
+   Is = [switch_on_term(1,external(1),external(2),external(6),fail)|...].
+```
+
+  `inlined_instructions/2` decompiles predicates at the code offset in
+  its first argument.
+
+  For example, given the program
+
+```
+?- [user].
+:- use_module(library(clpz)).
+
+all_eq(Vs, E) :- maplist(#=(E), Vs).
+
+```
+
+  we inspect the code of `all_eqs/2` using `wam_instructions/2`,
+  revealing:
+
+```
+?- wam_instructions(all_eq/2, Is),
+   maplist(portray_clause, Is).
+put_structure('$aux',2,x(3)).
+set_local_value(x(2)).
+set_void(1).
+set_constant('$index_ptr'(115334)).
+get_variable(x(4),1).
+put_structure(:,2,x(1)).
+set_constant(user).
+set_local_value(x(3)).
+get_variable(x(5),2).
+put_value(x(4),2).
+execute(maplist,2).
+   Is = [put_structure('$aux',2,x(3)),set_local_value(x(2)),set_void(1),set_constant('$index_ptr'(115334)),get_variable(x(4),1),put_structure(:,2,x(1)),set_constant(user),set_local_value(x(3)),get_variable(x(5),2),put_value(x(4),2),execute(maplist,2)].
+```
+
+  The `'$index_ptr(115334)` functor gives a code offset to an inlined
+  predicate compiled for the use of maplist/2. `inlined_instructions/2`
+  can be used to decompile its source code:
+
+```
+?- inlined_instructions(115334, Is),
+   maplist(portray_clause, Is).
+allocate(1).
+get_level(y(1)).
+get_variable(x(5),2).
+put_value(x(3),2).
+get_variable(x(6),3).
+put_value(x(5),3).
+put_unsafe_value(1,4).
+deallocate.
+jmp_by_execute(1).
+try_me_else(8).
+call(integer,1).
+neck_cut.
+get_variable(x(5),1).
+put_value(x(2),1).
+get_variable(x(6),2).
+put_value(x(5),2).
+jmp_by_execute(7).
+try_me_else(12).
+allocate(3).
+get_level(y(1)).
+get_variable(y(3),1).
+get_variable(y(2),2).
+call_default(true,0).
+call(var,1).
+cut(y(1)).
+put_unsafe_value(3,1).
+put_unsafe_value(2,2).
+deallocate.
+execute_default(is,2).
+default_retry_me_else(4).
+call(integer,1).
+neck_cut.
+execute(=:=,2).
+default_trust_me(0).
+allocate(2).
+get_variable(y(1),1).
+get_variable(y(2),3).
+put_value(y(2),1).
+call_default(is,2).
+put_unsafe_value(2,1).
+put_unsafe_value(1,2).
+deallocate.
+execute_default(clpz_equal,2).
+default_retry_me_else(4).
+call(integer,1).
+neck_cut.
+jmp_by_execute(29).
+try_me_else(12).
+allocate(3).
+get_level(y(1)).
+get_variable(y(3),1).
+get_variable(y(2),2).
+call_default(true,0).
+call(var,1).
+cut(y(1)).
+put_unsafe_value(3,1).
+put_unsafe_value(2,2).
+deallocate.
+execute_default(is,2).
+default_trust_me(0).
+allocate(2).
+get_variable(y(2),1).
+get_variable(y(1),3).
+put_value(y(1),1).
+call_default(is,2).
+put_unsafe_value(2,1).
+put_unsafe_value(1,2).
+deallocate.
+execute_default(clpz_equal,2).
+default_trust_me(0).
+execute_default(clpz_equal,2).
+   Is = [allocate(1),get_level(y(1)),get_variable(x(5),2),put_value(x(3),2),get_variable(x(6),3),put_value(x(5),3),put_unsafe_value(1,4),deallocate,jmp_by_execute(1),try_me_else(8),call(integer,1),neck_cut,get_variable(x(5),1),put_value(x(2),1),get_variable(x(6),2),put_value(x(5),2),jmp_by_execute(7),try_me_else(12),allocate(3),get_level(...),...].
+```
+*/
+
 
 :- use_module(library(error)).
 
+%% wam_instructions(+PI, -Instrs)
+%
+%  _Instrs_ are the WAM instructions corresponding to predicate indicator _PI_.
 
 wam_instructions(Clause, Listing) :-
     (  nonvar(Clause) ->
@@ -13,6 +166,16 @@ wam_instructions(Clause, Listing) :-
     ;  throw(error(instantiation_error, wam_instructions/2))
     ).
 
+%% inlined_instructions(+IndexPtr, -Instrs)
+%
+%  _Instrs_ are the WAM instructions corresponding to code offset _IndexPtr_.
+
+inlined_instructions(IndexPtr, Listing) :-
+    must_be(integer, IndexPtr),
+    (  IndexPtr >= 0 ->
+       '$inlined_instructions'(IndexPtr, Listing)
+    ;  throw(error(domain_error(not_less_than_zero, IndexPtr), inlined_instructions/2))
+    ).
 
 fetch_instructions(Module, Name, Arity, Listing) :-
     must_be(atom, Module),

src/lib/dif.pl

@@ -1,8 +1,13 @@
+/**
+Provides predicate `dif/2`. `dif/2` is a constraint that is true only if both of its
+arguments are different terms.
+*/
+
 :- module(dif, [dif/2]).
 
 :- use_module(library(atts)).
 :- use_module(library(dcgs)).
-:- use_module(library(lists), [append/3]).
+:- use_module(library(lists), [append/3, maplist/3]).
 
 :- attribute dif/1.
 
@@ -18,6 +23,34 @@ dif_set_variables([Var|Vars], X, Y) :-
     put_dif_att(Var, X, Y),
     dif_set_variables(Vars, X, Y).
 
+remove_goal([], _, []).
+remove_goal([G0|G0s], Goal0, Goals) :-
+    (   G0 == Goal0 ->
+        remove_goal(G0s, Goal0, Goals)
+    ;   Goals = [G0|Goals1],
+        remove_goal(G0s, Goal0, Goals1)
+    ).
+
+vars_remove_goal([], _).
+vars_remove_goal([Var|Vars], Goal0) :-
+    (  get_atts(Var, +dif(Goals0)) ->
+       remove_goal(Goals0, Goal0, Goals),
+       (   Goals = [] ->
+           put_atts(Var, -dif(_))
+       ;   put_atts(Var, +dif(Goals))
+       )
+    ;  true
+    ),
+    vars_remove_goal(Vars, Goal0).
+
+reinforce_goal(Goal0, Goal) :-
+    Goal = (
+        term_variables(Goal0, Vars),
+        dif:vars_remove_goal(Vars, Goal0),
+        Goal0 = (L \== R),
+        dif:dif(L, R)
+    ).
+
 append_goals([], _).
 append_goals([Var|Vars], Goals) :-
     (   get_atts(Var, +dif(VarGoals)) ->
@@ -29,31 +62,46 @@ append_goals([Var|Vars], Goals) :-
     append_goals(Vars, Goals).
 
 verify_attributes(Var, Value, Goals) :-
-    (   get_atts(Var, +dif(Goals)) ->
+    (   get_atts(Var, +dif(Goals0)) ->
 	    term_variables(Value, ValueVars),
-	    append_goals(ValueVars, Goals)
+	    append_goals(ValueVars, Goals0),
+        maplist(reinforce_goal, Goals0, Goals)
     ;   Goals = []
     ).
 
-% Probably the world's worst dif/2 implementation. I'm open to
-% suggestions for improvement.
-
+%% dif(?X, ?Y).
+%
+% True iff X and Y are different terms. Unlike `\=/2`, `dif/2` is more declarative because if X and Y can
+% unify but they're not yet equal, the decision is delayed, and prevents X and Y to become equal later.
+% Examples:
+%
+% ```
+% ?- dif(a, a).
+%    false.
+% ?- dif(a, b).
+%    true.
+% ?- dif(X, b).
+%    dif:dif(X,b).
+% ?- dif(X, b), X = b.
+%    false.
+% ```
 dif(X, Y) :-
     X \== Y,
     (   X \= Y -> true
-    ;   (   term_variables(X, XVars),
-            term_variables(Y, YVars),
-            dif_set_variables(XVars, X, Y),
-            dif_set_variables(YVars, X, Y)
-        )
+    ;   term_variables(dif(X,Y), Vars),
+        dif_set_variables(Vars, X, Y)
     ).
 
-gather_dif_goals([]) --> [].
-gather_dif_goals([(X \== Y) | Goals]) -->
-    [dif:dif(X, Y)],
-    gather_dif_goals(Goals).
+gather_dif_goals(_, []) --> [].
+gather_dif_goals(V, [(X \== Y) | Goals]) -->
+    (  { term_variables(X-Y, [V0 | _]),
+         V == V0 } ->
+       [dif:dif(X, Y)]
+    ;  []
+    ),
+    gather_dif_goals(V, Goals).
 
 attribute_goals(X) -->
     { get_atts(X, +dif(Goals)) },
-    gather_dif_goals(Goals),
+    gather_dif_goals(X, Goals),
     { put_atts(X, -dif(_)) }.

src/lib/error.pl

@@ -1,44 +1,49 @@
 /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
-   Written 2018-2022 by Markus Triska (triska@metalevel.at)
+   Written 2018-2025 by Markus Triska (triska@metalevel.at)
    I place this code in the public domain. Use it in any way you want.
 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
 
 :- module(error, [must_be/2,
                   can_be/2,
+                  instantiation_error/0,
+                  domain_error/2,
+                  type_error/2,
+                  representation_error/1,
+                  resource_error/1,
                   instantiation_error/1,
                   domain_error/3,
-                  type_error/3
+                  type_error/3,
+                  call_with_error_context/2
                   ]).
 
 
-:- meta_predicate check_(1, ?, ?).
+:- meta_predicate(check_(1, ?, ?)).
 
 
-/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
-   must_be(Type, Term)
-
-   This predicate is intended for type-checks of built-in predicates.
-
-   It asserts that Term is:
-
-       1) instantiated *and*
-       2) instantiated to an instance of the given Type.
-
-   It corresponds to usage mode +Term.
-
-   Currently, the following types are supported:
-
-       - atom
-       - boolean
-       - character
-       - chars
-       - in_character
-       - integer
-       - list
-       - octet_character
-       - octet_chars
-       - term
-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
+%% must_be(Type, Term)
+%
+% This predicate is intended for type-checks of built-in predicates.
+%
+% It asserts that Term is:
+%
+%     1) instantiated *and*
+%     2) instantiated to an instance of the given Type.
+%
+% It corresponds to usage mode +Term.
+%
+% Currently, the following types are supported:
+%
+%     - atom
+%     - boolean
+%     - character
+%     - chars
+%     - in_character
+%     - integer
+%     - list
+%     - octet_character
+%     - octet_chars
+%     - pair
+%     - term
 
 must_be(Type, Term) :-
         must_be_(type, Type),
@@ -84,12 +89,13 @@ must_be_(octet_chars, Cs) :-
 must_be_(list, Term)    :- check_(error:ilist, list, Term).
 must_be_(type, Term)    :- check_(error:type, type, Term).
 must_be_(boolean, Term) :- check_(error:boolean, boolean, Term).
+must_be_(pair, Term)    :- check_(error:pair, pair, Term).
 must_be_(term, Term)    :-
-        (   \+ ground(Term) ->
-            instantiation_error(must_be/2)
-        ;   \+ acyclic_term(Term) ->
-            type_error(term, Term, must_be/2)
-        ;   true
+        (   acyclic_term(Term) ->
+            (   ground(Term) ->  true
+            ;   instantiation_error(must_be/2)
+            )
+        ;   type_error(term, Term, must_be/2)
         ).
 
 % We cannot use maplist(must_be(character), Cs), because library(lists)
@@ -106,6 +112,9 @@ check_(Pred, Type, Term) :-
         ;   type_error(Type, Term, must_be/2)
         ).
 
+
+pair(_-_).
+
 boolean(B) :- ( B == true ; B == false ).
 
 character(C) :-
@@ -141,20 +150,18 @@ type(var).
 type(boolean).
 type(term).
 type(not_less_than_zero).
+type(pair).
 
-/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
-   can_be(Type, Term)
-
-   This predicate is intended for type-checks of built-in predicates.
-
-   It asserts that there is a substitution which, if applied to Term,
-   makes it an instance of Type.
-
-   It corresponds to usage mode ?Term.
-
-   It supports the same types as must_be/2.
-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
-
+%% can_be(Type, Term)
+%
+% This predicate is intended for type-checks of built-in predicates.
+%
+% It asserts that there is a substitution which, if applied to Term,
+% makes it an instance of Type.
+%
+% It corresponds to usage mode ?Term.
+%
+% It supports the same types as must_be/2.
 
 can_be(Type, Term) :-
         must_be(type, Type),
@@ -192,6 +199,7 @@ can_(octet_chars, Cs) :-
         ).
 can_(list, Term)    :- list_or_partial_list(Term).
 can_(boolean, Term) :- boolean(Term).
+can_(pair, Term)    :- pair(Term).
 can_(term, Term)    :-
         (   acyclic_term(Term) ->
             true
@@ -212,6 +220,30 @@ list_or_partial_list(Ls) :-
 
 /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    Shorthands for throwing ISO errors.
+
+   The variants without context promote the use of
+   call_with_error_context/2.
+- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
+
+instantiation_error :-
+    throw(error(instantiation_error, [])).
+
+domain_error(Type, Term) :-
+    throw(error(domain_error(Type, Term), [])).
+
+type_error(Type, Term) :-
+    throw(error(type_error(Type, Term), [])).
+
+representation_error(Flag) :-
+    throw(error(representation_error(Flag), [])).
+
+resource_error(Resource) :-
+    throw(error(resource_error(Resource), [])).
+
+/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
+   The variants *with* context would not have been needed if
+   call_with_error_context/2 had been found earlier. In the future,
+   we may be able to remove them.
 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
 
 instantiation_error(Context) :-
@@ -222,3 +254,20 @@ domain_error(Type, Term, Context) :-
 
 type_error(Type, Term, Context) :-
     throw(error(type_error(Type, Term), Context)).
+
+/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
+   call_with_error_context/2
+
+   See https://github.com/mthom/scryer-prolog/discussions/2839 .
+- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
+
+%% call_with_error_context(+Goal, +Pair)
+%
+% Call _Goal_ with error context _Pair_.
+%
+% Examples of error contexts: `predicate-PI`, `file-Filename` etc.
+
+:- meta_predicate(call_with_error_context(0,+)).
+call_with_error_context(G_0, Pair) :-
+   must_be(pair, Pair),
+   catch(G_0, error(E,Pairs), throw(error(E,[Pair|Pairs]))).

src/lib/ffi.pl

@@ -0,0 +1,218 @@
+:- module(ffi, [use_foreign_module/2, foreign_struct/2, with_locals/2, allocate/4, deallocate/3, read_ptr/3, array_type/3]).
+
+/** Foreign Function Interface
+
+This module contains predicates used to call native code (exposed by the C ABI).
+It uses [libffi](https://sourceware.org/libffi/) under the hood. The bridge is very simple
+and is very unsafe and should be used with care. FFI isn't the only way to communicate with
+the outside world in Prolog: sockets, pipes and HTTP may be good enough for your use case.
+
+The main predicate is `use_foreign_module/2`. It takes a library name (which depending on the
+operating system could be a `.so`, `.dylib` or `.dll` file). and a list of functions. Each
+function is defined by its name, a list of the type of the arguments, and the return argument.
+
+Types available are: `sint8`/`i8`, `uint8`/`u8`, `sint16`/`i16`, `uint16`/`u16`, `sint32`/`i32`, `uint32`/`u32`, `sint64`/`i64`,
+`uint64`/`u64`, `f32`, `f64`, `cstr`, `void`, `bool`, `ptr` and custom structs, which can be defined
+with `foreign_struct/2`.
+
+After that, each function on the lists maps to a predicate created in the ffi module which
+are used to call the native code.
+The predicate takes the functor name after the function name. Then, the arguments are the input
+arguments followed by a return argument. However, functions with return type `void` or `bool`
+don't have that return argument. Predicates with `void` always succeed and `bool` predicates depend
+on the return value on the native side.
+
+```
+ffi:FUNCTION_NAME(+InputArg1, ..., +InputArgN, -ReturnArg). % for all return types except void and bool
+ffi:FUNCTION_NAME(+InputArg1, ..., +InputArgN). % for void and bool
+```
+
+## Notes regarding cstr
+
+- When using `cstr` as an argument type the string will be deallocated once the function returns.
+- When using `cstr` as a return type the string will be copied and won't be deallocated.
+
+
+## Example
+
+For example, let's see how to define a function from the [raylib](https://www.raylib.com/) library.
+
+```
+?- use_foreign_module("./libraylib.so", ['InitWindow'([sint32, sint32, cstr], void)]).
+```
+
+This creates a `'InitWindow'` predicate under the ffi module. Now, we can call it:
+
+```
+?- ffi:'InitWindow'(800, 600, "Scryer Prolog + Raylib").
+```
+
+And a new window should pop up!
+*/
+
+:- use_module(library(lists)).
+:- use_module(library(error)).
+:- use_module(library(format)).
+:- use_module(library(dcgs)).
+:- use_module(library(iso_ext)).
+
+%% foreign_struct(+Name, +Elements).
+%
+% Defines a new struct type with name Name, composed of the elements Elements, which is a list
+% of other types.
+%
+% The name of the types doesn't matter, but the order of Elements must match the ones in the
+% native code.
+%
+% Example:
+%
+% ```
+% ?- foreign_struct(color, [uint8, uint8, uint8, uint8]).
+% ```
+foreign_struct(Name, Elements) :-
+    '$define_foreign_struct'(Name, Elements).
+
+
+%% use_foreign_module(+LibName, +Predicates)
+%
+% - LibName the path to the shared library to load/bind
+% - Predicates list of function definitions
+%
+%   Each function definition is a functor of arity 2.
+%   The functor name is the name of the function to bind,
+%   the first argument is the list of arguments of the function,
+%   the second argument is the return type of the function.
+%
+%   This will define a predicate in the ffi module with the defined name,
+%   for void and bool return type functions the arity will match the length of the arguments list,
+%   for other return types there will be an additional out parameter.
+%
+use_foreign_module(LibName, Predicates) :-
+    '$load_foreign_lib'(LibName, Predicates),
+    maplist(assert_predicate, Predicates).
+
+assert_predicate(PredicateDefinition) :-
+    PredicateDefinition =.. [Name, Inputs, void],
+    length(Inputs, NumInputs),
+    functor(Head, Name, NumInputs),
+    term_variables(Head, TermList),
+    Body = (
+	'$foreign_call'(Name, TermList, _),!
+    ),
+    Predicate = (Head:-Body),
+    assertz(ffi:Predicate).
+
+assert_predicate(PredicateDefinition) :-
+    PredicateDefinition =.. [Name, Inputs, bool],
+    length(Inputs, NumInputs),
+    functor(Head, Name, NumInputs),
+    term_variables(Head, TermList),
+    Body = (
+	'$foreign_call'(Name, TermList, 1),!
+    ),
+    Predicate = (Head:-Body),
+    assertz(ffi:Predicate).
+
+assert_predicate(PredicateDefinition) :-
+    PredicateDefinition =.. [Name, Inputs, Return],
+    \+ member(Return, [void, bool]),
+    length(Inputs, NumInputs),
+    NumArgs is NumInputs + 1,
+    functor(Head, Name, NumArgs),
+    term_variables(Head, TermList),
+    Body = (
+	lists:append(TermListInputs, [TermListReturn], TermList),
+	'$foreign_call'(Name, TermListInputs, TermListReturn),!
+    ),
+    Predicate = (Head:-Body),
+    assertz(ffi:Predicate).
+
+%% allocate(+Allocator, +Type, +Args, -Ptr)
+%
+% Using the Allocator allocate Type initialized with Args and 
+% unify Ptr with a pointer to that allocation.
+%
+allocate(Allocator, Type, Args, Ptr) :-
+    must_be(var, Ptr),
+    must_be(atom, Type),
+    must_be(atom, Allocator),
+    '$ffi_allocate'(Allocator, Type, Args, Ptr).
+
+
+%% read_ptr(+Type, +Ptr, -Value)
+%
+% Read a value of Type from the pointer Ptr and unify the read value with Value
+%
+% For type cstr take read a nul-terminated utf-8 string starting at Ptr.
+%
+read_ptr(Type, Ptr, Value) :-
+    must_be(atom, Type),
+    must_be(integer, Ptr),
+    '$ffi_read_ptr'(Type, Ptr, Value).
+
+%% deallocate(+Allocator, +Type, +Ptr)
+%
+% Deallocate the allocation at Ptr of Type allocated with Allocator
+%
+deallocate(Allocator, Type, Ptr) :-
+    must_be(atom, Allocator),
+    must_be(integer, Ptr),
+    '$ffi_deallocate'(Allocator, Type, Ptr).
+
+:- dynamic(is_array_type_defined/1).
+
+%% array_type(+ElemType, +Len, -ArrayType)
+%
+% unify the ffi type for an array of lenth Len with element type ElemType with ArrayType
+%
+array_type(ElemType, Len, ArrayType) :-
+    (Len =< 0 -> domain_error(greater_than_zero, Len, array_type/3); true),
+    phrase(format_("$[~a;~d]", [ElemType, Len]), ArrayTypeName),
+    atom_chars(ArrayType, ArrayTypeName),
+    (is_array_type_defined(ArrayType) -> true
+    ;   length(Fields, Len),
+        maplist(=(ElemType), Fields),
+        foreign_struct(ArrayType, Fields),
+        assertz(is_array_type_defined(ArrayType))
+    ).
+
+:- meta_predicate(with_locals(?, 0)).
+
+%% with_locals(+Locals, :Goal)
+%
+%  Allocate the Locals, evaluate the Goal and deallocate the Locals.
+%  The Locals will also be cleandup when Goal fails or throws an error.
+%
+%  Locals is a list of local variable definitions let(-Ptr, +Type, +Args).
+%  Ptr will be unified with the pointer to the local of Type initialized with Args.
+%
+with_locals(Locals, Goal) :-
+    verify_locals(Locals),
+    setup_call_cleanup(
+        allocate_locals(Locals),
+        Goal,
+        deallocate_locals(Locals)
+    ).
+
+verify_locals(Locals) :-
+    must_be(list, Locals),
+    ( maplist(verify_local, Locals) -> true
+    ; domain_error(locals_decl_list, Locals, [verify_locals/1])
+    ).
+
+verify_local(let(Var, Type, Init)) :-
+    must_be(var, Var),
+    must_be(atom, Type),
+    ground(Init).
+
+allocate_locals([]).
+allocate_locals([let(Var, Type, Init) | Ls]) :-
+    allocate(rust, Type, Init , Var),
+    (catch(allocate_locals(Ls), E, (deallocate_locals([let(Var, Type, Init)]), throw(E))) -> true
+    ; deallocate_locals([let(Var, Type, Init)]), false
+    ).
+
+deallocate_locals([]).
+deallocate_locals([let(Var, Type, _) | Ls]) :-
+    deallocate(rust, Type, Var),
+    deallocate_locals(Ls).

src/lib/files.pl

@@ -1,5 +1,24 @@
+/** Predicates for reasoning about files and directories.
+
+In this library, directories and files are represented as
+_lists of characters_. This is an ideal representation:
+
+* Lists of characters can be conveniently reasoned about with DCGs
+   and built-in Prolog predicates from `library(lists)`. This alone
+   is already a very compelling argument to use them.
+* Other Scryer libraries such as `library(http/http_open)` also already
+   use lists of characters to represent paths.
+* File names are mostly ephemeral, so it is good for efficiency
+   that they can quickly allocated transiently on the heap, leaving the
+   atom table mostly unaffected. Indexing is almost never needed
+   for file names. If needed, it should be added to the engine.
+* The previous point is also good for security, since the system
+   leaves little trace of which files were even accessed.
+* Scryer Prolog represents lists of characters extremely compactly.
+*/
+
 /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
-   Written 2020, 2022 by Markus Triska (triska@metalevel.at)
+   Written 2020-2025 by Markus Triska (triska@metalevel.at)
    Part of Scryer Prolog.
 
    Predicates for reasoning about files and directories.
@@ -51,8 +70,9 @@
                   file_exists/1,
                   directory_exists/1,
                   delete_file/1,
-				  rename_file/2,
-				  delete_directory/1,
+		          rename_file/2,
+		          file_copy/2,
+		          delete_directory/1,
                   make_directory/1,
                   make_directory_path/1,
                   working_directory/2,
@@ -67,41 +87,85 @@
 :- use_module(library(charsio)).
 :- use_module(library(dcgs)).
 
+%% directory_files(+Directory, -Files).
+%
+% True if `Files` are the files *and* directories available at a specific
+% `Directory` in the current system.
+
 directory_files(Directory, Files) :-
         must_be(chars, Directory),
         can_be(list, Files),
         '$directory_files'(Directory, Files).
 
+%% file_size(+File, ?Size).
+%
+% True iff `Size` is the size (in bytes) of `File`. The file must exist.
+
 file_size(File, Size) :-
         file_must_exist(File, file_size/2),
         can_be(integer, Size),
         '$file_size'(File, Size).
 
+%% file_exists(+File).
+%
+% True iff `File` is a file that exists in the current system.
+
 file_exists(File) :-
         must_be(chars, File),
         '$file_exists'(File).
 
+%% directory_exists(+Directory).
+%
+% True iff `Directory` is a directory that exists in the current system.
+
 directory_exists(Directory) :-
         must_be(chars, Directory),
         '$directory_exists'(Directory).
 
+%% make_directory(+Directory).
+%
+% Creates a new directory named `Directory`.
+% If you want to create a nested directory, use `make_directory_path/1`.
+
 make_directory(Directory) :-
         must_be(chars, Directory),
         '$make_directory'(Directory).
 
+%% make_directory_path(+Directory).
+%
+% Similar to `make_directory/1` but recursively creates directories if they're missing.
+% Equivalent to mkdir -p in Unix.
 make_directory_path(Directory) :-
         must_be(chars, Directory),
         '$make_directory_path'(Directory).
 
+%% delete_file(+File).
+%
+% Succeeds if deletes File from the current system.
 delete_file(File) :-
         file_must_exist(File, delete_file/1),
         '$delete_file'(File).
 
+%% rename_file(+File, +Renamed).
+%
+% Succeeds if File is renamed to Renamed
 rename_file(File, Renamed) :-
         file_must_exist(File, rename_file/2),
         must_be(chars, Renamed),
         '$rename_file'(File, Renamed).
 
+%% file_copy(+File, +Copied).
+%
+% Succeeds if File is copied to Copied
+file_copy(File, Copied) :-
+	file_must_exist(File, file_copy/2),
+	must_be(chars, Copied),
+	'$file_copy'(File, Copied).
+
+%% delete_directory(+Directory).
+%
+% Succeeds if Directory is deleted from the current system.
+% Directory must be empty.
 delete_directory(Directory) :-
         directory_must_exist(Directory, delete_directory/1),
         must_be(chars, Directory),
@@ -117,31 +181,31 @@ directory_must_exist(Directory, Context) :-
         ;   throw(error(existence_error(directory, Directory), Context))
         ).
 
-/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
-   Dir0 is the current working directory, and the working directory
-   is changed to Dir.
-
-   Use working_directory(Ds, Ds) to determine the current working directory,
-   and leave it as is.
-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
+%% working_directory(Dir0, Dir).
+%
+% Dir0 is the current working directory, and the working directory
+% is changed to Dir.
+% 
+% Use `working_directory/2` to determine the current working directory,
+% and leave it as is.
 
 working_directory(Dir0, Dir) :-
         can_be(list, Dir0),
         can_be(list, Dir),
         '$working_directory'(Dir0, Dir).
 
-/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
-   True iff Cs is the canonical, absolute path of Ps.
-
-   All intermediate components are normalized, and all symbolic links
-   are resolved.
-
-   The predicate fails in the following situations, though not
-   necessarily *only* in these cases:
-
-      1. Ps is a path that does not exist.
-      2. A non-final component in Ps is not a directory.
-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
+%% path_canonical(Ps, Cs).
+%
+% True iff Cs is the canonical, absolute path of Ps.
+%
+% All intermediate components are normalized, and all symbolic links
+% are resolved.
+%
+% The predicate fails in the following situations, though not
+% necessarily *only* in these cases:
+%
+% 1. Ps is a path that does not exist.
+% 2. A non-final component in Ps is not a directory.
 
 path_canonical(Ps, Cs) :-
         must_be(chars, Ps),
@@ -155,12 +219,28 @@ path_canonical(Ps, Cs) :-
    For two time stamps A and B, if A precedes B, then A @< B holds.
 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
 
+%% file_modification_time(+File, ?T).
+%
+% For a file `File` that must exist, `T` is the modification time.
+%
+% T is a time stamp compatible with `library(time)`.
 file_modification_time(File, T) :-
         file_time_(File, modification, T).
 
+%% file_access_time(+File, ?T).
+%
+% For a file `File` that must exist, `T` is the access time.
+%
+% T is a time stamp compatible with `library(time)`.
+
 file_access_time(File, T) :-
         file_time_(File, access, T).
 
+%% file_creation_time(+File, ?T).
+%
+% For a file `File` that must exist, `T` is the creation time.
+%
+% T is a time stamp compatible with `library(time)`.
 file_creation_time(File, T) :-
         file_time_(File, creation, T).
 
@@ -170,29 +250,31 @@ file_time_(File, Which, T) :-
         read_from_chars(T0, T).
 
 
-/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
-   path_segments(Ps, Segments): True iff Segments are the segments of Ps.
-
-   Segments is the list of components of the path Ps that are
-   separated by the platform-specific directory separator. Each
-   segment is a list of characters.
-
-   At least one of the arguments must be instantiated.
-
-   Examples:
-
-      ?- path_segments("/hello/there", Segments).
-         Segments = [[],"hello","there"].
-
-      ?- path_segments(Path, ["hello","there"]).
-         Path = "hello/there".
-
-
-   To obtain the platform-specific directory separator, you can use:
-
-      ?- path_segments(Separator, ["",""]).
-         Separator = "/".
-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
+%% path_segments(?Ps, ?Segments).
+%
+% True iff Segments are the segments of Ps.
+%
+% Segments is the list of components of the path Ps that are
+% separated by the platform-specific directory separator. Each
+% segment is a list of characters.
+%
+% At least one of the arguments must be instantiated.
+%
+% Examples:
+%
+% ```
+% ?- path_segments("/hello/there", Segments).
+%    Segments = [[],"hello","there"].
+% ?- path_segments(Path, ["hello","there"]).
+%    Path = "hello/there".
+% ```
+% 
+% To obtain the platform-specific directory separator, you can use:
+%
+% ```
+% ?- path_segments(Separator, ["",""]).
+%    Separator = "/".
+% ```
 
 path_segments(Path, Segments) :-
         '$directory_separator'(Sep),

src/lib/format.pl

@@ -1,83 +1,17 @@
 /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
-   Written 2020, 2021, 2022 by Markus Triska (triska@metalevel.at)
+   Written 2020-2025 by Markus Triska (triska@metalevel.at)
    Part of Scryer Prolog.
-
-   This library provides the nonterminal format_//2 to describe
-   formatted strings. format/[2,3] are provided for impure output.
-
-   Usage:
-   ======
-
-   phrase(format_(FormatString, Arguments), Ls)
-
-   format_//2 describes a list of characters Ls that are formatted
-   according to FormatString. FormatString is a string (i.e.,
-   a list of characters) that specifies the layout of Ls.
-   The characters in FormatString are used literally, except
-   for the following tokens with special meaning:
-
-     ~w    use the next available argument from Arguments here
-     ~q    use the next argument here, formatted as by writeq/1
-     ~a    use the next argument here, which must be an atom
-     ~s    use the next argument here, which must be a string
-     ~d    use the next argument here, which must be an integer
-     ~f    use the next argument here, a floating point number
-     ~Nf   where N is an integer: format the float argument
-           using N digits after the decimal point
-     ~Nd   like ~d, placing the last N digits after a decimal point;
-           if N is 0 or omitted, no decimal point is used.
-     ~ND   like ~Nd, separating digits to the left of the decimal point
-           in groups of three, using the character "," (comma)
-     ~NU   like ~ND, using "_" (underscore) to separate groups of digits
-     ~NL   format an integer so that at most N digits appear on a line.
-           If N is 0 or omitted, it defaults to 72.
-     ~Nr   where N is an integer between 2 and 36: format the
-           next argument, which must be an integer, in radix N.
-           The characters "a" to "z" are used for radices 10 to 36.
-           If N is omitted, it defaults to 8 (octal).
-     ~NR   like ~Nr, except that "A" to "Z" are used for radices > 9
-     ~|    place a tab stop at this position
-     ~N|   where N is an integer: place a tab stop at text column N
-     ~N+   where N is an integer: place a tab stop N characters
-           after the previous tab stop (or start of line)
-     ~t    distribute spaces evenly between the two closest tab stops
-     ~`Ct  like ~t, use character C instead of spaces to fill the space
-     ~n    newline
-     ~Nn   N newlines
-     ~i    ignore the next argument
-     ~~    the literal ~
-
-   Instead of ~N, you can write ~* to use the next argument from Arguments
-   as the numeric argument.
-
-   The predicate format/2 is like format_//2, except that it outputs
-   the text on the terminal instead of describing it declaratively.
-
-   format/3, used as format(Stream, FormatString, Arguments), outputs
-   the described string to the given Stream. If Stream is a binary
-   stream, then the code of each emitted character must be in 0..255.
-
-   If at all possible, format_//2 should be used, to stress pure parts
-   that enable easy testing etc. If necessary, you can emit the list Ls
-   with maplist(put_char, Ls) or, much faster, with format("~s", [Ls]).
-   Ideally, however, you use phrase_to_file/[2,3] or phrase_to_stream/2
-   from library(pio) to write the described list directly to a file
-   or stream, respectively: phrase_to_stream(format_(..., [...]), S).
-   The advantage of this is that an ideal implementation writes
-   the characters as they become known, without manifesting the list.
-
-   The entire library only works if the Prolog flag double_quotes
-   is set to chars, the default value in Scryer Prolog. This should
-   also stay that way, to encourage a sensible environment.
-
-   Example:
-
-   ?- phrase(format_("~s~n~`.t~w!~12|", ["hello",there]), Cs).
-   %@    Cs = "hello\n......there!".
-
    I place this code in the public domain. Use it in any way you want.
 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
 
+/** This library provides the nonterminal `format_//2` to describe
+    formatted strings. `format/[2,3]` are provided for _impure_ output.
+
+    The entire library only works if the Prolog flag `double_quotes`
+    is set to `chars`, the default value in Scryer Prolog. This should
+    also stay that way, to encourage a sensible environment.
+*/
+
 :- module(format, [format_//2,
                    format/2,
                    format/3,
@@ -94,13 +28,122 @@
 :- use_module(library(between)).
 :- use_module(library(pio)).
 
+%% format_(+FormatString, +Arguments)//
+%
+% Usage:
+%
+% ```
+% phrase(format_(FormatString, Arguments), Ls)
+% ```
+%
+% `format_//2` describes a list of characters Ls that are formatted
+% according to FormatString. FormatString is a string (i.e., a list of
+% characters) that specifies the layout of Ls. The characters in
+% FormatString are used literally, except for the following tokens
+% with special meaning:
+%
+% | `~q`     |  use the next argument here, formatted as by `writeq/1`        |
+% | `~a`     |  use the next argument here, which must be an atom             |
+% | `~s`     |  use the next argument here, which must be a string            |
+% | `~d`     |  use the next argument here, which must be an integer          |
+% | `~f`     |  use the next argument here, a floating point number           |
+% | `~Nf`    |  where N is an integer: format the float argument              |
+% |          |  using N digits after the decimal point                        |
+% | `~Nd`    |  like ~d, placing the last N digits after a decimal point;     |
+% |          |  if N is 0 or omitted, no decimal point is used.               |
+% | `~ND`    |  like ~Nd, separating digits to the left of the decimal point  |
+% |          |  in groups of three, using the character "," (comma)           |
+% | `~NU`    |  like ~ND, using "_" (underscore) to separate groups of digits |
+% | `~NL`    |  format an integer so that at most N digits appear on a line.  |
+% |          |  If N is 0 or omitted, it defaults to 72.                      |
+% | `~Nr`    |  where N is an integer between 2 and 36: format the            |
+% |          |  next argument, which must be an integer, in radix N.          |
+% |          |  The characters "a" to "z" are used for radices 10 to 36.      |
+% |          |  If N is omitted, it defaults to 8 (octal).                    |
+% | `~NR`    |  like ~Nr, except that "A" to "Z" are used for radices > 9     |
+% | `~|`     |  place a tab stop at this position                             |
+% | `~N|`    |  where N is an integer: place a tab stop at text column N      |
+% | `~N+`    |  where N is an integer: place a tab stop N characters          |
+% |          |  after the previous tab stop (or start of line)                |
+% | `~t`     |  distribute spaces evenly between the two closest tab stops    |
+% | ``~`Ct`` |  like ~t, use character C instead of spaces to fill the space  |
+% | `~n`     |  newline                                                       |
+% | `~Nn`    |  N newlines                                                    |
+% | `~i`     |  ignore the next argument                                      |
+% | `~~`     |  the literal ~                                                 |
+% | `~w`     |  format like `write/1` would; consider using `~q`, `~d`, etc.  |
+%
+% Instead of `~N`, you can write `~*` to use the next argument from
+% Arguments as the numeric argument.
+%
+% Example:
+%
+% ```
+% ?- phrase(format_("~s~n~`.t~w!~12|", ["hello",there]), Cs).
+%    Cs = "hello\n......there!".
+% ```
+
 format_(Fs, Args) -->
-        { must_be(list, Fs),
-          must_be(list, Args),
-          unique_variable_names(Args, VNs),
-          phrase(cells(Fs,Args,0,[],VNs), Cells) },
+        { format_args_cells(Fs, Args, Cells) },
         format_cells(Cells).
 
+format_args_cells(Fs, Args, Cells) :-
+        must_be(chars, Fs),
+        must_be(list, Args),
+        unique_variable_names(fabricated, Args, VNs),
+        phrase(cells(Fs,Args,0,[],VNs), Cells).
+
+unique_variable_names(Type, Term, VNs) :-
+        term_variables(Term, Vs),
+        foldl(var_name(Type), Vs, VNs, 0, _).
+
+var_name(Type, V, Name=V, Num0, Num) :-
+        charsio:fabricate_var_name(Type, Name, Num0),
+        Num is Num0 + 1.
+
+user:goal_expansion(format_(Fs,Args,Cs0,Cs),
+                    format:format_cells(Cells, Cs0, Cs)) :-
+        catch(format_args_cells(Fs,Args,Cells),
+              E,
+              % no partial evaluation for uses of format_//2 that
+              % cannot be compiled statically, for example those where
+              % the argument list is a variable, or where ~*n occurs
+              % in the format string, or a domain error occurs
+              (   (   E = error(instantiation_error,_)
+                  ;   E = error(domain_error(_,_), _)
+                  ) ->
+                  false
+              ;   throw(E)
+              )).
+
+/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
+   Partial evaluation of goals involving conditions that can be
+   checked at compilation time. This is especially useful for the
+   common case of conditions that test a numeric argument against 0.
+   It is currently used for the goals of ~d.
+- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
+
+goal_pe(G0, G) :- var(G0), !, G = G0.
+goal_pe((A0,B0), (A,B)) :- !, goal_pe(A0, A), goal_pe(B0, B).
+goal_pe((Body0 ; Else0), Body) :-
+        nonvar(Body0),
+        Body0 = ( If -> Then0 ),
+        !,
+        (   ground(If) ->
+            (   If ->
+                goal_pe(Then0, Body)
+            ;   goal_pe(Else0, Body)
+            )
+        ;   goal_pe(Then0, Then),
+            goal_pe(Else0, Else),
+            Body = ( If -> Then ; Else )
+        ).
+goal_pe(Goal, Goal).
+
+/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
+   format_cells//1 is an interpreter for cells, describing a string.
+- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
+
 format_cells([]) --> [].
 format_cells([Cell|Cells]) -->
         format_cell(Cell),
@@ -136,6 +179,7 @@ format_element(glue(Fill,Num)) -->
         { length(Ls, Num),
           maplist(=(Fill), Ls) },
         seq(Ls).
+format_element(goal(_)) --> [].
 
 elements_gluevars([], N, N) --> [].
 elements_gluevars([E|Es], N0, N) -->
@@ -143,9 +187,11 @@ elements_gluevars([E|Es], N0, N) -->
         elements_gluevars(Es, N1, N).
 
 element_gluevar(chars(Cs), N0, N) -->
-        { length(Cs, L),
+        { must_be(chars, Cs),
+          length(Cs, L),
           N is N0 + L }.
 element_gluevar(glue(_,V), N, N) --> [V].
+element_gluevar(goal(G), N, N)   --> { G }.
 
 /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    Our key datastructure is a list of cells and newlines.
@@ -153,12 +199,16 @@ element_gluevar(glue(_,V), N, N) --> [V].
    From and To denote the positions of surrounding tab stops.
 
    Elements is a list of elements that occur in a cell,
-   namely terms of the form chars(Cs) and glue(Char, Var).
+   namely terms of the form chars(Cs), glue(Char, Var)
+   and goal(G).
+
    "glue" elements (TeX terminology) are evenly stretched
    to fill the remaining whitespace in the cell. For each
    glue element, the character Char is used for filling,
    and Var is a free variable that is used when the
-   available space is distributed.
+   available space is distributed. Goals are dynamically
+   executed to obtain characters. In this way, format strings
+   can be parsed and compiled statically when possible.
 
    newline is used if ~n occurs in a format string.
    It is used because a newline character does not
@@ -172,54 +222,55 @@ cells([], Args, Tab, Es, _) --> !,
 cells([~,~|Fs], Args, Tab, Es, VNs) --> !,
         cells(Fs, Args, Tab, [chars("~")|Es], VNs).
 cells([~,w|Fs], [Arg|Args], Tab, Es, VNs) --> !,
-        { write_term_to_chars(Arg, [numbervars(true),variable_names(VNs)], Chars) },
-        cells(Fs, Args, Tab, [chars(Chars)|Es], VNs).
+        { G = write_term_to_chars(Arg, [numbervars(true),variable_names(VNs)], Chars) },
+        cells(Fs, Args, Tab, [chars(Chars),goal(G)|Es], VNs).
 cells([~,q|Fs], [Arg|Args], Tab, Es, VNs) --> !,
-        { write_term_to_chars(Arg, [quoted(true),numbervars(true),variable_names(VNs)], Chars) },
-        cells(Fs, Args, Tab, [chars(Chars)|Es], VNs).
+        { G = write_term_to_chars(Arg, [quoted(true),numbervars(true),variable_names(VNs)], Chars) },
+        cells(Fs, Args, Tab, [chars(Chars),goal(G)|Es], VNs).
 cells([~,a|Fs], [Arg|Args], Tab, Es, VNs) --> !,
-        { atom_chars(Arg, Chars) },
-        cells(Fs, Args, Tab, [chars(Chars)|Es], VNs).
+        { G = atom_chars(Arg, Chars) },
+        cells(Fs, Args, Tab, [chars(Chars),goal(G)|Es], VNs).
 cells([~|Fs0], Args0, Tab, Es, VNs) -->
         { numeric_argument(Fs0, Num, [d|Fs], Args0, [Arg0|Args]) },
         !,
-        { Arg is Arg0, % evaluate compound expression
-          must_be(integer, Arg),
-          number_chars(Arg, Cs0) },
-        (   { Num =:= 0 } -> { Cs = Cs0 }
-        ;   { length(Cs0, L),
-              (   L =< Num ->
-                  Delta is Num - L,
-                  length(Zs, Delta),
-                  maplist(=('0'), Zs),
-                  phrase(("0.",seq(Zs),seq(Cs0)), Cs)
-              ;   BeforeComma is L - Num,
-                  length(Bs, BeforeComma),
-                  append(Bs, Ds, Cs0),
-                  phrase((seq(Bs),".",seq(Ds)), Cs)
-              ) }
-        ),
-        cells(Fs, Args, Tab, [chars(Cs)|Es], VNs).
+        { G0 = ( Arg is Arg0, % evaluate compound expression
+                 must_be(integer, Arg),
+                 number_chars(Arg, Cs0),
+                 (   Num =:= 0 -> Cs = Cs0
+                 ;   length(Cs0, L),
+                     (   L =< Num ->
+                         Delta is Num - L,
+                         length(Zs, Delta),
+                         maplist(=('0'), Zs),
+                         phrase(("0.",seq(Zs),seq(Cs0)), Cs)
+                     ;   BeforeComma is L - Num,
+                         length(Bs, BeforeComma),
+                         append(Bs, Ds, Cs0),
+                         phrase((seq(Bs),".",seq(Ds)), Cs)
+                     )
+                 )),
+          goal_pe(G0, G) },
+        cells(Fs, Args, Tab, [chars(Cs),goal(G)|Es], VNs).
 cells([~|Fs0], Args0, Tab, Es, VNs) -->
         { numeric_argument(Fs0, Num, ['D'|Fs], Args0, [Arg|Args]) },
         !,
-        { separate_digits_fractional(Arg, ',', Num, Cs) },
-        cells(Fs, Args, Tab, [chars(Cs)|Es], VNs).
+        { G = separate_digits_fractional(Arg, ',', Num, Cs) },
+        cells(Fs, Args, Tab, [chars(Cs),goal(G)|Es], VNs).
 cells([~|Fs0], Args0, Tab, Es, VNs) -->
         { numeric_argument(Fs0, Num, ['U'|Fs], Args0, [Arg|Args]) },
         !,
-        { separate_digits_fractional(Arg, '_', Num, Cs) },
-        cells(Fs, Args, Tab, [chars(Cs)|Es], VNs).
+        { G = separate_digits_fractional(Arg, '_', Num, Cs) },
+        cells(Fs, Args, Tab, [chars(Cs),goal(G)|Es], VNs).
 cells([~|Fs0], Args0, Tab, Es, VNs) -->
         { numeric_argument(Fs0, Num0, ['L'|Fs], Args0, [Arg|Args]) },
         !,
-        { (   Num0 =:= 0 ->
-              Num = 72
-          ;   Num = Num0
-          ),
-          phrase(format_("~d", [Arg]), Cs0),
-          phrase(split_lines_width(Cs0, Num), Cs) },
-        cells(Fs, Args, Tab, [chars(Cs)|Es], VNs).
+        { G = ((   Num0 =:= 0 ->
+                   Num = 72
+               ;   Num = Num0
+               ),
+               phrase(format_("~d", [Arg]), Cs0),
+               phrase(split_lines_width(Cs0, Num), Cs) ) },
+        cells(Fs, Args, Tab, [chars(Cs),goal(G)|Es], VNs).
 cells([~,i|Fs], [_|Args], Tab, Es, VNs) --> !,
         cells(Fs, Args, Tab, Es, VNs).
 cells([~,n|Fs], Args, Tab, Es, VNs) --> !,
@@ -234,59 +285,40 @@ cells([~|Fs0], Args0, Tab, Es, VNs) -->
         cells(Fs, Args, 0, [], VNs).
 cells([~,s|Fs], [Arg|Args], Tab, Es, VNs) --> !,
         cells(Fs, Args, Tab, [chars(Arg)|Es], VNs).
-cells([~,f|Fs], [Arg|Args], Tab, Es, VNs) --> !,
-        { format_number_chars(Arg, Chars) },
-        cells(Fs, Args, Tab, [chars(Chars)|Es], VNs).
+cells([~,f|Fs], Args, Tab, Es, VNs) --> !,
+        cells([~,'6',f|Fs], Args, Tab, Es, VNs).
 cells([~|Fs0], Args0, Tab, Es, VNs) -->
         { numeric_argument(Fs0, Num, [f|Fs], Args0, [Arg|Args]) },
         !,
-        { format_number_chars(Arg, Cs0),
-          phrase(upto_what(Bs, .), Cs0, Cs),
-          (   Num =:= 0 -> Chars = Bs
-          ;   (   Cs = ['.'|Rest] ->
-                  length(Rest, L),
-                  (   Num < L ->
-                      length(Ds, Num),
-                      append(Ds, _, Rest)
-                  ;   Num =:= L ->
-                      Ds = Rest
-                  ;   Num > L,
-                      Delta is Num - L,
-                      % we should look into the float with
-                      % greater accuracy here, and use the
-                      % actual digits instead of 0.
-                      length(Zs, Delta),
-                      maplist(=('0'), Zs),
-                      append(Rest, Zs, Ds)
-                  )
-              ;   length(Ds, Num),
-                  maplist(=('0'), Ds)
-              ),
-              append(Bs, ['.'|Ds], Chars)
-          ) },
-        cells(Fs, Args, Tab, [chars(Chars)|Es], VNs).
+        { G = phrase(float_with_n_decimal_digits(Arg, Num), Chars) },
+        cells(Fs, Args, Tab, [chars(Chars),goal(G)|Es], VNs).
 cells([~,r|Fs], Args, Tab, Es, VNs) --> !,
         cells([~,'8',r|Fs], Args, Tab, Es, VNs).
 cells([~|Fs0], Args0, Tab, Es, VNs) -->
         { numeric_argument(Fs0, Num, [r|Fs], Args0, [Arg|Args]) },
         !,
-        { integer_to_radix(Arg, Num, lowercase, Cs) },
-        cells(Fs, Args, Tab, [chars(Cs)|Es], VNs).
+        { G = integer_to_radix(Arg, Num, lowercase, Cs) },
+        cells(Fs, Args, Tab, [chars(Cs),goal(G)|Es], VNs).
 cells([~,'R'|Fs], Args, Tab, Es, VNs) --> !,
         cells([~,'8','R'|Fs], Args, Tab, Es, VNs).
 cells([~|Fs0], Args0, Tab, Es, VNs) -->
         { numeric_argument(Fs0, Num, ['R'|Fs], Args0, [Arg|Args]) },
         !,
-        { integer_to_radix(Arg, Num, uppercase, Cs) },
-        cells(Fs, Args, Tab, [chars(Cs)|Es], VNs).
+        { G = integer_to_radix(Arg, Num, uppercase, Cs) },
+        cells(Fs, Args, Tab, [chars(Cs),goal(G)|Es], VNs).
 cells([~,'`',Char,t|Fs], Args, Tab, Es, VNs) --> !,
         cells(Fs, Args, Tab, [glue(Char,_)|Es], VNs).
 cells([~,t|Fs], Args, Tab, Es, VNs) --> !,
         cells(Fs, Args, Tab, [glue(' ',_)|Es], VNs).
 cells([~,'|'|Fs], Args, Tab0, Es, VNs) --> !,
-        { phrase(elements_gluevars(Es, 0, Width), _),
-          Tab is Tab0 + Width },
-        cell(Tab0, Tab, Es),
+        (   { ground(Tab0), Es = [chars(Cs)], ground(Cs) } ->
+            { length(Cs, Width),
+              Tab is Tab0 + Width },
+            cell(Tab0, Tab, Es)
+        ;   { G = (phrase(elements_gluevars(Es, 0, Width), _),
+                   Tab is Tab0 + Width) },
+            cell(Tab0, Tab, [goal(G)|Es])
+        ),
         cells(Fs, Args, Tab, [], VNs).
 cells([~|Fs0], Args0, Tab, Es, VNs) -->
         { numeric_argument(Fs0, Num, ['|'|Fs], Args0, Args) },
@@ -296,8 +328,12 @@ cells([~|Fs0], Args0, Tab, Es, VNs) -->
 cells([~|Fs0], Args0, Tab0, Es, VNs) -->
         { numeric_argument(Fs0, Num, [+|Fs], Args0, Args) },
         !,
-        { Tab is Tab0 + Num },
-        cell(Tab0, Tab, Es),
+        (   { ground(Tab0+Num) } ->
+            { Tab is Tab0 + Num },
+            cell(Tab0, Tab, Es)
+        ;   { G = (Tab is Tab0 + Num) },
+            cell(Tab0, Tab, [goal(G)|Es])
+        ),
         cells(Fs, Args, Tab, [], VNs).
 cells([~|Cs], Args, _, _, _) -->
         (   { Args == [] } ->
@@ -309,18 +345,27 @@ cells(Fs0, Args, Tab, Es, VNs) -->
           Fs1 = [_|_] },
         cells(Fs, Args, Tab, [chars(Fs1)|Es], VNs).
 
-format_number_chars(N0, Chars) :-
-        N is N0, % evaluate compound expression
-        number_chars(N, Chars).
+float_with_n_decimal_digits(F, N) -->
+   {  Fr is abs(float_fractional_part(F)),
+      FrR0 is round(Fr*10^N),
+      I0 is truncate(F),
+      (  FrR0 >= 10^N
+      -> I is I0+truncate(sign(F)), FrR = FrR0
+      ;  I = I0, FrR is FrR0+10^N
+      )
+   },
+   ( { I=0, F<0, FrR>10^N } -> "-0" ; { number_chars(I, Is) }, seq(Is) ),
+   ".",
+   ( { FrR = 1 } -> "0" ; { number_chars(FrR, ['1'|FrRs]) }, seq(FrRs) ).
 
-n_newlines(0) --> !.
 n_newlines(N0) --> { N0 > 0, N is N0 - 1 }, [newline], n_newlines(N).
+n_newlines(0)  --> [].
 
 /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
-?- phrase(upto_what(Cs, ~), "abc~test", Rest).
-Cs = [a,b,c], Rest = [~,t,e,s,t].
-?- phrase(upto_what(Cs, ~), "abc", Rest).
-Cs = [a,b,c], Rest = [].
+?- phrase(format:upto_what(Cs, ~), "abc~test", Rest).
+   Cs = "abc", Rest = "~test".
+?- phrase(format:upto_what(Cs, ~), "abc", Rest).
+   Cs = "abc", Rest = [].
 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
 
 separate_digits_fractional(Arg, Sep, Num, Cs) :-
@@ -414,13 +459,39 @@ digits(uppercase, "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ").
    Impure I/O, implemented as a small wrapper over format_//2.
 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
 
-format(Fs, Args) :-
-        current_output(Stream),
-        format(Stream, Fs, Args).
+%% format(+Fs, +Args)
+%
+%  The predicate `format/2` is like `format_//2`, except that it
+%  outputs the text on the terminal instead of describing it
+%  declaratively as a list of characters.
+%
+%  If at all possible, `format_//2` should be used, to stress pure
+%  parts that enable easy testing etc. If necessary, you can emit the
+%  described list of characters `Ls` with `maplist(put_char, Ls)` or,
+%  much faster, with `format("~s", [Ls])`. Ideally, however, you use
+%  `phrase_to_file/[2,3]` or `phrase_to_stream/2` from `library(pio)`
+%  to write the described list directly to a file or stream,
+%  respectively: `phrase_to_stream(format_(..., [...]), S)`. The
+%  advantage of this is that an ideal implementation writes the
+%  characters as they become known, without manifesting the list.
 
-format(Stream, Fs, Args) :-
-        phrase_to_stream(format_(Fs, Args), Stream),
-        flush_output(Stream).
+format(_, _) :- not_used.
+
+user:goal_expansion(format(Fs, Args),
+                    (   current_output(Stream),
+                        format(Stream, Fs, Args))).
+
+%% format(Stream, FormatString, Arguments)
+%
+%  Output the described string to the given Stream. If Stream is a
+%  binary stream, then the code of each emitted character must be in
+%  0..255.
+
+format(_, _, _) :- not_used.
+
+user:goal_expansion(format(Stream, Fs, Args),
+                    (   pio:phrase_to_stream(format:format_(Fs, Args), Stream),
+                        flush_output(Stream))).
 
 /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
 ?- phrase(format:cells("hello", [], 0, [], []), Cs).
@@ -433,9 +504,9 @@ format(Stream, Fs, Args) :-
 ?- phrase(format:cells("~`at~50|", [], 0, [], []), Cs),
    phrase(format:format_cells(Cs), Ls).
 ?- phrase(format:cells("~ta~t~tb~tc~21|", [], 0, [], []), Cs).
-Cs = [cell(0,21,[glue(' ',_A),chars("a"),glue(' ',_B),glue(' ',_C),chars("b"),glue(' ',_D),chars("c ...")])]
+   Cs = [cell(0,21,[glue(' ',_A),chars("a"),glue(' ',_B),glue(' ',_C),chars("b"),glue(' ',_D),chars("c")])].
 ?- phrase(format:cells("~ta~t~4|", [], 0, [], []), Cs).
-Cs = [cell(0,4,[glue(' ',_A),chars("a"),glue(' ',_B)])]
+   Cs = [cell(0,4,[glue(' ',_A),chars("a"),glue(' ',_B)])].
 
 ?- phrase(format:format_cell(cell(0,1,[glue(a,_94)])), Ls).
 
@@ -486,11 +557,14 @@ aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
 
    In the eventual library organization, portray_clause/1 and
    related predicates may be placed in their own dedicated library.
-
-   portray_clause/1 is useful for printing solutions in such a way
-   that they can be read back with read/1.
 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
 
+
+%% portray_clause(+Term)
+%
+%  `portray_clause/1` is useful for printing solutions in such a way
+%  that they can be read back with `read/1`.
+
 portray_clause(Term) :-
         current_output(Out),
         portray_clause(Out, Term).
@@ -500,20 +574,29 @@ portray_clause(Stream, Term) :-
         flush_output(Stream).
 
 portray_clause_(Term) -->
-        { unique_variable_names(Term, VNs) },
+        { unique_variable_names(numbervars, Term, VNs) },
         portray_(Term, VNs), ".\n".
 
-unique_variable_names(Term, VNs) :-
-        term_variables(Term, Vs),
-        foldl(var_name, Vs, VNs, 0, _).
-
-var_name(V, Name=V, Num0, Num) :-
-        charsio:fabricate_var_name(numbervars, Name, Num0),
-        Num is Num0 + 1.
-
 literal(Lit, VNs) -->
-        { write_term_to_chars(Lit, [quoted(true),variable_names(VNs)], Ls) },
-        seq(Ls).
+        { write_term_to_chars(Lit, [quoted(true),variable_names(VNs),double_quotes(true)], Ls) },
+        (   { nonvar(Lit),
+              \+ number(Lit),
+              functor(Lit, F, A),
+              current_op(Pri, _, F),
+              (   A =:= 0
+              ;   Pri >= 1000
+              ) } ->
+            "(", seq(Ls), ")"
+        ;   seq(Ls)
+        ).
+
+literal_(Lit, VNs) -->
+        { phrase(literal(Lit, VNs), Ls) },
+        seq(Ls),
+        (   { phrase((...,[Last]), Ls), char_type(Last, graphic_token) } ->
+            " "
+        ;   ""
+        ).
 
 portray_(Var, VNs) --> { var(Var) }, !, literal(Var, VNs).
 portray_((Head :- Body), VNs) --> !,
@@ -522,7 +605,7 @@ portray_((Head :- Body), VNs) --> !,
 portray_((Head --> Body), VNs) --> !,
         literal(Head, VNs), " -->\n",
         body_(Body, 0, 3, VNs).
-portray_(Any, VNs) --> literal(Any, VNs).
+portray_(Any, VNs) --> literal_(Any, VNs).
 
 
 body_(Var, C, I, VNs) --> { var(Var) }, !,
@@ -547,7 +630,7 @@ body_((A;B), C, I, VNs) --> !,
         body_(A, C1, C1, VNs), "\n",
         else_branch(B, I, VNs).
 body_(Goal, C, I, VNs) -->
-        indent_to(C, I), literal(Goal, VNs).
+        indent_to(C, I), literal_(Goal, VNs).
 
 
 % True iff Body has the shape ( If -> Then ; Else ).

src/lib/freeze.pl

@@ -1,9 +1,12 @@
 :- module(freeze, [freeze/2]).
 
+/** Provides the constraint `freeze/2`.
+*/
+
 :- use_module(library(atts)).
 :- use_module(library(dcgs)).
 
-:- meta_predicate freeze(-, 0).
+:- meta_predicate(freeze(-, 0)).
 
 :- attribute frozen/1.
 
@@ -19,6 +22,15 @@ verify_attributes(Var, Other, Goals) :-
         ).
 verify_attributes(_, _, []).
 
+%% freeze(Var, Goal)
+%
+%  Schedules Goal to be executed when Var is instantiated. This can
+%  be useful to observe the exact moment a variable becomes bound to a
+%  more concrete term, for example when creating animations of search
+%  processes. Higher-level constructs such as `phrase_from_file/2` can
+%  also be implemented with `freeze/2`, by scheduling a goal that
+%  reads additional data from a file as soon as it is needed.
+
 freeze(X, Goal) :-
     put_atts(Fresh, frozen(Goal)),
     Fresh = X.
@@ -26,5 +38,5 @@ freeze(X, Goal) :-
 attribute_goals(Var) -->
     { get_atts(Var, frozen(Goals)),
       put_atts(Var, -frozen(_)) },
-    [freeze(Var, Goals)].
+    [freeze:freeze(Var, Goals)].
 

src/lib/gensym.pl

@@ -19,14 +19,14 @@ gensym(Base, Unique) :-
     must_be(var, Unique),
     atom_si(Base),
     gensym_key(Base, BaseKey),
-    (  bb_get(BaseKey, UniqueID0) ->
-       UniqueID is UniqueID0 + 1,
-       bb_put(BaseKey, UniqueID),
-       append_id(Base, UniqueID, Unique)
-    ;  bb_put(BaseKey, 1),
-       append_id(Base, 1, Unique)
-    ).
+    (  bb_get(BaseKey, UniqueID0) -> true
+    ;  UniqueID0 = 0
+    ),
+    UniqueID is UniqueID0 + 1,
+    append_id(Base, UniqueID, Unique),
+    bb_put(BaseKey, UniqueID).
 
 reset_gensym(Base) :-
     atom_si(Base),
-    bb_put(Base, 0).
+    gensym_key(Base, BaseKey),
+    bb_put(BaseKey, 0).

src/lib/http/http_open.pl

@@ -1,34 +1,39 @@
 /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    Written 2022 by Adrián Arroyo Calle (adrian.arroyocalle@gmail.com)
    Part of Scryer Prolog.
+*/
 
-   http_open(+Address, -Stream, +Options)
-   ======================================
+/** Make HTTP requests.
 
-   Yields Stream to read the body of an HTTP reply from Address.
-   Address is a list of characters, and includes the method. Both HTTP
-   and HTTPS are supported.
-
-   Options supported:
-
-     * method(+Method): Sets the HTTP method of the call. Method can be get (default), head, delete, post, put or patch.
-     * data(+Data): Data to be sent in the request. Useful for POST, PUT and PATCH operations.
-     * size(-Size): Unifies with the value of the Content-Length header
-     * request_headers(+RequestHeaders): Headers to be used in the request
-     * headers(-ListHeaders): Unifies with a list with all headers returned in the response
-     * status_code(-Code): Unifies with the status code of the request (200, 201, 404, ...)
-
-   Example:
-
-       ?- http_open("https://github.com/mthom/scryer-prolog", S, []).
-       %@    S = '$stream'(0x7fcfc9e00f00).
-
-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
+This library contains the predicate `http_open/3` which allows you to perform HTTP(S) calls.
+Useful for making API calls, or parsing websites. It uses Hyper underneath.
+*/
 
 :- module(http_open, [http_open/3]).
 
 :- use_module(library(lists)).
 
+%% http_open(+Address, -Stream, +Options).
+%
+% Yields Stream to read the body of an HTTP reply from Address.
+% Address is a list of characters, and includes the method. Both HTTP
+% and HTTPS are supported.
+%
+% Options supported:
+%
+%   * `method(+Method)`: Sets the HTTP method of the call. Method can be `get` (default), `head`, `delete`, `post`, `put` or `patch`.
+%   * `data(+Data)`: Data to be sent in the request. Useful for POST, PUT and PATCH operations.
+%   * `size(-Size)`: Unifies with the value of the Content-Length header
+%   * `request_headers(+RequestHeaders)`: Headers to be used in the request
+%   * `headers(-ListHeaders)`: Unifies with a list with all headers returned in the response
+%   * `status_code(-Code)`: Unifies with the status code of the request (200, 201, 404, ...)
+%
+% Example:
+%
+% ```
+% ?- http_open("https://www.example.com", S, []), get_n_chars(S, N, HTML).
+%    S = '$stream'(0x7fb548001be8), N = 1256, HTML = "<!doctype html>\n<ht ...".
+% ```
 http_open(Address, Response, Options) :-
     parse_http_options(Options, OptionValues),
     ( member(method(Method), OptionValues) -> true; Method = get),

src/lib/http/http_server.pl

@@ -1,63 +1,74 @@
 /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    Written in December 2020 by Adrián Arroyo (adrian.arroyocalle@gmail.com)
    Updated in March 2022 by Adrián Arroyo to use the Hyper backend
-   Part of Scryer Prolog
-
-   This library provides an starting point to build HTTP server based applications.
-   It is based on Hyper, which allows for HTTP/1.0, HTTP/1.1 and HTTP/2. However,
-   some advanced features that Hyper provides are still not accesible.
-
-   Usage
-   ==========
-   The main predicate of the library is http_listen/2, which needs a port number
-    (usually 80) and a list of handlers. A handler is a compound term with the functor
-   as one HTTP method (in lowercase) and followed by a Route Match and a predicate
-   which will handle the call.
-
-   text_handler(Request, Response) :-
-    http_status_code(Response, 200),
-    http_body(Response, text("Welcome to Scryer Prolog!")).
-
-   parameter_handler(User, Request, Response) :-
-    http_body(Response, text(User)).
-
-   http_listen(7890, [
-        get(echo, text_handler),           % GET /echo
-        post(user/User, parameter_handler(User)) % POST /user/<User>
-   ]).
-
-   Every handler predicate will have at least 2-arity, with Request and Response.
-   Although you can work directly with http_request and http_response terms, it is
-   recommeded to use the helper predicates, which are easier to understand and cleaner:
-   - http_headers(Response/Request, Headers)
-   - http_status_code(Responde, StatusCode)
-   - http_body(Response/Request, text(Body))
-   - http_body(Response/Request, binary(Body))
-   - http_body(Request, form(Form))
-   - http_body(Response, file(Filename))
-   - http_redirect(Response, Url)
-   - http_query(Request, QueryName, QueryValue)
-
-   Some things that are still missing:
-   - Read forms in multipart format
-   - HTTP Basic Auth
-   - Session handling via cookies
-   - HTML Templating
-
+   Part of Scryer Prolog.
    I place this code in the public domain. Use it in any way you want.
-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
+*/
+
+/** This library provides an starting point to build HTTP server based applications.
+It is based on [Warp](https://github.com/seanmonstar/warp), which allows for HTTP/1.0, HTTP/1.1 and HTTP/2. However,
+some advanced features that Warp provides are still not accesible.
+
+## Usage
+
+The main predicate of the library is `http_listen/2`, which needs a port number
+(usually 80) and a list of handlers. A handler is a compound term with the functor
+as one HTTP method (in lowercase) and followed by a Route Match and a predicate
+which will handle the call.
+
+```
+:- use_module(library(http/http_server)).
+
+text_handler(Request, Response) :-
+  http_status_code(Response, 200),
+  http_body(Response, text("Welcome to Scryer Prolog!")).
+
+parameter_handler(User, Request, Response) :-
+  http_body(Response, text(User)).
+
+run:-
+  http_listen(7890, [
+    get(echo, text_handler),                 % GET /echo
+    post(user/User, parameter_handler(User)) % POST /user/<User>
+  ]).
+```
+
+Every handler predicate will have at least 2-arity, with Request and Response.
+Although you can work directly with `http_request` and `http_response` terms, it is
+recommeded to use the helper predicates, which are easier to understand and cleaner:
+
+   - `http_headers(Response/Request, Headers)`
+   - `http_status_code(Responde, StatusCode)`
+   - `http_body(Response/Request, text(Body))`
+   - `http_body(Response/Request, binary(Body))`
+   - `http_body(Request, form(Form))`
+   - `http_body(Response, file(Filename))`
+   - `http_redirect(Response, Url)`
+   - `http_query(Request, QueryName, QueryValue)`
+
+Some things that are still missing:
+
+   - Read forms in multipart format
+   - Session handling via cookies
+   - HTML Templating (but you can use [Teruel](https://github.com/aarroyoc/teruel/), [Marquete](https://github.com/aarroyoc/marquete/) or [Djota](https://github.com/aarroyoc/djota) for that)
+*/
 
 
 :- module(http_server, [
 	      http_listen/2,
+	      http_listen/3,
 	      http_headers/2,
 	      http_status_code/2,
 	      http_body/2,
 	      http_redirect/2,
-	      http_query/3
+	      http_query/3,
+	      http_basic_auth/4
 ]).
 
-:- meta_predicate http_listen(?, :).
+:- meta_predicate(http_listen(?, :)).
+:- meta_predicate(http_listen(?, :, ?)).
+
+:- meta_predicate(http_basic_auth(:, :, ?, ?)).
 
 :- use_module(library(charsio)).
 :- use_module(library(crypto)).
@@ -68,22 +79,60 @@
 :- use_module(library(pio)).
 :- use_module(library(time)).
 
+%% http_listen(+Port, +Handlers).
+%
+% Equivalent to `http_listen(Port, Handlers, [])`.
 http_listen(Port, Module:Handlers0) :-
     must_be(integer, Port),
     must_be(list, Handlers0),
     maplist(module_qualification(Module), Handlers0, Handlers),
-    http_listen_(Port, Handlers).
+    http_listen_(Port, Handlers, []).
+
+%% http_listen(+Port, +Handlers, +Options).
+%
+% Listens for HTTP connections on port Port. Each handler on the list Handlers should be of the form: `HttpVerb(PathUnification, Predicate)`.
+% For example: `get(user/User, get_info(User))` will match an HTTP request that is a GET, the path unifies with /user/User (where User is a variable)
+% and it will call `get_info` with three arguments: an `http_request` term, an `http_response` term and User.
+%
+% The following options are supported:
+%
+% - `tls_key(+Key)` - a TLS key for HTTPS (string)
+% - `tls_cert(+Cert)` - a TLS cert for HTTPS (string)
+% - `content_length_limit(+Limit)` - maximum length (in bytes) for the incoming bodies. By default, 32KB.
+%
+% In order to have a HTTPS server (instead of plain HTTP), both `tls_key` and `tls_cert` options must be provided.
+http_listen(Port, Module:Handlers0, Options) :-
+    must_be(integer, Port),
+    must_be(list, Handlers0),
+    must_be(list, Options),
+    maplist(module_qualification(Module), Handlers0, Handlers),
+    http_listen_(Port, Handlers, Options).
 
 module_qualification(M, H0, H) :-
     H0 =.. [Method, Path, Goal],
     H =.. [Method, Path, M:Goal].
 
-http_listen_(Port, Handlers) :-
+http_listen_(Port, Handlers, Options) :-
+    parse_options(Options, TLSKey, TLSCert, ContentLengthLimit),
     phrase(format_("0.0.0.0:~d", [Port]), Addr),
-    '$http_listen'(Addr, HttpListener),!,
+    '$http_listen'(Addr, HttpListener, TLSKey, TLSCert, ContentLengthLimit),!,
     format("Listening at ~s\n", [Addr]),
     http_loop(HttpListener, Handlers).
 
+parse_options(Options, TLSKey, TLSCert, ContentLengthLimit) :-
+    member_option_default(tls_key, Options, "", TLSKey),
+    member_option_default(tls_cert, Options, "", TLSCert),
+    member_option_default(content_length_limit, Options, 32768, ContentLengthLimit),
+    must_be(integer, ContentLengthLimit).
+
+member_option_default(Key, List, _Default, Value) :-
+    X =.. [Key, Value],
+    member(X, List).
+member_option_default(Key, List, Default, Default) :-
+    X =.. [Key, _],
+    \+ member(X, List).
+	
+
 http_loop(HttpListener, Handlers) :-
     '$http_accept'(HttpListener, RequestMethod, RequestPath, RequestHeaders, RequestQuery, RequestStream, ResponseHandle),
     current_time(Time),
@@ -105,42 +154,49 @@ http_loop(HttpListener, Handlers) :-
 	)
     ; (
 	'$http_answer'(ResponseHandle, 404, [], ResponseStream),
-	call_cleanup(format(ResponseStream, "Not Found"), close(ResponseStream)))
+	call_cleanup(format(ResponseStream, "Not Found", []), close(ResponseStream)))
     ),
     http_loop(HttpListener, Handlers).
 
 send_response(ResponseHandle, http_response(StatusCode0, text(ResponseText), ResponseHeaders0)) :-
     default(StatusCode0, 200, StatusCode),
     maplist(map_header_kv_2, ResponseHeaders, ResponseHeaders0),
-    '$http_answer'(ResponseHandle, StatusCode, ResponseHeaders, ResponseStream),
-    call_cleanup(
-	format(ResponseStream, "~s", [ResponseText]),
-	close(ResponseStream)
+    '$http_answer'(ResponseHandle, StatusCode, ResponseHeaders, ResponseStream0),
+    open(stream(ResponseStream0), write, ResponseStream, [type(text)]),
+    catch(
+	call_cleanup(format(ResponseStream, "~s", [ResponseText]),close(ResponseStream)),
+	error(existence_error(stream, _), _),
+	true
     ).
 
 send_response(ResponseHandle, http_response(StatusCode0, bytes(ResponseBytes), ResponseHeaders0)) :-
     default(StatusCode0, 200, StatusCode),
     maplist(map_header_kv_2, ResponseHeaders, ResponseHeaders0),
     '$http_answer'(ResponseHandle, StatusCode, ResponseHeaders, ResponseStream),
-    call_cleanup(
-	format(ResponseStream, "~s", [ResponseBytes]),
-	close(ResponseStream)
+    catch(
+        call_cleanup(format(ResponseStream, "~s", [ResponseBytes]),close(ResponseStream)),
+	error(existence_error(stream, _), _),
+	true
     ).
 
 send_response(ResponseHandle, http_response(StatusCode0, file(Filename), ResponseHeaders0)) :-
     default(StatusCode0, 200, StatusCode),
     maplist(map_header_kv_2, ResponseHeaders, ResponseHeaders0),
     '$http_answer'(ResponseHandle, StatusCode, ResponseHeaders, ResponseStream),
-    call_cleanup(
-	setup_call_cleanup(
-	    open(Filename, read, FileStream, [type(binary)]),
-	    (
-		get_n_chars(FileStream, _, FileCs),
-		format(ResponseStream, "~s", [FileCs])
+    catch(
+	call_cleanup(
+	    setup_call_cleanup(
+		open(Filename, read, FileStream, [type(binary)]),
+		(
+		    get_n_chars(FileStream, _, FileCs),
+		    format(ResponseStream, "~s", [FileCs])
+		),
+		close(FileStream)
 	    ),
-	    close(FileStream)
+	    close(ResponseStream)
 	),
-        close(ResponseStream)
+	error(existence_error(stream, _), _),
+	true
     ).
 
 
@@ -206,9 +262,21 @@ string_without(Not, [Char|String]) -->
 string_without(_, []) -->
     [].
 
+%% http_headers(?Request_Response, ?Headers).
+%
+% True iff `Request_Response` is a request or response with headers Headers. Can be used both to get headers (usually in from a request)
+% and to add headers (usually in a response).
 http_headers(http_request(Headers, _, _), Headers).
 http_headers(http_response(_, _, Headers), Headers).
 
+%% http_body(?Request_Response, ?Body).
+%
+% True iff Body is the body of the request or response. A body can be of the following types:
+%
+%  * `bytes(Bytes)` for both requests and responses, interprets the body as bytes
+%  * `text(Bytes)` for both requests and responses, interprets the body as text
+%  * `form(Form)` only for requests, interprets the body as an `application/x-www-form-urlencoded` form.
+%  * `file(File)` only for responses, interprets the body as the content of a file (useful to send static files).
 http_body(http_request(_, stream(StreamBody), _), bytes(BytesBody)) :- get_n_chars(StreamBody, _, BytesBody).
 http_body(http_request(_, stream(StreamBody), _), text(TextBody)) :- get_n_chars(StreamBody, _, TextBody).
 http_body(http_request(Headers, stream(StreamBody), _), form(FormBody)) :- 
@@ -218,31 +286,38 @@ http_body(http_request(Headers, stream(StreamBody), _), form(FormBody)) :-
 http_body(http_request(_, Body, _), Body).
 http_body(http_response(_, Body, _), Body).
 
+%% http_status_code(?Response, ?StatusCode).
+%
+% True iff the status code of the response Response unifies with StatusCode.
 http_status_code(http_response(StatusCode, _, _), StatusCode).
+
+%% http_redirect(-Response, +Uri).
+%
+% True iff Response is a response that redirects the user to the uri Uri.
 http_redirect(http_response(307, text("Moved Temporarily"), ["Location"-Uri]), Uri).
+
+%% http_query(+Request, ?Key, ?Value).
+%
+% True iff there's a query in request Request with key Key and value Value.
 http_query(http_request(_, _, Queries), Key, Value) :- member(Key-Value, Queries).
 
 parse_queries([Key-Value|Queries]) -->
     string_without("=", Key0),
-    {
-        phrase(url_decode(Key), Key0)
-    },
     "=",
     string_without("&", Value0),
-    {
-        phrase(url_decode(Value), Value0)
-    },
     "&",
-    parse_queries(Queries).
+    parse_queries(Queries),
+    {
+	phrase(url_decode(Key), Key0),
+	phrase(url_decode(Value), Value0)
+    }.
 
 parse_queries([Key-Value]) -->
     string_without("=", Key0),
-    {
-        phrase(url_decode(Key), Key0)
-    },
     "=",
     string_without(" ", Value0),
     {
+	phrase(url_decode(Key), Key0),
         phrase(url_decode(Value), Value0)  
     }.
 
@@ -253,9 +328,13 @@ parse_queries([]) -->
 url_decode([Char|Chars]) -->
     [Char],
     {
-        Char \= '%'
+        Char \= '%',
+	Char \= (+)
     },
     url_decode(Chars).
+url_decode([' '|Chars]) -->
+    "+",
+    url_decode(Chars).
 url_decode([Char|Chars]) -->
     "%",
     [A],
@@ -313,3 +392,49 @@ url_decode([Char|Chars]) -->
     url_decode(Chars).
 
 url_decode([]) --> [].
+
+%% http_basic_auth(+LoginPredicate, +Handler, +Request, -Response)
+%
+% Metapredicate that wraps an existing Handler with an HTTP Basic Auth flow.
+% Checks if a given user + password is authorized to execute that handler, returning 401
+% if it's not satisfied.
+% 
+% `LoginPredicate` must be a predicate of arity 2 that takes a User and a Password.
+% `Handler` will have, in addition to the Request and Response arguments, a User argument
+% containing the User given in the authentication.
+%
+% Example:
+%
+% ```
+% main :-
+%    http_listen(8800,[get('/', http_basic_auth(login, inside_handler("data")))]).
+%
+% login(User, Pass) :-
+%    User = "aarroyoc",
+%    Pass = "123456".
+%
+% inside_handler(Data, User, Request, Response) :-
+%    http_body(Response, text(User)).
+% ```
+http_basic_auth(LoginPredicate, Handler, Request, Response) :-
+    http_headers(Request, Headers),
+    member("authorization"-AuthorizationStr, Headers),
+    append("Basic ", Coded, AuthorizationStr),
+    chars_base64(UserPass, Coded, []),
+    append(User, [':'|Password], UserPass),
+    (
+	call(LoginPredicate, User, Password) ->
+	call(Handler, User, Request, Response)
+    ;   http_basic_auth_unauthorized_response(Response)
+    ).
+
+http_basic_auth(_LoginPredicate, _Handler, Request, Response) :-
+    http_headers(Request, Headers),
+    \+ member("authorization"-_, Headers),
+    http_basic_auth_unauthorized_response(Response).
+
+http_basic_auth_unauthorized_response(Response) :-
+    http_status_code(Response, 401),
+    http_headers(Response, ["www-authenticate"-"Basic realm=\"Scryer Prolog\", charset=\"UTF-8\""]),
+    http_body(Response, text("Unauthorized")).
+	

src/lib/iso_ext.pl

@@ -1,46 +1,114 @@
+/** Useful general predicates that are not ISO standard yet
+
+Predicates available here are similar to the ones defined in builtin.pl,
+but they're not part of the ISO Prolog standard at the moment.
+*/
+
 :- module(iso_ext, [bb_b_put/2,
                     bb_get/2,
                     bb_put/2,
                     call_cleanup/2,
                     call_with_inference_limit/3,
+                    call_residue_vars/2,
                     forall/2,
                     partial_string/1,
                     partial_string/3,
                     partial_string_tail/2,
                     setup_call_cleanup/3,
+                    succ/2,
                     call_nth/2,
+                    countall/2,
                     copy_term_nat/2,
-		    asserta/2,
-		    assertz/2]).
+		    copy_term/3]).
 
 :- use_module(library(error), [can_be/2,
+                               must_be/2,
                                domain_error/3,
                                instantiation_error/1,
                                type_error/3]).
 
 :- use_module(library(lists), [maplist/3]).
 
+:- use_module(library('$project_atts')).
+
 :- meta_predicate(forall(0, 0)).
 
+%% forall(Generate, Test).
+%
+% For all bindings possible by Generate, Test must be true.
+%
+% In this example, it checks that all numbers are even:
+%
+% ```
+% ?- Ns = [2,4,6], forall(member(N, Ns), 0 is N mod 2).
+%    Ns = [2,4,6].
+% ```
 forall(Generate, Test) :-
     \+ (Generate, \+ Test).
 
-%% (non-)backtrackable global variables.
+% (non-)backtrackable global variables.
 
+%% bb_put(+Key, +Value).
+%
+% Sets a global variable named Key (must be an atom) with value Value.
+% The global variable isn't backtrackable. Check `bb_b_put/2` for the
+% backtrackable version.
+%
+% ```
+% ?- bb_put(city, "Valladolid").
+%    true.
+% ?- bb_get(city, X).
+%    X = "Valladolid".
+% ```
+%
+% In this example one can understand the difference between `bb_put/2` and
+% `bb_b_put/2`:
+%
+% ```
+% ?- bb_put(city, "Valladolid"), (bb_put(city, "Salamanca"), false);(bb_get(city, X)).
+%    X = "Salamanca".
+% ?- bb_put(city, "Valladolid"), (bb_b_put(city, "Salamanca"), false);(bb_get(city, X)).
+%    X = "Valladolid".
+% ```
 bb_put(Key, Value) :-
     (  atom(Key) ->
        '$store_global_var'(Key, Value)
     ;  type_error(atom, Key, bb_put/2)
     ).
 
-%% backtrackable global variables.
+% backtrackable global variables.
 
+%% bb_b_put(+Key, +Value).
+%
+% Sets a global variable named Key (must be an atom) with value Value.
+% The global variable is backtrackable. Check `bb_put/2` for the
+% non-backtrackable version.
+%
+% ```
+% ?- bb_b_put(city, "Valladolid").
+%    true.
+% ?- bb_get(city, X).
+%    X = "Valladolid".
+% ```
+%
+% In this example one can understand the difference between `bb_put/2` and
+% `bb_b_put/2`:
+%
+% ```
+% ?- bb_put(city, "Valladolid"), (bb_put(city, "Salamanca"), false);(bb_get(city, X)).
+%    X = "Salamanca".
+% ?- bb_put(city, "Valladolid"), (bb_b_put(city, "Salamanca"), false);(bb_get(city, X)).
+%    X = "Valladolid".
+% ```
 bb_b_put(Key, Value) :-
     (  atom(Key) ->
        '$store_backtrackable_global_var'(Key, Value)
     ;  type_error(atom, Key, bb_b_put/2)
     ).
 
+%% bb_get(+Key, -Value).
+%
+% Gets the value Value of a global variable named Key (must be an atom)
 bb_get(Key, Value) :-
     (  atom(Key) ->
        '$fetch_global_var'(Key, Value)
@@ -48,21 +116,51 @@ bb_get(Key, Value) :-
     ).
 
 
+%% succ(?I, ?S).
+%
+% True iff S is the successor of the non-negative integer I.
+% At least one of the arguments must be instantiated.
+
+succ(I, S) :-
+    can_be(not_less_than_zero, I),
+    can_be(not_less_than_zero, S),
+    (   integer(S) ->
+        S > 0,
+        I is S-1
+    ;   integer(I) ->
+        S is I+1
+    ;   instantiation_error(succ/2)
+    ).
+
+
 % setup_call_cleanup.
 
 :- meta_predicate(call_cleanup(0, 0)).
 
+%% call_cleanup(Goal, Cleanup).
+%
+% Executes Goal and then, either on success or failure, executes Cleanup.
+% The success or failure of Cleanup is ignored and choice points created inside are destroyed.
 call_cleanup(G, C) :- setup_call_cleanup(true, G, C).
 
 :- meta_predicate(setup_call_cleanup(0, 0, 0)).
 
 :- non_counted_backtracking setup_call_cleanup/3.
 
+%% setup_call_cleanup(Setup, Goal, Cleanup).
+%
+% If Setup succeeds, Cleanup will be called after the execution of Goal. Goal itself can succeed or not.
+%
+% In this example, we use the predicate to always close an open file:
+%
+% ```
+% ?- setup_call_cleanup(open(File, read, Stream), do_something_with_stream(Stream), close(Stream)).
+% ```
 setup_call_cleanup(S, G, C) :-
     '$get_b_value'(B),
     '$call_with_inference_counting'(call(S)),
     '$set_cp_by_default'(B),
-    '$get_current_block'(Bb),
+    '$get_current_scc_block'(Bb),
     (  C = _:CC,
        var(CC) ->
        instantiation_error(setup_call_cleanup/3)
@@ -75,17 +173,16 @@ setup_call_cleanup(S, G, C) :-
 
 scc_helper(C, G, Bb) :-
     '$get_cp'(Cp),
-    '$install_scc_cleaner'(C, NBb),
+    '$install_scc_cleaner'(C),
     '$call_with_inference_counting'(call(G)),
     (  '$check_cp'(Cp) ->
-       '$reset_block'(Bb),
+       '$reset_scc_block'(Bb),
        run_cleaners_without_handling(Cp)
     ;  true
-    ;  '$reset_block'(NBb),
-       '$fail'
+    ;  '$fail'
     ).
 scc_helper(_, _, Bb) :-
-    '$reset_block'(Bb),
+    '$reset_scc_block'(Bb),
     '$push_ball_stack',
     run_cleaners_with_handling,
     '$pop_from_ball_stack',
@@ -99,7 +196,7 @@ scc_helper(_, _, _) :-
 
 run_cleaners_with_handling :-
     '$get_scc_cleaner'(C),
-    '$get_level'(B),
+    '$get_cp'(B),
     catch(C, _, true),
     '$set_cp_by_default'(B),
     run_cleaners_with_handling.
@@ -110,7 +207,7 @@ run_cleaners_with_handling :-
 
 run_cleaners_without_handling(Cp) :-
     '$get_scc_cleaner'(C),
-    '$get_level'(B),
+    '$get_cp'(B),
     call(C),
     '$set_cp_by_default'(B),
     run_cleaners_without_handling(Cp).
@@ -120,30 +217,14 @@ run_cleaners_without_handling(Cp) :-
 
 % call_with_inference_limit
 
-:- non_counted_backtracking end_block/4.
-
-end_block(_, Bb, NBb, _L) :-
-    '$clean_up_block'(NBb),
-    '$reset_block'(Bb).
-end_block(B, _Bb, NBb, L) :-
-    '$install_inference_counter'(B, L, _),
-    '$reset_block'(NBb),
-    '$fail'.
-
-:- non_counted_backtracking handle_ile/3.
-
-handle_ile(B, inference_limit_exceeded(B), inference_limit_exceeded) :-
-    !,
-    '$pop_ball_stack'.
-handle_ile(B, _, _) :-
-    '$remove_call_policy_check'(B),
-    '$pop_from_ball_stack',
-    '$unwind_stack'.
-
 :- meta_predicate(call_with_inference_limit(0, ?, ?)).
 
 :- non_counted_backtracking call_with_inference_limit/3.
 
+%% call_with_inference_limit(Goal, Limit, Result).
+%
+% Similar to `call(Goal)` but it limits the number of inferences for each solution of Goal.
+% Calls to it may be nested, but only the last limit will be in power.
 call_with_inference_limit(G, L, R) :-
     (  integer(L) ->
        (  L < 0 ->
@@ -159,8 +240,6 @@ call_with_inference_limit(G, L, R) :-
     call_with_inference_limit(G, L, R, Bb, B),
     '$remove_call_policy_check'(B).
 
-install_inference_counter(B, L, Count0) :-
-    '$install_inference_counter'(B, L, Count0).
 
 :- meta_predicate(call_with_inference_limit(0,?,?,?,?)).
 
@@ -168,36 +247,57 @@ install_inference_counter(B, L, Count0) :-
 
 call_with_inference_limit(G, L, R, Bb, B) :-
     '$install_new_block'(NBb),
-    '$install_inference_counter'(B, L, Count0),
+    '$install_inference_counter'(NBb, L, Count0),
     '$call_with_inference_counting'(call(G)),
     '$inference_level'(R, B),
-    '$remove_inference_counter'(B, Count1),
+    '$remove_inference_counter'(NBb, Count1),
     Diff is L - (Count1 - Count0),
-    end_block(B, Bb, NBb, Diff).
-call_with_inference_limit(_, _, R, Bb, B) :-
-    '$reset_block'(Bb),
-    '$remove_inference_counter'(B, _),
-    (  '$get_ball'(Ball),
-       '$push_ball_stack',
-       '$get_level'(Cp),
-       '$set_cp_by_default'(Cp)
-    ;  '$remove_call_policy_check'(B),
+    (  '$clean_up_block'(NBb),
+       '$reset_block'(Bb)
+    ;  '$install_inference_counter'(NBb, Diff, _),
+       '$reset_block'(NBb),
        '$fail'
+    ).
+call_with_inference_limit(_, _, R, Bb, B) :-
+    (  '$inference_limit_exceeded' ->
+       R = inference_limit_exceeded
+    ;  true
     ),
-    handle_ile(B, Ball, R).
-
-partial_string(String, L, L0) :-
-    (  String == [] ->
-       L = L0
-    ;  catch(atom_chars(Atom, String),
-             error(E, _),
-             throw(error(E, partial_string/3))),
-       '$create_partial_string'(Atom, L, L0)
+    '$get_current_block'(NBb),
+    '$remove_inference_counter'(NBb, _),
+    '$reset_block'(Bb),
+    '$remove_call_policy_check'(B),
+    (  '$get_ball'(_),
+       '$push_ball_stack',
+       '$get_cp'(Cp),
+       '$set_cp_by_default'(Cp),
+       '$pop_from_ball_stack',
+       '$unwind_stack'
+    ;  nonvar(R)
     ).
 
+%% partial_string(String, Ls0, Ls)
+%
+% Explicitly construct a partial string "manually". It can be used as an optimized append/3.
+% It's not recommended to use this predicate in application code.
+partial_string(String, Ls0, Ls) :-
+    must_be(chars, String),
+    (  String == [] ->
+       Ls0 = Ls
+    ;  '$create_partial_string'(String, Ls0, Ls)
+    ).
+
+%% partial_string(+String)
+%
+% Succeeds if String is a _partial string_. A partial string is a string composed of several smaller
+% strings, even just one. That means all strings in Scryer are partial strings.
 partial_string(String) :-
     '$is_partial_string'(String).
 
+%% partial_string_tail(+String, -Tail).
+%
+% Unifies Tail with the last section of the partial string.
+% It's not recommended to use this predicate in application code.
 partial_string_tail(String, Tail) :-
     (  partial_string(String) ->
        '$partial_string_tail'(String, Tail)
@@ -209,6 +309,9 @@ partial_string_tail(String, Tail) :-
 
 :- meta_predicate(call_nth(0, ?)).
 
+%% call_nth(Goal, N).
+%
+% Succeeds when Goal succeeded for the Nth time (there are at least N solutions)
 call_nth(Goal, N) :-
     can_be(integer, N),
     (   integer(N) ->
@@ -246,20 +349,60 @@ call_nth_nesting(C, ID) :-
     bb_put(ID, 0),
     bb_put(i_call_nth_counter, C).
 
+%% countall(G_0, N).
+%
+% countall(G_0, N) is true iff N unifies with the total number of
+% answers of call(G_0).
 
+:- meta_predicate(countall(0, ?)).
+
+countall(Goal, N) :-
+    can_be(integer, N),
+    (   integer(N) ->
+        (   N < 0 ->
+            domain_error(not_less_than_zero, N, countall/2)
+        ;   true
+        )
+    ;   true
+    ),
+    setup_call_cleanup(call_nth_nesting(C, ID),
+                       (   (   Goal,
+                               bb_get(ID, N0),
+                               N1 is N0 + 1,
+                               bb_put(ID, N1),
+                               false
+                           ;   bb_get(ID, N)
+                           )
+                       ),
+                       (   bb_get(i_call_nth_counter, C) ->
+                           C1 is C - 1,
+                           bb_put(i_call_nth_counter, C1)
+                       ;   true
+                       )).
+
+%% copy_term_nat(Source, Dest)
+%
+% Similar to `copy_term/2` but without attribute variables
 copy_term_nat(Source, Dest) :-
     '$copy_term_without_attr_vars'(Source, Dest).
 
+%% copy_term(+Term, -Copy, -Gs).
+%
+% Produce a deep copy of Term and unify it to Copy, without attributes.
+% Unify Gs with a list of goals that represent the attributes of Term.
+% Similar to `copy_term/2` but splitting the attributes.
+copy_term(Term, Copy, Gs) :-
+   can_be(list, Gs),
+   findall(Term-Rs, '$project_atts':term_residual_goals(Term,Rs), [Copy-Gs]),
+   (  var(Gs) ->
+      Gs = []
+   ;  true
+   ).
 
-asserta(Module, (Head :- Body)) :-
-    !,
-    '$asserta'(Module, Head, Body).
-asserta(Module, Fact) :-
-    '$asserta'(Module, Fact, true).
-
-assertz(Module, (Head :- Body)) :-
-    !,
-    '$assertz'(Module, Head, Body).
-assertz(Module, Fact) :-
-    '$assertz'(Module, Fact, true).
+:- meta_predicate(call_residue_vars(0, ?)).
 
+call_residue_vars(Goal, Vars) :-
+    can_be(list, Vars),
+    '$get_attr_var_queue_delim'(B),
+    call(Goal),
+    '$get_attr_var_queue_beyond'(B, Vars).

src/lib/lambda.pl

@@ -50,11 +50,13 @@ programming based on call/N.
 Lambda expressions are represented by ordinary Prolog terms.
 There are two kinds of lambda expressions:
 
+```
     Free+\X1^X2^ ..^XN^Goal
 
          \X1^X2^ ..^XN^Goal
+```
 
-The second is a shorthand for t+\X1^X2^..^XN^Goal.
+The second is a shorthand for `t+\X1^X2^..^XN^Goal`.
 
 Xi are the parameters.
 
@@ -70,20 +72,20 @@ currently not checked. Violations may lead to unexpected bindings.
 
 In the following example the parentheses around X>3 are necessary.
 
-==
+```
 ?- use_module(library(lambda)).
 ?- use_module(library(lists)).
 
 ?- maplist(\X^(X>3),[4,5,9]).
    true.
-==
+```
 
 In the following X is a variable that is shared by both instances of
 the lambda expression. The second query illustrates the cooperation of
 continuations and lambdas. The lambda expression is in this case a
 continuation expecting a further argument.
 
-==
+```
 ?- use_module(library(dif)).
    true.
 
@@ -92,11 +94,12 @@ continuation expecting a further argument.
 
 ?- Xs = [A,B], maplist(X+\dif(X), Xs).
    Xs = [A,B], dif:dif(X,A), dif:dif(X,B).
-==
+```
 
 The following queries are all equivalent. To see this, use
-the fact f(x,y).
-==
+the fact `f(x,y)`.
+
+```
 ?- call(f,A1,A2).
 ?- call(\X^f(X),A1,A2).
 ?- call(\X^Y^f(X,Y), A1,A2).
@@ -105,41 +108,41 @@ the fact f(x,y).
 ?- call(f(A1),A2).
 ?- f(A1,A2).
    A1 = x, A2 = y.
-==
+```
 
 Further discussions
-http://www.complang.tuwien.ac.at/ulrich/Prolog-inedit/ISO-Hiord
+[http://www.complang.tuwien.ac.at/ulrich/Prolog-inedit/ISO-Hiord](http://www.complang.tuwien.ac.at/ulrich/Prolog-inedit/ISO-Hiord)
 
 @tbd Static expansion similar to apply_macros.
 @author Ulrich Neumerkel
 */
 
-:- meta_predicate ^(?,0,?).
-:- meta_predicate ^(?,1,?,?).
-:- meta_predicate ^(?,2,?,?,?).
-:- meta_predicate ^(?,3,?,?,?,?).
-:- meta_predicate ^(?,4,?,?,?,?,?).
-:- meta_predicate ^(?,5,?,?,?,?,?,?).
-:- meta_predicate ^(?,6,?,?,?,?,?,?,?).
-:- meta_predicate ^(?,7,?,?,?,?,?,?,?,?).
-:- meta_predicate \(0).
-:- meta_predicate \(1,?).
-:- meta_predicate \(2,?,?).
-:- meta_predicate \(3,?,?,?).
-:- meta_predicate \(4,?,?,?,?).
-:- meta_predicate \(5,?,?,?,?,?).
-:- meta_predicate \(6,?,?,?,?,?,?).
-:- meta_predicate \(7,?,?,?,?,?,?,?).
-:- meta_predicate +\(?,0).
-:- meta_predicate +\(?,1,?).
-:- meta_predicate +\(?,2,?,?).
-:- meta_predicate +\(?,3,?,?,?).
-:- meta_predicate +\(?,4,?,?,?,?).
-:- meta_predicate +\(?,5,?,?,?,?,?).
-:- meta_predicate +\(?,6,?,?,?,?,?,?).
-:- meta_predicate +\(?,7,?,?,?,?,?,?,?).
+:- meta_predicate(^(?,0,?)).
+:- meta_predicate(^(?,1,?,?)).
+:- meta_predicate(^(?,2,?,?,?)).
+:- meta_predicate(^(?,3,?,?,?,?)).
+:- meta_predicate(^(?,4,?,?,?,?,?)).
+:- meta_predicate(^(?,5,?,?,?,?,?,?)).
+:- meta_predicate(^(?,6,?,?,?,?,?,?,?)).
+:- meta_predicate(^(?,7,?,?,?,?,?,?,?,?)).
+:- meta_predicate(\(0)).
+:- meta_predicate(\(1,?)).
+:- meta_predicate(\(2,?,?)).
+:- meta_predicate(\(3,?,?,?)).
+:- meta_predicate(\(4,?,?,?,?)).
+:- meta_predicate(\(5,?,?,?,?,?)).
+:- meta_predicate(\(6,?,?,?,?,?,?)).
+:- meta_predicate(\(7,?,?,?,?,?,?,?)).
+:- meta_predicate(+\(?,0)).
+:- meta_predicate(+\(?,1,?)).
+:- meta_predicate(+\(?,2,?,?)).
+:- meta_predicate(+\(?,3,?,?,?)).
+:- meta_predicate(+\(?,4,?,?,?,?)).
+:- meta_predicate(+\(?,5,?,?,?,?,?)).
+:- meta_predicate(+\(?,6,?,?,?,?,?,?)).
+:- meta_predicate(+\(?,7,?,?,?,?,?,?,?)).
 
-:- meta_predicate no_hat_call(0).
+:- meta_predicate(no_hat_call(0)).
 
 ^(V1,C_0,V1) :-
    no_hat_call(C_0).

src/lib/lists.pl

@@ -1,3 +1,7 @@
+/**
+List manipulation predicates
+*/
+
 :- module(lists, [member/2, select/3, append/2, append/3, foldl/4, foldl/5,
 		          memberchk/2, reverse/2, length/2, maplist/2,
 		          maplist/3, maplist/4, maplist/5, maplist/6,
@@ -38,17 +42,17 @@
 :- use_module(library(error)).
 
 
-:- meta_predicate maplist(1, ?).
-:- meta_predicate maplist(2, ?, ?).
-:- meta_predicate maplist(3, ?, ?, ?).
-:- meta_predicate maplist(4, ?, ?, ?, ?).
-:- meta_predicate maplist(5, ?, ?, ?, ?, ?).
-:- meta_predicate maplist(6, ?, ?, ?, ?, ?, ?).
-:- meta_predicate maplist(7, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate maplist(8, ?, ?, ?, ?, ?, ?, ?, ?).
+:- meta_predicate(maplist(1, ?)).
+:- meta_predicate(maplist(2, ?, ?)).
+:- meta_predicate(maplist(3, ?, ?, ?)).
+:- meta_predicate(maplist(4, ?, ?, ?, ?)).
+:- meta_predicate(maplist(5, ?, ?, ?, ?, ?)).
+:- meta_predicate(maplist(6, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(maplist(7, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(maplist(8, ?, ?, ?, ?, ?, ?, ?, ?)).
 
-:- meta_predicate foldl(3, ?, ?, ?).
-:- meta_predicate foldl(4, ?, ?, ?, ?).
+:- meta_predicate(foldl(3, ?, ?, ?)).
+:- meta_predicate(foldl(4, ?, ?, ?, ?)).
 
 :- use_module(library(error)).
 
@@ -57,6 +61,20 @@
 resource_error(Resource, Context) :-
    throw(error(resource_error(Resource), Context)).
 
+%% length(?Xs, ?N).
+%
+% Relates a list to its length (number of elements). It can be used to count the elements of a current list or
+% to create a list full of free variables with N length.
+%
+% ```
+% ?- length("abc", 3).
+%    true.
+% ?- length("abc", N).
+%    N = 3.
+% ?- length(Xs, 3).
+%    Xs = [_A,_B,_C].
+% ```
+
 length(Xs0, N) :-
    '$skip_max_list'(M, N, Xs0,Xs),
    !,
@@ -74,7 +92,7 @@ length(_, N) :-
 
 length_rundown(Xs, 0) :- !, Xs = [].
 length_rundown(Vs, N) :-
-    \+ \+ '$project_atts':copy_term(Vs,Vs,[]), % unconstrained
+    '$unattributed_var'(Vs), % unconstrained
     !,
     '$det_length_rundown'(Vs, N).
 length_rundown([_|Xs], N) :- % force unification
@@ -82,7 +100,7 @@ length_rundown([_|Xs], N) :- % force unification
     length(Xs, N1). % maybe some new info on Xs
 
 failingvarskip(Xs) :-
-    \+ \+ '$project_atts':copy_term(Xs,Xs,[]), % unconstrained
+    '$unattributed_var'(Xs), % unconstrained
     !.
 failingvarskip([_|Xs0]) :- % force unification
     '$skip_max_list'(_, _, Xs0,Xs),
@@ -95,28 +113,71 @@ length_addendum([_|Xs], N, M) :-
     M1 is M + 1,
     length_addendum(Xs, N, M1).
 
+%% member(?X, ?Xs).
+%
+% Succeeds when X unifies with an item of the list Xs, which can be at any position.
+%
+% ```
+% ?- member(X, "hello world").
+%    X = h
+% ;  ... .
+% ```
 
-member(X, [X|_]).
-member(X, [_|Xs]) :- member(X, Xs).
+member(X, [L|Ls]) :-
+        member_(Ls, L, X).
 
+member_(_, X, X).
+member_([L|Ls], _, X) :-
+        member_(Ls, L, X).
 
+%% select(X, Xs0, Xs1).
+%
+% Succeeds when the list Xs1 is the list Xs0 without the item X
+%
+% ```
+% ?- select(c, "abcd", X).
+%    X = "abd"
+% ;  false.
+% ```
 select(X, [X|Xs], Xs).
 select(X, [Y|Xs], [Y|Ys]) :- select(X, Xs, Ys).
 
-
+%% append(+XsXs, ?Xs).
+%
+% Concatenates a list of lists
+%
+% ```
+% ?- append([[1, 2], [3]], Xs).
+%    Xs = [1,2,3].
+% ```
 append([], []).
 append([L0|Ls0], Ls) :-
     append(L0, Rest, Ls),
     append(Ls0, Rest).
 
-
+%% append(Xs0, Xs1, Xs).
+%
+% List Xs is the concatenation of Xs0 and Xs1
+%
+% ```
+% ?- append([1,2,3], [4,5,6], Xs).
+%    Xs = [1,2,3,4,5,6].
+% ```
 append([], R, R).
 append([X|L], R, [X|S]) :- append(L, R, S).
 
-
+%% memberchk(?X, +Xs).
+%
+% This predicate is similar to `member/2`, but it only provides a single answer
 memberchk(X, Xs) :- member(X, Xs), !.
 
-
+%% reverse(?Xs, ?Ys).
+%
+% Xs is the Ys list in reverse order
+%
+%     ?- reverse([1,2,3], [3,2,1]).
+%        true.
+%
 reverse(Xs, Ys) :-
     (  nonvar(Xs) -> reverse(Xs, Ys, [], Xs)
     ;  reverse(Ys, Xs, [], Ys)
@@ -126,81 +187,141 @@ reverse([], [], YsRev, YsRev).
 reverse([_|Xs], [Y1|Ys], YsPreludeRev, Xss) :-
     reverse(Xs, Ys, [Y1|YsPreludeRev], Xss).
 
+%% maplist(+Predicate, ?Xs0).
+%
+% This is a metapredicate that applies predicate to each element of the list Xs0
+%
+% ```
+% ?- maplist(write, [1,2,3]).
+% 123   true.
+% ```
 maplist(_, []).
 maplist(Cont1, [E1|E1s]) :-
     call(Cont1, E1),
     maplist(Cont1, E1s).
 
+%% maplist(+Predicate, ?Xs0, ?Xs1).
+%
+% This is a metapredicate that applies predicate to each element of the lists Xs0 and Xs1.
+%
+% ```
+% ?- maplist(length, ["hello", "prolog", "marseille"], Xs1).
+%    Xs1 = [5,6,9].
+% ```
 maplist(_, [], []).
 maplist(Cont2, [E1|E1s], [E2|E2s]) :-
     call(Cont2, E1, E2),
     maplist(Cont2, E1s, E2s).
 
+%% maplist(+Predicate, ?Xs0, ?Xs1, ?Xs2).
+%
+% This is a metapredicate that applies predicate to each element of the lists Xs0, Xs1 and Xs2.
 maplist(_, [], [], []).
 maplist(Cont3, [E1|E1s], [E2|E2s], [E3|E3s]) :-
     call(Cont3, E1, E2, E3),
     maplist(Cont3, E1s, E2s, E3s).
 
+%% maplist(+Predicate, ?Xs0, ?Xs1, ?Xs2, ?Xs3).
+%
+% This is a metapredicate that applies predicate to each element of the lists Xs0, Xs1, Xs2 and Xs3.
 maplist(_, [], [], [], []).
 maplist(Cont, [E1|E1s], [E2|E2s], [E3|E3s], [E4|E4s]) :-
     call(Cont, E1, E2, E3, E4),
     maplist(Cont, E1s, E2s, E3s, E4s).
 
-
+%% maplist(+Predicate, ?Xs0, ?Xs1, ?Xs2, ?Xs3, ?Xs4).
+%
+% This is a metapredicate that applies predicate to each element of the lists Xs0, Xs1, Xs2, Xs3 and Xs4.
 maplist(_, [], [], [], [], []).
 maplist(Cont, [E1|E1s], [E2|E2s], [E3|E3s], [E4|E4s], [E5|E5s]) :-
     call(Cont, E1, E2, E3, E4, E5),
     maplist(Cont, E1s, E2s, E3s, E4s, E5s).
 
-
+%% maplist(+Predicate, ?Xs0, ?Xs1, ?Xs2, ?Xs3, ?Xs4, ?Xs5).
+%
+% This is a metapredicate that applies predicate to each element of the lists Xs0, Xs1, Xs2, Xs3, Xs4 and Xs5.
 maplist(_, [], [], [], [], [], []).
 maplist(Cont, [E1|E1s], [E2|E2s], [E3|E3s], [E4|E4s], [E5|E5s], [E6|E6s]) :-
     call(Cont, E1, E2, E3, E4, E5, E6),
     maplist(Cont, E1s, E2s, E3s, E4s, E5s, E6s).
 
-
+%% maplist(+Predicate, ?Xs0, ?Xs1, ?Xs2, ?Xs3, ?Xs4, ?Xs5, ?Xs6).
+%
+% This is a metapredicate that applies predicate to each element of the lists Xs0, Xs1, Xs2, Xs3, Xs4, Xs5 and Xs6.
 maplist(_, [], [], [], [], [], [], []).
 maplist(Cont, [E1|E1s], [E2|E2s], [E3|E3s], [E4|E4s], [E5|E5s], [E6|E6s], [E7|E7s]) :-
     call(Cont, E1, E2, E3, E4, E5, E6, E7),
     maplist(Cont, E1s, E2s, E3s, E4s, E5s, E6s, E7s).
 
-
+%% maplist(+Predicate, ?Xs0, ?Xs1, ?Xs2, ?Xs3, ?Xs4, ?Xs5, ?Xs6, ?Xs7).
+%
+% This is a metapredicate that applies predicate to each element of the lists Xs0, Xs1, Xs2, Xs3, Xs4, Xs5, Xs6 and Xs7.
 maplist(_, [], [], [], [], [], [], [], []).
 maplist(Cont, [E1|E1s], [E2|E2s], [E3|E3s], [E4|E4s], [E5|E5s], [E6|E6s], [E7|E7s], [E8|E8s]) :-
     call(Cont, E1, E2, E3, E4, E5, E6, E7, E8),
     maplist(Cont, E1s, E2s, E3s, E4s, E5s, E6s, E7s, E8s).
 
-
+%% sum_list(+Xs, -Sum).
+%
+% Takes a lists of numbers and unifies Sum with the result of summing all the elements of the list.
+%
+% ```
+% ?- sum_list([2,2,2], 6).
+%    true.
+% ```
 sum_list(Ls, S) :-
         foldl(lists:sum_, Ls, 0, S).
 
 sum_(L, S0, S) :- S is S0 + L.
 
 
-
+%% same_length(?Xs, ?Ys).
+%
+% Succeeds if Xs and Ys are lists of the same length
 same_length([], []).
 same_length([_|As], [_|Bs]) :-
         same_length(As, Bs).
 
+%% foldl(+Predicate, ?Ls, +A0, ?A).
+%
+% foldl, sometimes called reduce, is a metapredicate that takes a predicate, a list of items
+% and a starting value, and outputs a single value. The predicate _Predicate_ must be able to take the current
+% element of the list, the previous value of the computation and the next value of the computation.
+%
+% For example, if we define sum_ as:
+%
+% ```
+% sum_(L, S0, S) :- S is S0 + L.
+% ```
+%
+% Then we can define `sum_list/2` as the following:
+%
+% ```
+% sum_list(Ls, S) :- foldl(sum_, Ls, 0, S).
+% ```
 
-foldl(Goal_3, Ls, A0, A) :-
-        foldl_(Ls, Goal_3, A0, A).
-
-foldl_([], _, A, A).
-foldl_([L|Ls], G_3, A0, A) :-
+foldl(_, [], A, A).
+foldl(G_3, [L|Ls], A0, A) :-
         call(G_3, L, A0, A1),
-        foldl_(Ls, G_3, A1, A).
+        foldl(G_3, Ls, A1, A).
 
+%% foldl(+Predicate, ?Ls0, ?Ls1, +A0, ?A).
+%
+% Same as `foldl/4` but with an extra list
 
-foldl(Goal_4, Xs, Ys, A0, A) :-
-        foldl_(Xs, Ys, Goal_4, A0, A).
-
-
-foldl_([], [], _, A, A).
-foldl_([X|Xs], [Y|Ys], G_4, A0, A) :-
+foldl(_, [], [], A, A).
+foldl(G_4, [X|Xs], [Y|Ys], A0, A) :-
         call(G_4, X, Y, A0, A1),
-        foldl_(Xs, Ys, G_4, A1, A).
+        foldl(G_4, Xs, Ys, A1, A).
 
+%% transpose(?Ls, ?Ts).
+%
+% If Ls is a list of lists, Ts contains the transposition
+%
+% ```
+% ?- transpose([[1,1],[2,2]], Ts).
+%    Ts = [[1,2],[1,2]].
+% ```
 transpose(Ls, Ts) :-
         lists_transpose(Ls, Ts).
 
@@ -214,7 +335,14 @@ transpose_(_, Fs, Lists0, Lists) :-
 
 list_first_rest([L|Ls], L, Ls).
 
-
+%% list_to_set(+Ls0, -Set).
+%
+% Takes a list Ls0 and returns a list Set that doesn't contain any repeated element
+%
+% ```
+% ?- list_to_set([2,3,4,4,1,2], Set).
+%    Set = [2,3,4,1].
+% ```
 list_to_set(Ls0, Ls) :-
         maplist(lists:with_var, Ls0, LVs0),
         keysort(LVs0, LVs),
@@ -242,7 +370,14 @@ unify_same(E-V, Prev-Var, E-V) :-
         ;   true
         ).
 
-
+%% nth0(?N, ?Ls, ?E).
+%
+% Succeeds if in the N position of the list Ls, we found the element E. The elements start counting from zero.
+%
+% ```
+% ?- nth0(2, [1,2,3,4], 3).
+%    true.
+% ```
 nth0(N, Es0, E) :-
    nonvar(N),
    '$skip_max_list'(Skip, N, Es0,Es1),
@@ -261,7 +396,6 @@ nth0(N, Es0, E) :-
 
 skipn(N0, Es0,Es) :-
    N0>0,
-   !, % should not be necessary #1028
    N1 is N0-1,
    Es0 = [_|Es1],
    skipn(N1, Es1,Es).
@@ -277,6 +411,14 @@ nth0_el(N0,N, _,E, [E0|Es0]) :-
    N1 is N0+1,
    nth0_el(N1,N, E0,E, Es0).
 
+%% nth1(?N, ?Ls, ?E).
+%
+% Succeeds if in the N position of the list Ls, we found the element E. The elements start counting from one.
+%
+% ```
+% ?- nth1(2, [1,2,3,4], 2).
+%    true.
+% ```
 nth1(N, Es0, E) :-
    N \== 0,
    nth0(N, [_|Es0], E),
@@ -284,13 +426,20 @@ nth1(N, Es0, E) :-
 
 skipn(N0, Es0,Es, Xs0,Xs) :-
    N0>0,
-   !, % should not be necessary #1028
    N1 is N0-1,
    Es0 = [E|Es1],
    Xs0 = [E|Xs1],
    skipn(N1, Es1,Es, Xs1,Xs).
 skipn(0, Es,Es, Xs,Xs).
 
+%% nth0(?N, ?Ls, ?E, ?Rs).
+%
+% Succeeds if in the N position of the list Ls, we found the element E and the rest of the list is Rs. The elements start counting from zero.
+%
+% ```
+% ?- nth0(2, [1,2,3,4], 3, [1,2,4]).
+%    true.
+% ```
 nth0(N, Es0, E, Es) :-
    integer(N),
    N >= 0,
@@ -315,45 +464,58 @@ nth0_elx(N0,N, E0,E, [E1|Es0], [E0|Es]) :-
 
 % p.p.8.5
 
+%% nth1(?N, ?Ls, ?E, ?Rs).
+%
+% Succeeds if in the N position of the list Ls, we found the element E and the rest of the list is Rs. The elements start counting from one.
+%
+% ```
+% ?- nth1(2, [1,2,3,4], 2, [1,3,4]).
+%    true.
+% ```
 nth1(N, Es0, E, Es) :-
    N \== 0,
    nth0(N, [_|Es0], E, [_|Es]),
    N \== 0.
 
-
+%% list_max(+Xs, -Max).
+%
+% Takes a list Xs and unifies with the maximum value of the list
 list_max([N|Ns], Max) :-
     foldl(lists:list_max_, Ns, N, Max).
 
 list_max_(N, Max0, Max) :-
     Max is max(N, Max0).
 
+%% list_min(+Xs, -Min).
+%
+% Takes a list Xs and unifies with the minimum value of the list
 list_min([N|Ns], Min) :-
     foldl(lists:list_min_, Ns, N, Min).
 
 list_min_(N, Min0, Min) :-
     Min is min(N, Min0).
 
-%!  permutation(?Xs, ?Ys) is nondet.
+%% permutation(?Xs, ?Ys) is nondet.
 %
-%   True when Xs is a permutation of Ys. This can solve for Ys given
-%   Xs or Xs given Ys, or  even   enumerate  Xs and Ys together. The
-%   predicate  permutation/2  is  primarily   intended  to  generate
-%   permutations. Note that a list of  length N has N! permutations,
-%   and  unbounded  permutation  generation   becomes  prohibitively
-%   expensive, even for rather short lists (10! = 3,628,800).
+% True when Xs is a permutation of Ys. This can solve for Ys given
+% Xs or Xs given Ys, or  even   enumerate  Xs and Ys together. The
+% predicate  `permutation/2`  is  primarily   intended  to  generate
+% permutations. Note that a list of  length N has N! permutations,
+% and  unbounded  permutation  generation   becomes  prohibitively
+% expensive, even for rather short lists (10! = 3,628,800).
 %
-%   The example below illustrates that Xs   and Ys being proper lists
-%   is not a sufficient condition to use the above replacement.
+% The example below illustrates that Xs   and Ys being proper lists
+% is not a sufficient condition to use the above replacement.
 %
-%     ==
-%     ?- permutation([1,2], [X,Y]).
-%     X = 1, Y = 2 ;
-%     X = 2, Y = 1 ;
-%     false.
-%     ==
+% ```
+% ?- permutation([1,2], [X,Y]).
+%    X = 1, Y = 2
+% ;  X = 2, Y = 1
+% ;  false.
+% ```
 %
-%   @error  type_error(list, Arg) if either argument is not a proper
-%           or partial list.
+%   Throws `type_error(list, Arg)` if either argument is not a proper
+%   or partial list.
 
 permutation(Xs, Ys) :-
     '$skip_max_list'(Xlen, _, Xs, XTail),

src/lib/numerics/quadtests.pl

@@ -0,0 +1,187 @@
+% Efforts toward literate tests with quads
+ 
+:- use_module(library(iso_ext)).
+:- use_module(library(pio)).
+:- use_module(library(lists)).
+:- use_module(library(dcgs)).
+:- use_module(library(format)).
+:- use_module(library(reif)).
+:- use_module(library(debug)).
+:- use_module(library(lambda)).
+:- use_module(library(error)).
+:- use_module(library(time)).
+
+:- use_module(testutils).
+:- use_module(special_functions).
+
+portray_term(Stream) :-
+    read_term(Stream, Term, []),
+    portray_clause(Term).
+
+?- check_module_quads(special_functions, _).
+% Checking 11 quads ..
+% CHECKING.. (?-A=0.6174468790806071,erf(A,A),B is-A,erf(B,B)).
+% CHECKING.. (?-try_falsify(odd_t(erf,real(A)))).
+% CHECKING.. (?-witness(odd_t(erf,real(A),false))).
+% CHECKING.. (?-witness((real(A),erf(A,B),erf(-A,C),abs(B+C)>0))).
+% CHECKING.. (?-length(A,B)).
+% CHECKING.. (?-real(A),erf(A,B),erfc(A,C),abs(B+C-1)>epsilon).
+% CHECKING.. (?-try_falsify(δ_inverses_t(40*epsilon,erf,inverf,interval(-2,2,A)))).
+% CHECKING.. (?-try_falsify(δ_inverses_t(40*epsilon,erfc,inverfc,interval(-2,2,A)))).
+% CHECKING.. (?-A=10,B is A+1,gamma(B,C),int_realfact(A,D)).
+% CHECKING.. (?-gamma_P_Q(1.2,2.3,A,B),abs(A+B-1)<epsilon).
+% CHECKING.. (?-A=1.5,B=0.7,invgammp(A,B,C),gamma_P_Q(A,C,D,E),abs(B-D)<epsilon).
+   true.
+
+check_module_quads(Module, Quads) :-
+    use_module(Module),
+    read_quads(Module, Quads),
+    zip(Qs, ADs, Quads),
+    length(Qs, NQ),
+    format("% Checking ~d quads ..~n", [NQ]),
+    maplist(check_qu_ad, Qs, ADs).
+
+read_quads(Module, Quads) :-
+    module_terms(Module, Terms),
+    terms_quads(Terms, Quads).
+
+module_terms(Module, Terms) :-
+    module_file(Module, File),
+    setup_call_cleanup(
+        open(File, read, Stream, [type(text)]),
+        read_terms(Stream, Terms),
+        close(Stream)
+    ).
+
+module_file(Module, File) :- atom_concat(Module, '.pl', File).
+
+% Given a list of terms, filter out the predicate clauses.
+% TODO: Arg 1 is really a list of Term-VarNames _pairs_;
+%       it would be very nice to find a less unsightly
+%       name than 'TermVN' for these!
+terms_quads([Term|Terms], Quads) :-
+    (   term_type(Term, clause) -> terms_quads(Terms, Quads)
+    ;   Quads = [Term|Quads_],
+        terms_quads(Terms, Quads_)
+    ).
+terms_quads([], []).
+
+term_type(Term-_, Type) :-
+    (   Term = (?- _) -> Type = query
+    ;   Term = (_,_) -> Type = answer_description
+    ;   Term = (_;_) -> Type = answer_description
+    ;   Term == true -> Type = answer_description
+    ;   Term == false -> Type = answer_description
+    ;   Type = clause
+    ).
+
+?- term_type(test("erf is odd",try_falsify(odd_t(erf,real(_L))))-_, Type).
+   Type = clause.
+
+read_terms(Stream, Terms) :-
+    read_terms_(Stream, [], Terms).
+read_terms_(Stream, Terms0, Terms) :-
+    Options = [variable_names(VarNames)],
+    read_term(Stream, Term, Options),
+    (   Term = end_of_file -> reverse(Terms0, Terms)
+    ;   read_terms_(Stream, [Term-VarNames|Terms0], Terms)
+    ).
+
+%% zip(+Xs, +Ys, ?XYs)
+%% zip(?Xs, ?Ys, +XYs)
+%
+% List XYs interleaves same-length lists Xs and Ys.
+zip([X|Xs], [Y|Ys], [X,Y|XYs]) :-
+    zip(Xs, Ys, XYs).
+zip([], [], []).
+
+?- zip(Xs, Ys, XYs). % MGQ does not terminate
+   error('$interrupt_thrown',repl/0).
+
+% The following suggested by Ulrich via Quad Works chat
+?- zip(Xs, Ys, XYs), false. % loops
+
+?- zip(X, [4,5,6], [1,4,2,5,3,6]).
+   X = [1,2,3].
+
+?- zip([1,2,3], Y, [1,4,2,5,3,6]).
+   Y = [4,5,6].
+
+?- zip([1,2,3], [4,5,6], Z).
+   Z = [1,4,2,5,3,6].
+
+?- zip(Xs, Ys, [1,4,2,5,3,6]).
+   Xs = [1,2,3], Ys = [4,5,6].
+
+% 3. Demonstrate checking 1 quad, the top two elements of a QAs list.
+check_qu_ad(Q-QVN, A-AVN) :-
+    Q = ?-(G),
+    phrase(portray_clause_(Q), LitQ), % NB: LitQ terminates w/ newline
+    format("% CHECKING.. ",[]),
+    (   A == true -> call(G)
+    ;   A == false -> (   call(G) -> false
+                      ;   true
+                      )
+    ;   phrase(unconj(A), As) ->
+        (   length(As, N),
+            n_answers(N, A, AVN, ADs),
+            n_answers(N, G, QVN, Answers),
+            maplist(contains, ADs, Answers)
+        )
+    ;   % Otherwise, we have the ',' case of a solitary answer
+        call(G),
+        call(A),
+        QVN == AVN
+    ),
+    format("~s", [LitQ]).
+
+% Answer-description AD (qua set-of-bindings) contains Answer.
+contains(AD, Answer) :- append(Answer, _, AD).
+
+?- contains(['Xs'=[C],'L'=1,'_A'=C,'_B'=D], ['Xs'=[A],'L'=1]).
+   C = A.
+
+?- check_qu_ad((?-length(_F,_G))-['Xs'=_F,'L'=_G],(_H=[],_I=0;_H=[_J],_I=1;_H=[_J,_K],_I=2;...)-['Xs'=_H,'L'=_I,'_A'=_J,'_B'=_K]).
+% CHECKING.. (?-length(A,B)).
+   _F = [_A,_B], _G = 2, _H = [_J,_K], _I = 2.
+
+% Unravel the nested (;)/1 applications of multiple-AD structures.
+unconj(Conj) --> { Conj = (Elt;Conj_) },
+                 [Elt],
+                 unconj(Conj_).
+unconj(...) --> [].
+
+?- phrase(unconj((_H=[],_I=0;_H=[_J],_I=1;_H=[_J,_K],_I=2;...)), List).
+   List = [(_H=[],_I=0),(_H=[_J],_I=1),(_H=[_J,_K],_I=2)].
+
+empty_anstack :-
+    (   retract('$anstack'(_)), fail
+    ;   asserta('$anstack'([]))
+    ).
+
+push(VN) :-
+    retract('$anstack'(As)),
+    asserta('$anstack'([VN|As])).
+
+backtrack(N) :-
+    (   '$anstack'(Ans),
+        length(Ans, N) -> true
+    ;   fail
+    ).
+
+n_answers(N, G, VN, ADs) :-
+    must_be(integer, N),
+    (   N > 0 -> n_answers_(N, G, VN, ADs)
+    ;   domain_error(not_less_than_zero, N, n_answers/4)
+    ).
+
+n_answers_(N, G, VN, ADs) :-
+    empty_anstack,
+    call(G), push(VN),
+    backtrack(N),
+    !,
+    retract('$anstack'(As)),
+    reverse(As, ADs).
+    
+?- n_answers(3, length(Xs, L), ('Xs'=Xs,'Len'=L), ADs).
+   Xs = [_A,_B], L = 2, ADs = [('Xs'=[],'Len'=0),('Xs'=[_C],'Len'=1),('Xs'=[_D,_E],'Len'=2)].

src/lib/numerics/special_functions.pl

@@ -0,0 +1,257 @@
+/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
+   Written 2025 by David C. Norris (david@precisionmethods.guru)
+   As with all things floating-point, use at your own risk.
+   Part of Scryer Prolog.
+- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
+
+/** Special math functions in the Error, Gamma and Beta families
+
+The underlying Rust implementations come from the
+[puruspe](https://docs.rs/puruspe/latest/puruspe/) crate.
+*/
+
+:- module(special_functions, [
+              erf/2
+              ,erfc/2
+              ,inverf/2
+              ,inverfc/2
+              ,gamma/2
+              ,gamma/3
+              ,gamma_P_Q/4
+              ,invgammp/3
+              ,log_gamma/2
+              ,beta/3
+              ,betai/4
+              ,invbetai/4
+              ,test/2
+              ,test_special_functions/0
+              ,try_falsify/1
+              ,witness/1
+          ]).
+
+:- use_module(library(numerics/testutils)).
+
+%% erf(+Xexpr, -Erf)
+%
+% X is Xexpr ∈ ℝ, Erf is erf(X).
+%
+% [DLMF §7.2.1](https://dlmf.nist.gov/7.2#E1),
+% [`puruspe::error::erf`](https://docs.rs/puruspe/latest/puruspe/error/fn.erf.html)
+erf(Xexpr, Erf) :-
+    X is Xexpr,
+    builtins:must_be_number(X, erf/2),
+    '$erf'(X, Erf).
+
+% Demonstrate the roots of x - erf(x))
+?- X0 = 0.6174468790806071, erf(X0, X0), _X0 is -X0, erf(_X0, _X0).
+   X0 = 0.6174468790806071, _X0 = -0.6174468790806071.
+
+% erf is an odd function:
+?- try_falsify(odd_t(erf, real(_))).
+   false.
+
+% Another way to say the same thing..
+?- witness(odd_t(erf, real(_), false)).
+   false.
+
+% ..and yet one more:
+?- witness((real(X), erf(X,Erf), erf(-X,_Erf), abs(Erf+_Erf) > 0)).
+   false.
+
+%% erfc(+X, -Erfc)
+%
+% Erfc is erfc(X) for X ∈ ℝ.
+%
+% [DLMF §7.2.2](https://dlmf.nist.gov/7.2#E2),
+% [`puruspe::error::erfc`](https://docs.rs/puruspe/latest/puruspe/error/fn.erfc.html)
+erfc(X, Erfc) :-
+    builtins:must_be_number(X, erfc/2),
+    '$erfc'(X, Erfc).
+
+?- real(X), erf(X, Erf), erfc(X, Erfc), abs(Erf+Erfc-1) > epsilon.
+   false.
+
+%% inverf(+ErfX, -X)
+%
+% X is erf⁻¹(ErfX) for ErfX ∈ (-1,1).
+inverf(ErfX, X) :-
+    builtins:must_be_number(ErfX, inverf/2),
+    '$inverf'(ErfX, X).
+
+?- try_falsify(δ_inverses_t(40*epsilon, erf, inverf, interval(-2,2,_))).
+   false.
+
+%% inverfc(+ErfcX, -X)
+%
+% X is erfc⁻¹(ErfcX) for ErfcX ∈ (0,2).
+inverfc(ErfcX, X) :-
+    builtins:must_be_number(ErfcX, inverfc/2),
+    '$inverfc'(ErfcX, X).
+
+?- try_falsify(δ_inverses_t(40*epsilon, erfc, inverfc, interval(-2,2,_))).
+   false.
+
+%% gamma(+X, -Gamma)
+%
+% Gamma is Γ(X), the [ordinary] gamma function evaluated at X ∈ ℝ.
+%
+% [DLMF §5.2.1](https://dlmf.nist.gov/5.2#E1)
+% [`puruspe::gamma::gamma`](https://docs.rs/puruspe/latest/puruspe/gamma/fn.gamma.html)
+gamma(X, Gamma) :-
+    builtins:must_be_number(X, gamma/2),
+    '$gamma'(X, Gamma).
+
+% Γ(n+1) ≡ n!
+?- N = 10, N1 is N+1, gamma(N1, ΓN1), int_realfact(N, Γ11).
+   N = 10, N1 = 11, ΓN1 = 3628800.0, Γ11 = 3628800.0.
+
+
+%% gamma(+A, +X, -Gamma)
+%
+% Gamma is Γ(A,X), the upper incomplete gamma function, where A > 0
+% is the shape parameter and X ≥ 0 is the lower limit of integration.
+%
+% [DLMF §8.2.2](https://dlmf.nist.gov/8.2#E2),
+gamma(A, X, Gamma) :-
+    builtins:must_be_number(A, gammq/3),
+    builtins:must_be_number(X, gammq/3),
+    '$gammq'(A, X, Q),
+    gamma(A, GammaA),
+    Gamma is Q*GammaA.
+
+%% gamma_P_Q(+A, +X, -P, -Q)
+%
+% For shape parameter A > 0 and lower limit of integration X ≥ 0,
+%
+% * P is γ(A,X)/Γ(X), the regularized _lower_ incomplete gamma function, and
+% 
+% * Q is Γ(A,X)/Γ(X), the regularized _upper_ incomplete gamma function.
+%
+% [DLMF §8.2.4](https://dlmf.nist.gov/8.2#E4),
+% [`puruspe::gammp::gammp`](https://docs.rs/puruspe/latest/puruspe/gamma/fn.gammp.html),
+% [`puruspe::gammp::gammq`](https://docs.rs/puruspe/latest/puruspe/gamma/fn.gammq.html)
+gamma_P_Q(A, X, P, Q) :-
+    builtins:must_be_number(A, gamma_P_Q/4),
+    builtins:must_be_number(X, gamma_P_Q/4),
+    '$gammp'(A, X, P),
+    '$gammq'(A, X, Q).
+
+% P + Q ≈ 1
+?- gamma_P_Q(1.2, 2.3, P, Q), abs(P + Q - 1) < epsilon.
+   P = 0.8621845438106976, Q = 0.1378154561893024.
+
+%% invgammp(+A, +P, -X)
+%
+% Given shape parameter A > 0 and probability P ∈ [0,1),
+%
+% X is the unique solution of P = P(A,X), where P(-,-) is the
+% regularized lower incomplete gamma function.
+%
+% [`puruspe::gamma::invgammp`](https://docs.rs/puruspe/latest/puruspe/gamma/fn.invgammp.html)
+invgammp(A, P, X) :-
+    builtins:must_be_number(A, invgammp/3),
+    builtins:must_be_number(P, invgammp/3),
+    '$invgammp'(P, A, X).
+
+?- A = 1.5, P = 0.7, invgammp(A, P, X), gamma_P_Q(A, X, P_, _), abs(P-P_) < epsilon.
+   A = 1.5, P = 0.7, X = 1.8324353915624363, P_ = 0.7000000000000001.
+
+%% log_gamma(+X, -LogGamma)
+%
+% LogGamma is ln(Γ(X)), the natural logarithm of Γ(X).
+%
+% [`puruspe::gamma::ln_gamma`](https://docs.rs/puruspe/latest/puruspe/gamma/fn.ln_gamma.html)
+log_gamma(X, LnGamma) :-
+    builtins:must_be_number(X, log_gamma/2),
+    '$ln_gamma'(X, LnGamma).
+
+%% beta(+X, +Y, -B)
+%
+% B is B(X,Y) ≡ Γ(X)*Γ(Y)/Γ(X+Y)
+%
+% [DLMF  §5.12.1](https://dlmf.nist.gov/5.12#E1)
+% [`puruspe::beta::beta`](https://docs.rs/puruspe/latest/puruspe/beta/fn.beta.html)
+beta(X, Y, B) :-
+    builtins:must_be_number(X, beta/3),
+    builtins:must_be_number(Y, beta/3),
+    '$beta'(X, Y, B).
+
+%% betai(+A, +B, +X, -Ix)
+%
+% Given:
+%
+% * shape parameters A > 0 and B > 0,
+% * upper limit of integration X ∈ [0,1],
+%
+% Ix is Iₓ(A,B) ≡ B(X;A,B)/B(A,B), the regularized incomplete beta function;
+%
+% [DLMF  §8.17.2](https://dlmf.nist.gov/8.17#E2),
+% [`puruspe::beta::betai`](https://docs.rs/puruspe/latest/puruspe/beta/fn.betai.html)
+betai(A, B, X, Ix) :-
+    builtins:must_be_number(A, betai/4),
+    builtins:must_be_number(B, betai/4),
+    builtins:must_be_number(X, betai/4),
+    '$betai'(A, B, X, Ix).
+
+%% invbetai(+A, +B, +P, -X)
+%
+% Given:
+%
+% * shape parameters A > 0 and B > 0,
+% * probability P ∈ [0,1],
+%
+% X ∈ [0,1] is the unique solution of P = Iₓ(A,B) ≡ B(X;A,B)/B(A,B).
+%
+% [`puruspe::beta::invbetai`](https://docs.rs/puruspe/latest/puruspe/beta/fn.invbetai.html)
+invbetai(A, B, P, X) :-
+    builtins:must_be_number(A, invbetai/4),
+    builtins:must_be_number(B, invbetai/4),
+    builtins:must_be_number(P, invbetai/4),
+    '$invbetai'(P, A, B, X).
+
+
+% ============================== TESTS ==============================
+
+%% test_special_functions
+%
+% Run all tests defined in this module.  (These tests _succeed_ when
+% they find counterexamples, so the 'desirable' result is `false`.)
+test_special_functions :-
+    format("Seeking counterexamples to assertions:~n", []),
+    test(T, G), format("% ~s ~n", [T]),
+    call(G).
+
+% We default to 1M falsification attempts per assertion, and -- more
+% importantly -- use the (unexported) testutils:try_falsify_/2, to
+% avoid testutils:reproducibly/0 fixing an RNG seed.  Thus we obtain
+% truly pseudorandom tests untainted by seed-hacking impropriety.
+:- meta_predicate(try_falsify(1)).
+try_falsify(G) :- testutils:try_falsify_(10^6, G).
+
+% A unary witness/1 predicate similarly renders queries more concise.
+:- meta_predicate(witness(0)).
+witness(G) :- testutils:witness(10^6, G).
+
+:- discontiguous(test/2).
+
+%% test(+Name, ?Goal)
+%
+% Tests have the signature established by @bakaq's test_framework,
+% each with a user-facing string Name, and a Goal which serves as an
+% _assertion_ by succeeding iff the Name'd desirable property holds.
+test("erf is odd", try_falsify(odd_t(erf, real(_)))).
+
+test("pos root of erf(x)-x", \+ (X0 = 0.6174468790806071, erf(X0, X0))).
+
+test("erfc ≈ 1 - erf", try_falsify(erf_plus_erfc_unity_t(real(_)))).
+
+erf_plus_erfc_unity_t(Any, T) :-
+    call_free(Any, X), erf(X, Erf), erfc(X, Erfc),
+    (   abs(Erf + Erfc - 1) < epsilon -> T = true
+    ;   T = false
+    ).
+
+test("inverf ≈ erf⁻¹",
+     try_falsify(δ_inverses_t(40*epsilon, erf, inverf, interval(-2,2,_)))).
+
+test("('false' is good)", false).

src/lib/numerics/testutils.pl

@@ -0,0 +1,147 @@
+/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
+   Written 2025 by David C. Norris (david@precisionmethods.guru)
+   As with all things floating-point, use at your own risk.
+   Part of Scryer Prolog.
+- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
+
+/** Utility predicates for testing numerics
+
+*/
+
+:- module(testutils, [
+              try_falsify/2
+              ,witness/2
+              ,real/1
+              ,posreal/1
+              ,interval/3
+              ,call_free/2
+              ,odd_t/3
+              ,δ_inverses_t/5
+              ,int_realfact/2
+          ]).
+
+:- use_module(library(lists)).
+:- use_module(library(random)).
+:- use_module(library(error)).
+:- use_module(library(format)).
+
+%% real(-X)
+%
+% X = tan(U) for uniform U ~ U[-π,π).
+real(X) :-
+    random(U),
+    X is tan(pi*(U - 0.5)).
+
+%% posreal(-X)
+%
+% X = 1/U for uniform U ~ U[0,1).
+posreal(X) :-
+    random(U),
+    X is 1/U - 1.
+
+%% interval(+A, +B, -X)
+%
+% X ~ U[A,B).
+interval(A, B, X) :-
+    random(U),
+    X is A + (B-A)*U.
+
+:- meta_predicate(try_falsify(+, 1)).
+
+% NOTE: NON-reproducible tests would in fact be MORE STRINGENT,
+%       creating opportunities to detect rare error cases over time.
+reproducibly :- set_random(seed(2025)).
+
+%% try_falsify(+IntExpr, ?G_1)
+%
+% Make (N is IntExpr) attempts to falsify the partial goal G/1,
+% reporting the first counterexample found.
+try_falsify(N, G_1) :- reproducibly -> try_falsify_(N, G_1).
+try_falsify_(IntExpr, G_1) :- N is IntExpr, must_be(integer, N), N > 0,
+                      (   call(G_1, false) -> counterexample(G_1)
+                      ;   N_ is N - 1,
+                          try_falsify_(N_, G_1)
+                      ).
+
+%% call_free(?G, -V)
+%
+% Call goal G, the single free variable of which is bound to V.
+call_free(G, V) :- term_variables(G, [V]), call(G).
+
+:- meta_predicate(odd_t(2, 0, ?)).
+
+:- meta_predicate(witness(+, 0)).
+
+%% witness(+N, ?OhNo)
+%
+% Make N attempts to satisfy the goal OhNo, reporting the first
+% counterexample found.
+witness(N, OhNo) :- N > 0,
+                    (   call(OhNo) -> counterexample(OhNo)
+                    ;   N_ is N - 1,
+                        witness(N_, OhNo)
+                    ).
+
+% The goal below is obviously _designed_ to succeed 10% of the time,
+% giving us plenty of chances to see 'counterexamples':
+?- witness(10, (random(X), format("% X = ~f~n", [X]), X > 0.9)).
+% X = 0.8084510175379878
+% X = 0.3173976152322231
+% X = 0.6016479707924276
+% X = 0.2782828276608216
+% X = 0.5737059731916494
+% X = 0.4300520177737992
+% X = 0.5411038305190763
+% X = 0.6901983097300066
+% X = 0.05013429563996663
+% X = 0.7321078902421636
+   false.
+% X = 0.6401599325865659
+% X = 0.7024957773981413
+% X = 0.5798797162672757
+% X = 0.3107013295892864
+% X = 0.3792925200660049
+% X = 0.3424593598278811
+% X = 0.0589899862019303
+% X = 0.9692529829158145
+% COUNTEREXAMPLE: user:random(0.9692529829158145),(current_output(user_output),pio:phrase_to_stream(format:format_([%, ,X, ,=, ,~,f,~,n],[0.9692529829158145]),user_output),flush_output(user_output)),0.9692529829158145>0.9
+   X = 0.9692529829158145.
+
+%% odd_t(+F_2, +Any, ?T)
+%
+% T is the truth-value from testing that function F_2 is
+% [odd](https://en.wikipedia.org/wiki/Even_and_odd_functions) at a
+% value X obtained via call_free(Any, X).
+odd_t(F, Any, T) :-
+    (   call_free(Any, X),
+        _X is -X,
+        call(F, X, Fx),
+        call(F, _X, _Fx),
+        _Fx is -Fx -> T = true
+    ;   T = false
+    ).
+
+:- meta_predicate(δ_inverses_t(?, 2, 2, 0, ?)).
+
+%% δ_inverses_t(+Δ, +F_2, +Finv_2, +Any, ?T)
+%
+% For given functions F_2 and Finv_2, T is the truth-value from
+% testing that Finv_2 ∘ F_2 is within Δ of the identity at a value X
+% obtained via call_free(Any, X).
+δ_inverses_t(Δ, F, Finv, Any, T) :-
+    (   call_free(Any, X),
+        call(F, X, Fx),
+        call(Finv, Fx, X_),
+        (   abs(X - X_) < Δ -> T = true
+        ;   T = false
+        )
+    ).
+
+%% int_realfact(+N, -FactN)
+%
+% FactN is floating-point N!
+int_realfact(N, FactN) :-
+    N > 0, N_ is N - 1, int_realfact(N_, FactN_), FactN is N*FactN_.
+int_realfact(0, 1.0).
+
+counterexample(G) :- format("% COUNTEREXAMPLE: ~w~n", [G]).

src/lib/ops_and_meta_predicates.pl

@@ -3,11 +3,16 @@
 % module resolution operator.
 :- op(600, xfy, :).
 
-:- op(1199, fx, meta_predicate).
-
-/* this is an implementation specific declarative operator used to implement call_with_inference_limit/3
-   and setup_call_cleanup/3. switches to the default trust_me and retry_me_else. Indexing choice
-   instructions are unchanged. */
+% Implementation specific declarative operator used to implement
+% call_with_inference_limit/3 and setup_call_cleanup/3. Compiler switches
+% to the default trust_me, retry_me_else and some other instructions for all
+% predicates that are marked with it. Indexing choice instructions are unchanged.
+%
+% Implementation details:
+% Default instructions are not subject to inference counting, so their
+% execution will not be considered if they happen to be called by
+% call_with_inference_limit/3.
+%
 :- op(700, fx, non_counted_backtracking).
 
 % arithmetic operators.
@@ -19,7 +24,6 @@
 :- op(200, xfy, ^).
 :- op(500, yfx, /\).
 :- op(500, yfx, \/).
-:- op(500, yfx, xor).
 :- op(400, yfx, div).
 :- op(400, yfx, //).
 :- op(400, yfx, rdiv).
@@ -60,70 +64,70 @@
 :- op(1200, xfx, -->).
 
 % meta_predicate declarations for call/{1, 66}.
-:- meta_predicate call(0).
-:- meta_predicate call(1, ?).
-:- meta_predicate call(2, ?, ?).
-:- meta_predicate call(3, ?, ?, ?).
-:- meta_predicate call(4, ?, ?, ?, ?).
-:- meta_predicate call(5, ?, ?, ?, ?, ?).
-:- meta_predicate call(6, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(7, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(8, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(9, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(10, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(11, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(12, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(13, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(14, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(15, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(16, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(17, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(18, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(19, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(20, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(21, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(22, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(23, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(24, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(25, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(26, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(27, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(28, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(29, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(30, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(31, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(32, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(33, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(34, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(35, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(36, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(37, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(38, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(39, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(40, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(41, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(42, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(43, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(44, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(45, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(46, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(47, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(48, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(49, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(50, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(51, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(52, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(53, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(54, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(55, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(56, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(57, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(58, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(59, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(60, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(60, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(61, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(62, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(63, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(64, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
-:- meta_predicate call(65, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?).
+:- meta_predicate(call(0)).
+:- meta_predicate(call(1, ?)).
+:- meta_predicate(call(2, ?, ?)).
+:- meta_predicate(call(3, ?, ?, ?)).
+:- meta_predicate(call(4, ?, ?, ?, ?)).
+:- meta_predicate(call(5, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(6, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(7, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(8, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(9, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(10, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(11, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(12, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(13, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(14, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(15, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(16, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(17, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(18, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(19, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(20, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(21, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(22, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(23, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(24, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(25, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(26, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(27, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(28, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(29, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(30, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(31, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(32, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(33, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(34, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(35, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(36, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(37, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(38, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(39, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(40, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(41, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(42, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(43, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(44, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(45, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(46, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(47, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(48, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(49, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(50, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(51, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(52, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(53, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(54, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(55, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(56, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(57, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(58, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(59, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(60, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(60, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(61, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(62, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(63, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(64, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).
+:- meta_predicate(call(65, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)).

src/lib/ordsets.pl

@@ -54,39 +54,38 @@
 
 :- use_module(library(lists)).
 
-/** <module> Ordered set manipulation
+/** Ordered set manipulation
+
 Ordered sets are lists with unique elements sorted to the standard order
-of terms (see sort/2). Exploiting ordering,   many of the set operations
+of terms (see `sort/2`). Exploiting ordering,   many of the set operations
 can be expressed in order N rather  than N^2 when dealing with unordered
 sets that may contain duplicates. The library(ordsets) is available in a
 number of Prolog implementations. Our  predicates   are  designed  to be
-compatible  with  common  practice   in    the   Prolog  community.  The
-implementation is incomplete and  relies   partly  on  library(oset), an
-older ordered set library distributed  with SWI-Prolog. New applications
-are advised to use library(ordsets).
+compatible  with  common  practice   in    the   Prolog  community.
 Some  of  these  predicates  match    directly   to  corresponding  list
 operations. It is advised to use the  versions from this library to make
-clear you are operating on ordered sets.   An exception is member/2. See
-ord_memberchk/2.
+clear you are operating on ordered sets.   An exception is `member/2`. See
+`ord_memberchk/2`.
+
 The ordsets library is based  on  the   standard  order  of  terms. This
 implies it can handle  all  Prolog   terms,  including  variables.  Note
 however, that the ordering is not stable  if   a  term inside the set is
 further instantiated. Also  note  that   variable  ordering  changes  if
 variables in the set are unified with each   other  or a variable in the
-set is unified with a variable that  is `older' than the newest variable
+set is unified with a variable that  is _older_ than the newest variable
 in the set. In  practice,  this  implies   that  it  is  allowed  to use
 member(X, OrdSet) on an ordered set that holds  variables only if X is a
 fresh variable. In other cases one should   cease  using it as an ordset
 because the order it relies on may have been changed.
 */
 
-%!  is_ordset(@Term) is semidet.
+%% is_ordset(@Term) is semidet.
 %
-%   True if Term is an ordered set.   All predicates in this library
-%   expect ordered sets as input arguments.  Failing to fullfil this
-%   assumption results in undefined   behaviour.  Typically, ordered
-%   sets are created by predicates  from   this  library,  sort/2 or
-%   setof/3.
+% True if Term is an ordered set.   All predicates in this library
+% expect ordered sets as input arguments.  Failing to fullfil this
+% assumption results in undefined   behaviour.  Typically, ordered
+% sets are created by predicates  from   this  library,  `sort/2` or
+% `setof/3`.
 
 is_ordset(Term) :-
     '$skip_max_list'(_, _, Term, Tail), Tail == [], %% is_list(Term),
@@ -102,37 +101,35 @@ is_ordset3([H2|T], H) :-
     is_ordset3(T, H2).
 
 
-%!  ord_empty(?List) is semidet.
+%% ord_empty(?List) is semidet.
 %
-%   True when List is the  empty   ordered  set. Simply unifies list
-%   with the empty list. Not part of Quintus.
+% True when List is the  empty   ordered  set. Simply unifies list
+% with the empty list. Not part of Quintus.
 
 ord_empty([]).
 
 
-%!  ord_seteq(+Set1, +Set2) is semidet.
+%% ord_seteq(+Set1, +Set2) is semidet.
 %
-%   True if Set1 and Set2  have  the   same  elements.  As  both are
-%   canonical sorted lists, this is the same as ==/2.
-%
-%   @compat sicstus
+% True if Set1 and Set2  have  the   same  elements.  As  both are
+% canonical sorted lists, this is the same as `==/2`.
 
 ord_seteq(Set1, Set2) :-
     Set1 == Set2.
 
 
-%!  list_to_ord_set(+List, -OrdSet) is det.
+%% list_to_ord_set(+List, -OrdSet) is det.
 %
-%   Transform a list into an ordered set.  This is the same as
-%   sorting the list.
+% Transform a list into an ordered set.  This is the same as
+% sorting the list.
 
 list_to_ord_set(List, Set) :-
     sort(List, Set).
 
 
-%!  ord_intersect(+Set1, +Set2) is semidet.
+%% ord_intersect(+Set1, +Set2) is semidet.
 %
-%   True if both ordered sets have a non-empty intersection.
+% True if both ordered sets have a non-empty intersection.
 
 ord_intersect([H1|T1], L2) :-
     ord_intersect_(L2, H1, T1).
@@ -148,31 +145,29 @@ ord_intersect__(>, H1, T1,  _H2, T2) :-
     ord_intersect_(T2, H1, T1).
 
 
-%!  ord_disjoint(+Set1, +Set2) is semidet.
+%% ord_disjoint(+Set1, +Set2) is semidet.
 %
-%   True if Set1 and Set2  have  no   common  elements.  This is the
-%   negation of ord_intersect/2.
+% True if Set1 and Set2  have  no   common  elements.  This is the
+% negation of `ord_intersect/2`.
 
 ord_disjoint(Set1, Set2) :-
     \+ ord_intersect(Set1, Set2).
 
 
-%!  ord_intersect(+Set1, +Set2, -Intersection)
+%% ord_intersect(+Set1, +Set2, -Intersection)
 %
-%   Intersection  holds  the  common  elements  of  Set1  and  Set2.
+% Intersection  holds  the  common  elements  of  Set1  and  Set2.
 %
-%   @deprecated Use ord_intersection/3
+% This predicate is *deprecated*. Use `ord_intersection/3`
 
 ord_intersect(Set1, Set2, Intersection) :-
     oset_int(Set1, Set2, Intersection).
 
 
-%!  ord_intersection(+PowerSet, -Intersection)
+%% ord_intersection(+PowerSet, -Intersection)
 %
-%   Intersection of a powerset. True when Intersection is an ordered
-%   set holding all elements common to all sets in PowerSet.
-%
-%   @compat sicstus
+% Intersection of a powerset. True when Intersection is an ordered
+% set holding all elements common to all sets in PowerSet.
 
 ord_intersection(PowerSet, Intersection) :-
     key_by_length(PowerSet, Pairs),
@@ -190,10 +185,10 @@ l_int([_-H|T], S0, S) :-
     l_int(T, S1, S).
 
 
-%!  ord_intersection(+Set1, +Set2, -Intersection) is det.
+%% ord_intersection(+Set1, +Set2, -Intersection) is det.
 %
-%   Intersection holds the common elements of Set1 and Set2.  Uses
-%   ord_disjoint/2 if Intersection is bound to `[]` on entry.
+% Intersection holds the common elements of Set1 and Set2.  Uses
+% `ord_disjoint/2` if Intersection is bound to `[]` on entry.
 
 ord_intersection(Set1, Set2, Intersection) :-
     (   Intersection == []
@@ -202,13 +197,11 @@ ord_intersection(Set1, Set2, Intersection) :-
     ).
 
 
-%!  ord_intersection(+Set1, +Set2, ?Intersection, ?Difference) is det.
+%% ord_intersection(+Set1, +Set2, ?Intersection, ?Difference) is det.
 %
-%   Intersection  and  difference   between    two   ordered   sets.
-%   Intersection is the intersection between   Set1  and Set2, while
-%   Difference is defined by ord_subtract(Set2, Set1, Difference).
-%
-%   @see ord_intersection/3 and ord_subtract/3.
+% Intersection  and  difference   between    two   ordered   sets.
+% Intersection is the intersection between   Set1  and Set2, while
+% Difference is defined by `ord_subtract(Set2, Set1, Difference)`.
 
 ord_intersection([], L, [], L) :- !.
 ord_intersection([_|_], [], [], []) :- !.
@@ -224,35 +217,35 @@ ord_intersection2(>, H1, T1, H2, T2, Intersection, [H2|HDiff]) :-
     ord_intersection([H1|T1], T2, Intersection, HDiff).
 
 
-%!  ord_add_element(+Set1, +Element, ?Set2) is det.
+%% ord_add_element(+Set1, +Element, ?Set2) is det.
 %
-%   Insert  an  element  into  the  set.    This   is  the  same  as
-%   ord_union(Set1, [Element], Set2).
+% Insert  an  element  into  the  set.    This   is  the  same  as
+% `ord_union(Set1, [Element], Set2)`.
 
 ord_add_element(Set1, Element, Set2) :-
     oset_addel(Set1, Element, Set2).
 
 
-%!  ord_del_element(+Set, +Element, -NewSet) is det.
+%% ord_del_element(+Set, +Element, -NewSet) is det.
 %
-%   Delete an element from an  ordered  set.   This  is  the same as
-%   ord_subtract(Set, [Element], NewSet).
+% Delete an element from an  ordered  set.   This  is  the same as
+% `ord_subtract(Set, [Element], NewSet)`.
 
 ord_del_element(Set, Element, NewSet) :-
     oset_delel(Set, Element, NewSet).
 
 
-%!  ord_selectchk(+Item, ?Set1, ?Set2) is semidet.
+%% ord_selectchk(+Item, ?Set1, ?Set2) is semidet.
 %
-%   Selectchk/3,  specialised  for  ordered  sets.    Is  true  when
-%   select(Item, Set1, Set2) and Set1, Set2   are  both sorted lists
-%   without duplicates. This implementation is only expected to work
-%   for Item ground and either Set1 or Set2 ground. The "chk" suffix
-%   is meant to remind you of   memberchk/2,  which also expects its
-%   first  argument  to  be  ground.    ord_selectchk(X,  S,  T)  =>
-%   ord_memberchk(X, S) & \+ ord_memberchk(X, T).
+% `selectchk/3`,  specialised  for  ordered  sets.    Is  true  when
+% select(Item, Set1, Set2) and Set1, Set2   are  both sorted lists
+% without duplicates. This implementation is only expected to work
+% for Item ground and either Set1 or Set2 ground. The "chk" suffix
+% is meant to remind you of   `memberchk/2`,  which also expects its
+% first  argument  to  be  ground.    `ord_selectchk(X,  S,  T)  =>
+% ord_memberchk(X, S) & \+ ord_memberchk(X, T).`
 %
-%   @author Richard O'Keefe
+% Author: Richard O'Keefe
 
 ord_selectchk(Item, [X|Set1], [X|Set2]) :-
     X @< Item,
@@ -266,19 +259,19 @@ ord_selectchk(Item, [Item|Set1], Set1) :-
     ).
 
 
-%!  ord_memberchk(+Element, +OrdSet) is semidet.
+%% ord_memberchk(+Element, +OrdSet) is semidet.
 %
-%   True if Element is a member of   OrdSet, compared using ==. Note
-%   that _enumerating_ elements of an ordered  set can be done using
-%   member/2.
+% True if Element is a member of   OrdSet, compared using ==. Note
+% that _enumerating_ elements of an ordered  set can be done using
+% `member/2`.
 %
-%   Some Prolog implementations also provide  ord_member/2, with the
-%   same semantics as ord_memberchk/2.  We   believe  that  having a
-%   semidet ord_member/2 is unacceptably inconsistent with the *_chk
-%   convention.  Portable  code  should    use   ord_memberchk/2  or
-%   member/2.
+% Some Prolog implementations also provide  `ord_member/2`, with the
+% same semantics as `ord_memberchk/2`.  We   believe  that  having a
+% semidet `ord_member/2` is unacceptably inconsistent with the \*\_chk
+% convention.  Portable  code  should    use   `ord_memberchk/2`  or
+% `member/2`.
 %
-%   @author Richard O'Keefe
+% Author: Richard O'Keefe
 
 ord_memberchk(Item, [X1,X2,X3,X4|Xs]) :-
     !,
@@ -303,9 +296,9 @@ ord_memberchk(Item, [X1]) :-
     Item == X1.
 
 
-%!  ord_subset(+Sub, +Super) is semidet.
+%% ord_subset(+Sub, +Super) is semidet.
 %
-%   Is true if all elements of Sub are in Super
+% Is true if all elements of Sub are in Super
 
 ord_subset([], _).
 ord_subset([H1|T1], [H2|T2]) :-
@@ -319,22 +312,20 @@ ord_subset_(=, _, T1, T2) :-
     ord_subset(T1, T2).
 
 
-%!  ord_subtract(+InOSet, +NotInOSet, -Diff) is det.
+%% ord_subtract(+InOSet, +NotInOSet, -Diff) is det.
 %
-%   Diff is the set holding all elements of InOSet that are not in
-%   NotInOSet.
+% Diff is the set holding all elements of InOSet that are not in
+% NotInOSet.
 
 ord_subtract(InOSet, NotInOSet, Diff) :-
     oset_diff(InOSet, NotInOSet, Diff).
 
 
-%!  ord_union(+SetOfSets, -Union) is det.
+%% ord_union(+SetOfSets, -Union) is det.
 %
-%   True if Union is the  union  of   all  elements  in the superset
-%   SetOfSets. Each member of SetOfSets must  be an ordered set, the
-%   sets need not be ordered in any way.
-%
-%   @author Copied from YAP, probably originally by Richard O'Keefe.
+% True if Union is the  union  of   all  elements  in the superset
+% SetOfSets. Each member of SetOfSets must  be an ordered set, the
+% sets need not be ordered in any way.
 
 ord_union([], []).
 ord_union([Set|Sets], Union) :-
@@ -355,18 +346,18 @@ ord_union_all(N, Sets0, Union, Sets) :-
     ).
 
 
-%!  ord_union(+Set1, +Set2, ?Union) is det.
+%% ord_union(+Set1, +Set2, ?Union) is det.
 %
-%   Union is the union of Set1 and Set2
+% Union is the union of Set1 and Set2
 
 ord_union(Set1, Set2, Union) :-
     oset_union(Set1, Set2, Union).
 
 
-%!  ord_union(+Set1, +Set2, -Union, -New) is det.
+%% ord_union(+Set1, +Set2, -Union, -New) is det.
 %
-%   True iff ord_union(Set1, Set2, Union) and
-%   ord_subtract(Set2, Set1, New).
+% True iff `ord_union(Set1, Set2, Union)` and
+% `ord_subtract(Set2, Set1, New)`.
 
 ord_union([], Set2, Set2, Set2).
 ord_union([H|T], Set2, Union, New) :-
@@ -390,26 +381,26 @@ ord_union_2([H|T], H2, T2, Union, New) :-
     ord_union(Order, H, T, H2, T2, Union, New).
 
 
-%!  ord_symdiff(+Set1, +Set2, ?Difference) is det.
+%% ord_symdiff(+Set1, +Set2, ?Difference) is det.
 %
-%   Is true when Difference is the  symmetric difference of Set1 and
-%   Set2. I.e., Difference contains all elements that are not in the
-%   intersection of Set1 and Set2. The semantics  is the same as the
-%   sequence below (but the actual   implementation  requires only a
-%   single scan).
+% Is true when Difference is the  symmetric difference of Set1 and
+% Set2. I.e., Difference contains all elements that are not in the
+% intersection of Set1 and Set2. The semantics  is the same as the
+% sequence below (but the actual   implementation  requires only a
+% single scan).
 %
-%     ==
-%           ord_union(Set1, Set2, Union),
-%           ord_intersection(Set1, Set2, Intersection),
-%           ord_subtract(Union, Intersection, Difference).
-%     ==
+% ```
+% ord_union(Set1, Set2, Union),
+% ord_intersection(Set1, Set2, Intersection),
+% ord_subtract(Union, Intersection, Difference).
+% ```
 %
-%   For example:
+% For example:
 %
-%     ==
-%     ?- ord_symdiff([1,2], [2,3], X).
-%     X = [1,3].
-%     ==
+% ```
+% ?- ord_symdiff([1,2], [2,3], X).
+% X = [1,3].
+% ```
 
 ord_symdiff([], Set2, Set2).
 ord_symdiff([H1|T1], Set2, Difference) :-
@@ -457,7 +448,7 @@ ord_symdiff(>, H1, T1, H2, Set2, [H2|Difference]) :-
 */
 
 
-/** <module> Ordered set manipulation
+/* Ordered set manipulation
 
 This library defines set operations on sets represented as ordered
 lists.

src/lib/os.pl

@@ -12,37 +12,80 @@
    Public domain code.
 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
 
+/** Predicates for reasoning about the operating system (OS) environment.
+
+This includes predicates about environment variables, calls to shell and
+finding out the PID of the running system.
+*/
+
 :- module(os, [getenv/2,
                setenv/2,
                unsetenv/1,
                shell/1,
                shell/2,
-               pid/1]).
+               pid/1,
+	       raw_argv/1,
+	       argv/1]).
 
 :- use_module(library(error)).
 :- use_module(library(charsio)).
 :- use_module(library(lists)).
 :- use_module(library(si)).
 
+%% getenv(+Key, -Value).
+%
+% True iff Value contains the value of the environment variable Key.
+% Example:
+%
+% ```
+% ?- getenv("LANG", Ls).
+%    Ls = "en_US.UTF-8".
+% ```
 getenv(Key, Value) :-
         must_be_env_var(Key),
         '$getenv'(Key, Value).
 
+%% setenv(+Key, +Value).
+%
+% Sets the environment variable Key to Value
 setenv(Key, Value) :-
         must_be_env_var(Key),
         must_be_chars(Value),
         '$setenv'(Key, Value).
 
+%% unsetenv(+Key).
+%
+% Unsets the environment variable Key
 unsetenv(Key) :-
         must_be_env_var(Key),
         '$unsetenv'(Key).
 
+%% shell(+Command)
+%
+% Equivalent to `shell(Command, 0)`.
 shell(Command) :- shell(Command, 0).
+
+%% shell(+Command, -Status).
+%
+% True iff executes Command in a shell of the operating system and the exit code is Status.
+% Keep in mind the shell syntax is dependant on the operating system, so it should be
+% used very carefully.
+%
+% Example (using Linux and fish shell):
+%
+% ```
+% ?- shell("echo $SHELL", Status).
+% /bin/fish
+%    Status = 0.
+% ```
 shell(Command, Status) :-
     must_be_chars(Command),
     can_be(integer, Status),
     '$shell'(Command, Status).
 
+%% pid(-PID).
+%
+% True iff PID is the process identification number of current Scryer Prolog instance.
 pid(PID) :-
         can_be(integer, PID),
         '$pid'(PID).
@@ -69,3 +112,34 @@ permitted('_').
 must_be_chars(Cs) :-
         must_be(list, Cs),
         maplist(must_be(character), Cs).
+
+%% raw_argv(-Argv)
+%
+% True iff Argv is the list of arguments that this program was started with (usually passed via command line).
+% In contrast to `argv/1`, this version includes every argument, without any postprocessing, just as the operating
+% system reports it to the system. This includes-flags of Scryer itself, which are not needed in general.
+raw_argv(Argv) :-
+    can_be(list, Argv),
+    '$argv'(Argv).
+
+%% argv(-Argv)
+%
+% True if Argv is the list of arguments that this program was started with (usually passed via command line).
+% In this version, only arguments specific to the program are passed. To differentiate between the system
+% arguments and the program arguments, we use `--` as a separator.
+%
+% Example:
+%
+% ```
+% % Call with scryer-prolog -f -- -t hello
+% ?- argv(X).
+%     X = ["-t", "hello"].
+% ```
+argv(Argv) :-
+    can_be(list, Argv),
+    '$argv'(Argv0),
+    ( append(_, ["--"|Argv1], Argv0) ->
+        Argv = Argv1
+    ;
+        Argv = []
+    ).

src/lib/pairs.pl

@@ -1,3 +1,10 @@
+/** Reasoning about pairs.
+
+    Pairs are Prolog terms with principal functor `(-)/2`. A pair
+    often has the form `Key-Value`. The predicates of this library
+    relate pairs to keys and values.
+*/
+
 :- module(pairs, [pairs_keys_values/3,
 		  pairs_keys/2,
 		  pairs_values/2,
@@ -5,14 +12,27 @@
 		  map_list_to_pairs/3]).
 
 
-:- meta_predicate map_list_to_pairs(2, ?, ?).
+:- meta_predicate(map_list_to_pairs(2, ?, ?)).
+
+%% pairs_keys_values(?Pairs, ?Keys, ?Values)
+%
+%  The first argument is a list of Pairs, the second the corresponding
+%  Keys, and the third argument the corresponding values.
 
 pairs_keys_values([], [], []).
 pairs_keys_values([A-B|ABs], [A|As], [B|Bs]) :-
         pairs_keys_values(ABs, As, Bs).
 
+%% pairs_keys(?Pairs, ?Keys)
+%
+%  Same as `pairs_keys_values(Pairs, Keys, _)`.
+
 pairs_keys(Ps, Ks) :- pairs_keys_values(Ps, Ks, _).
 
+%% pairs_values(?Pairs, ?Values)
+%
+%  Same as `pairs_keys_values(Pairs, _, Values)`.
+
 pairs_values(Ps, Vs) :- pairs_keys_values(Ps, _, Vs).
 
 map_list_to_pairs(Pred, Ls, Ps) :-

src/lib/pio.pl

@@ -1,16 +1,15 @@
-/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
-   Pure I/O
-   ========
+/** Pure I/O.
 
    Our goal is to encourage the use of definite clause grammars (DCGs)
-   for describing strings. The predicates phrase_from_file/[2,3],
-   phrase_to_file/[2,3] and phrase_to_stream/2 let us apply DCGs
+   for describing strings. The predicates `phrase_from_file/[2,3]`,
+   `phrase_to_file/[2,3]` and `phrase_to_stream/2` let us apply DCGs
    transparently to files and streams, and therefore decouple side-effects
    from declarative descriptions.
-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
+*/
 
 :- module(pio, [phrase_from_file/2,
                 phrase_from_file/3,
+                phrase_from_stream/2,
                 phrase_to_file/2,
                 phrase_to_file/3,
                 phrase_to_stream/2
@@ -19,26 +18,43 @@
 :- use_module(library(dcgs)).
 :- use_module(library(error)).
 :- use_module(library(freeze)).
-:- use_module(library(iso_ext), [setup_call_cleanup/3, partial_string/3]).
-:- use_module(library(lists), [member/2, maplist/2]).
+:- use_module(library(gensym)).
+:- use_module(library(iso_ext), [
+    bb_get/2, bb_put/2, setup_call_cleanup/3, partial_string/3, partial_string_tail/2
+]).
+:- use_module(library(lists), [append/3, length/2, member/2, maplist/2]).
 :- use_module(library(charsio), [get_n_chars/3]).
 
 :- meta_predicate(phrase_from_file(2, ?)).
 :- meta_predicate(phrase_from_file(2, ?, ?)).
+:- meta_predicate(phrase_from_stream(2, ?)).
 :- meta_predicate(phrase_to_file(2, ?)).
 :- meta_predicate(phrase_to_file(2, ?, ?)).
 :- meta_predicate(phrase_to_stream(2, ?)).
 
-/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
-   phrase_from_file(GRBody, File)
 
-   True if grammar rule body GRBody covers the contents of File,
-   represented as a list of characters.
-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
+%% phrase_from_stream(+GRBody, +Stream)
+%
+%  True if grammar rule body GRBody covers the contents of the stream,
+%  represented as a list of characters.
+
+phrase_from_stream(GRBody, Stream) :-
+    stream_property(Stream, reposition(Reposition)),
+    stream_to_lazy_list(Reposition, Stream, Ls),
+    phrase(GRBody, Ls).
+
+%% phrase_from_file(+GRBody, +File)
+%
+%  True if grammar rule body GRBody covers the contents of File,
+%  represented as a list of characters.
 
 phrase_from_file(NT, File) :-
     phrase_from_file(NT, File, []).
 
+%% phrase_from_file(+GRBody, +File, +Options)
+%
+%  Like `phrase_from_file/2`, using Options to open the file.
+
 phrase_from_file(NT, File, Options) :-
     (   var(File) -> instantiation_error(phrase_from_file/3)
     ;   must_be(list, Options),
@@ -48,43 +64,145 @@ phrase_from_file(NT, File, Options) :-
             member(Type, [text,binary])
         ;   Type = text
         ),
-        setup_call_cleanup(open(File, read, Stream, [reposition(true)|Options]),
-                           (   stream_to_lazy_list(Stream, Xs),
-                               phrase(NT, Xs) ),
-                           close(Stream))
-   ).
+        setup_call_cleanup(
+            open(File, read, Stream, [reposition(true)|Options]),
+            phrase_from_stream(NT, Stream),
+            close(Stream)
+        )
+    ).
 
+% How many chars to read from stream and buffer in each step
+chars_to_read(4096).
 
-stream_to_lazy_list(Stream, Xs) :-
-        stream_property(Stream, position(Pos)),
-        freeze(Xs, reader_step(Stream, Pos, Xs)).
+stream_to_lazy_list(Reposition, Stream, Ls) :-
+    get_stream_buffer_position(Reposition, Stream, Pos),
+    freeze(Ls, render_step(Reposition, Stream, Pos, Ls)).
 
-reader_step(Stream, Pos, Xs0) :-
-        set_stream_position(Stream, Pos),
-        (   at_end_of_stream(Stream)
-        ->  Xs0 = []
-        ;   get_n_chars(Stream, 4096, Cs),
-            partial_string(Cs, Xs0, Xs),
-            stream_to_lazy_list(Stream, Xs)
-        ).
+render_step(Reposition, Stream, Pos, Ls) :-
+    set_stream_buffer_position(Reposition, Stream, Pos),
+    (   buffer_at_end_of_stream(Reposition, Stream) ->
+        Ls = []
+    ;   chars_to_read(CharsToRead),
+        buffer_get_n_chars(Reposition, Stream, CharsToRead, Chars),
+        partial_string(Chars, Ls, Ls0),
+        stream_to_lazy_list(Reposition, Stream, Ls0)
+    ).
 
-/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
-   phrase_to_stream(+GRBody, +Stream)
+buffer_at_end_of_stream(true, Stream) :- at_end_of_stream(Stream).
+buffer_at_end_of_stream(false, Stream) :-
+    stream_bufferids(Stream, _, BufferPosId, _),
+    bb_get(BufferPosId, Pos),
+    Pos = eof.
 
-   Emit the list of characters described by the grammar rule body
-   GRBody to Stream.
+get_stream_buffer_position(true, Stream, Pos) :-
+    stream_property(Stream, position(Pos)).
+get_stream_buffer_position(false, Stream, Pos) :-
+    stream_bufferids(Stream, _, BufferPosId, _),
+    bb_get(BufferPosId, Pos).
 
-   An ideal implementation of phrase_to_stream/2 writes each character
-   as soon as it becomes known and no choice-points remain, and thus
-   avoids the manifestation of the entire string in memory. See #691
-   for more information.
+set_stream_buffer_position(true, Stream, Pos) :-
+    set_stream_position(Stream, Pos).
+set_stream_buffer_position(false, Stream, Pos) :-
+    stream_bufferids(Stream, _, BufferPosId, _),
+    bb_put(BufferPosId, Pos).
 
-   The current preliminary implementation is provided so that Prolog
-   programmers can already get used to describing output with DCGs,
-   and then writing it to a file when necessary. This simple
-   implementation suffices as long as the entire contents can be
-   represented in memory, and thus covers a large number of use cases.
-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
+buffer_get_n_chars(true, Stream, N, Chars) :-
+    get_n_chars(Stream, N, Chars).
+buffer_get_n_chars(false, Stream, N, Chars) :-
+    stream_bufferids(Stream, BufferId, BufferPosId, BufferLenId),
+    buffer_prepare_for_n(Stream, BufferId, BufferPosId, BufferLenId, N),
+    bb_get(BufferId, Buffer),
+    bb_get(BufferPosId, BufferPos),
+    (   BufferPos = eof ->
+        Chars = []
+    ;   string_get_n_chars(Buffer, BufferPos, N, Chars),
+        length(Chars, NChars),
+        (   NChars = 0 ->
+            BufferPos1 = eof
+        ;   BufferPos1 is BufferPos + NChars
+        ),
+        bb_put(BufferPosId, BufferPos1)
+    ).
+
+buffer_prepare_for_n(Stream, BufferId, BufferPosId, BufferLenId, N) :-
+    bb_get(BufferPosId, BufferPos),
+    bb_get(BufferLenId, BufferLen),
+    (   BufferLen < BufferPos + N ->
+        bb_get(BufferId, Buffer),
+        (
+            (   var(Buffer) ->
+                BufferTail = Buffer
+            ;   partial_string_last_tail(Buffer, BufferTail)
+            ) ->
+            (   at_end_of_stream(Stream) ->
+                BufferTail = [],
+                bb_put(BufferId, Buffer)
+            ;   chars_to_read(CharsToRead),
+                get_n_chars(Stream, CharsToRead, Chars),
+                length(Chars, NChars),
+                partial_string(Chars, BufferTail, _),
+                bb_put(BufferId, Buffer),
+                BufferLen1 is BufferLen + NChars,
+                bb_put(BufferLenId, BufferLen1),
+                buffer_prepare_for_n(Stream, BufferId, BufferPosId, BufferLenId, N)
+            )
+        ;   true
+        )
+    ;   true
+    ).
+
+partial_string_last_tail(PartialString, PartialStringTail) :-
+    partial_string_tail(PartialString, PartialStringTail0),
+    (   var(PartialStringTail0) ->
+        PartialStringTail = PartialStringTail0
+    ;   partial_string_last_tail(PartialStringTail0, PartialStringTail)
+    ).
+
+string_get_n_chars(String, Pos, N, Chars) :-
+    '$skip_max_list'(_, Pos, String, String1),
+    '$skip_max_list'(N1, N, String1, _),
+    length(Chars, N1),
+    append(Chars, _, String1).
+
+stream_bufferids(Stream, BufferId, BufferPosId, BufferLenId) :-
+    (   bb_get(streams_buffers, _) ->
+        true
+    ;   bb_put(streams_buffers, [])
+    ),
+    bb_get(streams_buffers, StreamsBuffers),
+    (   member(
+            stream_buffer(Stream, BufferId, BufferPosId, BufferLenId),
+            StreamsBuffers
+        ) ->
+        true
+    ;   gensym(buffer, BufferId),
+        gensym(buffer_pos, BufferPosId),
+        gensym(buffer_len, BufferLenId),
+        bb_put(
+            streams_buffers,
+            [stream_buffer(Stream, BufferId, BufferPosId, BufferLenId)|StreamsBuffers]
+        ),
+        bb_put(BufferId, _),
+        bb_put(BufferPosId, 0),
+        bb_put(BufferLenId, 0)
+    ).
+
+%% phrase_to_stream(+GRBody, +Stream)
+%
+%  Emit the list of characters described by the grammar rule body
+%  GRBody to Stream.
+%
+%  An ideal implementation of `phrase_to_stream/2` writes each
+%  character as soon as it becomes known and no choice-points remain,
+%  and thus avoids the manifestation of the entire string in memory.
+%  See [#691](https://github.com/mthom/scryer-prolog/issues/691) for
+%  more information.
+%
+%  The current preliminary implementation is provided so that Prolog
+%  programmers can already get used to describing output with DCGs,
+%  and then writing it to a file when necessary. This simple
+%  implementation suffices as long as the entire contents can be
+%  represented in memory, and thus covers a large number of use cases.
 
 phrase_to_stream(GRBody, Stream) :-
         phrase(GRBody, Cs),
@@ -101,14 +219,18 @@ phrase_to_stream(GRBody, Stream) :-
         % maplist(put_char(Stream), Cs). It also works for binary streams.
         '$put_chars'(Stream, Cs).
 
-/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
-   phrase_to_file(+GRBody, +File), writing the string described
-   by GRBody to File.
-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
+%% phrase_to_file(+GRBody, +File)
+%
+%  Write the string described by GRBody to File.
 
 phrase_to_file(GRBody, File) :-
         phrase_to_file(GRBody, File, []).
 
+
+%% phrase_to_file(+GRBody, +File, +Options)
+%
+%  Like `phrase_to_file/2`, using Options to open the file.
+
 phrase_to_file(GRBody, File, Options) :-
         setup_call_cleanup(open(File, write, Stream, Options),
                            phrase_to_stream(GRBody, Stream),

src/lib/process.pl

@@ -0,0 +1,224 @@
+:- module(process, [
+    process_create/3, 
+    process_id/2,
+    process_release/1, 
+    process_wait/2, 
+    process_wait/3, 
+    process_kill/1
+]).
+
+:- use_module(library(error)).
+:- use_module(library(iso_ext)).
+:- use_module(library(lists), [member/2, maplist/2, maplist/3, append/2]).
+:- use_module(library(reif), [tfilter/3, memberd_t/3]).
+
+
+%% process_create(+Exe, +Args:list, +Options).
+%
+% Create a new process by executing the executable Exe and passing it the Arguments Args.
+%
+% Note: On windows please take note of [windows argument splitting](https://doc.rust-lang.org/std/process/index.html#windows-argument-splitting).
+% 
+% Options is a list consisting of the following options:
+%
+%  * `cwd(+Path)`           Set the processes working directory to `Path`
+%  * `process(-Process)`    `Process` will be assigned a process handle for the spawned process
+%  * `env(+List)`           Don't inherit environment variables and set the variables defined in `List`
+%  * `environment(+List)`   Inherit environment variables and set/override the variables defined in `List`
+%  * `stdin(Spec)`, `stdout(Spec)` or `stderr(Spec)` defines how to redirect the spawned processes io streams
+%
+%  The elements of `List` in `env(List)`/`environment(List)` List must be string pairs using `=/2`.
+%  `env/1` and `environment/1` may not be both specified.
+%
+%  The following stdio `Spec` are available:
+%  
+%  * `std`          inherit the current processes original stdio streams (does currently not account for stdio being changed by `set_input` or `set_output`)
+%  * `file(+Path)`  attach the strea to the file at `Path`
+%  * `null`         discards writes and behaves as eof for read. Equivalent to using `file(/dev/null)`
+%  * `pipe(-Steam)` create a new pipe and assigne one end to the created process and the other end to `Stream`
+%
+% Specifying an option multiple times is an error, when an option is not specified the following defaults apply:
+%
+% - `cwd(".")`
+% - `environment([])`
+% - `stdin(std)`, `stdout(std)`, `stderr(std)`
+%
+process_create(Exe, Args, Options) :- call_with_error_context(process_create_(Exe, Args, Options), predicate-process_create/3).
+
+process_create_(Exe, Args, Options) :-
+    must_be(chars, Exe),
+    must_be(list, Args),
+    maplist(must_be(chars), Args),
+    must_be(list, Options),
+    check_options(
+        [
+            option([stdin], valid_stdio, stdin(std), stdin(Stdin)),
+            option([stdout], valid_stdio, stdout(std), stdout(Stdout)),
+            option([stderr], valid_stdio, stderr(std), stderr(Stderr)),
+            option([env, environment], valid_env, environment([]), Env),
+            option([process], valid_uninit_process, process(_), process(Process)),
+            option([cwd], valid_cwd, cwd("."), cwd(Cwd))
+        ],
+        Options,
+        process_create_option
+    ),
+    Stdin =.. Stdin1,
+    Stdout =.. Stdout1,
+    Stderr =.. Stderr1,
+    simplify_env(Env, Env1),
+    '$process_create'(Exe, Args, Stdin1, Stdout1, Stderr1, Env1, Cwd, Process).
+
+%% process_id(+Process, -Pid).
+%
+process_id(Process, Pid) :- call_with_error_context(process_id_(Process, Pid), predicate-process_id/2).
+
+process_id_(Process, Pid) :-
+    valid_process(Process),
+    must_be(var, Pid),
+    '$process_id'(Process, Pid).
+
+%% process_wait(+Process, Status).
+%
+% See `process_create/3` with `Options = []`
+%
+process_wait(Process, Status) :- call_with_error_context(process_wait(Process, Status, []), predicate-process_wait/2).
+
+
+%% process_wait(+Process, Status, Options).
+%
+% Wait for the process behind the process handle `Process` to exit.
+%
+% When the process exits regulary `Status` will be unified with `exit(Exit)` where `Exit` is the processes exit code.
+% When the process exits was killed `Status` will be unified with `killed(Signal)` where `Signal` is the signal number that killed the process.
+% When the process doesn't exit before the timeout `Status` will be unified with `timeout`.
+%
+% `Options` is a a list of the following options
+%
+%  * timeout(Timeout) supported values for `Timeout` are 0 or `infinite`
+%
+% Each options may be specified at most once, when an option is not specified the following defaults apply:
+%
+% - timeout(infinite)
+%
+process_wait(Process, Status, Options) :- call_with_error_context(process_wait_(Process, Status, Options), predicate-process_wait/3).
+
+process_wait_(Process, Status, Options) :- 
+    valid_process(Process),
+    check_options(
+        [
+            option([timeout], valid_timeout, timeout(infinite), timeout(Timeout))
+        ],
+        Options,
+        process_wait_option
+    ),
+    '$process_wait'(Process, Exit, Timeout),
+    Exit = Status.
+
+valid_timeout(timeout(infinite)).
+valid_timeout(timeout(0)).
+
+
+%% process_kill(+Process).
+%
+% Kill the process using the process handle `Process`.
+% On Unix this sends SIGKILL.
+%
+% Only works for processes spawned with `process_create/3` that have not yet been release with `process_release/1`
+% 
+process_kill(Process) :- call_with_error_context(process_kill_(Process), predicate-process_kill/1).
+
+process_kill_(Process) :- 
+    valid_process(Process),
+    '$process_kill'(Process).
+
+%% process_release(+Process)
+%
+% wait for the process to exit (if not already) and release process handle `Process`
+%
+% It's an error if `Process` is not a valid process handle
+%
+process_release(Process) :- call_with_error_context(process_release_(Process), predicate-process_release/1).
+
+process_release_(Process) :- 
+    valid_process(Process),
+    process_wait(Process, _),
+    '$process_release'(Process).
+
+
+must_be_known_options(Valid, Options, Domain) :- call_with_error_context(must_be_known_options_(Valid, [], Options, Domain),predicate-must_be_known_options/3).
+
+must_be_known_options_(_, _,  [], _).
+must_be_known_options_(Valid, Found, [X|XS], Domain) :-
+    ( functor(X, Option, 1) -> true
+    ; domain_error(Domain, X, [])
+    ) ,
+    ( member(Option, Found) -> domain_error(non_duplicate_options, Option , [])
+    ; member(Option, Valid) -> true 
+    ; domain_error(Domain, Option, [])
+    ),
+    must_be_known_options_(Valid, [Option | Found], XS, Domain).
+
+check_options(KnownOptions, Options, Domain) :- call_with_error_context(check_options_(KnownOptions, Options, Domain), predicate-check_options/3).
+
+check_options_(KnownOptions, Options, Domain) :-
+    maplist(option_names, KnownOptions, Namess),
+    append(Namess, Names),
+    must_be_known_options(Names, Options, Domain),
+    extract_options(KnownOptions, Options).
+
+option_names(option(Names,_,_,_), Names).
+
+extract_options(KnownOptions, Options) :- call_with_error_context(extract_options_(KnownOptions, Options), predicate-extract_options/2).
+
+extract_options_([], _).
+extract_options_([X | XS], Options) :- 
+    option(Kinds, Pred, Default, Choice) = X,
+    tfilter(find_option(Kinds), Options, Solutions),
+    ( Solutions = [] -> Choice = Default
+    ; Solutions = [Provided] -> functor(Pred, Name, Arity), ArityP1 is Arity+1, call_with_error_context(call(Pred, Provided),predicate-Name/ArityP1), Choice = Provided 
+    ; domain_error(non_conflicting_options, Solutions, [])
+    ),
+    extract_options_(XS, Options).
+
+find_option(Names, Found, T) :- 
+    functor(Found, Name, 1), 
+    memberd_t(Name, Names, T).
+
+valid_stdio(IO) :- arg(1, IO, Arg), 
+    ( var(Arg) -> instantiation_error([]) 
+    ; valid_stdio_(Arg) -> true 
+    ; domain_error(stdio_spec, Arg, [])
+    ).
+
+valid_stdio_(std).
+valid_stdio_(null).
+valid_stdio_(pipe(Stream)) :- must_be(var, Stream).
+valid_stdio_(file(Path)) :- must_be(chars, Path).
+
+valid_env(env(E)) :- 
+    must_be(list, E),
+    ( valid_env_(E) -> true 
+    ; domain_error(process_create_option, env(E), [])
+    ).
+valid_env(environment(E)) :- 
+    must_be(list, E),
+    ( valid_env_(E) -> true 
+    ; domain_error(process_create_option, environment(E), [])
+    ).
+
+valid_env_([]).
+valid_env_([N=V|ES]) :- 
+    must_be(chars, N), 
+    must_be(chars, V),
+    valid_env_(ES).
+
+valid_uninit_process(process(Process)) :- must_be(var, Process).
+
+valid_process(Process) :- var(Process) -> instantiation_error([]) ; true.
+
+valid_cwd(cwd(Cwd)) :- must_be(chars, Cwd).
+
+simplify_env(E, [Kind, Envs1]) :- E =.. [Kind, Envs], simplify_env_(Envs, Envs1).
+
+simplify_env_([],[]).
+simplify_env_([N=V|E],[[N, V]|E1]) :- simplify_env_(E, E1).

src/lib/random.pl

@@ -1,24 +1,36 @@
-:- module(random, [maybe/0, random/1, random_integer/3, set_random/1]).
+/**
+This library provides probabilistic predicates and random number generators.
 
-/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
-   To retain desirable declarative properties, predicates that internally
-   use random numbers should be equipped with an argument that specifies
-   the random seed. This makes everything completely reproducible.
-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
+To retain desirable declarative properties, predicates that internally
+use random numbers should be equipped with an argument that specifies
+the random seed. This makes everything completely reproducible.
+*/
+
+:- module(random, [maybe/0, random/1, random_integer/3, set_random/1]).
 
 :- use_module(library(error)).
 
-% succeeds with probability 0.5.
+%% maybe.
+% 
+% Succeeds with probability 0.5.
 maybe :- '$maybe'.
 
-% The higher the precision, the slower it gets.
-random_number_precision(64).
-
+%% random(-R).
+%
+% Generates a random floating number between 0 (inclusive) and 1 (exclusive).
 random(R) :-
     var(R),
-    random_number_precision(N),
-    rnd(N, R).
+    N is 2^50,
+    '$random_integer'(0, N, K),
+    R is K/N.
 
+%% random_integer(+Lower, +Upper, -R).
+%
+% Generates a random integer number between Lower (inclusive) and Upper (exclusive).
+%
+% Throws `instantiation_error` if Lower or Upper are variables.
+%
+% Throws `type_error` if Lower or Upper aren't integers.
 random_integer(Lower, Upper, R) :-
     var(R),
     (   (var(Lower) ; var(Upper)) ->
@@ -27,25 +39,14 @@ random_integer(Lower, Upper, R) :-
         type_error(integer, Lower, random_integer/3)
     ;   \+ integer(Upper) ->
         type_error(integer, Upper, random_integer/3)
-    ;   Upper > Lower,
-        random(R0),
-        R is floor((Upper - Lower) * R0 + Lower)
+    ;   Lower < Upper,
+        '$random_integer'(Lower, Upper, R)
     ).
 
-rnd(N, R) :-
-    rnd_(N, 0, R).
-
-rnd_(0, R, R) :- !.
-rnd_(N, R0, R) :-
-    maybe,
-    !,
-    N1 is N - 1,
-    rnd_(N1, R0, R).
-rnd_(N, R0, R) :-
-    N1 is N - 1,
-    R1 is R0 + 1.0 / 2.0 ^ N,
-    rnd_(N1, R1, R).
-
+%% set_random(+Seed).
+%
+% Sets a seed that will be used for subsequent random generations in this library.
+% It's necessary to set a seed to provide reproducible executions using this library.
 set_random(Seed) :-
     (   nonvar(Seed) ->
         (  Seed = seed(S) ->

src/lib/reif.pl

@@ -1,3 +1,16 @@
+/** Predicates from [*Indexing dif/2*](https://arxiv.org/abs/1607.01590).
+
+Example:
+
+```
+?- tfilter(=(a), [X,Y], Es).
+   X = a, Y = a, Es = "aa"
+;  X = a, Es = "a", dif:dif(a,Y)
+;  Y = a, Es = "a", dif:dif(a,X)
+;  Es = [], dif:dif(a,X), dif:dif(a,Y).
+```
+*/
+
 :- module(reif, [if_/3, (=)/3, (',')/3, (;)/3, cond_t/3, dif/3,
 		         memberd_t/3, tfilter/3, tmember/2, tmember_t/3,
 		         tpartition/4]).

src/lib/sgml.pl

@@ -2,57 +2,70 @@
    Predicates for parsing HTML and XML documents.
    Written 2020-2022 by Markus Triska (triska@metalevel.at)
    Part of Scryer Prolog.
-
-   Currently, two predicates are provided:
-
-     -  load_html(+Source, -Es, +Options)
-     -  load_xml(+Source, -Es, +Options)
-
-   These predicates parse HTML and XML documents, respectively.
-
-   Source must be one of:
-
-     - a list of characters with the document contents
-     - stream(S), specifying a stream S from which to read the content
-     - file(Name), where Name is a list of characters specifying a file name.
-
-   Es is unified with the abstract syntax tree of the parsed document,
-   represented as a list of elements where each is of the form:
-
-     * a list of characters, representing text
-     * element(Name, Attrs, Children)
-       - Name, an atom, is the name of the tag
-       - Attrs is a list of Key=Value pairs:
-         Key is an atom, and Value is a list of characters
-       - Children is a list of elements as specified here.
-
-   Currently, Options are ignored. In the future, more options may be
-   provided to control parsing.
-
-   Example:
-
-      ?- load_html("<html><head><title>Hello!</title></head></html>", Es, []).
-
-   Yielding:
-
-         Es = [element(html,[],
-                [element(head,[],
-                  [element(title,[],
-                    ["Hello!"])]),
-                 element(body,[],[])])].
-
-   library(xpath) provides convenient reasoning about parsed documents.
-   For example, to fetch the title of the document above, we can use:
-
-      ?- load_html("<html><head><title>Hello!</title></head></html>", Es, []),
-         xpath(Es, //title(text), T).
-
-   Yielding T = "Hello!".
-
-   Use http_open/3 from library(http/http_open) to read answers from
-   web servers via streams.
 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
 
+/** Predicates for parsing HTML and XML documents.
+
+Currently, two predicates are provided:
+
+  -  `load_html(+Source, -Es, +Options)`
+  -  `load_xml(+Source, -Es, +Options)`
+
+These predicates parse HTML and XML documents, respectively.
+
+Source must be one of:
+
+  - a list of characters with the document contents
+  - `stream(S)`, specifying a stream S from which to read the content
+  - `file(Name)`, where Name is a list of characters specifying a file name.
+
+Es is unified with the abstract syntax tree of the parsed document,
+represented as a list of elements where each is of the form:
+
+  * a list of characters, representing text
+
+  * `element(Name, Attrs, Children)`
+
+    - `Name`, an atom, is the name of the tag
+
+    - `Attrs` is a list of `Key=Value` pairs:
+      `Key` is an atom, and `Value` is a list of characters
+
+    - `Children` is a list of elements as specified here.
+
+Currently, Options are ignored. In the future, more options may be
+provided to control parsing.
+
+Example:
+
+```
+   ?- load_html("<html><head><title>Hello!</title></head></html>", Es, []).
+```
+
+Yielding:
+
+```
+      Es = [element(html,[],
+             [element(head,[],
+               [element(title,[],
+                 ["Hello!"])]),
+              element(body,[],[])])].
+```
+
+`library(xpath)` provides convenient reasoning about parsed documents.
+For example, to fetch the title of the document above, we can use:
+
+```
+   ?- load_html("<html><head><title>Hello!</title></head></html>", Es, []),
+      xpath(Es, //title(text), T).
+```
+
+Yielding `T = "Hello!"`.
+
+Use `http_open/3` from `library(http/http_open)` to read answers from
+web servers via streams.
+*/
+
 :- module(sgml, [load_html/3,
                  load_xml/3]).
 
@@ -83,14 +96,14 @@ is_sgml_source([]).
 is_sgml_source([C|Cs]) :- must_be(chars, [C|Cs]).
 
 load_structure_([], [], _, _).
-load_structure_([C|Cs], [E], Options, What) :-
-        load_(What, [C|Cs], E, Options).
-load_structure_(file(Fs), [E], Options, What) :-
+load_structure_([C|Cs], [E|Es], Options, What) :-
+        load_(What, [C|Cs], [E|Es], Options).
+load_structure_(file(Fs), [E|Es], Options, What) :-
         once(phrase_from_file(seq(Cs), Fs)),
-        load_(What, Cs, E, Options).
-load_structure_(stream(Stream), [E], Options, What) :-
+        load_(What, Cs, [E|Es], Options).
+load_structure_(stream(Stream), [E|Es], Options, What) :-
         get_n_chars(Stream, _, Cs),
-        load_(What, Cs, E, Options).
+        load_(What, Cs, [E|Es], Options).
 
 load_(html, Cs, E, Options) :- '$load_html'(Cs, E, Options).
 load_(xml, Cs, E, Options) :- '$load_xml'(Cs, E, Options).

src/lib/si.pl

@@ -1,34 +1,46 @@
-/* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
 
-   Safe type tests
-   ===============
+/** Safe type tests.
 
-   "si" stands for "sufficiently instantiated".
+   "si" stands for "sufficiently instantiated". It can also be read as
+   "safe inference", so possibly also other predicates are candidates
+   for this library.
 
-   These predicates:
+   A safe type test:
 
-    - throw instantiation errors if the argument is
+    - throws an *instantiation error* if the argument is
       not sufficiently instantiated to make a sound decision
-    - succeed if the argument is of the specified type
-    - fail otherwise.
+    - *succeeds* if the argument is of the specified type
+    - *fails* otherwise.
 
-   For instance, atom_si(A) yields an *instantiation error* if A is a
+   For instance, `atom_si(A)` yields an *instantiation error* if `A` is a
    variable. This is logically sound, since in that case the argument
    is not sufficiently instantiated to make any decision.
 
-   The definitions are taken from:
+   The definitions are taken from [Safer type tests in Prolog](https://stackoverflow.com/questions/27306453/safer-type-tests-in-prolog).
 
-       https://stackoverflow.com/questions/27306453/safer-type-tests-in-prolog
+   Examples:
 
-   "si" can also be read as "safe inference", so possibly also other
-   predicates are candidates for this library.
-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
+```
+?- chars_si(Cs).
+   error(instantiation_error,list_si/1).
+?- chars_si([h|Cs]).
+   error(instantiation_error,list_si/1).
+?- chars_si("hello").
+   true.
+?- chars_si(hello).
+   false.
+```
+*/
 
 :- module(si, [atom_si/1,
                integer_si/1,
                atomic_si/1,
                list_si/1,
-               chars_si/1]).
+               character_si/1,
+               term_si/1,
+               chars_si/1,
+               dif_si/2,
+               when_si/2]).
 
 :- use_module(library(lists)).
 
@@ -53,6 +65,65 @@ list_si(L0) :-
    ;  throw(error(instantiation_error, list_si/1))
    ).
 
-chars_si(Cs) :-
-   list_si(Cs),
-   '$is_partial_string'(Cs).
+character_si(Ch) :-
+   functor(Ch,Ch,0),
+   atom(Ch),
+   atom_length(Ch,1).
+
+term_si(Term) :-
+   (   ground(Term) -> acyclic_term(Term)
+   ;   throw(error(instantiation_error, term_si/1))
+   ).
+
+chars_si(Chs0) :-
+   '$skip_max_list'(_,_, Chs0,Chs),
+   ( nonvar(Chs) -> Chs == [] ; true ), % fails for infinite lists too
+   failnochars(Chs0, Uninstantiated),
+   (  nonvar(Uninstantiated)
+   -> throw(error(instantiation_error, chars_si/1))
+   ;  true
+   ).
+
+failnochars(Chs0, U) :-
+   (  var(Chs0) -> U = true
+   ;  Chs0 == [] -> true
+   ;  Chs0 = [Ch|Chs1],
+      (  nonvar(Ch) -> atom(Ch), atom_length(Ch,1)
+      ;  U = true
+      ),
+      failnochars(Chs1, U)
+   ).
+
+dif_si(X, Y) :-
+   X \== Y,
+   ( X \= Y -> true
+   ; throw(error(instantiation_error,dif_si/2))
+   ).
+
+:- meta_predicate(when_si(+, 0)).
+
+%% when_si(Condition, Goal).
+%
+% Executes Goal when Condition becomes true. Throws an instantiation error if
+% it can't decide.
+when_si(Condition, Goal) :-
+    % Taken from https://stackoverflow.com/a/40449516
+    (   when_condition_si(Condition) ->
+        (   Condition ->
+            Goal
+        ;   throw(error(instantiation_error,when_si/2))
+        )
+    ;   throw(error(domain_error(when_condition_si, Condition),_))
+    ).
+
+when_condition_si(Cond) :-
+    var(Cond), !, throw(error(instantiation_error,when_condition_si/2)).
+when_condition_si(ground(_)).
+when_condition_si(nonvar(_)).
+when_condition_si((A, B)) :-
+    when_condition_si(A),
+    when_condition_si(B).
+when_condition_si((A ; B)) :-
+    when_condition_si(A),
+    when_condition_si(B).
+

src/lib/simplex.pl

@@ -77,9 +77,9 @@ thesis project, for example.
 A *linear programming problem* or simply *linear program* (LP)
 consists of:
 
-  - a set of _linear_ **constraints**
-  - a set of **variables**
-  - a _linear_ **objective function**.
+  - a set of _linear_ *constraints*
+  - a set of *variables*
+  - a _linear_ *objective function*.
 
 The goal is to assign values to the variables so as to _maximize_ (or
 minimize) the value of the objective function while satisfying all
@@ -107,10 +107,10 @@ non-negativity constraints should therefore be stated explicitly.
 This is the "radiation therapy" example, taken from _Introduction to
 Operations Research_ by Hillier and Lieberman.
 
-[**Prolog DCG notation**](https://www.metalevel.at/prolog/dcg) is
+[*Prolog DCG notation*](https://www.metalevel.at/prolog/dcg) is
 used to _implicitly_ thread the state through posting the constraints:
 
-==
+```
 :- use_module(library(simplex)).
 :- use_module(library(dcgs)).
 
@@ -125,15 +125,15 @@ post_constraints -->
 	constraint([0.6*x1, 0.4*x2] >= 6),
 	constraint([x1] >= 0),
 	constraint([x2] >= 0).
-==
+```
 
 An example query:
 
-==
+```
 ?- radiation(S), variable_value(S, x1, Val1),
 		 variable_value(S, x2, Val2).
    S = solved(...), Val1 = 15 rdiv 2, Val2 = 9 rdiv 2.
-==
+```
 
 ## Example 2     {#simplex-ex-2}
 
@@ -143,7 +143,7 @@ Here is an instance of the knapsack problem described above, where `C
 variables, `x(1)` and `x(2)` that denote how many items to take of
 each type.
 
-==
+```
 :- use_module(library(simplex)).
 
 knapsack(S) :-
@@ -155,15 +155,15 @@ knapsack_constraints(S) :-
 	constraint([6*x(1), 4*x(2)] =< 8, S0, S1),
 	constraint([x(1)] =< 1, S1, S2),
 	constraint([x(2)] =< 2, S2, S).
-==
+```
 
 An example query yields:
 
-==
+```
 ?- knapsack(S), variable_value(S, x(1), X1),
 	        variable_value(S, x(2), X2).
    S = solved(...), X1 = 1 rdiv 1, X2 = 1 rdiv 2.
-==
+```
 
 That is, we are to take the one item of the first type, and half of one of
 the items of the other type to maximize the total value of items in the
@@ -171,23 +171,23 @@ knapsack.
 
 If items can not be split, integrality constraints have to be imposed:
 
-==
+```
 knapsack_integral(S) :-
 	knapsack_constraints(S0),
 	constraint(integral(x(1)), S0, S1),
 	constraint(integral(x(2)), S1, S2),
 	maximize([7*x(1), 4*x(2)], S2, S).
-==
+```
 
 Now the result is different:
 
-==
+```
 ?- knapsack_integral(S), variable_value(S, x(1), X1),
 			 variable_value(S, x(2), X2).
 
 X1 = 0
 X2 = 2
-==
+```
 
 That is, we are to take only the _two_ items of the second type.
 Notice in particular that always choosing the remaining item with best
@@ -207,7 +207,7 @@ The task is to find a _minimal_ number of these coins that amount to
 111 units in total. We introduce variables `c(1)`, `c(5)` and `c(20)`
 denoting how many coins to take of the respective type:
 
-==
+```
 :- use_module(library(simplex)).
 
 coins(S) :-
@@ -226,16 +226,16 @@ coins -->
 	constraint(integral(c(5))),
 	constraint(integral(c(20))),
 	minimize([c(1), c(5), c(20)]).
-==
+```
 
 An example query:
 
-==
+```
 ?- coins(S), variable_value(S, c(1), C1),
 	     variable_value(S, c(5), C5),
 	     variable_value(S, c(20), C20).
    S = solved(...), C1 = 1 rdiv 1, C5 = 2 rdiv 1, C20 = 5 rdiv 1.
-==
+```
 
 @author [Markus Triska](https://www.metalevel.at)
 */

src/lib/sockets.pl

@@ -1,4 +1,9 @@
-
+/**
+Predicates for handling network sockets, both as a server and as a client.
+As a server, you should open a socket an call `socket_server_accept/4` to get a stream for each connection.
+As a client, you should just open a socket and you will receive a stream.
+In both cases, with a stream, you can use the usual predicates to read and write to the stream.
+*/
 :- module(sockets, [socket_client_open/3,
                     socket_server_open/2,
                     socket_server_accept/4,
@@ -7,6 +12,18 @@
 
 :- use_module(library(error)).
 
+%% socket_client_open(+Addr, -Stream, +Options).
+%
+% Open a socket to a server, returning a stream. Addr must satisfy `Addr = Address:Port`.
+%
+% The following options are available:
+%
+%  * `alias(+Alias)`: Set an alias to the stream
+%  * `eof_action(+Action)`: Defined what happens if the end of the stream is reached. Values: `error`, `eof_code` and `reset`.
+%  * `reposition(+Boolean)`: Specifies whether repositioning is required for the stream. `false` is the default.
+%  * `type(+Type)`: Type can be `text` or `binary`. Defines the type of the stream, if it's optimized for plain text
+%    or just binary
+%
 socket_client_open(Addr, Stream, Options) :-
     (  var(Addr) ->
        throw(error(instantiation_error, socket_client_open/3))
@@ -27,7 +44,11 @@ socket_client_open(Addr, Stream, Options) :-
                                   socket_client_open/3),
     '$socket_client_open'(Address, Port, Stream, Alias, EOFAction, Reposition, Type).
 
-
+%% socket_server_open(+Addr, -ServerSocket).
+%
+% Open a server socket, returning a ServerSocket. Use that ServerSocket to accept incoming connections in
+% `socket_server_accept/4`. Addr must satisfy `Addr = Address:Port`. Depending on the operating system
+% configuration, some ports might be reserved for superusers.
 socket_server_open(Addr, ServerSocket) :-
     must_be(var, ServerSocket),
     (  ( integer(Addr) ; var(Addr) ) ->
@@ -39,7 +60,19 @@ socket_server_open(Addr, ServerSocket) :-
        '$socket_server_open'(Address, Port, ServerSocket)
     ).
 
-
+%% socket_server_accept(+ServerSocket, -Client, -Stream, +Options).
+%
+% Given a ServerSocket and a list of Options, accepts a incoming connection, returning data from the Client and
+% a Stream to read or write data.
+%
+% The following options are available:
+%
+%  * `alias(+Alias)`: Set an alias to the stream
+%  * `eof_action(+Action)`: Defined what happens if the end of the stream is reached. Values: `error`, `eof_code` and `reset`.
+%  * `reposition(+Boolean)`: Specifies whether repositioning is required for the stream. `false` is the default.
+%  * `type(+Type)`: Type can be `text` or `binary`. Defines the type of the stream, if it's optimized for plain text
+%    or just binary
+% 
 socket_server_accept(ServerSocket, Client, Stream, Options) :-
     must_be(var, Client),
     must_be(var, Stream),
@@ -48,10 +81,14 @@ socket_server_accept(ServerSocket, Client, Stream, Options) :-
                                   socket_server_accept/4),
     '$socket_server_accept'(ServerSocket, Client, Stream, Alias, EOFAction, Reposition, Type).
 
-
+%% socket_server_close(+ServerSocket).
+%
+% Stops listening on that ServerSocket. It's recommended to always close a ServerSocket once it's no longer needed
 socket_server_close(ServerSocket) :-
     '$socket_server_close'(ServerSocket).
 
-
+%% current_hostname(-HostName).
+%
+% Returns the current hostname of the computer in which Scryer Prolog is executing right now
 current_hostname(HostName) :-
     '$current_hostname'(HostName).

src/lib/tabling.pl

@@ -1,3 +1,29 @@
+/** Tabling, also called SLG resolution.
+
+    SLG resolution is an alternative execution strategy that sometimes
+    helps to improve termination and performance characteristics of Prolog
+    predicates.
+
+    To enable this execution strategy for a Prolog predicate, add a
+    `(table)/1` directive, using the prefix operator `table` that this
+    module defines. For example, to enable tabling for the predicate
+    `p/2`, use:
+
+```
+:- use_module(library(tabling)).
+
+:- table p/2.
+
+...
+```
+
+    The possibility to apply different execution strategies is one of
+    the greatest attractions of pure Prolog code, and one of the
+    strongest arguments for keeping to the pure core of Prolog as far
+    as possible.
+
+    Scryer Prolog implements tabling as described by Desouter et al. in [*Tabling as a Library with Delimited Control*](https://www.ijcai.org/Proceedings/16/Papers/619.pdf).
+*/
 
 :- module(tabling,
 	  [ start_tabling/2,		% +Wrapper, :Worker.
@@ -67,7 +93,7 @@ table_and_status_for_variant(V,T,S) :-
   tbd_table_status(T,S).
 
 
-:- meta_predicate start_tabling(?, :).
+:- meta_predicate(start_tabling(?, :)).
 
 start_tabling(Wrapper,Worker) :-
   put_new_trie_table_link,
@@ -138,7 +164,9 @@ activate(Wrapper,Worker,T) :-
 
 delim(Wrapper,Worker,Table) :-
 %   debug(tabling, 'ACT: ~p on ~p', [Wrapper, Table]),
-    reset(Worker,SourceCall,Continuation),
+    catch(reset(Worker,SourceCall,Continuation),
+          _,
+          fail),
    ( Continuation = none ->
      (	 add_answer(Table,Wrapper)
      ->	 true %debug(tabling, 'ADD: ~p', [Wrapper])

src/lib/tabling/batched_worklist.pl

@@ -49,12 +49,20 @@
 :- use_module(library(tabling/double_linked_list)).
 
 :- use_module(library(atts)).
+:- use_module(library(dcgs)).
 :- use_module(library(lists)).
 
 :- attribute executing_all_work/1, worklist_presence/1, wkl_answer_cluster/1, wkl_suspension_cluster/1, wkl_answer_cluster_pointer_flag/1.
 
 verify_attributes(_, _, []).
 
+attribute_goals(X) -->
+    { put_atts(X, -executing_all_work(_)),
+      put_atts(X, -worklist_presence(_)),
+      put_atts(X, -wkl_answer_cluster(_)),
+      put_atts(X, -wkl_suspension_cluster(_)),
+      put_atts(X, -wkl_answer_cluster_pointer_flag(_)) }.
+
 /** <module> Tabling Worklist management
 
 A batched worklist: a worklist that  clusters suspensions and answers as

src/lib/tabling/double_linked_list.pl

@@ -49,9 +49,15 @@
 	  ]).
 
 :- use_module(library(atts)).
+:- use_module(library(dcgs)).
 
 :- attribute dll_element/1, dll_next/1, dll_prev/1.
 
+attribute_goals(X) -->
+     { put_atts(X, -dll_element(_)),
+       put_atts(X, -dll_next(_)),
+       put_atts(X, -dll_prev(_)) }.
+
 % A circular double linked list
 % =============================
 

src/lib/tabling/global_worklist.pl

@@ -9,12 +9,15 @@
 	  ]).
 
 :- use_module(library(atts)).
+:- use_module(library(dcgs)).
 :- use_module(library(iso_ext)).
 
 :- attribute table_global_worklist/1.
 
 verify_attributes(_, _, []).
 
+attribute_goals(X) --> { put_atts(X, -table_global_worklist(_)) }.
+
 put_new_global_worklist :-
   (  bb_get(table_global_worklist_initialized, _) ->
      true

src/lib/tabling/table_data_structure.pl

@@ -56,6 +56,7 @@
 :- use_module(library(tabling/batched_worklist)).
 
 :- use_module(library(atts)).
+:- use_module(library(dcgs)).
 :- use_module(library(gensym)).
 :- use_module(library(iso_ext)).
 
@@ -63,6 +64,10 @@
 
 verify_attributes(_, _, []).
 
+attribute_goals(X) -->
+    { put_atts(X, -table_status(_)),
+      put_atts(X, -newly_created_table_identifiers(_)) }.
+
 % This file defines the table datastructure.
 %
 % The table datastructure contains the following sub-structures:

src/lib/tabling/table_link_manager.pl

@@ -43,6 +43,7 @@
 	  ]).
 
 :- use_module(library(atts)).
+:- use_module(library(dcgs)).
 :- use_module(library(lists)).
 :- use_module(library(iso_ext)).
 :- use_module(library(terms)).
@@ -53,6 +54,9 @@
 
 verify_attributes(_, _, []).
 
+attribute_goals(X) -->
+    { put_atts(X, -trie_table_link(_)) }.
+
 % This file defines a call pattern trie.
 %
 % This data structure keeps the relation between a variant and the

src/lib/tabling/trie.pl

@@ -45,12 +45,17 @@
 
 :- use_module(library(assoc)).
 :- use_module(library(atts)).
+:- use_module(library(dcgs)).
 :- use_module(library(lists)).
 
 :- attribute maybe_just/1, children/1.
 
 verify_attributes(_, _, []).
 
+attribute_goals(X) -->
+    { put_atts(X, -maybe_just(_)),
+      put_atts(X, -children(_)) }.
+
 % Implementation of a prefix tree, a.k.a. trie %
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 

src/lib/time.pl

@@ -1,48 +1,18 @@
 /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
-   Written 2020, 2021 by Markus Triska (triska@metalevel.at)
+   Written 2020-2024 by Markus Triska (triska@metalevel.at)
    Part of Scryer Prolog.
-
-   This library provides predicates for reasoning about time.
-
-   current_time(T) yields the current system time in an opaque form,
-   called a time stamp. Use format_time//2 to describe strings that
-   contain attributes of the time stamp.
-
-   The nonterminal format_time//2 describes a list of characters that
-   are formatted according to a format string. Usage:
-
-     phrase(format_time(FormatString, TimeStamp), Cs)
-
-   TimeStamp represents a moment in time in an opaque form, as for
-   example obtained by current_time/1.
-
-   FormatString is a list of characters that are interpreted literally,
-   except for the following specifiers (and possibly more in the future):
-
-     %Y    year of the time stamp. Example: 2020.
-     %m    month number (01-12), zero-padded to 2 digits
-     %d    day number (01-31), zero-padded to 2 digits
-     %H    hour number (00-24), zero-padded to 2 digits
-     %M    minute number (00-59), zero-padded to 2 digits
-     %S    second number (00-60), zero-padded to 2 digits
-     %b    abbreviated month name, always 3 letters
-     %a    abbreviated weekday name, always 3 letters
-     %A    full weekday name
-     %j    day of the year (001-366), zero-padded to 3 digits
-     %%    the literal %
-
-   Example:
-
-     ?- current_time(T), phrase(format_time("%d.%m.%Y (%H:%M:%S)", T), Cs).
-        T = [...], Cs = "11.06.2020 (00:24:32)".
-
-   sleep(S) sleeps for S seconds (a floating point number).
-
-   time(Goal) reports the execution time of Goal.
-
 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
 
-:- module(time, [max_sleep_time/1, sleep/1, time/1, current_time/1, format_time//2]).
+/** This library provides predicates for reasoning about time.
+*/
+
+:- module(time, [max_sleep_time/1,
+                 sleep/1,
+                 time/1,
+                 current_time/1,
+                 format_time//2,
+                 statistics/2
+                ]).
 
 :- use_module(library(format)).
 :- use_module(library(iso_ext)).
@@ -51,10 +21,51 @@
 :- use_module(library(lists)).
 :- use_module(library(charsio), [read_from_chars/2]).
 
+
+%% current_time(-T)
+%
+%  Yields the current system time _T_ in an opaque form, called a
+%  _time stamp_. Use `format_time//2` to describe strings that contain
+%  attributes of the time stamp.
+
 current_time(T) :-
         '$current_time'(T0),
         read_from_chars(T0, T).
 
+%% format_time(FormatString, TimeStamp)//
+%
+% The nonterminal format_time//2 describes a list of characters that
+% are formatted according to a format string. Usage:
+%
+% ```
+%    phrase(format_time(FormatString, TimeStamp), Cs)
+% ```
+%
+% TimeStamp represents a moment in time in an opaque form, as for
+% example obtained by `current_time/1`.
+%
+% FormatString is a list of characters that are interpreted literally,
+% except for the following specifiers (and possibly more in the future):
+%
+% |  `%Y` |  year of the time stamp. Example: 2020.                |
+% |  `%m` |  month number (01-12), zero-padded to 2 digits         |
+% |  `%d` |  day number (01-31), zero-padded to 2 digits           |
+% |  `%H` |  hour number (00-24), zero-padded to 2 digits          |
+% |  `%M` |  minute number (00-59), zero-padded to 2 digits        |
+% |  `%S` |  second number (00-60), zero-padded to 2 digits        |
+% |  `%b` |  abbreviated month name, always 3 letters              |
+% |  `%a` |  abbreviated weekday name, always 3 letters            |
+% |  `%A` |  full weekday name                                     |
+% |  `%j` |  day of the year (001-366), zero-padded to 3 digits    |
+% |  `%%` |  the literal `%`                                       |
+%
+% Example:
+%
+% ```
+%    ?- current_time(T), phrase(format_time("%d.%m.%Y (%H:%M:%S)", T), Cs).
+%       T = [...], Cs = "11.06.2020 (00:24:32)".
+% ```
+
 format_time([], _) --> [].
 format_time(['%','%'|Fs], T) --> !, "%", format_time(Fs, T).
 format_time(['%',Spec|Fs], T) --> !,
@@ -65,8 +76,17 @@ format_time(['%',Spec|Fs], T) --> !,
         format_time(Fs, T).
 format_time([F|Fs], T) --> [F], format_time(Fs, T).
 
+%% max_sleep_time(T)
+%
+%  The maximum admissible time span for `sleep/1`.
+
 max_sleep_time(0xfffffffffffffbff).
 
+
+%% sleep(S)
+%
+%  Sleeps for S seconds (a floating point number or integer).
+
 sleep(T) :-
     builtins:must_be_number(T, sleep),
     (   T < 0 ->
@@ -77,12 +97,21 @@ sleep(T) :-
     ).
 
 
-% '$cpu_now' can be replaced by statistics/2 once that is implemented.
+%% statistics(?Keyword, ?List)
+%
+%  Preliminary support for statistics/2, yielding timing information.
+%  The only supported `Keyword` is `runtime`. The first element of
+%  `List` is the CPU time in milliseconds, the second element is
+%  currently not supported.
 
-:- meta_predicate time(0).
+statistics(runtime, [T,unsupported]) :-
+        '$cpu_now'(T0),
+        T is T0*1000.
+
+:- meta_predicate(time(0)).
 
 :- dynamic(time_id/1).
-:- dynamic(time_state/2).
+:- dynamic(time_state/3).
 
 time_next_id(N) :-
         (   retract(time_id(N0)) ->
@@ -91,10 +120,15 @@ time_next_id(N) :-
         ),
         asserta(time_id(N)).
 
+
+%% time(Goal)
+%
+%  Reports the execution time of Goal.
+
 time(Goal) :-
-        '$cpu_now'(T0),
+        cputime_inferences(T0, I0),
         time_next_id(ID),
-        setup_call_cleanup(asserta(time_state(ID, T0)),
+        setup_call_cleanup(asserta(time_state(ID, T0, I0)),
                            (   call_cleanup(catch(Goal, E, (report_time(ID),throw(E))),
                                             Det = true),
                                time_true(ID),
@@ -104,49 +138,72 @@ time(Goal) :-
                            ;   report_time(ID),
                                false
                            ),
-                           retract(time_state(ID, _))).
+                           retract(time_state(ID, _, _))).
+
+cputime_inferences(T, I) :-
+        '$cpu_now'(T),
+        '$inference_count'(I).
 
 time_true(ID) :-
         report_time(ID).
 time_true(ID)  :-
         % on backtracking, update the stored CPU time for this ID
-        retract(time_state(ID, _)),
-        '$cpu_now'(T0),
-        asserta(time_state(ID, T0)),
+        retract(time_state(ID, _, _)),
+        cputime_inferences(T0, I0),
+        asserta(time_state(ID, T0, I0)),
         false.
 
 report_time(ID) :-
-        time_state(ID, T0),
-        '$cpu_now'(T),
+        time_state(ID, T0, I0),
+        cputime_inferences(T, I),
         Time is T - T0,
+        Inferences0 is I - I0,
+        % we must subtract the number of inferences that time/1 itself takes;
+        % this may have to be adapted if the implementation changes,
+        % so that (for example) true/1 takes exactly 1 inference.
         (   bb_get('$answer_count', 0) ->
+            Inferences is Inferences0 - 60,
             Pre = "   ", Post = ""
-        ;   Pre = "", Post = "   "
+        ;   Inferences is Inferences0 - 9,
+            Pre = "", Post = "   "
         ),
-        format("~s% CPU time: ~3fs~n~s", [Pre,Time,Post]).
+        phrase((Pre,"% CPU time: ", format_("~3f", [Time]), "s, ",
+                format_("~U", [Inferences])," inference",s_if_necessary(Inferences),"\n",
+                Post), Cs),
+        format("~s", [Cs]).
+
+s_if_necessary(Inferences) -->
+        { compare(C, 1, Inferences) },
+        s_(C).
+
+s_(=) --> "".
+s_(<) --> "s".
+s_(>) --> " (exception?)".
 
 /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
 ?- time((true;false)).
-%@    % CPU time: 0.006s
-%@    true
-%@ ;  % CPU time: 0.001s
-%@    false.
+   % CPU time: 0.000s, 1 inference
+   true
+;  % CPU time: 0.000s, 0 inference (exception?)
+   false.
 
 :- time(use_module(library(clpz))).
-%@    % CPU time: 3.711s
-%@    true.
+   % CPU time: 0.343s, 409_874 inferences
+   true.
 
 :- time(use_module(library(lists))).
-%@    % CPU time: 0.006s
-%@    true.
+   % CPU time: 0.000s, 19 inferences
+   true.
 
 ?- time(member(X, "abc")).
-%@    % CPU time: 0.005s
-%@    X = a
-%@ ;  % CPU time: 0.000s
-%@    X = b
-%@ ;  % CPU time: 0.000s
-%@    X = c
-%@ ;  % CPU time: 0.000s
-%@    false.
+   % CPU time: 0.000s, 1 inference
+   X = a
+;  % CPU time: 0.000s, 3 inferences
+   X = b
+;  % CPU time: 0.000s, 3 inferences
+   X = c.
+
+?- time((repeat,false)).
+   % CPU time: 2.726s, 53_330_502 inferences
+   error('$interrupt_thrown',repl/0).
 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */

src/lib/ugraphs.pl

@@ -53,7 +53,7 @@
             connect_ugraph/3            % +Graph1, -Start, -Graph
           ]).
 
-/** <module> Graph manipulation library
+/** Graph manipulation library
 
 The S-representation of a graph is  a list of (vertex-neighbours) pairs,
 where the pairs are in standard order   (as produced by keysort) and the
@@ -61,55 +61,56 @@ neighbours of each vertex are also  in   standard  order (as produced by
 sort). This form is convenient for many calculations.
 
 A   new   UGraph   from    raw    data     can    be    created    using
-vertices_edges_to_ugraph/3.
+`vertices_edges_to_ugraph/3`.
 
 Adapted to support some of  the   functionality  of  the SICStus ugraphs
 library by Vitor Santos Costa.
 
 Ported from YAP 5.0.1 to SWI-Prolog by Jan Wielemaker.
 
-@author R.A.O'Keefe
-@author Vitor Santos Costa
-@author Jan Wielemaker
-@license BSD-2 or Artistic 2.0
+Ported from SWI-Prolog to Scryer by [Adrián Arroyo Calle](https://adrianistan.eu)
+
+License: BSD-2 or Artistic 2.0
 */
 
 :- use_module(library(lists)).
 :- use_module(library(pairs)).
 :- use_module(library(ordsets)).
 
-%!  vertices(+Graph, -Vertices)
+%% vertices(+Graph, -Vertices)
 %
-%   Unify Vertices with all vertices appearing in Graph. Example:
+% Unify Vertices with all vertices appearing in Graph. Example:
 %
-%       ?- vertices([1-[3,5],2-[4],3-[],4-[5],5-[]], L).
-%       L = [1, 2, 3, 4, 5]
+% ```
+% ?- vertices([1-[3,5],2-[4],3-[],4-[5],5-[]], L).
+%    L = [1, 2, 3, 4, 5]
+% ```
 
 vertices([], []) :- !.
 vertices([Vertex-_|Graph], [Vertex|Vertices]) :-
     vertices(Graph, Vertices).
 
 
-%!  vertices_edges_to_ugraph(+Vertices, +Edges, -UGraph) is det.
+%% vertices_edges_to_ugraph(+Vertices, +Edges, -UGraph) is det.
 %
-%   Create a UGraph from Vertices and edges.   Given  a graph with a
-%   set of Vertices and a set of   Edges,  Graph must unify with the
-%   corresponding S-representation. Note that   the vertices without
-%   edges will appear in Vertices but not  in Edges. Moreover, it is
-%   sufficient for a vertice to appear in Edges.
+% Create a UGraph from Vertices and edges.   Given  a graph with a
+% set of Vertices and a set of   Edges,  Graph must unify with the
+% corresponding S-representation. Note that   the vertices without
+% edges will appear in Vertices but not  in Edges. Moreover, it is
+% sufficient for a vertice to appear in Edges.
 %
-%   ==
-%   ?- vertices_edges_to_ugraph([],[1-3,2-4,4-5,1-5], L).
-%   L = [1-[3,5], 2-[4], 3-[], 4-[5], 5-[]]
-%   ==
+% ```
+% ?- vertices_edges_to_ugraph([],[1-3,2-4,4-5,1-5], L).
+%    L = [1-[3,5], 2-[4], 3-[], 4-[5], 5-[]]
+% ```
 % 
-%   In this case all  vertices  are   defined  implicitly.  The next
-%   example shows three unconnected vertices:
+% In this case all  vertices  are   defined  implicitly.  The next
+% example shows three unconnected vertices:
 %
-%   ==
-%   ?- vertices_edges_to_ugraph([6,7,8],[1-3,2-4,4-5,1-5], L).
-%   L = [1-[3,5], 2-[4], 3-[], 4-[5], 5-[], 6-[], 7-[], 8-[]]
-%   ==
+% ```
+% ?- vertices_edges_to_ugraph([6,7,8],[1-3,2-4,4-5,1-5], L).
+%    L = [1-[3,5], 2-[4], 3-[], 4-[5], 5-[], 6-[], 7-[], 8-[]]
+% ```
 
 vertices_edges_to_ugraph(Vertices, Edges, Graph) :-
     sort(Edges, EdgeSet),
@@ -119,15 +120,15 @@ vertices_edges_to_ugraph(Vertices, Edges, Graph) :-
     p_to_s_group(VertexSet, EdgeSet, Graph).
 
 
-%!  add_vertices(+Graph, +Vertices, -NewGraph)
+%% add_vertices(+Graph, +Vertices, -NewGraph)
 %
-%   Unify NewGraph with a new  graph  obtained   by  adding  the list of
-%   Vertices to Graph. Example:
+% Unify NewGraph with a new  graph  obtained   by  adding  the list of
+% Vertices to Graph. Example:
 %
-%   ```
-%   ?- add_vertices([1-[3,5],2-[]], [0,1,2,9], NG).
-%   NG = [0-[], 1-[3,5], 2-[], 9-[]]
-%   ```
+% ```
+% ?- add_vertices([1-[3,5],2-[]], [0,1,2,9], NG).
+%    NG = [0-[], 1-[3,5], 2-[], 9-[]]
+% ```
 
 % replace with real msort/2 when available
 msort_(List, Sorted) :-
@@ -159,23 +160,18 @@ add_empty_vertices([], []).
 add_empty_vertices([V|G], [V-[]|NG]) :-
     add_empty_vertices(G, NG).
 
-%!  del_vertices(+Graph, +Vertices, -NewGraph) is det.
+%% del_vertices(+Graph, +Vertices, -NewGraph) is det.
 %
-%   Unify NewGraph with a new graph obtained by deleting the list of
-%   Vertices and all the edges that start from  or go to a vertex in
-%   Vertices to the Graph. Example:
+% Unify NewGraph with a new graph obtained by deleting the list of
+% Vertices and all the edges that start from  or go to a vertex in
+% Vertices to the Graph. Example:
 %
-%   ==
-%   ?- del_vertices([1-[3,5],2-[4],3-[],4-[5],5-[],6-[],7-[2,6],8-[]],
-%                   [2,1],
-%                   NL).
-%   NL = [3-[],4-[5],5-[],6-[],7-[6],8-[]]
-%   ==
-%
-%   @compat Upto 5.6.48 the argument order was (+Vertices, +Graph,
-%   -NewGraph). Both YAP and SWI-Prolog have changed the argument
-%   order for compatibility with recent SICStus as well as
-%   consistency with del_edges/3.
+% ```
+% ?- del_vertices([1-[3,5],2-[4],3-[],4-[5],5-[],6-[],7-[2,6],8-[]],
+%               [2,1],
+%               NL).
+%    NL = [3-[],4-[5],5-[],6-[],7-[6],8-[]]
+% ```
 
 del_vertices(Graph, Vertices, NewGraph) :-
     sort(Vertices, V1),             % JW: was msort
@@ -204,32 +200,32 @@ split_on_del_vertices(>, V, Edges, [_|Vs], Vs, V1, [V-NEdges|NG], NG) :-
     ord_subtract(Edges, V1, NEdges).
 split_on_del_vertices(=, _, _, [_|Vs], Vs, _, NG, NG).
 
-%!  add_edges(+Graph, +Edges, -NewGraph)
+%% add_edges(+Graph, +Edges, -NewGraph)
 %
-%   Unify NewGraph with a new graph obtained by adding the list of Edges
-%   to Graph. Example:
+% Unify NewGraph with a new graph obtained by adding the list of Edges
+% to Graph. Example:
 %
-%   ```
-%   ?- add_edges([1-[3,5],2-[4],3-[],4-[5],
-%                 5-[],6-[],7-[],8-[]],
-%                [1-6,2-3,3-2,5-7,3-2,4-5],
-%                NL).
-%   NL = [1-[3,5,6], 2-[3,4], 3-[2], 4-[5],
-%         5-[7], 6-[], 7-[], 8-[]]
-%   ```
+% ```
+% ?- add_edges([1-[3,5],2-[4],3-[],4-[5],
+%             5-[],6-[],7-[],8-[]],
+%            [1-6,2-3,3-2,5-7,3-2,4-5],
+%            NL).
+%    NL = [1-[3,5,6], 2-[3,4], 3-[2], 4-[5],
+%     5-[7], 6-[], 7-[], 8-[]]
+% ```
 
 add_edges(Graph, Edges, NewGraph) :-
     p_to_s_graph(Edges, G1),
     ugraph_union(Graph, G1, NewGraph).
 
-%!  ugraph_union(+Graph1, +Graph2, -NewGraph)
+%% ugraph_union(+Graph1, +Graph2, -NewGraph)
 %
-%   NewGraph is the union of Graph1 and Graph2. Example:
+% NewGraph is the union of Graph1 and Graph2. Example:
 %
-%   ```
-%   ?- ugraph_union([1-[2],2-[3]],[2-[4],3-[1,2,4]],L).
-%   L = [1-[2], 2-[3,4], 3-[1,2,4]]
-%   ```
+% ```
+% ?- ugraph_union([1-[2],2-[3]],[2-[4],3-[1,2,4]],L).
+%    L = [1-[2], 2-[3,4], 3-[1,2,4]]
+% ```
 
 ugraph_union(Set1, [], Set1) :- !.
 ugraph_union([], Set2, Set2) :- !.
@@ -245,25 +241,25 @@ ugraph_union(<, Head1, Tail1, Head2, Tail2, [Head1|Union]) :-
 ugraph_union(>, Head1, Tail1, Head2, Tail2, [Head2|Union]) :-
     ugraph_union([Head1|Tail1], Tail2, Union).
 
-%!  del_edges(+Graph, +Edges, -NewGraph)
+%% del_edges(+Graph, +Edges, -NewGraph)
 %
-%   Unify NewGraph with a new graph  obtained   by  removing the list of
-%   Edges from Graph. Notice that no vertices are deleted. Example:
+% Unify NewGraph with a new graph  obtained   by  removing the list of
+% Edges from Graph. Notice that no vertices are deleted. Example:
 %
-%   ```
-%   ?- del_edges([1-[3,5],2-[4],3-[],4-[5],5-[],6-[],7-[],8-[]],
-%                [1-6,2-3,3-2,5-7,3-2,4-5,1-3],
-%                NL).
-%   NL = [1-[5],2-[4],3-[],4-[],5-[],6-[],7-[],8-[]]
-%   ```
+% ```
+% ?- del_edges([1-[3,5],2-[4],3-[],4-[5],5-[],6-[],7-[],8-[]],
+%            [1-6,2-3,3-2,5-7,3-2,4-5,1-3],
+%            NL).
+%    NL = [1-[5],2-[4],3-[],4-[],5-[],6-[],7-[],8-[]]
+% ```
 
 del_edges(Graph, Edges, NewGraph) :-
     p_to_s_graph(Edges, G1),
     graph_subtract(Graph, G1, NewGraph).
 
-%!  graph_subtract(+Set1, +Set2, ?Difference)
+%% graph_subtract(+Set1, +Set2, ?Difference)
 %
-%   Is based on ord_subtract
+% Is based on `ord_subtract/3`
 
 graph_subtract(Set1, [], Set1) :- !.
 graph_subtract([], _, []).
@@ -279,12 +275,14 @@ graph_subtract(<, Head1, Tail1, Head2, Tail2, [Head1|Difference]) :-
 graph_subtract(>, Head1, Tail1, _,     Tail2, Difference) :-
     graph_subtract([Head1|Tail1], Tail2, Difference).
 
-%!  edges(+Graph, -Edges)
+%% edges(+Graph, -Edges)
 %
-%   Unify Edges with all edges appearing in Graph. Example:
+% Unify Edges with all edges appearing in Graph. Example:
 %
-%       ?- edges([1-[3,5],2-[4],3-[],4-[5],5-[]], L).
-%       L = [1-3, 1-5, 2-4, 4-5]
+% ```
+% ?- edges([1-[3,5],2-[4],3-[],4-[5],5-[]], L).
+%    L = [1-3, 1-5, 2-4, 4-5]
+% ```
 
 edges(Graph, Edges) :-
     s_to_p_graph(Graph, Edges).
@@ -324,15 +322,15 @@ s_to_p_graph([], _, P_Graph, P_Graph) :- !.
 s_to_p_graph([Neib|Neibs], Vertex, [Vertex-Neib|P], Rest_P) :-
     s_to_p_graph(Neibs, Vertex, P, Rest_P).
 
-%!  transitive_closure(+Graph, -Closure)
+%% transitive_closure(+Graph, -Closure)
 %
-%   Generate the graph Closure  as  the   transitive  closure  of Graph.
-%   Example:
+% Generate the graph Closure  as  the   transitive  closure  of Graph.
+% Example:
 %
-%   ```
-%   ?- transitive_closure([1-[2,3],2-[4,5],4-[6]],L).
-%   L = [1-[2,3,4,5,6], 2-[4,5,6], 4-[6]]
-%   ```
+% ```
+% ?- transitive_closure([1-[2,3],2-[4,5],4-[6]],L).
+%    L = [1-[2,3,4,5,6], 2-[4,5,6], 4-[6]]
+% ```
 
 transitive_closure(Graph, Closure) :-
     warshall(Graph, Graph, Closure).
@@ -354,23 +352,18 @@ warshall([X-Neibs|G], V, Y, [X-Neibs|NewG]) :-
     warshall(G, V, Y, NewG).
 warshall([], _, _, []).
 
-%!  transpose_ugraph(Graph, NewGraph) is det.
+%% transpose_ugraph(Graph, NewGraph) is det.
 %
-%   Unify NewGraph with a new graph obtained from Graph by replacing
-%   all edges of the form V1-V2 by edges of the form V2-V1. The cost
-%   is O(|V|*log(|V|)). Notice that an undirected   graph is its own
-%   transpose. Example:
+% Unify NewGraph with a new graph obtained from Graph by replacing
+% all edges of the form V1-V2 by edges of the form V2-V1. The cost
+% is O(|V|\*log(|V|)). Notice that an undirected   graph is its own
+% transpose. Example:
 %
-%     ==
-%     ?- transpose([1-[3,5],2-[4],3-[],4-[5],
-%                   5-[],6-[],7-[],8-[]], NL).
-%     NL = [1-[],2-[],3-[1],4-[2],5-[1,4],6-[],7-[],8-[]]
-%     ==
-%
-%   @compat  This  predicate  used  to   be  known  as  transpose/2.
-%   Following  SICStus  4,  we  reserve    transpose/2   for  matrix
-%   transposition    and    renamed    ugraph    transposition    to
-%   transpose_ugraph/2.
+% ```
+% ?- transpose([1-[3,5],2-[4],3-[],4-[5],
+%               5-[],6-[],7-[],8-[]], NL).
+%    NL = [1-[],2-[],3-[1],4-[2],5-[1,4],6-[],7-[],8-[]]
+% ```
 
 transpose_ugraph(Graph, NewGraph) :-
     edges(Graph, Edges),
@@ -382,13 +375,15 @@ flip_edges([], []).
 flip_edges([Key-Val|Pairs], [Val-Key|Flipped]) :-
     flip_edges(Pairs, Flipped).
 
-%!  compose(+LeftGraph, +RightGraph, -NewGraph)
+%% compose(+LeftGraph, +RightGraph, -NewGraph)
 %
-%   Compose NewGraph by connecting the  _drains_   of  LeftGraph  to the
-%   _sources_ of RightGraph. Example:
+% Compose NewGraph by connecting the  _drains_   of  LeftGraph  to the
+% _sources_ of RightGraph. Example:
 %
-%       ?- compose([1-[2],2-[3]],[2-[4],3-[1,2,4]],L).
-%       L = [1-[4], 2-[1,2,4], 3-[]]
+% ```
+% ?- compose([1-[2],2-[3]],[2-[4],3-[1,2,4]],L).
+%    L = [1-[4], 2-[1,2,4], 3-[]]
+% ```
 
 compose(G1, G2, Composition) :-
     vertices(G1, V1),
@@ -423,21 +418,17 @@ compose1(=, V1, Vs1, V1, N2, G2, SoFar, Comp) :-
     ord_union(N2, SoFar, Next),
     compose1(Vs1, G2, Next, Comp).
 
-%!  top_sort(+Graph, -Sorted) is semidet.
-%!  top_sort(+Graph, -Sorted, ?Tail) is semidet.
+%% top_sort(+Graph, -Sorted) is semidet.
 %
-%   Sorted is a  topological  sorted  list   of  nodes  in  Graph. A
-%   toplogical sort is possible  if  the   graph  is  connected  and
-%   acyclic. In the example we show   how  topological sorting works
-%   for a linear graph:
+% Sorted is a  topological  sorted  list   of  nodes  in  Graph. A
+% toplogical sort is possible  if  the   graph  is  connected  and
+% acyclic. In the example we show   how  topological sorting works
+% for a linear graph:
 %
-%   ==
-%   ?- top_sort([1-[2], 2-[3], 3-[]], L).
-%   L = [1, 2, 3]
-%   ==
-%
-%   The  predicate  top_sort/3  is  a  difference  list  version  of
-%   top_sort/2.
+% ```
+% ?- top_sort([1-[2], 2-[3], 3-[]], L).
+%    L = [1, 2, 3]
+% ```
 
 top_sort(Graph, Sorted) :-
     vertices_and_zeros(Graph, Vertices, Counts0),
@@ -445,6 +436,11 @@ top_sort(Graph, Sorted) :-
     select_zeros(Counts1, Vertices, Zeros),
     top_sort(Zeros, Sorted, Graph, Vertices, Counts1).
 
+%% top_sort(+Graph, -Sorted, ?Tail) is semidet.
+%
+% The  predicate  `top_sort/3`  is  a  difference  list  version  of
+% `top_sort/2`.
+
 top_sort(Graph, Sorted0, Sorted) :-
     vertices_and_zeros(Graph, Vertices, Counts0),
     count_edges(Graph, Vertices, Counts0, Counts1),
@@ -520,17 +516,21 @@ decr_list(Neibs, [_|Vertices], [N|Counts1], [N|Counts2], Zi, Zo) :-
     decr_list(Neibs, Vertices, Counts1, Counts2, Zi, Zo).
 
 
-%!  neighbors(+Vertex, +Graph, -Neigbours) is det.
-%!  neighbours(+Vertex, +Graph, -Neigbours) is det.
+
+%% neighbours(+Vertex, +Graph, -Neigbours) is det.
 %
-%   Neigbours is a sorted list of  the   neighbours  of Vertex in Graph.
-%   Example:
+% Neigbours is a sorted list of  the   neighbours  of Vertex in Graph.
+% Example:
 %
-%   ```
-%   ?- neighbours(4,[1-[3,5],2-[4],3-[],
-%                    4-[1,2,7,5],5-[],6-[],7-[],8-[]], NL).
-%   NL = [1,2,7,5]
-%   ```
+% ```
+% ?- neighbours(4,[1-[3,5],2-[4],3-[],
+%                4-[1,2,7,5],5-[],6-[],7-[],8-[]], NL).
+%    NL = [1,2,7,5]
+% ```
+
+%% neighbors(+Vertex, +Graph, -Neigbours) is det.
+%
+% Same as `neighbours/3`.
 
 neighbors(Vertex, Graph, Neig) :-
     neighbours(Vertex, Graph, Neig).
@@ -542,24 +542,24 @@ neighbours(V,[_|G],Neig) :-
     neighbours(V,G,Neig).
 
 
-%!  connect_ugraph(+UGraphIn, -Start, -UGraphOut) is det.
+%% connect_ugraph(+UGraphIn, -Start, -UGraphOut) is det.
 %
-%   Adds Start as an additional vertex that is connected to all vertices
-%   in UGraphIn. This can be used to   create  an topological sort for a
-%   not connected graph. Start is before any   vertex in UGraphIn in the
-%   standard order of terms.  No vertex in UGraphIn can be a variable.
+% Adds Start as an additional vertex that is connected to all vertices
+% in UGraphIn. This can be used to   create  an topological sort for a
+% not connected graph. Start is before any   vertex in UGraphIn in the
+% standard order of terms.  No vertex in UGraphIn can be a variable.
 %
-%   Can be used to order a not-connected graph as follows:
+% Can be used to order a not-connected graph as follows:
 %
-%   ```
-%   top_sort_unconnected(Graph, Vertices) :-
-%       (   top_sort(Graph, Vertices)
-%       ->  true
-%       ;   connect_ugraph(Graph, Start, Connected),
-%           top_sort(Connected, Ordered0),
-%           Ordered0 = [Start|Vertices]
-%       ).
-%   ```
+% ```
+% top_sort_unconnected(Graph, Vertices) :-
+%   (   top_sort(Graph, Vertices)
+%   ->  true
+%   ;   connect_ugraph(Graph, Start, Connected),
+%       top_sort(Connected, Ordered0),
+%       Ordered0 = [Start|Vertices]
+%   ).
+% ```
 
 connect_ugraph([], 0, []) :- !.
 connect_ugraph(Graph, Start, [Start-Vertices|Graph]) :-
@@ -567,12 +567,12 @@ connect_ugraph(Graph, Start, [Start-Vertices|Graph]) :-
     Vertices = [First|_],
     before(First, Start).
 
-%!  before(+Term, -Before) is det.
+%% before(+Term, -Before) is det.
 %
-%   Unify Before to a term that comes   before  Term in the standard
-%   order of terms.
+% Unify Before to a term that comes   before  Term in the standard
+% order of terms.
 %
-%   @error instantiation_error if Term is unbound.
+% Throws `instantiation_error` if Term is unbound.
 
 before(X, _) :-
     var(X),
@@ -585,21 +585,22 @@ before(Number, Start) :-
 before(_, 0).
 
 
-%!  complement(+UGraphIn, -UGraphOut)
+%% complement(+UGraphIn, -UGraphOut)
 %
-%   UGraphOut is a ugraph with an  edge   between  all vertices that are
-%   _not_ connected in UGraphIn and  all   edges  from UGraphIn removed.
-%   Example:
+% UGraphOut is a ugraph with an  edge   between  all vertices that are
+% _not_ connected in UGraphIn and  all   edges  from UGraphIn removed.
+% Example:
 %
-%   ```
-%   ?- complement([1-[3,5],2-[4],3-[],
-%                  4-[1,2,7,5],5-[],6-[],7-[],8-[]], NL).
-%   NL = [1-[2,4,6,7,8],2-[1,3,5,6,7,8],3-[1,2,4,5,6,7,8],
-%         4-[3,5,6,8],5-[1,2,3,4,6,7,8],6-[1,2,3,4,5,7,8],
-%         7-[1,2,3,4,5,6,8],8-[1,2,3,4,5,6,7]]
-%   ```
-%
-%   @tbd Simple two-step algorithm. You could be smarter, I suppose.
+% ```
+% ?- complement([1-[3,5],2-[4],3-[],
+%              4-[1,2,7,5],5-[],6-[],7-[],8-[]], NL).
+%    NL = [1-[2,4,6,7,8],2-[1,3,5,6,7,8],3-[1,2,4,5,6,7,8],
+%     4-[3,5,6,8],5-[1,2,3,4,6,7,8],6-[1,2,3,4,5,7,8],
+%     7-[1,2,3,4,5,6,8],8-[1,2,3,4,5,6,7]]
+% ```
+
+
+% TODO: Simple two-step algorithm. You could be smarter, I suppose.
 
 complement(G, NG) :-
     vertices(G,Vs),
@@ -611,13 +612,15 @@ complement([V-Ns|G], Vs, [V-INs|NG]) :-
     ord_subtract(Vs,Ns1,INs),
     complement(G, Vs, NG).
 
-%!  reachable(+Vertex, +UGraph, -Vertices)
+%% reachable(+Vertex, +UGraph, -Vertices)
 %
-%   True when Vertices is  an  ordered   set  of  vertices  reachable in
-%   UGraph, including Vertex.  Example:
+% True when Vertices is  an  ordered   set  of  vertices  reachable in
+% UGraph, including Vertex.  Example:
 %
-%       ?- reachable(1,[1-[3,5],2-[4],3-[],4-[5],5-[]],V).
-%       V = [1, 3, 5]
+% ```
+% ?- reachable(1,[1-[3,5],2-[4],3-[],4-[5],5-[]],V).
+%    V = [1, 3, 5]
+% ```
 
 reachable(N, G, Rs) :-
     reachable([N], G, [N], Rs).

src/lib/uuid.pl

@@ -1,25 +1,32 @@
 /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    Written in February 2021 by Adrián Arroyo (adrian.arroyocalle@gmail.com)
    Part of Scryer-Prolog
-   This library provides reasoning about UUID (only version 4 right now).
-   There are three predicates:
-    * uuidv4/1, to generate a new UUIDv4
-    * uuidv4_string/1, to generate a new UUIDv4 in string hex representation
-    * uuid_string/2, to converte between UUID list of bytes and UUID hex representation
-
-   Examples:
-   ?- uuidv4(X).
-      X = [42,147,248,242,117,196,79,2,129,159|...].
-   ?- uuidv4_string(X).
-      X = "428499fc-76e3-4240- ...".
-   ?- uuidv4(X), uuid_string(X, S).
-      X = [173,12,244,152,139,118,64,139,137,4|...], S = "ad0cf498-8b76-408b- ...".
-   ?- uuid_string(X, "61ae692e-eaf6-4199-8dd3-9f01db70a20b").
-      X = [97,174,105,46,234,246,65,153,141,211|...].
-
    I place this code in the public domain. Use it in any way you want.
 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
 
+/**
+This library provides reasoning and working with [UUID](https://en.wikipedia.org/wiki/Universally_unique_identifier)
+(only version 4 right now).
+
+There are three predicates:
+
+ * `uuidv4/1`, to generate a new UUIDv4
+ * `uuidv4_string/1`, to generate a new UUIDv4 in string hex representation
+ * `uuid_string/2`, to converte between UUID list of bytes and UUID hex representation
+
+Examples:
+
+```
+?- uuidv4(X).
+   X = [42,147,248,242,117,196,79,2,129,159|...].
+?- uuidv4_string(X).
+   X = "428499fc-76e3-4240- ...".
+?- uuidv4(X), uuid_string(X, S).
+   X = [173,12,244,152,139,118,64,139,137,4|...], S = "ad0cf498-8b76-408b- ...".
+?- uuid_string(X, "61ae692e-eaf6-4199-8dd3-9f01db70a20b").
+   X = [97,174,105,46,234,246,65,153,141,211|...].
+*/
+
 :- module(uuid, [
     uuidv4/1,
     uuidv4_string/1,
@@ -39,6 +46,10 @@ clock_seq_hi_and_res_clock_seq_low - 2
 node - 6
 UUID v4 can be generated from a set of 16 random bytes: https://www.rfc-archive.org/getrfc.php?rfc=4122#gsc.tab=0 (section 4.4)
 */
+
+%% uuidv4(-Uuid).
+%
+% Generates a new UUID v4 (random). It unifies with a list of bytes.
 uuidv4(Uuid) :-
     crypto_n_random_bytes(16, Bytes),
     Bytes = [B1, B2, B3, B4, B5, B6, B7, B8, B9, B10, B11, B12, B13, B14, B15, B16],
@@ -52,8 +63,15 @@ uuidv4(Uuid) :-
     byte_bits(NewTimeHi, NewBitsTimeHi),
     Uuid = [B1, B2, B3, B4, B5, B6, NewTimeHi, B8, NewClockSeqHi0, B10, B11, B12, B13, B14, B15, B16].
 
+%% uuidv4_string(-UuidString).
+%
+% Generates a new UUID v4 (random). It unifies with a string representation of the UUID.
+% It is equivalent of calling `uuidv4/1` followed by `uuid_string/2`.
 uuidv4_string(String) :- uuidv4(Uuid), uuid_string(Uuid, String).
 
+%% uuid_string(?UuidBytes, ?UuidString).
+%
+% Translates between the bytes representation and the string representation of the same UUID.
 uuid_string(Uuid, String) :-
     Uuid = [B1, B2, B3, B4, B5, B6, B7, B8, B9, B10, B11, B12, B13, B14, B15, B16],
     phrase(uuid_([S1, S2, S3, S4, S5]), String),