[standalone] Add demo for global variables

Needed in a future commit.

.clang-format

@@ -0,0 +1,32 @@
+BasedOnStyle: LLVM
+
+Language: Cpp
+Standard: Cpp11
+
+AccessModifierOffset: -1
+AlignEscapedNewlines: Left
+AlwaysBreakAfterReturnType: None
+AlwaysBreakTemplateDeclarations: Yes
+AllowAllParametersOfDeclarationOnNextLine: false
+AllowShortFunctionsOnASingleLine: Inline
+BinPackParameters: false
+BreakBeforeBinaryOperators: NonAssignment
+BreakBeforeTernaryOperators: true
+BreakConstructorInitializers: AfterColon
+BreakInheritanceList: AfterColon
+ColumnLimit: 120
+ConstructorInitializerAllOnOneLineOrOnePerLine: true
+ContinuationIndentWidth: 4
+DerivePointerAlignment: false
+IndentCaseLabels: true
+IndentPPDirectives: None
+IndentWidth: 4
+MaxEmptyLinesToKeep: 1
+PointerAlignment: Left
+ReflowComments: true
+SortIncludes: false
+SortUsingDeclarations: true
+SpaceAfterTemplateKeyword: false
+SpacesBeforeTrailingComments: 2
+TabWidth: 4
+UseTab: Never

.clippy.toml

@@ -0,0 +1 @@
+doc-valid-idents = ["CPython", "NumPy", ".."]

.gitignore

@@ -1,2 +1,4 @@
 __pycache__
 /target
+/nac3standalone/demo/linalg/target
+nix/windows/msys2

.pre-commit-config.yaml

@@ -0,0 +1,24 @@
+# See https://pre-commit.com for more information
+# See https://pre-commit.com/hooks.html for more hooks
+
+default_stages: [commit]
+
+repos:
+  - repo: local
+    hooks:
+      - id: nac3-cargo-fmt
+        name: nac3 cargo format
+        entry: nix
+        language: system
+        types: [file, rust]
+        pass_filenames: false
+        description: Runs cargo fmt on the codebase.
+        args: [develop, -c, cargo, fmt, --all]
+      - id: nac3-cargo-clippy
+        name: nac3 cargo clippy
+        entry: nix
+        language: system
+        types: [file, rust]
+        pass_filenames: false
+        description: Runs cargo clippy on the codebase.
+        args: [develop, -c, cargo, clippy, --tests]

Cargo.toml

@@ -1,6 +1,14 @@
 [workspace]
 members = [
+  "nac3ld",
+  "nac3ast",
+  "nac3parser",
   "nac3core",
   "nac3standalone",
-  "nac3embedded",
+  "nac3artiq",
+  "runkernel",
 ]
+resolver = "2"
+
+[profile.release]
+debug = true

README.md

@@ -1,34 +1,62 @@
-# nac3 compiler
+<div align="center">
+
+![icon](https://git.m-labs.hk/M-Labs/nac3/raw/branch/master/nac3.svg)
+
+</div>
+
+# NAC3
+NAC3 is a major, backward-incompatible rewrite of the compiler for the [ARTIQ](https://m-labs.hk/artiq) physics experiment control and data acquisition system. It features greatly improved compilation speeds, a much better type system, and more predictable and transparent operation.
+
+NAC3 has a modular design and its applicability reaches beyond ARTIQ. The ``nac3core`` module does not contain anything specific to ARTIQ, and can be used in any project that requires compiling Python to machine code.
+
+**WARNING: NAC3 is currently experimental software and several important features are not implemented yet.**
+
+## Packaging
+
+NAC3 is packaged using the [Nix](https://nixos.org) Flakes system. Install Nix 2.8+ and enable flakes by adding ``experimental-features = nix-command flakes`` to ``nix.conf`` (e.g. ``~/.config/nix/nix.conf``).
+
+## Try NAC3
+
+### Linux
+
+After setting up Nix as above, use ``nix shell git+https://github.com/m-labs/artiq.git?ref=nac3`` to get a shell with the NAC3 version of ARTIQ. See the ``examples`` directory in ARTIQ (``nac3`` Git branch) for some samples of NAC3 kernel code.
+
+### Windows
+
+Install [MSYS2](https://www.msys2.org/), and open "MSYS2 CLANG64". Edit ``/etc/pacman.conf`` to add:
+```
+[artiq]
+SigLevel = Optional TrustAll
+Server = https://msys2.m-labs.hk/artiq-nac3
+```
+
+Then run the following commands:
+```
+pacman -Syu
+pacman -S mingw-w64-clang-x86_64-artiq
+```
+
+## For developers
 
 This repository contains:
-- nac3core: Core compiler library, containing type-checking, static analysis (in
-    the future) and code generation.
-- nac3embedded: Integration with CPython runtime.
-- nac3standalone: Standalone compiler tool.
+- ``nac3ast``: Python abstract syntax tree definition (based on RustPython).
+- ``nac3parser``: Python parser (based on RustPython).
+- ``nac3core``: Core compiler library, containing type-checking and code generation.
+- ``nac3standalone``: Standalone compiler tool (core language only).
+- ``nac3ld``: Minimalist RISC-V and ARM linker.
+- ``nac3artiq``: Integration with ARTIQ and implementation of ARTIQ-specific extensions to the core language.
+- ``runkernel``: Simple program that runs compiled ARTIQ kernels on the host and displays RTIO operations. Useful for testing without hardware.
 
-The core compiler would know nothing about symbol resolution, host variables
-etc. The nac3embedded/nac3standalone library would provide (implement) the
-symbol resolver to the core compiler for resolving the type and value for
-unknown symbols. The core compiler would only type check classes and functions
-requested by the nac3embedded/nac3standalone lib (the API should allow the
-caller to specify which methods should be compiled). After type checking, the
-compiler would analyse the set of functions/classes that are used and perform
-code generation.
+Use ``nix develop`` in this repository to enter a development shell.
+If you are using a different shell than bash you can use e.g. ``nix develop --command fish``.
 
-value could be integer values, boolean values, bytes (for memcpy), function ID
-(full name + concrete type)
+Build NAC3 with ``cargo build --release``. See the demonstrations in ``nac3artiq`` and ``nac3standalone``.
 
-## Current Plan
-
-Type checking:
-
-- [x] Basic interface for symbol resolver.
-- [x] Track location information in context object (for diagnostics).
-- [ ] Refactor old expression and statement type inference code. (anto)
-- [ ] Error diagnostics utilities. (pca)
-- [ ] Move tests to external files, write scripts for testing. (pca)
-- [ ] Implement function type checking (instantiate bounded type parameters),
-    loop unrolling, type inference for lists with virtual objects. (pca)
+### Pre-Commit Hooks
 
+You are strongly recommended to use the provided pre-commit hooks to automatically reformat files and check for non-optimal Rust practices using Clippy. Run `pre-commit install` to install the hook and `pre-commit` will automatically run `cargo fmt` and `cargo clippy` for you.
 
+Several things to note:
 
+- If `cargo fmt` or `cargo clippy` returns an error, the pre-commit hook will fail. You should fix all errors before trying to commit again.
+- If `cargo fmt` reformats some files, the pre-commit hook will also fail. You should review the changes and, if satisfied, try to commit again.

flake.lock

@@ -0,0 +1,27 @@
+{
+  "nodes": {
+    "nixpkgs": {
+      "locked": {
+        "lastModified": 1727348695,
+        "narHash": "sha256-J+PeFKSDV+pHL7ukkfpVzCOO7mBSrrpJ3svwBFABbhI=",
+        "owner": "NixOS",
+        "repo": "nixpkgs",
+        "rev": "1925c603f17fc89f4c8f6bf6f631a802ad85d784",
+        "type": "github"
+      },
+      "original": {
+        "owner": "NixOS",
+        "ref": "nixos-unstable",
+        "repo": "nixpkgs",
+        "type": "github"
+      }
+    },
+    "root": {
+      "inputs": {
+        "nixpkgs": "nixpkgs"
+      }
+    }
+  },
+  "root": "root",
+  "version": 7
+}

flake.nix

@@ -0,0 +1,212 @@
+{
+  description = "The third-generation ARTIQ compiler";
+
+  inputs.nixpkgs.url = github:NixOS/nixpkgs/nixos-unstable;
+
+  outputs = { self, nixpkgs }:
+    let
+      pkgs = import nixpkgs { system = "x86_64-linux"; };
+      pkgs32 = import nixpkgs { system = "i686-linux"; };
+    in rec {
+      packages.x86_64-linux = rec {
+        llvm-nac3 = pkgs.callPackage ./nix/llvm {};
+        llvm-tools-irrt = pkgs.runCommandNoCC "llvm-tools-irrt" {}
+          ''
+          mkdir -p $out/bin
+          ln -s ${pkgs.llvmPackages_14.clang-unwrapped}/bin/clang $out/bin/clang-irrt
+          ln -s ${pkgs.llvmPackages_14.llvm.out}/bin/llvm-as $out/bin/llvm-as-irrt
+          '';
+        demo-linalg-stub = pkgs.rustPlatform.buildRustPackage {
+          name = "demo-linalg-stub";
+          src = ./nac3standalone/demo/linalg;
+          cargoLock = {
+            lockFile = ./nac3standalone/demo/linalg/Cargo.lock;
+          };
+          doCheck = false;
+        };
+        demo-linalg-stub32 = pkgs32.rustPlatform.buildRustPackage {
+          name = "demo-linalg-stub32";
+          src = ./nac3standalone/demo/linalg;
+          cargoLock = {
+            lockFile = ./nac3standalone/demo/linalg/Cargo.lock;
+          };
+          doCheck = false;
+        };
+        nac3artiq = pkgs.python3Packages.toPythonModule (
+          pkgs.rustPlatform.buildRustPackage rec {
+            name = "nac3artiq";
+            outputs = [ "out" "runkernel" "standalone" ];
+            src = self;
+            cargoLock = {
+              lockFile = ./Cargo.lock;
+            };
+            passthru.cargoLock = cargoLock;
+            nativeBuildInputs = [ pkgs.python3 (pkgs.wrapClangMulti pkgs.llvmPackages_14.clang) llvm-tools-irrt pkgs.llvmPackages_14.llvm.out llvm-nac3 ];
+            buildInputs = [ pkgs.python3 llvm-nac3 ];
+            checkInputs = [ (pkgs.python3.withPackages(ps: [ ps.numpy ps.scipy ])) ];
+            checkPhase =
+              ''
+              echo "Checking nac3standalone demos..."
+              pushd nac3standalone/demo
+              patchShebangs .
+              export DEMO_LINALG_STUB=${demo-linalg-stub}/lib/liblinalg.a
+              export DEMO_LINALG_STUB32=${demo-linalg-stub32}/lib/liblinalg.a
+              ./check_demos.sh -i686
+              popd
+              echo "Running Cargo tests..."
+              cargoCheckHook
+              '';
+            installPhase =
+              ''
+              PYTHON_SITEPACKAGES=$out/${pkgs.python3Packages.python.sitePackages}
+              mkdir -p $PYTHON_SITEPACKAGES
+              cp target/x86_64-unknown-linux-gnu/release/libnac3artiq.so $PYTHON_SITEPACKAGES/nac3artiq.so
+
+              mkdir -p $runkernel/bin
+              cp target/x86_64-unknown-linux-gnu/release/runkernel $runkernel/bin
+
+              mkdir -p $standalone/bin
+              cp target/x86_64-unknown-linux-gnu/release/nac3standalone $standalone/bin
+              '';
+          }
+        );
+        python3-mimalloc = pkgs.python3 // rec {
+          withMimalloc = pkgs.python3.buildEnv.override({ makeWrapperArgs = [ "--set LD_PRELOAD ${pkgs.mimalloc}/lib/libmimalloc.so" ]; });
+          withPackages = f: let packages = f pkgs.python3.pkgs; in withMimalloc.override { extraLibs = packages; };
+        };
+
+        # LLVM PGO support
+        llvm-nac3-instrumented = pkgs.callPackage ./nix/llvm {
+          stdenv = pkgs.llvmPackages_14.stdenv;
+          extraCmakeFlags = [ "-DLLVM_BUILD_INSTRUMENTED=IR" ];
+        };
+        nac3artiq-instrumented = pkgs.python3Packages.toPythonModule (
+          pkgs.rustPlatform.buildRustPackage {
+            name = "nac3artiq-instrumented";
+            src = self;
+            inherit (nac3artiq) cargoLock;
+            nativeBuildInputs = [ pkgs.python3 packages.x86_64-linux.llvm-tools-irrt llvm-nac3-instrumented ];
+            buildInputs = [ pkgs.python3 llvm-nac3-instrumented ];
+            cargoBuildFlags = [ "--package" "nac3artiq" "--features" "init-llvm-profile" ];
+            doCheck = false;
+            configurePhase =
+              ''
+              export CARGO_TARGET_X86_64_UNKNOWN_LINUX_GNU_RUSTFLAGS="-C link-arg=-L${pkgs.llvmPackages_14.compiler-rt}/lib/linux -C link-arg=-lclang_rt.profile-x86_64"
+              '';
+            installPhase =
+              ''
+              TARGET_DIR=$out/${pkgs.python3Packages.python.sitePackages}
+              mkdir -p $TARGET_DIR
+              cp target/x86_64-unknown-linux-gnu/release/libnac3artiq.so $TARGET_DIR/nac3artiq.so
+              '';
+          }
+        );
+        nac3artiq-profile = pkgs.stdenvNoCC.mkDerivation {
+          name = "nac3artiq-profile";
+          srcs = [
+            (pkgs.fetchFromGitHub {
+              owner = "m-labs";
+              repo = "sipyco";
+              rev = "939f84f9b5eef7efbf7423c735d1834783b6140e";
+              sha256 = "sha256-15Nun4EY35j+6SPZkjzZtyH/ncxLS60KuGJjFh5kSTc=";
+            })
+            (pkgs.fetchFromGitHub {
+              owner = "m-labs";
+              repo = "artiq";
+              rev = "923ca3377d42c815f979983134ec549dc39d3ca0";
+              sha256 = "sha256-oJoEeNEeNFSUyh6jXG8Tzp6qHVikeHS0CzfE+mODPgw=";
+            })
+          ];
+          buildInputs = [
+            (python3-mimalloc.withPackages(ps: [ ps.numpy ps.scipy ps.jsonschema ps.lmdb nac3artiq-instrumented ]))
+            pkgs.llvmPackages_14.llvm.out
+          ];
+          phases = [ "buildPhase" "installPhase" ];
+          buildPhase =
+            ''
+            srcs=($srcs)
+            sipyco=''${srcs[0]}
+            artiq=''${srcs[1]}
+            export PYTHONPATH=$sipyco:$artiq
+            python -m artiq.frontend.artiq_ddb_template $artiq/artiq/examples/nac3devices/nac3devices.json > device_db.py
+            cp $artiq/artiq/examples/nac3devices/nac3devices.py .
+            python -m artiq.frontend.artiq_compile nac3devices.py
+            '';
+          installPhase =
+            ''
+            mkdir $out
+            llvm-profdata merge -o $out/llvm.profdata /build/llvm/build/profiles/*
+            '';
+        };
+        llvm-nac3-pgo = pkgs.callPackage ./nix/llvm {
+          stdenv = pkgs.llvmPackages_14.stdenv;
+          extraCmakeFlags = [ "-DLLVM_PROFDATA_FILE=${nac3artiq-profile}/llvm.profdata" ];
+        };
+        nac3artiq-pgo = pkgs.python3Packages.toPythonModule (
+          pkgs.rustPlatform.buildRustPackage {
+            name = "nac3artiq-pgo";
+            src = self;
+            inherit (nac3artiq) cargoLock;
+            nativeBuildInputs = [ pkgs.python3 packages.x86_64-linux.llvm-tools-irrt llvm-nac3-pgo ];
+            buildInputs = [ pkgs.python3 llvm-nac3-pgo ];
+            cargoBuildFlags = [ "--package" "nac3artiq" ];
+            cargoTestFlags = [ "--package" "nac3ast" "--package" "nac3parser" "--package" "nac3core" "--package" "nac3artiq" ];
+            installPhase =
+              ''
+              TARGET_DIR=$out/${pkgs.python3Packages.python.sitePackages}
+              mkdir -p $TARGET_DIR
+              cp target/x86_64-unknown-linux-gnu/release/libnac3artiq.so $TARGET_DIR/nac3artiq.so
+              '';
+          }
+        );
+      };
+
+      packages.x86_64-w64-mingw32 = import ./nix/windows { inherit pkgs; };
+
+      devShells.x86_64-linux.default = pkgs.mkShell {
+        name = "nac3-dev-shell";
+        buildInputs = with pkgs; [
+          # build dependencies
+          packages.x86_64-linux.llvm-nac3
+          (pkgs.wrapClangMulti llvmPackages_14.clang) llvmPackages_14.llvm.out  # for running nac3standalone demos
+          packages.x86_64-linux.llvm-tools-irrt
+          cargo
+          rustc
+          # runtime dependencies
+          lld_14                           # for running kernels on the host
+          (packages.x86_64-linux.python3-mimalloc.withPackages(ps: [ ps.numpy ps.scipy ]))
+          # development tools
+          cargo-insta
+          clippy
+          pre-commit
+          rustfmt
+        ];
+        shellHook =
+          ''
+          export DEMO_LINALG_STUB=${packages.x86_64-linux.demo-linalg-stub}/lib/liblinalg.a
+          export DEMO_LINALG_STUB32=${packages.x86_64-linux.demo-linalg-stub32}/lib/liblinalg.a
+          '';
+      };
+      devShells.x86_64-linux.msys2 = pkgs.mkShell {
+        name = "nac3-dev-shell-msys2";
+        buildInputs = with pkgs; [
+          curl
+          pacman
+          fakeroot
+          packages.x86_64-w64-mingw32.wine-msys2
+        ];
+      };
+
+      hydraJobs = {
+        inherit (packages.x86_64-linux) llvm-nac3 nac3artiq nac3artiq-pgo;
+        llvm-nac3-msys2 = packages.x86_64-w64-mingw32.llvm-nac3;
+        nac3artiq-msys2 = packages.x86_64-w64-mingw32.nac3artiq;
+        nac3artiq-msys2-pkg = packages.x86_64-w64-mingw32.nac3artiq-pkg;
+      };
+  };
+
+  nixConfig = {
+    extra-trusted-public-keys = "nixbld.m-labs.hk-1:5aSRVA5b320xbNvu30tqxVPXpld73bhtOeH6uAjRyHc=";
+    extra-substituters = "https://nixbld.m-labs.hk";
+  };
+}

nac3.svg

@@ -0,0 +1,56 @@
+<?xml version="1.0" encoding="UTF-8" standalone="no"?>
+<svg
+   id="a"
+   width="128"
+   height="128"
+   viewBox="0 0 95.99999 95.99999"
+   version="1.1"
+   sodipodi:docname="nac3.svg"
+   inkscape:version="1.1.1 (3bf5ae0d25, 2021-09-20)"
+   xmlns:inkscape="http://www.inkscape.org/namespaces/inkscape"
+   xmlns:sodipodi="http://sodipodi.sourceforge.net/DTD/sodipodi-0.dtd"
+   xmlns="http://www.w3.org/2000/svg"
+   xmlns:svg="http://www.w3.org/2000/svg">
+  <defs
+     id="defs11" />
+  <sodipodi:namedview
+     id="namedview9"
+     pagecolor="#ffffff"
+     bordercolor="#666666"
+     borderopacity="1.0"
+     inkscape:pageshadow="2"
+     inkscape:pageopacity="0.0"
+     inkscape:pagecheckerboard="0"
+     inkscape:document-units="mm"
+     showgrid="false"
+     units="px"
+     width="128px"
+     inkscape:zoom="5.9448568"
+     inkscape:cx="60.472441"
+     inkscape:cy="60.556547"
+     inkscape:window-width="2560"
+     inkscape:window-height="1371"
+     inkscape:window-x="0"
+     inkscape:window-y="32"
+     inkscape:window-maximized="1"
+     inkscape:current-layer="a" />
+  <rect
+     x="40.072601"
+     y="-26.776209"
+     width="55.668747"
+     height="55.668747"
+     transform="matrix(0.71803815,0.69600374,-0.71803815,0.69600374,0,0)"
+     style="fill:#be211e;stroke:#000000;stroke-width:4.37375px;stroke-linecap:round;stroke-linejoin:round"
+     id="rect2" />
+  <line
+     x1="38.00692"
+     y1="63.457153"
+     x2="57.993061"
+     y2="63.457153"
+     style="fill:none;stroke:#000000;stroke-width:4.37269px;stroke-linecap:round;stroke-linejoin:round"
+     id="line4" />
+  <path
+     d="m 48.007301,57.843329 c -1.943097,0 -3.877522,-0.41727 -5.686157,-1.246007 -3.218257,-1.474616 -5.650382,-4.075418 -6.849639,-7.323671 -2.065624,-5.588921 -1.192751,-10.226647 2.575258,-13.827 0.611554,-0.584909 1.518048,-0.773041 2.323689,-0.488206 0.80673,0.286405 1.369495,0.998486 1.447563,1.827234 0.237469,2.549302 2.439719,5.917376 4.28414,6.55273 0.396859,0.13506 0.820953,-0.05859 1.097084,-0.35222 0.339254,-0.360754 0.451065,-0.961893 -1.013597,-3.191372 -2.089851,-3.181137 -4.638728,-8.754903 -0.262407,-15.069853 0.494457,-0.713491 1.384673,-1.068907 2.256469,-0.909156 0.871795,0.161332 1.583757,0.806404 1.752251,1.651189 0.716448,3.591862 2.962357,6.151755 5.199306,8.023138 1.935503,1.61861 4.344688,3.867387 5.435687,7.096643 2.283183,6.758017 -1.202511,14.114988 -8.060822,16.494025 -1.467083,0.509226 -2.98513,0.762536 -4.498836,0.762536 z M 39.358865,40.002192 c -0.304711,0.696206 -0.541636,2.080524 -0.56865,2.237454 -0.330316,1.918771 0.168305,3.803963 0.846157,5.539951 0.856828,2.19436 2.437543,3.942467 4.583411,4.925713 2.143691,0.981675 4.554131,1.097816 6.789992,0.322666 4.571485,-1.586549 6.977584,-6.532238 5.363036,-11.02597 v -5.27e-4 C 55.455481,39.447968 54.023463,38.162043 52.221335,36.65432 50.876945,35.529534 49.409662,33.987726 48.417983,32.135555 48.01343,31.37996 47.79547,30.34303 47.76669,29.413263 c -0.187481,0.669514 -0.212441,2.325923 -0.150396,2.93691 0.179209,1.764456 1.333476,3.644546 2.340611,5.171243 1.311568,1.988179 2.72058,6.037272 0.459681,8.367985 -1.54192,1.58953 -4.038511,2.052034 -5.839973,1.38492 -2.398314,-0.888147 -3.942744,-2.690627 -4.941118,-4.768029 -0.121194,-0.25217 -0.532464,-1.174187 -0.276619,-2.5041 z"
+     id="path6"
+     style="stroke-width:1.09317" />
+</svg>

nac3artiq/Cargo.toml

@@ -0,0 +1,21 @@
+[package]
+name = "nac3artiq"
+version = "0.1.0"
+authors = ["M-Labs"]
+edition = "2021"
+
+[lib]
+name = "nac3artiq"
+crate-type = ["cdylib"]
+
+[dependencies]
+itertools = "0.13"
+pyo3 = { version = "0.21", features = ["extension-module", "gil-refs"] }
+parking_lot = "0.12"
+tempfile = "3.13"
+nac3core = { path = "../nac3core" }
+nac3ld = { path = "../nac3ld" }
+
+[features]
+init-llvm-profile = []
+no-escape-analysis = ["nac3core/no-escape-analysis"]

nac3artiq/demo/demo.py

@@ -0,0 +1,26 @@
+from min_artiq import *
+
+
+@nac3
+class Demo:
+    core: KernelInvariant[Core]
+    led0: KernelInvariant[TTLOut]
+    led1: KernelInvariant[TTLOut]
+
+    def __init__(self):
+        self.core = Core()
+        self.led0 = TTLOut(self.core, 18)
+        self.led1 = TTLOut(self.core, 19)
+
+    @kernel
+    def run(self):
+        self.core.reset()
+        while True:
+            with parallel:
+                self.led0.pulse(100.*ms)
+                self.led1.pulse(100.*ms)
+            self.core.delay(100.*ms)
+
+
+if __name__ == "__main__":
+    Demo().run()

nac3artiq/demo/device_db.py

@@ -0,0 +1,16 @@
+# python demo.py
+# artiq_run module.elf
+
+device_db = {
+    "core": {
+        "type": "local",
+        "module": "artiq.coredevice.core",
+        "class": "Core",
+        "arguments": {
+            "host": "kc705",
+            "ref_period": 1e-9,
+            "ref_multiplier": 8,
+            "target": "rv32g"
+        }
+    },
+}

nac3artiq/demo/embedding_map.py

@@ -0,0 +1,66 @@
+class EmbeddingMap:
+    def __init__(self):
+        self.object_inverse_map = {}
+        self.object_map = {}
+        self.string_map = {}
+        self.string_reverse_map = {}
+        self.function_map = {}
+        self.attributes_writeback = []
+
+        # preallocate exception names
+        self.preallocate_runtime_exception_names(["RuntimeError",
+                                          "RTIOUnderflow",
+                                          "RTIOOverflow",
+                                          "RTIODestinationUnreachable",
+                                          "DMAError",
+                                          "I2CError",
+                                          "CacheError",
+                                          "SPIError",
+                                          "0:ZeroDivisionError",
+                                          "0:IndexError",
+                                          "0:ValueError",
+                                          "0:RuntimeError",
+                                          "0:AssertionError",
+                                          "0:KeyError",
+                                          "0:NotImplementedError",
+                                          "0:OverflowError",
+                                          "0:IOError",
+                                          "0:UnwrapNoneError"])
+
+    def preallocate_runtime_exception_names(self, names):
+        for i, name in enumerate(names):
+            if ":" not in name:
+                name = "0:artiq.coredevice.exceptions." + name
+            exn_id = self.store_str(name)
+            assert exn_id == i
+
+    def store_function(self, key, fun):
+        self.function_map[key] = fun
+        return key
+
+    def store_object(self, obj):
+        obj_id = id(obj)
+        if obj_id in self.object_inverse_map:
+            return self.object_inverse_map[obj_id]
+        key = len(self.object_map) + 1
+        self.object_map[key] = obj
+        self.object_inverse_map[obj_id] = key
+        return key
+
+    def store_str(self, s):
+        if s in self.string_reverse_map:
+            return self.string_reverse_map[s]
+        key = len(self.string_map)
+        self.string_map[key] = s
+        self.string_reverse_map[s] = key
+        return key
+
+    def retrieve_function(self, key):
+        return self.function_map[key]
+
+    def retrieve_object(self, key):
+        return self.object_map[key]
+
+    def retrieve_str(self, key):
+        return self.string_map[key]
+

nac3artiq/demo/list_cmp.py

@@ -0,0 +1,24 @@
+from min_artiq import *
+from numpy import int32
+
+
+@nac3
+class EmptyList:
+    core: KernelInvariant[Core]
+
+    def __init__(self):
+        self.core = Core()
+
+    @rpc
+    def get_empty(self) -> list[int32]:
+        return []
+
+    @kernel
+    def run(self):
+        a: list[int32] = self.get_empty()
+        if a != []:
+            raise ValueError
+
+
+if __name__ == "__main__":
+    EmptyList().run()

nac3artiq/demo/min_artiq.py

@@ -0,0 +1,300 @@
+from inspect import getfullargspec
+from functools import wraps
+from types import SimpleNamespace
+from numpy import int32, int64
+from typing import Generic, TypeVar
+from math import floor, ceil
+
+import nac3artiq
+from embedding_map import EmbeddingMap
+
+
+__all__ = [
+    "Kernel", "KernelInvariant", "virtual", "ConstGeneric",
+    "Option", "Some", "none", "UnwrapNoneError",
+    "round64", "floor64", "ceil64",
+    "extern", "kernel", "portable", "nac3",
+    "rpc", "ms", "us", "ns",
+    "print_int32", "print_int64",
+    "Core", "TTLOut",
+    "parallel", "sequential"
+]
+
+
+T = TypeVar('T')
+
+class Kernel(Generic[T]):
+    pass
+
+class KernelInvariant(Generic[T]):
+    pass
+
+# The virtual class must exist before nac3artiq.NAC3 is created.
+class virtual(Generic[T]):
+    pass
+
+class Option(Generic[T]):
+    _nac3_option: T
+
+    def __init__(self, v: T):
+        self._nac3_option = v
+
+    def is_none(self):
+        return self._nac3_option is None
+    
+    def is_some(self):
+        return not self.is_none()
+    
+    def unwrap(self):
+        if self.is_none():
+            raise UnwrapNoneError()
+        return self._nac3_option
+
+    def __repr__(self) -> str:
+        if self.is_none():
+            return "none"
+        else:
+            return "Some({})".format(repr(self._nac3_option))
+    
+    def __str__(self) -> str:
+        if self.is_none():
+            return "none"
+        else:
+            return "Some({})".format(str(self._nac3_option))
+
+def Some(v: T) -> Option[T]:
+    return Option(v)
+
+none = Option(None)
+
+class _ConstGenericMarker:
+    pass
+
+def ConstGeneric(name, constraint):
+    return TypeVar(name, _ConstGenericMarker, constraint)
+
+def round64(x):
+    return round(x)
+
+def floor64(x):
+    return floor(x)
+
+def ceil64(x):
+    return ceil(x)
+
+
+import device_db
+core_arguments = device_db.device_db["core"]["arguments"]
+
+artiq_builtins = {
+    "none": none,
+    "virtual": virtual,
+    "_ConstGenericMarker": _ConstGenericMarker,
+    "Option": Option,
+}
+compiler = nac3artiq.NAC3(core_arguments["target"], artiq_builtins)
+allow_registration = True
+# Delay NAC3 analysis until all referenced variables are supposed to exist on the CPython side.
+registered_functions = set()
+registered_classes = set()
+
+def register_function(fun):
+    assert allow_registration
+    registered_functions.add(fun)
+
+def register_class(cls):
+    assert allow_registration
+    registered_classes.add(cls)
+
+
+def extern(function):
+    """Decorates a function declaration defined by the core device runtime."""
+    register_function(function)
+    return function
+
+
+def rpc(arg=None, flags={}):
+    """Decorates a function or method to be executed on the host interpreter."""
+    if arg is None:
+        def inner_decorator(function):
+            return rpc(function, flags)
+        return inner_decorator
+    register_function(arg)
+    return arg
+
+def kernel(function_or_method):
+    """Decorates a function or method to be executed on the core device."""
+    register_function(function_or_method)
+    argspec = getfullargspec(function_or_method)
+    if argspec.args and argspec.args[0] == "self":
+        @wraps(function_or_method)
+        def run_on_core(self, *args, **kwargs):
+            fake_method = SimpleNamespace(__self__=self, __name__=function_or_method.__name__)
+            self.core.run(fake_method, *args, **kwargs)
+    else:
+        @wraps(function_or_method)
+        def run_on_core(*args, **kwargs):
+            raise RuntimeError("Kernel functions need explicit core.run()")
+    return run_on_core
+
+
+def portable(function):
+    """Decorates a function or method to be executed on the same device (host/core device) as the caller."""
+    register_function(function)
+    return function
+
+
+def nac3(cls):
+    """
+    Decorates a class to be analyzed by NAC3.
+    All classes containing kernels or portable methods must use this decorator.
+    """
+    register_class(cls)
+    return cls
+
+
+ms = 1e-3
+us = 1e-6
+ns = 1e-9
+
+@extern
+def rtio_init():
+    raise NotImplementedError("syscall not simulated")
+
+
+@extern
+def rtio_get_counter() -> int64:
+    raise NotImplementedError("syscall not simulated")
+
+
+@extern
+def rtio_output(target: int32, data: int32):
+    raise NotImplementedError("syscall not simulated")
+
+
+@extern
+def rtio_input_timestamp(timeout_mu: int64, channel: int32) -> int64:
+    raise NotImplementedError("syscall not simulated")
+
+
+@extern
+def rtio_input_data(channel: int32) -> int32:
+    raise NotImplementedError("syscall not simulated")
+
+
+# These is not part of ARTIQ and only available in runkernel. Defined here for convenience.
+@extern
+def print_int32(x: int32):
+    raise NotImplementedError("syscall not simulated")
+
+
+@extern
+def print_int64(x: int64):
+    raise NotImplementedError("syscall not simulated")
+
+
+@nac3
+class Core:
+    ref_period: KernelInvariant[float]
+
+    def __init__(self):
+        self.ref_period = core_arguments["ref_period"]
+
+    def run(self, method, *args, **kwargs):
+        global allow_registration
+
+        embedding = EmbeddingMap()
+
+        if allow_registration:
+            compiler.analyze(registered_functions, registered_classes)
+            allow_registration = False
+
+        if hasattr(method, "__self__"):
+            obj = method.__self__
+            name = method.__name__
+        else:
+            obj = method
+            name = ""
+
+        compiler.compile_method_to_file(obj, name, args, "module.elf", embedding)
+
+    @kernel
+    def reset(self):
+        rtio_init()
+        at_mu(rtio_get_counter() + int64(125000))
+
+    @kernel
+    def break_realtime(self):
+        min_now = rtio_get_counter() + int64(125000)
+        if now_mu() < min_now:
+            at_mu(min_now)
+
+    @portable
+    def seconds_to_mu(self, seconds: float) -> int64:
+        return int64(round(seconds/self.ref_period))
+
+    @portable
+    def mu_to_seconds(self, mu: int64) -> float:
+        return float(mu)*self.ref_period
+
+    @kernel
+    def delay(self, dt: float):
+        delay_mu(self.seconds_to_mu(dt))
+
+
+@nac3
+class TTLOut:
+    core: KernelInvariant[Core]
+    channel: KernelInvariant[int32]
+    target_o: KernelInvariant[int32]
+
+    def __init__(self, core: Core, channel: int32):
+        self.core = core
+        self.channel = channel
+        self.target_o = channel << 8
+
+    @kernel
+    def output(self):
+        pass
+
+    @kernel
+    def set_o(self, o: bool):
+        rtio_output(self.target_o, 1 if o else 0)
+
+    @kernel
+    def on(self):
+        self.set_o(True)
+
+    @kernel
+    def off(self):
+        self.set_o(False)
+
+    @kernel
+    def pulse_mu(self, duration: int64):
+        self.on()
+        delay_mu(duration)
+        self.off()
+
+    @kernel
+    def pulse(self, duration: float):
+        self.on()
+        self.core.delay(duration)
+        self.off()
+
+@nac3
+class KernelContextManager:
+    @kernel
+    def __enter__(self):
+        pass
+
+    @kernel
+    def __exit__(self):
+        pass
+
+@nac3
+class UnwrapNoneError(Exception):
+    """raised when unwrapping a none value"""
+    artiq_builtin = True
+
+parallel = KernelContextManager()
+sequential = KernelContextManager()

nac3artiq/demo/nac3artiq.so

@@ -0,0 +1 @@
+../../target/release/libnac3artiq.so

nac3artiq/demo/str_abi.py

@@ -0,0 +1,26 @@
+from min_artiq import *
+from numpy import ndarray, zeros as np_zeros
+
+
+@nac3
+class StrFail:
+    core: KernelInvariant[Core]
+
+    def __init__(self):
+        self.core = Core()
+
+    @kernel
+    def hello(self, arg: str):
+        pass
+
+    @kernel
+    def consume_ndarray(self, arg: ndarray[str, 1]):
+        pass
+
+    def run(self):
+        self.hello("world")
+        self.consume_ndarray(np_zeros([10], dtype=str))
+
+
+if __name__ == "__main__":
+    StrFail().run()

nac3artiq/demo/string_attribute_issue337.py

@@ -0,0 +1,24 @@
+from min_artiq import *
+from numpy import int32
+
+
+@nac3
+class Demo:
+    core: KernelInvariant[Core]
+    attr1: KernelInvariant[str]
+    attr2: KernelInvariant[int32]
+
+
+    def __init__(self):
+        self.core = Core()
+        self.attr2 = 32
+        self.attr1 = "SAMPLE"
+
+    @kernel
+    def run(self):
+        print_int32(self.attr2)
+        self.attr1
+
+
+if __name__ == "__main__":
+    Demo().run()

nac3artiq/demo/support_class_attr_issue102.py

@@ -0,0 +1,40 @@
+from min_artiq import *
+from numpy import int32
+
+
+@nac3
+class Demo:
+    attr1: KernelInvariant[int32] = 2
+    attr2: int32 = 4
+    attr3: Kernel[int32]
+
+    @kernel
+    def __init__(self):
+        self.attr3 = 8
+
+
+@nac3
+class NAC3Devices:
+    core: KernelInvariant[Core]
+    attr4: KernelInvariant[int32] = 16
+    
+    def __init__(self):
+        self.core = Core()
+
+    @kernel
+    def run(self):
+        Demo.attr1  # Supported
+        # Demo.attr2  # Field not accessible on Kernel
+        # Demo.attr3  # Only attributes can be accessed in this way
+        # Demo.attr1 = 2  # Attributes are immutable
+
+        self.attr4  # Attributes can be accessed within class
+
+        obj = Demo()
+        obj.attr1  # Attributes can be accessed by class objects
+
+        NAC3Devices.attr4  # Attributes accessible for classes without __init__
+
+
+if __name__ == "__main__":
+    NAC3Devices().run()

nac3artiq/src/kernel.ld

@@ -0,0 +1,56 @@
+/* Force ld to make the ELF header as loadable. */
+PHDRS
+{
+    headers     PT_LOAD FILEHDR PHDRS ;
+    text        PT_LOAD ;
+    data        PT_LOAD ;
+    dynamic     PT_DYNAMIC ;
+    eh_frame    PT_GNU_EH_FRAME ;
+}
+
+SECTIONS
+{
+    /* Push back .text section enough so that ld.lld not complain */
+    . = SIZEOF_HEADERS;
+
+    .text :
+    {
+        *(.text .text.*)
+    } : text
+
+    .rodata :
+    {
+        *(.rodata .rodata.*)
+    }
+
+    .eh_frame :
+    {
+        KEEP(*(.eh_frame))
+    } : text
+
+    .eh_frame_hdr :
+    {
+        KEEP(*(.eh_frame_hdr))
+    } : text : eh_frame
+
+    .data :
+    {
+        *(.data)
+    } : data
+
+    .dynamic :
+    {
+        *(.dynamic)
+    } : data : dynamic
+
+    .bss (NOLOAD) : ALIGN(4)
+    {
+        __bss_start = .;
+        *(.sbss .sbss.* .bss .bss.*);
+        . = ALIGN(4);
+        _end = .;
+    }
+
+    . = ALIGN(0x1000);
+    _sstack_guard = .;
+}

nac3artiq/src/timeline.rs

@@ -0,0 +1,324 @@
+use itertools::Either;
+
+use nac3core::{
+    codegen::CodeGenContext,
+    inkwell::{
+        values::{BasicValueEnum, CallSiteValue},
+        AddressSpace, AtomicOrdering,
+    },
+};
+
+/// Functions for manipulating the timeline.
+pub trait TimeFns {
+    /// Emits LLVM IR for `now_mu`.
+    fn emit_now_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>) -> BasicValueEnum<'ctx>;
+
+    /// Emits LLVM IR for `at_mu`.
+    fn emit_at_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>, t: BasicValueEnum<'ctx>);
+
+    /// Emits LLVM IR for `delay_mu`.
+    fn emit_delay_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>, dt: BasicValueEnum<'ctx>);
+}
+
+pub struct NowPinningTimeFns64 {}
+
+// For FPGA design reasons, on VexRiscv with 64-bit data bus, the "now" CSR is split into two 32-bit
+// values that are each padded to 64-bits.
+impl TimeFns for NowPinningTimeFns64 {
+    fn emit_now_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>) -> BasicValueEnum<'ctx> {
+        let i64_type = ctx.ctx.i64_type();
+        let i32_type = ctx.ctx.i32_type();
+        let now = ctx
+            .module
+            .get_global("now")
+            .unwrap_or_else(|| ctx.module.add_global(i64_type, None, "now"));
+        let now_hiptr = ctx
+            .builder
+            .build_bit_cast(now, i32_type.ptr_type(AddressSpace::default()), "now.hi.addr")
+            .map(BasicValueEnum::into_pointer_value)
+            .unwrap();
+
+        let now_loptr = unsafe {
+            ctx.builder.build_gep(now_hiptr, &[i32_type.const_int(2, false)], "now.lo.addr")
+        }
+        .unwrap();
+
+        let now_hi = ctx
+            .builder
+            .build_load(now_hiptr, "now.hi")
+            .map(BasicValueEnum::into_int_value)
+            .unwrap();
+        let now_lo = ctx
+            .builder
+            .build_load(now_loptr, "now.lo")
+            .map(BasicValueEnum::into_int_value)
+            .unwrap();
+
+        let zext_hi = ctx.builder.build_int_z_extend(now_hi, i64_type, "").unwrap();
+        let shifted_hi =
+            ctx.builder.build_left_shift(zext_hi, i64_type.const_int(32, false), "").unwrap();
+        let zext_lo = ctx.builder.build_int_z_extend(now_lo, i64_type, "").unwrap();
+        ctx.builder.build_or(shifted_hi, zext_lo, "now_mu").map(Into::into).unwrap()
+    }
+
+    fn emit_at_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>, t: BasicValueEnum<'ctx>) {
+        let i32_type = ctx.ctx.i32_type();
+        let i64_type = ctx.ctx.i64_type();
+
+        let i64_32 = i64_type.const_int(32, false);
+        let time = t.into_int_value();
+
+        let time_hi = ctx
+            .builder
+            .build_int_truncate(
+                ctx.builder.build_right_shift(time, i64_32, false, "time.hi").unwrap(),
+                i32_type,
+                "",
+            )
+            .unwrap();
+        let time_lo = ctx.builder.build_int_truncate(time, i32_type, "time.lo").unwrap();
+        let now = ctx
+            .module
+            .get_global("now")
+            .unwrap_or_else(|| ctx.module.add_global(i64_type, None, "now"));
+        let now_hiptr = ctx
+            .builder
+            .build_bit_cast(now, i32_type.ptr_type(AddressSpace::default()), "now.hi.addr")
+            .map(BasicValueEnum::into_pointer_value)
+            .unwrap();
+
+        let now_loptr = unsafe {
+            ctx.builder.build_gep(now_hiptr, &[i32_type.const_int(2, false)], "now.lo.addr")
+        }
+        .unwrap();
+        ctx.builder
+            .build_store(now_hiptr, time_hi)
+            .unwrap()
+            .set_atomic_ordering(AtomicOrdering::SequentiallyConsistent)
+            .unwrap();
+        ctx.builder
+            .build_store(now_loptr, time_lo)
+            .unwrap()
+            .set_atomic_ordering(AtomicOrdering::SequentiallyConsistent)
+            .unwrap();
+    }
+
+    fn emit_delay_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>, dt: BasicValueEnum<'ctx>) {
+        let i64_type = ctx.ctx.i64_type();
+        let i32_type = ctx.ctx.i32_type();
+        let now = ctx
+            .module
+            .get_global("now")
+            .unwrap_or_else(|| ctx.module.add_global(i64_type, None, "now"));
+        let now_hiptr = ctx
+            .builder
+            .build_bit_cast(now, i32_type.ptr_type(AddressSpace::default()), "now.hi.addr")
+            .map(BasicValueEnum::into_pointer_value)
+            .unwrap();
+
+        let now_loptr = unsafe {
+            ctx.builder.build_gep(now_hiptr, &[i32_type.const_int(2, false)], "now.lo.addr")
+        }
+        .unwrap();
+
+        let now_hi = ctx
+            .builder
+            .build_load(now_hiptr, "now.hi")
+            .map(BasicValueEnum::into_int_value)
+            .unwrap();
+        let now_lo = ctx
+            .builder
+            .build_load(now_loptr, "now.lo")
+            .map(BasicValueEnum::into_int_value)
+            .unwrap();
+        let dt = dt.into_int_value();
+
+        let zext_hi = ctx.builder.build_int_z_extend(now_hi, i64_type, "").unwrap();
+        let shifted_hi =
+            ctx.builder.build_left_shift(zext_hi, i64_type.const_int(32, false), "").unwrap();
+        let zext_lo = ctx.builder.build_int_z_extend(now_lo, i64_type, "").unwrap();
+        let now_val = ctx.builder.build_or(shifted_hi, zext_lo, "now").unwrap();
+
+        let time = ctx.builder.build_int_add(now_val, dt, "time").unwrap();
+        let time_hi = ctx
+            .builder
+            .build_int_truncate(
+                ctx.builder
+                    .build_right_shift(time, i64_type.const_int(32, false), false, "")
+                    .unwrap(),
+                i32_type,
+                "time.hi",
+            )
+            .unwrap();
+        let time_lo = ctx.builder.build_int_truncate(time, i32_type, "time.lo").unwrap();
+
+        ctx.builder
+            .build_store(now_hiptr, time_hi)
+            .unwrap()
+            .set_atomic_ordering(AtomicOrdering::SequentiallyConsistent)
+            .unwrap();
+        ctx.builder
+            .build_store(now_loptr, time_lo)
+            .unwrap()
+            .set_atomic_ordering(AtomicOrdering::SequentiallyConsistent)
+            .unwrap();
+    }
+}
+
+pub static NOW_PINNING_TIME_FNS_64: NowPinningTimeFns64 = NowPinningTimeFns64 {};
+
+pub struct NowPinningTimeFns {}
+
+impl TimeFns for NowPinningTimeFns {
+    fn emit_now_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>) -> BasicValueEnum<'ctx> {
+        let i64_type = ctx.ctx.i64_type();
+        let now = ctx
+            .module
+            .get_global("now")
+            .unwrap_or_else(|| ctx.module.add_global(i64_type, None, "now"));
+        let now_raw = ctx
+            .builder
+            .build_load(now.as_pointer_value(), "now")
+            .map(BasicValueEnum::into_int_value)
+            .unwrap();
+
+        let i64_32 = i64_type.const_int(32, false);
+        let now_lo = ctx.builder.build_left_shift(now_raw, i64_32, "now.lo").unwrap();
+        let now_hi = ctx.builder.build_right_shift(now_raw, i64_32, false, "now.hi").unwrap();
+        ctx.builder.build_or(now_lo, now_hi, "now_mu").map(Into::into).unwrap()
+    }
+
+    fn emit_at_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>, t: BasicValueEnum<'ctx>) {
+        let i32_type = ctx.ctx.i32_type();
+        let i64_type = ctx.ctx.i64_type();
+        let i64_32 = i64_type.const_int(32, false);
+
+        let time = t.into_int_value();
+
+        let time_hi = ctx
+            .builder
+            .build_int_truncate(
+                ctx.builder.build_right_shift(time, i64_32, false, "").unwrap(),
+                i32_type,
+                "time.hi",
+            )
+            .unwrap();
+        let time_lo = ctx.builder.build_int_truncate(time, i32_type, "now_trunc").unwrap();
+        let now = ctx
+            .module
+            .get_global("now")
+            .unwrap_or_else(|| ctx.module.add_global(i64_type, None, "now"));
+        let now_hiptr = ctx
+            .builder
+            .build_bit_cast(now, i32_type.ptr_type(AddressSpace::default()), "now.hi.addr")
+            .map(BasicValueEnum::into_pointer_value)
+            .unwrap();
+
+        let now_loptr = unsafe {
+            ctx.builder.build_gep(now_hiptr, &[i32_type.const_int(1, false)], "now.lo.addr")
+        }
+        .unwrap();
+        ctx.builder
+            .build_store(now_hiptr, time_hi)
+            .unwrap()
+            .set_atomic_ordering(AtomicOrdering::SequentiallyConsistent)
+            .unwrap();
+        ctx.builder
+            .build_store(now_loptr, time_lo)
+            .unwrap()
+            .set_atomic_ordering(AtomicOrdering::SequentiallyConsistent)
+            .unwrap();
+    }
+
+    fn emit_delay_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>, dt: BasicValueEnum<'ctx>) {
+        let i32_type = ctx.ctx.i32_type();
+        let i64_type = ctx.ctx.i64_type();
+        let i64_32 = i64_type.const_int(32, false);
+        let now = ctx
+            .module
+            .get_global("now")
+            .unwrap_or_else(|| ctx.module.add_global(i64_type, None, "now"));
+        let now_raw = ctx
+            .builder
+            .build_load(now.as_pointer_value(), "")
+            .map(BasicValueEnum::into_int_value)
+            .unwrap();
+
+        let dt = dt.into_int_value();
+
+        let now_lo = ctx.builder.build_left_shift(now_raw, i64_32, "now.lo").unwrap();
+        let now_hi = ctx.builder.build_right_shift(now_raw, i64_32, false, "now.hi").unwrap();
+        let now_val = ctx.builder.build_or(now_lo, now_hi, "now_val").unwrap();
+        let time = ctx.builder.build_int_add(now_val, dt, "time").unwrap();
+        let time_hi = ctx
+            .builder
+            .build_int_truncate(
+                ctx.builder.build_right_shift(time, i64_32, false, "time.hi").unwrap(),
+                i32_type,
+                "now_trunc",
+            )
+            .unwrap();
+        let time_lo = ctx.builder.build_int_truncate(time, i32_type, "time.lo").unwrap();
+        let now_hiptr = ctx
+            .builder
+            .build_bit_cast(now, i32_type.ptr_type(AddressSpace::default()), "now.hi.addr")
+            .map(BasicValueEnum::into_pointer_value)
+            .unwrap();
+
+        let now_loptr = unsafe {
+            ctx.builder.build_gep(now_hiptr, &[i32_type.const_int(1, false)], "now.lo.addr")
+        }
+        .unwrap();
+        ctx.builder
+            .build_store(now_hiptr, time_hi)
+            .unwrap()
+            .set_atomic_ordering(AtomicOrdering::SequentiallyConsistent)
+            .unwrap();
+        ctx.builder
+            .build_store(now_loptr, time_lo)
+            .unwrap()
+            .set_atomic_ordering(AtomicOrdering::SequentiallyConsistent)
+            .unwrap();
+    }
+}
+
+pub static NOW_PINNING_TIME_FNS: NowPinningTimeFns = NowPinningTimeFns {};
+
+pub struct ExternTimeFns {}
+
+impl TimeFns for ExternTimeFns {
+    fn emit_now_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>) -> BasicValueEnum<'ctx> {
+        let now_mu = ctx.module.get_function("now_mu").unwrap_or_else(|| {
+            ctx.module.add_function("now_mu", ctx.ctx.i64_type().fn_type(&[], false), None)
+        });
+        ctx.builder
+            .build_call(now_mu, &[], "now_mu")
+            .map(CallSiteValue::try_as_basic_value)
+            .map(Either::unwrap_left)
+            .unwrap()
+    }
+
+    fn emit_at_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>, t: BasicValueEnum<'ctx>) {
+        let at_mu = ctx.module.get_function("at_mu").unwrap_or_else(|| {
+            ctx.module.add_function(
+                "at_mu",
+                ctx.ctx.void_type().fn_type(&[ctx.ctx.i64_type().into()], false),
+                None,
+            )
+        });
+        ctx.builder.build_call(at_mu, &[t.into()], "at_mu").unwrap();
+    }
+
+    fn emit_delay_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>, dt: BasicValueEnum<'ctx>) {
+        let delay_mu = ctx.module.get_function("delay_mu").unwrap_or_else(|| {
+            ctx.module.add_function(
+                "delay_mu",
+                ctx.ctx.void_type().fn_type(&[ctx.ctx.i64_type().into()], false),
+                None,
+            )
+        });
+        ctx.builder.build_call(delay_mu, &[dt.into()], "delay_mu").unwrap();
+    }
+}
+
+pub static EXTERN_TIME_FNS: ExternTimeFns = ExternTimeFns {};

nac3ast/Cargo.toml

@@ -0,0 +1,16 @@
+[package]
+name = "nac3ast"
+version = "0.1.0"
+authors = ["RustPython Team", "M-Labs"]
+edition = "2021"
+
+[features]
+default = ["constant-optimization", "fold"]
+constant-optimization = ["fold"]
+fold = []
+
+[dependencies]
+lazy_static = "1.5"
+parking_lot = "0.12"
+string-interner = "0.17"
+fxhash = "0.2"

nac3ast/Python.asdl

@@ -0,0 +1,127 @@
+-- ASDL's 4 builtin types are:
+-- identifier, int, string, constant
+
+module Python
+{
+    mod = Module(stmt* body, type_ignore* type_ignores)
+        | Interactive(stmt* body)
+        | Expression(expr body)
+        | FunctionType(expr* argtypes, expr returns)
+
+    stmt = FunctionDef(identifier name, arguments args,
+                       stmt* body, expr* decorator_list, expr? returns,
+                       string? type_comment, identifier* config_comment)
+          | AsyncFunctionDef(identifier name, arguments args,
+                             stmt* body, expr* decorator_list, expr? returns,
+                             string? type_comment, identifier* config_comment)
+
+          | ClassDef(identifier name,
+             expr* bases,
+             keyword* keywords,
+             stmt* body,
+             expr* decorator_list, identifier* config_comment)
+          | Return(expr? value, identifier* config_comment)
+
+          | Delete(expr* targets, identifier* config_comment)
+          | Assign(expr* targets, expr value, string? type_comment, identifier* config_comment)
+          | AugAssign(expr target, operator op, expr value, identifier* config_comment)
+          -- 'simple' indicates that we annotate simple name without parens
+          | AnnAssign(expr target, expr annotation, expr? value, bool simple, identifier* config_comment)
+
+          -- use 'orelse' because else is a keyword in target languages
+          | For(expr target, expr iter, stmt* body, stmt* orelse, string? type_comment, identifier* config_comment)
+          | AsyncFor(expr target, expr iter, stmt* body, stmt* orelse, string? type_comment, identifier* config_comment)
+          | While(expr test, stmt* body, stmt* orelse, identifier* config_comment)
+          | If(expr test, stmt* body, stmt* orelse, identifier* config_comment)
+          | With(withitem* items, stmt* body, string? type_comment, identifier* config_comment)
+          | AsyncWith(withitem* items, stmt* body, string? type_comment, identifier* config_comment)
+
+          | Raise(expr? exc, expr? cause, identifier* config_comment)
+          | Try(stmt* body, excepthandler* handlers, stmt* orelse, stmt* finalbody, identifier* config_comment)
+          | Assert(expr test, expr? msg, identifier* config_comment)
+
+          | Import(alias* names, identifier* config_comment)
+          | ImportFrom(identifier? module, alias* names, int level, identifier* config_comment)
+
+          | Global(identifier* names, identifier* config_comment)
+          | Nonlocal(identifier* names, identifier* config_comment)
+          | Expr(expr value, identifier* config_comment)
+          | Pass(identifier* config_comment)
+          | Break(identifier* config_comment)
+          | Continue(identifier* config_comment)
+
+          -- col_offset is the byte offset in the utf8 string the parser uses
+          attributes (int lineno, int col_offset, int? end_lineno, int? end_col_offset)
+
+          -- BoolOp() can use left & right?
+    expr = BoolOp(boolop op, expr* values)
+         | NamedExpr(expr target, expr value)
+         | BinOp(expr left, operator op, expr right)
+         | UnaryOp(unaryop op, expr operand)
+         | Lambda(arguments args, expr body)
+         | IfExp(expr test, expr body, expr orelse)
+         | Dict(expr?* keys, expr* values)
+         | Set(expr* elts)
+         | ListComp(expr elt, comprehension* generators)
+         | SetComp(expr elt, comprehension* generators)
+         | DictComp(expr key, expr value, comprehension* generators)
+         | GeneratorExp(expr elt, comprehension* generators)
+         -- the grammar constrains where yield expressions can occur
+         | Await(expr value)
+         | Yield(expr? value)
+         | YieldFrom(expr value)
+         -- need sequences for compare to distinguish between
+         -- x < 4 < 3 and (x < 4) < 3
+         | Compare(expr left, cmpop* ops, expr* comparators)
+         | Call(expr func, expr* args, keyword* keywords)
+         | FormattedValue(expr value, conversion_flag? conversion, expr? format_spec)
+         | JoinedStr(expr* values)
+         | Constant(constant value, string? kind)
+
+         -- the following expression can appear in assignment context
+         | Attribute(expr value, identifier attr, expr_context ctx)
+         | Subscript(expr value, expr slice, expr_context ctx)
+         | Starred(expr value, expr_context ctx)
+         | Name(identifier id, expr_context ctx)
+         | List(expr* elts, expr_context ctx)
+         | Tuple(expr* elts, expr_context ctx)
+
+         -- can appear only in Subscript
+         | Slice(expr? lower, expr? upper, expr? step)
+
+          -- col_offset is the byte offset in the utf8 string the parser uses
+          attributes (int lineno, int col_offset, int? end_lineno, int? end_col_offset)
+
+    expr_context = Load | Store | Del
+
+    boolop = And | Or
+
+    operator = Add | Sub | Mult | MatMult | Div | Mod | Pow | LShift
+                 | RShift | BitOr | BitXor | BitAnd | FloorDiv
+
+    unaryop = Invert | Not | UAdd | USub
+
+    cmpop = Eq | NotEq | Lt | LtE | Gt | GtE | Is | IsNot | In | NotIn
+
+    comprehension = (expr target, expr iter, expr* ifs, bool is_async)
+
+    excepthandler = ExceptHandler(expr? type, identifier? name, stmt* body)
+                    attributes (int lineno, int col_offset, int? end_lineno, int? end_col_offset)
+
+    arguments = (arg* posonlyargs, arg* args, arg? vararg, arg* kwonlyargs,
+                 expr?* kw_defaults, arg? kwarg, expr* defaults)
+
+    arg = (identifier arg, expr? annotation, string? type_comment)
+           attributes (int lineno, int col_offset, int? end_lineno, int? end_col_offset)
+
+    -- keyword arguments supplied to call (NULL identifier for **kwargs)
+    keyword = (identifier? arg, expr value)
+               attributes (int lineno, int col_offset, int? end_lineno, int? end_col_offset)
+
+    -- import name with optional 'as' alias.
+    alias = (identifier name, identifier? asname)
+
+    withitem = (expr context_expr, expr? optional_vars)
+
+    type_ignore = TypeIgnore(int lineno, string tag)
+}

nac3ast/asdl.py

@@ -0,0 +1,385 @@
+#-------------------------------------------------------------------------------
+# Parser for ASDL [1] definition files. Reads in an ASDL description and parses
+# it into an AST that describes it.
+#
+# The EBNF we're parsing here: Figure 1 of the paper [1]. Extended to support
+# modules and attributes after a product. Words starting with Capital letters
+# are terminals. Literal tokens are in "double quotes". Others are
+# non-terminals. Id is either TokenId or ConstructorId.
+#
+# module        ::= "module" Id "{" [definitions] "}"
+# definitions   ::= { TypeId "=" type }
+# type          ::= product | sum
+# product       ::= fields ["attributes" fields]
+# fields        ::= "(" { field, "," } field ")"
+# field         ::= TypeId ["?" | "*"] [Id]
+# sum           ::= constructor { "|" constructor } ["attributes" fields]
+# constructor   ::= ConstructorId [fields]
+#
+# [1] "The Zephyr Abstract Syntax Description Language" by Wang, et. al. See
+#     http://asdl.sourceforge.net/
+#-------------------------------------------------------------------------------
+from collections import namedtuple
+import re
+
+__all__ = [
+    'builtin_types', 'parse', 'AST', 'Module', 'Type', 'Constructor',
+    'Field', 'Sum', 'Product', 'VisitorBase', 'Check', 'check']
+
+# The following classes define nodes into which the ASDL description is parsed.
+# Note: this is a "meta-AST". ASDL files (such as Python.asdl) describe the AST
+# structure used by a programming language. But ASDL files themselves need to be
+# parsed. This module parses ASDL files and uses a simple AST to represent them.
+# See the EBNF at the top of the file to understand the logical connection
+# between the various node types.
+
+builtin_types = {'identifier', 'string', 'int', 'constant', 'bool', 'conversion_flag'}
+
+class AST:
+    def __repr__(self):
+        raise NotImplementedError
+
+class Module(AST):
+    def __init__(self, name, dfns):
+        self.name = name
+        self.dfns = dfns
+        self.types = {type.name: type.value for type in dfns}
+
+    def __repr__(self):
+        return 'Module({0.name}, {0.dfns})'.format(self)
+
+class Type(AST):
+    def __init__(self, name, value):
+        self.name = name
+        self.value = value
+
+    def __repr__(self):
+        return 'Type({0.name}, {0.value})'.format(self)
+
+class Constructor(AST):
+    def __init__(self, name, fields=None):
+        self.name = name
+        self.fields = fields or []
+
+    def __repr__(self):
+        return 'Constructor({0.name}, {0.fields})'.format(self)
+
+class Field(AST):
+    def __init__(self, type, name=None, seq=False, opt=False):
+        self.type = type
+        self.name = name
+        self.seq = seq
+        self.opt = opt
+
+    def __str__(self):
+        if self.seq:
+            extra = "*"
+        elif self.opt:
+            extra = "?"
+        else:
+            extra = ""
+
+        return "{}{} {}".format(self.type, extra, self.name)
+
+    def __repr__(self):
+        if self.seq:
+            extra = ", seq=True"
+        elif self.opt:
+            extra = ", opt=True"
+        else:
+            extra = ""
+        if self.name is None:
+            return 'Field({0.type}{1})'.format(self, extra)
+        else:
+            return 'Field({0.type}, {0.name}{1})'.format(self, extra)
+
+class Sum(AST):
+    def __init__(self, types, attributes=None):
+        self.types = types
+        self.attributes = attributes or []
+
+    def __repr__(self):
+        if self.attributes:
+            return 'Sum({0.types}, {0.attributes})'.format(self)
+        else:
+            return 'Sum({0.types})'.format(self)
+
+class Product(AST):
+    def __init__(self, fields, attributes=None):
+        self.fields = fields
+        self.attributes = attributes or []
+
+    def __repr__(self):
+        if self.attributes:
+            return 'Product({0.fields}, {0.attributes})'.format(self)
+        else:
+            return 'Product({0.fields})'.format(self)
+
+# A generic visitor for the meta-AST that describes ASDL. This can be used by
+# emitters. Note that this visitor does not provide a generic visit method, so a
+# subclass needs to define visit methods from visitModule to as deep as the
+# interesting node.
+# We also define a Check visitor that makes sure the parsed ASDL is well-formed.
+
+class VisitorBase(object):
+    """Generic tree visitor for ASTs."""
+    def __init__(self):
+        self.cache = {}
+
+    def visit(self, obj, *args):
+        klass = obj.__class__
+        meth = self.cache.get(klass)
+        if meth is None:
+            methname = "visit" + klass.__name__
+            meth = getattr(self, methname, None)
+            self.cache[klass] = meth
+        if meth:
+            try:
+                meth(obj, *args)
+            except Exception as e:
+                print("Error visiting %r: %s" % (obj, e))
+                raise
+
+class Check(VisitorBase):
+    """A visitor that checks a parsed ASDL tree for correctness.
+
+    Errors are printed and accumulated.
+    """
+    def __init__(self):
+        super(Check, self).__init__()
+        self.cons = {}
+        self.errors = 0
+        self.types = {}
+
+    def visitModule(self, mod):
+        for dfn in mod.dfns:
+            self.visit(dfn)
+
+    def visitType(self, type):
+        self.visit(type.value, str(type.name))
+
+    def visitSum(self, sum, name):
+        for t in sum.types:
+            self.visit(t, name)
+
+    def visitConstructor(self, cons, name):
+        key = str(cons.name)
+        conflict = self.cons.get(key)
+        if conflict is None:
+            self.cons[key] = name
+        else:
+            print('Redefinition of constructor {}'.format(key))
+            print('Defined in {} and {}'.format(conflict, name))
+            self.errors += 1
+        for f in cons.fields:
+            self.visit(f, key)
+
+    def visitField(self, field, name):
+        key = str(field.type)
+        l = self.types.setdefault(key, [])
+        l.append(name)
+
+    def visitProduct(self, prod, name):
+        for f in prod.fields:
+            self.visit(f, name)
+
+def check(mod):
+    """Check the parsed ASDL tree for correctness.
+
+    Return True if success. For failure, the errors are printed out and False
+    is returned.
+    """
+    v = Check()
+    v.visit(mod)
+
+    for t in v.types:
+        if t not in mod.types and not t in builtin_types:
+            v.errors += 1
+            uses = ", ".join(v.types[t])
+            print('Undefined type {}, used in {}'.format(t, uses))
+    return not v.errors
+
+# The ASDL parser itself comes next. The only interesting external interface
+# here is the top-level parse function.
+
+def parse(filename):
+    """Parse ASDL from the given file and return a Module node describing it."""
+    with open(filename) as f:
+        parser = ASDLParser()
+        return parser.parse(f.read())
+
+# Types for describing tokens in an ASDL specification.
+class TokenKind:
+    """TokenKind is provides a scope for enumerated token kinds."""
+    (ConstructorId, TypeId, Equals, Comma, Question, Pipe, Asterisk,
+     LParen, RParen, LBrace, RBrace) = range(11)
+
+    operator_table = {
+        '=': Equals, ',': Comma,    '?': Question, '|': Pipe,    '(': LParen,
+        ')': RParen, '*': Asterisk, '{': LBrace,   '}': RBrace}
+
+Token = namedtuple('Token', 'kind value lineno')
+
+class ASDLSyntaxError(Exception):
+    def __init__(self, msg, lineno=None):
+        self.msg = msg
+        self.lineno = lineno or '<unknown>'
+
+    def __str__(self):
+        return 'Syntax error on line {0.lineno}: {0.msg}'.format(self)
+
+def tokenize_asdl(buf):
+    """Tokenize the given buffer. Yield Token objects."""
+    for lineno, line in enumerate(buf.splitlines(), 1):
+        for m in re.finditer(r'\s*(\w+|--.*|.)', line.strip()):
+            c = m.group(1)
+            if c[0].isalpha():
+                # Some kind of identifier
+                if c[0].isupper():
+                    yield Token(TokenKind.ConstructorId, c, lineno)
+                else:
+                    yield Token(TokenKind.TypeId, c, lineno)
+            elif c[:2] == '--':
+                # Comment
+                break
+            else:
+                # Operators
+                try:
+                    op_kind = TokenKind.operator_table[c]
+                except KeyError:
+                    raise ASDLSyntaxError('Invalid operator %s' % c, lineno)
+                yield Token(op_kind, c, lineno)
+
+class ASDLParser:
+    """Parser for ASDL files.
+
+    Create, then call the parse method on a buffer containing ASDL.
+    This is a simple recursive descent parser that uses tokenize_asdl for the
+    lexing.
+    """
+    def __init__(self):
+        self._tokenizer = None
+        self.cur_token = None
+
+    def parse(self, buf):
+        """Parse the ASDL in the buffer and return an AST with a Module root.
+        """
+        self._tokenizer = tokenize_asdl(buf)
+        self._advance()
+        return self._parse_module()
+
+    def _parse_module(self):
+        if self._at_keyword('module'):
+            self._advance()
+        else:
+            raise ASDLSyntaxError(
+                'Expected "module" (found {})'.format(self.cur_token.value),
+                self.cur_token.lineno)
+        name = self._match(self._id_kinds)
+        self._match(TokenKind.LBrace)
+        defs = self._parse_definitions()
+        self._match(TokenKind.RBrace)
+        return Module(name, defs)
+
+    def _parse_definitions(self):
+        defs = []
+        while self.cur_token.kind == TokenKind.TypeId:
+            typename = self._advance()
+            self._match(TokenKind.Equals)
+            type = self._parse_type()
+            defs.append(Type(typename, type))
+        return defs
+
+    def _parse_type(self):
+        if self.cur_token.kind == TokenKind.LParen:
+            # If we see a (, it's a product
+            return self._parse_product()
+        else:
+            # Otherwise it's a sum. Look for ConstructorId
+            sumlist = [Constructor(self._match(TokenKind.ConstructorId),
+                                   self._parse_optional_fields())]
+            while self.cur_token.kind  == TokenKind.Pipe:
+                # More constructors
+                self._advance()
+                sumlist.append(Constructor(
+                                self._match(TokenKind.ConstructorId),
+                                self._parse_optional_fields()))
+            return Sum(sumlist, self._parse_optional_attributes())
+
+    def _parse_product(self):
+        return Product(self._parse_fields(), self._parse_optional_attributes())
+
+    def _parse_fields(self):
+        fields = []
+        self._match(TokenKind.LParen)
+        while self.cur_token.kind == TokenKind.TypeId:
+            typename = self._advance()
+            is_seq, is_opt = self._parse_optional_field_quantifier()
+            id = (self._advance() if self.cur_token.kind in self._id_kinds
+                                  else None)
+            fields.append(Field(typename, id, seq=is_seq, opt=is_opt))
+            if self.cur_token.kind == TokenKind.RParen:
+                break
+            elif self.cur_token.kind == TokenKind.Comma:
+                self._advance()
+        self._match(TokenKind.RParen)
+        return fields
+
+    def _parse_optional_fields(self):
+        if self.cur_token.kind == TokenKind.LParen:
+            return self._parse_fields()
+        else:
+            return None
+
+    def _parse_optional_attributes(self):
+        if self._at_keyword('attributes'):
+            self._advance()
+            return self._parse_fields()
+        else:
+            return None
+
+    def _parse_optional_field_quantifier(self):
+        is_seq, is_opt = False, False
+        if self.cur_token.kind == TokenKind.Question:
+            is_opt = True
+            self._advance()
+        if self.cur_token.kind == TokenKind.Asterisk:
+            is_seq = True
+            self._advance()
+        return is_seq, is_opt
+
+    def _advance(self):
+        """ Return the value of the current token and read the next one into
+            self.cur_token.
+        """
+        cur_val = None if self.cur_token is None else self.cur_token.value
+        try:
+            self.cur_token = next(self._tokenizer)
+        except StopIteration:
+            self.cur_token = None
+        return cur_val
+
+    _id_kinds = (TokenKind.ConstructorId, TokenKind.TypeId)
+
+    def _match(self, kind):
+        """The 'match' primitive of RD parsers.
+
+        * Verifies that the current token is of the given kind (kind can
+          be a tuple, in which the kind must match one of its members).
+        * Returns the value of the current token
+        * Reads in the next token
+        """
+        if (isinstance(kind, tuple) and self.cur_token.kind in kind or
+            self.cur_token.kind == kind
+            ):
+            value = self.cur_token.value
+            self._advance()
+            return value
+        else:
+            raise ASDLSyntaxError(
+                'Unmatched {} (found {})'.format(kind, self.cur_token.kind),
+                self.cur_token.lineno)
+
+    def _at_keyword(self, keyword):
+        return (self.cur_token.kind == TokenKind.TypeId and
+                self.cur_token.value == keyword)

nac3ast/asdl_rs.py

@@ -0,0 +1,609 @@
+#! /usr/bin/env python
+"""Generate Rust code from an ASDL description."""
+
+import os
+import sys
+import textwrap
+
+import json
+
+from argparse import ArgumentParser
+from pathlib import Path
+
+import asdl
+
+TABSIZE = 4
+AUTOGEN_MESSAGE = "// File automatically generated by {}.\n\n"
+
+builtin_type_mapping = {
+    'identifier': 'Ident',
+    'string': 'String',
+    'int': 'usize',
+    'constant': 'Constant',
+    'bool': 'bool',
+    'conversion_flag': 'ConversionFlag',
+}
+assert builtin_type_mapping.keys() == asdl.builtin_types
+
+def get_rust_type(name):
+    """Return a string for the C name of the type.
+
+    This function special cases the default types provided by asdl.
+    """
+    if name in asdl.builtin_types:
+        return builtin_type_mapping[name]
+    else:
+        return "".join(part.capitalize() for part in name.split("_"))
+
+def is_simple(sum):
+    """Return True if a sum is a simple.
+
+    A sum is simple if its types have no fields, e.g.
+    unaryop = Invert | Not | UAdd | USub
+    """
+    for t in sum.types:
+        if t.fields:
+            return False
+    return True
+
+def asdl_of(name, obj):
+    if isinstance(obj, asdl.Product) or isinstance(obj, asdl.Constructor):
+        fields = ", ".join(map(str, obj.fields))
+        if fields:
+            fields = "({})".format(fields)
+        return "{}{}".format(name, fields)
+    else:
+        if is_simple(obj):
+            types = " | ".join(type.name for type in obj.types)
+        else:
+            sep = "\n{}| ".format(" " * (len(name) + 1))
+            types = sep.join(
+                asdl_of(type.name, type) for type in obj.types
+            )
+        return "{} = {}".format(name, types)
+
+class EmitVisitor(asdl.VisitorBase):
+    """Visit that emits lines"""
+
+    def __init__(self, file):
+        self.file = file
+        self.identifiers = set()
+        super(EmitVisitor, self).__init__()
+
+    def emit_identifier(self, name):
+        name = str(name)
+        if name in self.identifiers:
+            return
+        self.emit("_Py_IDENTIFIER(%s);" % name, 0)
+        self.identifiers.add(name)
+
+    def emit(self, line, depth):
+        if line:
+            line = (" " * TABSIZE * depth) + line
+        self.file.write(line + "\n")
+
+class TypeInfo:
+    def __init__(self, name):
+        self.name = name
+        self.has_userdata = None
+        self.children = set()
+        self.boxed = False
+
+    def __repr__(self):
+        return f"<TypeInfo: {self.name}>"
+
+    def determine_userdata(self, typeinfo, stack):
+        if self.name in stack:
+            return None
+        stack.add(self.name)
+        for child, child_seq in self.children:
+            if child in asdl.builtin_types:
+                continue
+            childinfo = typeinfo[child]
+            child_has_userdata = childinfo.determine_userdata(typeinfo, stack)
+            if self.has_userdata is None and child_has_userdata is True:
+                self.has_userdata = True
+
+        stack.remove(self.name)
+        return self.has_userdata
+
+class FindUserdataTypesVisitor(asdl.VisitorBase):
+    def __init__(self, typeinfo):
+        self.typeinfo = typeinfo
+        super().__init__()
+
+    def visitModule(self, mod):
+        for dfn in mod.dfns:
+            self.visit(dfn)
+        stack = set()
+        for info in self.typeinfo.values():
+            info.determine_userdata(self.typeinfo, stack)
+
+    def visitType(self, type):
+        self.typeinfo[type.name] = TypeInfo(type.name)
+        self.visit(type.value, type.name)
+
+    def visitSum(self, sum, name):
+        info = self.typeinfo[name]
+        if is_simple(sum):
+            info.has_userdata = False
+        else:
+            if len(sum.types) > 1:
+                info.boxed = True
+            if sum.attributes:
+                # attributes means Located, which has the `custom: U` field
+                info.has_userdata = True
+        for variant in sum.types:
+            self.add_children(name, variant.fields)
+
+    def visitProduct(self, product, name):
+        info = self.typeinfo[name]
+        if product.attributes:
+            # attributes means Located, which has the `custom: U` field
+            info.has_userdata = True
+        if len(product.fields) > 2:
+            info.boxed = True
+        self.add_children(name, product.fields)
+
+    def add_children(self, name, fields):
+        self.typeinfo[name].children.update((field.type, field.seq) for field in fields)
+
+def rust_field(field_name):
+    if field_name == 'type':
+        return 'type_'
+    else:
+        return field_name
+
+class TypeInfoEmitVisitor(EmitVisitor):
+    def __init__(self, file, typeinfo):
+        self.typeinfo = typeinfo
+        super().__init__(file)
+
+    def has_userdata(self, typ):
+        return self.typeinfo[typ].has_userdata
+
+    def get_generics(self, typ, *generics):
+        if self.has_userdata(typ):
+            return [f"<{g}>" for g in generics]
+        else:
+            return ["" for g in generics]
+
+class StructVisitor(TypeInfoEmitVisitor):
+    """Visitor to generate typedefs for AST."""
+
+    def visitModule(self, mod):
+        for dfn in mod.dfns:
+            self.visit(dfn)
+
+    def visitType(self, type, depth=0):
+        self.visit(type.value, type.name, depth)
+
+    def visitSum(self, sum, name, depth):
+        if is_simple(sum):
+            self.simple_sum(sum, name, depth)
+        else:
+            self.sum_with_constructors(sum, name, depth)
+
+    def emit_attrs(self, depth):
+        self.emit("#[derive(Clone, Debug, PartialEq)]", depth)
+
+    def simple_sum(self, sum, name, depth):
+        rustname = get_rust_type(name)
+        self.emit_attrs(depth)
+        self.emit(f"pub enum {rustname} {{", depth)
+        for variant in sum.types:
+            self.emit(f"{variant.name},", depth + 1)
+        self.emit("}", depth)
+        self.emit("", depth)
+
+    def sum_with_constructors(self, sum, name, depth):
+        typeinfo = self.typeinfo[name]
+        generics, generics_applied = self.get_generics(name, "U = ()", "U")
+        enumname = rustname = get_rust_type(name)
+        # all the attributes right now are for location, so if it has attrs we
+        # can just wrap it in Located<>
+        if sum.attributes:
+            enumname = rustname + "Kind"
+        self.emit_attrs(depth)
+        self.emit(f"pub enum {enumname}{generics} {{", depth)
+        for t in sum.types:
+            self.visit(t, typeinfo, depth + 1)
+        self.emit("}", depth)
+        if sum.attributes:
+            self.emit(f"pub type {rustname}<U = ()> = Located<{enumname}{generics_applied}, U>;", depth)
+        self.emit("", depth)
+
+    def visitConstructor(self, cons, parent, depth):
+        if cons.fields:
+            self.emit(f"{cons.name} {{", depth)
+            for f in cons.fields:
+                self.visit(f, parent, "", depth + 1)
+            self.emit("},", depth)
+        else:
+            self.emit(f"{cons.name},", depth)
+
+    def visitField(self, field, parent, vis, depth):
+        typ = get_rust_type(field.type)
+        fieldtype = self.typeinfo.get(field.type)
+        if fieldtype and fieldtype.has_userdata:
+            typ = f"{typ}<U>"
+        # don't box if we're doing Vec<T>, but do box if we're doing Vec<Option<Box<T>>>
+        if fieldtype and fieldtype.boxed and (not field.seq or field.opt):
+            typ = f"Box<{typ}>"
+        if field.opt:
+            typ = f"Option<{typ}>"
+        if field.seq:
+            typ = f"Vec<{typ}>"
+        name = rust_field(field.name)
+        self.emit(f"{vis}{name}: {typ},", depth)
+
+    def visitProduct(self, product, name, depth):
+        typeinfo = self.typeinfo[name]
+        generics, generics_applied = self.get_generics(name, "U = ()", "U")
+        dataname = rustname = get_rust_type(name)
+        if product.attributes:
+            dataname = rustname + "Data"
+        self.emit_attrs(depth)
+        self.emit(f"pub struct {dataname}{generics} {{", depth)
+        for f in product.fields:
+            self.visit(f, typeinfo, "pub ", depth + 1)
+        self.emit("}", depth)
+        if product.attributes:
+            # attributes should just be location info
+            self.emit(f"pub type {rustname}<U = ()> = Located<{dataname}{generics_applied}, U>;", depth);
+        self.emit("", depth)
+
+
+class FoldTraitDefVisitor(TypeInfoEmitVisitor):
+    def visitModule(self, mod, depth):
+        self.emit("pub trait Fold<U> {", depth)
+        self.emit("type TargetU;", depth + 1)
+        self.emit("type Error;", depth + 1)
+        self.emit("fn map_user(&mut self, user: U) -> Result<Self::TargetU, Self::Error>;", depth + 2)
+        for dfn in mod.dfns:
+            self.visit(dfn, depth + 2)
+        self.emit("}", depth)
+
+    def visitType(self, type, depth):
+        name = type.name
+        apply_u, apply_target_u = self.get_generics(name, "U", "Self::TargetU")
+        enumname = get_rust_type(name)
+        self.emit(f"fn fold_{name}(&mut self, node: {enumname}{apply_u}) -> Result<{enumname}{apply_target_u}, Self::Error> {{", depth)
+        self.emit(f"fold_{name}(self, node)", depth + 1)
+        self.emit("}", depth)
+
+
+class FoldImplVisitor(TypeInfoEmitVisitor):
+    def visitModule(self, mod, depth):
+        self.emit("fn fold_located<U, F: Fold<U> + ?Sized, T, MT>(folder: &mut F, node: Located<T, U>, f: impl FnOnce(&mut F, T) -> Result<MT, F::Error>) -> Result<Located<MT, F::TargetU>, F::Error> {", depth)
+        self.emit("Ok(Located { custom: folder.map_user(node.custom)?, location: node.location, node: f(folder, node.node)? })", depth + 1)
+        self.emit("}", depth)
+        for dfn in mod.dfns:
+            self.visit(dfn, depth)
+
+    def visitType(self, type, depth=0):
+        self.visit(type.value, type.name, depth)
+
+    def visitSum(self, sum, name, depth):
+        apply_t, apply_u, apply_target_u = self.get_generics(name, "T", "U", "F::TargetU")
+        enumname = get_rust_type(name)
+        is_located = bool(sum.attributes)
+
+        self.emit(f"impl<T, U> Foldable<T, U> for {enumname}{apply_t} {{", depth)
+        self.emit(f"type Mapped = {enumname}{apply_u};", depth + 1)
+        self.emit("fn fold<F: Fold<T, TargetU = U> + ?Sized>(self, folder: &mut F) -> Result<Self::Mapped, F::Error> {", depth + 1)
+        self.emit(f"folder.fold_{name}(self)", depth + 2)
+        self.emit("}", depth + 1)
+        self.emit("}", depth)
+
+        self.emit(f"pub fn fold_{name}<U, F: Fold<U> + ?Sized>(#[allow(unused)] folder: &mut F, node: {enumname}{apply_u}) -> Result<{enumname}{apply_target_u}, F::Error> {{", depth)
+        if is_located:
+            self.emit("fold_located(folder, node, |folder, node| {", depth)
+            enumname += "Kind"
+        self.emit("match node {", depth + 1)
+        for cons in sum.types:
+            fields_pattern = self.make_pattern(cons.fields)
+            self.emit(f"{enumname}::{cons.name} {{ {fields_pattern} }} => {{", depth + 2)
+            self.gen_construction(f"{enumname}::{cons.name}", cons.fields, depth + 3)
+            self.emit("}", depth + 2)
+        self.emit("}", depth + 1)
+        if is_located:
+            self.emit("})", depth)
+        self.emit("}", depth)
+
+
+    def visitProduct(self, product, name, depth):
+        apply_t, apply_u, apply_target_u = self.get_generics(name, "T", "U", "F::TargetU")
+        structname = get_rust_type(name)
+        is_located = bool(product.attributes)
+
+        self.emit(f"impl<T, U> Foldable<T, U> for {structname}{apply_t} {{", depth)
+        self.emit(f"type Mapped = {structname}{apply_u};", depth + 1)
+        self.emit("fn fold<F: Fold<T, TargetU = U> + ?Sized>(self, folder: &mut F) -> Result<Self::Mapped, F::Error> {", depth + 1)
+        self.emit(f"folder.fold_{name}(self)", depth + 2)
+        self.emit("}", depth + 1)
+        self.emit("}", depth)
+
+        self.emit(f"pub fn fold_{name}<U, F: Fold<U> + ?Sized>(#[allow(unused)] folder: &mut F, node: {structname}{apply_u}) -> Result<{structname}{apply_target_u}, F::Error> {{", depth)
+        if is_located:
+            self.emit("fold_located(folder, node, |folder, node| {", depth)
+            structname += "Data"
+        fields_pattern = self.make_pattern(product.fields)
+        self.emit(f"let {structname} {{ {fields_pattern} }} = node;", depth + 1)
+        self.gen_construction(structname, product.fields, depth + 1)
+        if is_located:
+            self.emit("})", depth)
+        self.emit("}", depth)
+
+    def make_pattern(self, fields):
+        return ",".join(rust_field(f.name) for f in fields)
+
+    def gen_construction(self, cons_path, fields, depth):
+        self.emit(f"Ok({cons_path} {{", depth)
+        for field in fields:
+            name = rust_field(field.name)
+            self.emit(f"{name}: Foldable::fold({name}, folder)?,", depth + 1)
+        self.emit("})", depth)
+
+
+class FoldModuleVisitor(TypeInfoEmitVisitor):
+    def visitModule(self, mod):
+        depth = 0
+        self.emit('#[cfg(feature = "fold")]', depth)
+        self.emit("pub mod fold {", depth)
+        self.emit("use super::*;", depth + 1)
+        self.emit("use crate::fold_helpers::Foldable;", depth + 1)
+        FoldTraitDefVisitor(self.file, self.typeinfo).visit(mod, depth + 1)
+        FoldImplVisitor(self.file, self.typeinfo).visit(mod, depth + 1)
+        self.emit("}", depth)
+
+
+class ClassDefVisitor(EmitVisitor):
+
+    def visitModule(self, mod):
+        for dfn in mod.dfns:
+            self.visit(dfn)
+
+    def visitType(self, type, depth=0):
+        self.visit(type.value, type.name, depth)
+
+    def visitSum(self, sum, name, depth):
+        for cons in sum.types:
+            self.visit(cons, sum.attributes, depth)
+
+    def visitConstructor(self, cons, attrs, depth):
+        self.gen_classdef(cons.name, cons.fields, attrs, depth)
+
+    def visitProduct(self, product, name, depth):
+        self.gen_classdef(name, product.fields, product.attributes, depth)
+
+    def gen_classdef(self, name, fields, attrs, depth):
+        structname = "Node" + name
+        self.emit(f'#[pyclass(module = "_ast", name = {json.dumps(name)}, base = "AstNode")]', depth)
+        self.emit(f"struct {structname};", depth)
+        self.emit("#[pyimpl(flags(HAS_DICT, BASETYPE))]", depth)
+        self.emit(f"impl {structname} {{", depth)
+        self.emit(f"#[extend_class]", depth + 1)
+        self.emit("fn extend_class_with_fields(ctx: &PyContext, class: &PyTypeRef) {", depth + 1)
+        fields = ",".join(f"ctx.new_str({json.dumps(f.name)})" for f in fields)
+        self.emit(f'class.set_str_attr("_fields", ctx.new_list(vec![{fields}]));', depth + 2)
+        attrs = ",".join(f"ctx.new_str({json.dumps(attr.name)})" for attr in attrs)
+        self.emit(f'class.set_str_attr("_attributes", ctx.new_list(vec![{attrs}]));', depth + 2)
+        self.emit("}", depth + 1)
+        self.emit("}", depth)
+
+class ExtendModuleVisitor(EmitVisitor):
+
+    def visitModule(self, mod):
+        depth = 0
+        self.emit("pub fn extend_module_nodes(vm: &VirtualMachine, module: &PyObjectRef) {", depth)
+        self.emit("extend_module!(vm, module, {", depth + 1)
+        for dfn in mod.dfns:
+            self.visit(dfn, depth + 2)
+        self.emit("})", depth + 1)
+        self.emit("}", depth)
+
+    def visitType(self, type, depth):
+        self.visit(type.value, type.name, depth)
+
+    def visitSum(self, sum, name, depth):
+        for cons in sum.types:
+            self.visit(cons, depth)
+
+    def visitConstructor(self, cons, depth):
+        self.gen_extension(cons.name, depth)
+
+    def visitProduct(self, product, name, depth):
+        self.gen_extension(name, depth)
+
+    def gen_extension(self, name, depth):
+        self.emit(f"{json.dumps(name)} => Node{name}::make_class(&vm.ctx),", depth)
+
+
+class TraitImplVisitor(EmitVisitor):
+
+    def visitModule(self, mod):
+        for dfn in mod.dfns:
+            self.visit(dfn)
+
+    def visitType(self, type, depth=0):
+        self.visit(type.value, type.name, depth)
+
+    def visitSum(self, sum, name, depth):
+        enumname = get_rust_type(name)
+        if sum.attributes:
+            enumname += "Kind"
+
+
+        self.emit(f"impl NamedNode for ast::{enumname} {{", depth)
+        self.emit(f"const NAME: &'static str = {json.dumps(name)};", depth + 1)
+        self.emit("}", depth)
+        self.emit(f"impl Node for ast::{enumname} {{", depth)
+        self.emit("fn ast_to_object(self, _vm: &VirtualMachine) -> PyObjectRef {", depth + 1)
+        self.emit("match self {", depth + 2)
+        for variant in sum.types:
+            self.constructor_to_object(variant, enumname, depth + 3)
+        self.emit("}", depth + 2)
+        self.emit("}", depth + 1)
+        self.emit("fn ast_from_object(_vm: &VirtualMachine, _object: PyObjectRef) -> PyResult<Self> {", depth + 1)
+        self.gen_sum_fromobj(sum, name, enumname, depth + 2)
+        self.emit("}", depth + 1)
+        self.emit("}", depth)
+
+    def constructor_to_object(self, cons, enumname, depth):
+        fields_pattern = self.make_pattern(cons.fields)
+        self.emit(f"ast::{enumname}::{cons.name} {{ {fields_pattern} }} => {{", depth)
+        self.make_node(cons.name, cons.fields, depth + 1)
+        self.emit("}", depth)
+
+    def visitProduct(self, product, name, depth):
+        structname = get_rust_type(name)
+        if product.attributes:
+            structname += "Data"
+
+        self.emit(f"impl NamedNode for ast::{structname} {{", depth)
+        self.emit(f"const NAME: &'static str = {json.dumps(name)};", depth + 1)
+        self.emit("}", depth)
+        self.emit(f"impl Node for ast::{structname} {{", depth)
+        self.emit("fn ast_to_object(self, _vm: &VirtualMachine) -> PyObjectRef {", depth + 1)
+        fields_pattern = self.make_pattern(product.fields)
+        self.emit(f"let ast::{structname} {{ {fields_pattern} }} = self;", depth + 2)
+        self.make_node(name, product.fields, depth + 2)
+        self.emit("}", depth + 1)
+        self.emit("fn ast_from_object(_vm: &VirtualMachine, _object: PyObjectRef) -> PyResult<Self> {", depth + 1)
+        self.gen_product_fromobj(product, name, structname, depth + 2)
+        self.emit("}", depth + 1)
+        self.emit("}", depth)
+
+    def make_node(self, variant, fields, depth):
+        lines = []
+        self.emit(f"let _node = AstNode.into_ref_with_type(_vm, Node{variant}::static_type().clone()).unwrap();", depth)
+        if fields:
+            self.emit("let _dict = _node.as_object().dict().unwrap();", depth)
+        for f in fields:
+            self.emit(f"_dict.set_item({json.dumps(f.name)}, {rust_field(f.name)}.ast_to_object(_vm), _vm).unwrap();", depth)
+        self.emit("_node.into_object()", depth)
+
+    def make_pattern(self, fields):
+        return ",".join(rust_field(f.name) for f in fields)
+
+    def gen_sum_fromobj(self, sum, sumname, enumname, depth):
+        if sum.attributes:
+            self.extract_location(sumname, depth)
+
+        self.emit("let _cls = _object.class();", depth)
+        self.emit("Ok(", depth)
+        for cons in sum.types:
+            self.emit(f"if _cls.is(Node{cons.name}::static_type()) {{", depth)
+            self.gen_construction(f"{enumname}::{cons.name}", cons, sumname, depth + 1)
+            self.emit("} else", depth)
+
+        self.emit("{", depth)
+        msg = f'format!("expected some sort of {sumname}, but got {{}}",_vm.to_repr(&_object)?)'
+        self.emit(f"return Err(_vm.new_type_error({msg}));", depth + 1)
+        self.emit("})", depth)
+
+    def gen_product_fromobj(self, product, prodname, structname, depth):
+        if product.attributes:
+            self.extract_location(prodname, depth)
+
+        self.emit("Ok(", depth)
+        self.gen_construction(structname, product, prodname, depth + 1)
+        self.emit(")", depth)
+
+    def gen_construction(self, cons_path, cons, name, depth):
+        self.emit(f"ast::{cons_path} {{", depth)
+        for field in cons.fields:
+            self.emit(f"{rust_field(field.name)}: {self.decode_field(field, name)},", depth + 1)
+        self.emit("}", depth)
+
+    def extract_location(self, typename, depth):
+        row = self.decode_field(asdl.Field('int', 'lineno'), typename)
+        column = self.decode_field(asdl.Field('int', 'col_offset'), typename)
+        self.emit(f"let _location = ast::Location::new({row}, {column});", depth)
+
+    def wrap_located_node(self, depth):
+        self.emit(f"let node = ast::Located::new(_location, node);", depth)
+
+    def decode_field(self, field, typename):
+        name = json.dumps(field.name)
+        if field.opt and not field.seq:
+            return f"get_node_field_opt(_vm, &_object, {name})?.map(|obj| Node::ast_from_object(_vm, obj)).transpose()?"
+        else:
+            return f"Node::ast_from_object(_vm, get_node_field(_vm, &_object, {name}, {json.dumps(typename)})?)?"
+
+class ChainOfVisitors:
+    def __init__(self, *visitors):
+        self.visitors = visitors
+
+    def visit(self, object):
+        for v in self.visitors:
+            v.visit(object)
+            v.emit("", 0)
+
+
+def write_ast_def(mod, typeinfo, f):
+    f.write('pub use crate::location::Location;\n')
+    f.write('pub use crate::constant::*;\n')
+    f.write('\n')
+    f.write('type Ident = String;\n')
+    f.write('\n')
+    StructVisitor(f, typeinfo).emit_attrs(0)
+    f.write('pub struct Located<T, U = ()> {\n')
+    f.write('    pub location: Location,\n')
+    f.write('    pub custom: U,\n')
+    f.write('    pub node: T,\n')
+    f.write('}\n')
+    f.write('\n')
+    f.write('impl<T> Located<T> {\n')
+    f.write('    pub fn new(location: Location, node: T) -> Self {\n')
+    f.write('        Self { location, custom: (), node }\n')
+    f.write('    }\n')
+    f.write('}\n')
+    f.write('\n')
+
+    c = ChainOfVisitors(StructVisitor(f, typeinfo),
+                        FoldModuleVisitor(f, typeinfo))
+    c.visit(mod)
+
+
+def write_ast_mod(mod, f):
+    f.write('use super::*;\n')
+    f.write('\n')
+
+    c = ChainOfVisitors(ClassDefVisitor(f),
+                        TraitImplVisitor(f),
+                        ExtendModuleVisitor(f))
+    c.visit(mod)
+
+def main(input_filename, ast_mod_filename, ast_def_filename, dump_module=False):
+    auto_gen_msg = AUTOGEN_MESSAGE.format("/".join(Path(__file__).parts[-2:]))
+    mod = asdl.parse(input_filename)
+    if dump_module:
+        print('Parsed Module:')
+        print(mod)
+    if not asdl.check(mod):
+        sys.exit(1)
+
+    typeinfo = {}
+    FindUserdataTypesVisitor(typeinfo).visit(mod)
+
+    with ast_def_filename.open("w") as def_file, \
+         ast_mod_filename.open("w") as mod_file:
+        def_file.write(auto_gen_msg)
+        write_ast_def(mod, typeinfo, def_file)
+
+        mod_file.write(auto_gen_msg)
+        write_ast_mod(mod, mod_file)
+
+    print(f"{ast_def_filename}, {ast_mod_filename} regenerated.")
+
+if __name__ == "__main__":
+    parser = ArgumentParser()
+    parser.add_argument("input_file", type=Path)
+    parser.add_argument("-M", "--mod-file", type=Path, required=True)
+    parser.add_argument("-D", "--def-file", type=Path, required=True)
+    parser.add_argument("-d", "--dump-module", action="store_true")
+
+    args = parser.parse_args()
+    main(args.input_file, args.mod_file, args.def_file, args.dump_module)

nac3ast/src/constant.rs

@@ -0,0 +1,183 @@
+#[derive(Clone, Debug, PartialEq)]
+pub enum Constant {
+    None,
+    Bool(bool),
+    Str(String),
+    Bytes(Vec<u8>),
+    Int(i128),
+    Tuple(Vec<Constant>),
+    Float(f64),
+    Complex { real: f64, imag: f64 },
+    Ellipsis,
+}
+
+impl From<String> for Constant {
+    fn from(s: String) -> Constant {
+        Self::Str(s)
+    }
+}
+impl From<Vec<u8>> for Constant {
+    fn from(b: Vec<u8>) -> Constant {
+        Self::Bytes(b)
+    }
+}
+impl From<bool> for Constant {
+    fn from(b: bool) -> Constant {
+        Self::Bool(b)
+    }
+}
+impl From<i32> for Constant {
+    fn from(i: i32) -> Constant {
+        Self::Int(i128::from(i))
+    }
+}
+impl From<i64> for Constant {
+    fn from(i: i64) -> Constant {
+        Self::Int(i128::from(i))
+    }
+}
+
+/// Transforms a value prior to formatting it.
+#[derive(Copy, Clone, Debug, PartialEq)]
+#[repr(u8)]
+pub enum ConversionFlag {
+    /// Converts by calling `str(<value>)`.
+    Str = b's',
+    /// Converts by calling `ascii(<value>)`.
+    Ascii = b'a',
+    /// Converts by calling `repr(<value>)`.
+    Repr = b'r',
+}
+
+impl ConversionFlag {
+    #[must_use]
+    pub fn try_from_byte(b: u8) -> Option<Self> {
+        match b {
+            b's' => Some(Self::Str),
+            b'a' => Some(Self::Ascii),
+            b'r' => Some(Self::Repr),
+            _ => None,
+        }
+    }
+}
+
+#[cfg(feature = "constant-optimization")]
+#[derive(Default)]
+pub struct ConstantOptimizer {
+    _priv: (),
+}
+
+#[cfg(feature = "constant-optimization")]
+impl ConstantOptimizer {
+    #[inline]
+    #[must_use]
+    pub fn new() -> Self {
+        Self { _priv: () }
+    }
+}
+
+#[cfg(feature = "constant-optimization")]
+impl<U> crate::fold::Fold<U> for ConstantOptimizer {
+    type TargetU = U;
+    type Error = std::convert::Infallible;
+    #[inline]
+    fn map_user(&mut self, user: U) -> Result<Self::TargetU, Self::Error> {
+        Ok(user)
+    }
+    fn fold_expr(&mut self, node: crate::Expr<U>) -> Result<crate::Expr<U>, Self::Error> {
+        match node.node {
+            crate::ExprKind::Tuple { elts, ctx } => {
+                let elts =
+                    elts.into_iter().map(|x| self.fold_expr(x)).collect::<Result<Vec<_>, _>>()?;
+                let expr =
+                    if elts.iter().all(|e| matches!(e.node, crate::ExprKind::Constant { .. })) {
+                        let tuple = elts
+                            .into_iter()
+                            .map(|e| match e.node {
+                                crate::ExprKind::Constant { value, .. } => value,
+                                _ => unreachable!(),
+                            })
+                            .collect();
+                        crate::ExprKind::Constant { value: Constant::Tuple(tuple), kind: None }
+                    } else {
+                        crate::ExprKind::Tuple { elts, ctx }
+                    };
+                Ok(crate::Expr { node: expr, custom: node.custom, location: node.location })
+            }
+            _ => crate::fold::fold_expr(self, node),
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    #[cfg(feature = "constant-optimization")]
+    #[test]
+    fn test_constant_opt() {
+        use super::*;
+        use crate::fold::Fold;
+        use crate::*;
+
+        let location = Location::new(0, 0, FileName::default());
+        let custom = ();
+        let ast = Located {
+            location,
+            custom,
+            node: ExprKind::Tuple {
+                ctx: ExprContext::Load,
+                elts: vec![
+                    Located {
+                        location,
+                        custom,
+                        node: ExprKind::Constant { value: 1.into(), kind: None },
+                    },
+                    Located {
+                        location,
+                        custom,
+                        node: ExprKind::Constant { value: 2.into(), kind: None },
+                    },
+                    Located {
+                        location,
+                        custom,
+                        node: ExprKind::Tuple {
+                            ctx: ExprContext::Load,
+                            elts: vec![
+                                Located {
+                                    location,
+                                    custom,
+                                    node: ExprKind::Constant { value: 3.into(), kind: None },
+                                },
+                                Located {
+                                    location,
+                                    custom,
+                                    node: ExprKind::Constant { value: 4.into(), kind: None },
+                                },
+                                Located {
+                                    location,
+                                    custom,
+                                    node: ExprKind::Constant { value: 5.into(), kind: None },
+                                },
+                            ],
+                        },
+                    },
+                ],
+            },
+        };
+        let new_ast = ConstantOptimizer::new().fold_expr(ast).unwrap_or_else(|e| match e {});
+        assert_eq!(
+            new_ast,
+            Located {
+                location,
+                custom,
+                node: ExprKind::Constant {
+                    value: Constant::Tuple(vec![
+                        1.into(),
+                        2.into(),
+                        Constant::Tuple(vec![3.into(), 4.into(), 5.into(),])
+                    ]),
+                    kind: None
+                },
+            }
+        );
+    }
+}

nac3ast/src/fold_helpers.rs

@@ -0,0 +1,67 @@
+use crate::constant;
+use crate::fold::Fold;
+use crate::StrRef;
+
+pub(crate) trait Foldable<T, U> {
+    type Mapped;
+    fn fold<F: Fold<T, TargetU = U> + ?Sized>(
+        self,
+        folder: &mut F,
+    ) -> Result<Self::Mapped, F::Error>;
+}
+
+impl<T, U, X> Foldable<T, U> for Vec<X>
+where
+    X: Foldable<T, U>,
+{
+    type Mapped = Vec<X::Mapped>;
+    fn fold<F: Fold<T, TargetU = U> + ?Sized>(
+        self,
+        folder: &mut F,
+    ) -> Result<Self::Mapped, F::Error> {
+        self.into_iter().map(|x| x.fold(folder)).collect()
+    }
+}
+
+impl<T, U, X> Foldable<T, U> for Option<X>
+where
+    X: Foldable<T, U>,
+{
+    type Mapped = Option<X::Mapped>;
+    fn fold<F: Fold<T, TargetU = U> + ?Sized>(
+        self,
+        folder: &mut F,
+    ) -> Result<Self::Mapped, F::Error> {
+        self.map(|x| x.fold(folder)).transpose()
+    }
+}
+
+impl<T, U, X> Foldable<T, U> for Box<X>
+where
+    X: Foldable<T, U>,
+{
+    type Mapped = Box<X::Mapped>;
+    fn fold<F: Fold<T, TargetU = U> + ?Sized>(
+        self,
+        folder: &mut F,
+    ) -> Result<Self::Mapped, F::Error> {
+        (*self).fold(folder).map(Box::new)
+    }
+}
+
+macro_rules! simple_fold {
+    ($($t:ty),+$(,)?) => {
+        $(impl<T, U> $crate::fold_helpers::Foldable<T, U> for $t {
+            type Mapped = Self;
+            #[inline]
+            fn fold<F: Fold<T, TargetU = U> + ?Sized>(
+                self,
+                _folder: &mut F,
+            ) -> Result<Self::Mapped, F::Error> {
+                Ok(self)
+            }
+        })+
+    };
+}
+
+simple_fold!(usize, String, bool, StrRef, constant::Constant, constant::ConversionFlag);

nac3ast/src/impls.rs

@@ -0,0 +1,51 @@
+use crate::{Constant, ExprKind};
+
+impl<U> ExprKind<U> {
+    /// Returns a short name for the node suitable for use in error messages.
+    #[must_use]
+    pub fn name(&self) -> &'static str {
+        match self {
+            ExprKind::BoolOp { .. } | ExprKind::BinOp { .. } | ExprKind::UnaryOp { .. } => {
+                "operator"
+            }
+            ExprKind::Subscript { .. } => "subscript",
+            ExprKind::Await { .. } => "await expression",
+            ExprKind::Yield { .. } | ExprKind::YieldFrom { .. } => "yield expression",
+            ExprKind::Compare { .. } => "comparison",
+            ExprKind::Attribute { .. } => "attribute",
+            ExprKind::Call { .. } => "function call",
+            ExprKind::Constant { value, .. } => match value {
+                Constant::Str(_)
+                | Constant::Int(_)
+                | Constant::Float(_)
+                | Constant::Complex { .. }
+                | Constant::Bytes(_) => "literal",
+                Constant::Tuple(_) => "tuple",
+                Constant::Bool(_) | Constant::None => "keyword",
+                Constant::Ellipsis => "ellipsis",
+            },
+            ExprKind::List { .. } => "list",
+            ExprKind::Tuple { .. } => "tuple",
+            ExprKind::Dict { .. } => "dict display",
+            ExprKind::Set { .. } => "set display",
+            ExprKind::ListComp { .. } => "list comprehension",
+            ExprKind::DictComp { .. } => "dict comprehension",
+            ExprKind::SetComp { .. } => "set comprehension",
+            ExprKind::GeneratorExp { .. } => "generator expression",
+            ExprKind::Starred { .. } => "starred",
+            ExprKind::Slice { .. } => "slice",
+            ExprKind::JoinedStr { values } => {
+                if values.iter().any(|e| matches!(e.node, ExprKind::JoinedStr { .. })) {
+                    "f-string expression"
+                } else {
+                    "literal"
+                }
+            }
+            ExprKind::FormattedValue { .. } => "f-string expression",
+            ExprKind::Name { .. } => "name",
+            ExprKind::Lambda { .. } => "lambda",
+            ExprKind::IfExp { .. } => "conditional expression",
+            ExprKind::NamedExpr { .. } => "named expression",
+        }
+    }
+}

nac3ast/src/lib.rs

@@ -0,0 +1,30 @@
+#![deny(
+    future_incompatible,
+    let_underscore,
+    nonstandard_style,
+    rust_2024_compatibility,
+    clippy::all
+)]
+#![warn(clippy::pedantic)]
+#![allow(
+    clippy::missing_errors_doc,
+    clippy::missing_panics_doc,
+    clippy::module_name_repetitions,
+    clippy::too_many_lines,
+    clippy::wildcard_imports
+)]
+
+#[macro_use]
+extern crate lazy_static;
+
+mod ast_gen;
+mod constant;
+#[cfg(feature = "fold")]
+mod fold_helpers;
+mod impls;
+mod location;
+
+pub use ast_gen::*;
+pub use location::{FileName, Location};
+
+pub type Suite<U = ()> = Vec<Stmt<U>>;

nac3ast/src/location.rs

@@ -0,0 +1,116 @@
+//! Datatypes to support source location information.
+use crate::ast_gen::StrRef;
+use std::cmp::Ordering;
+use std::fmt;
+
+#[derive(Clone, Copy, Debug, Eq, PartialEq)]
+pub struct FileName(pub StrRef);
+impl Default for FileName {
+    fn default() -> Self {
+        FileName("unknown".into())
+    }
+}
+
+impl From<String> for FileName {
+    fn from(s: String) -> Self {
+        FileName(s.into())
+    }
+}
+
+/// A location somewhere in the sourcecode.
+#[derive(Clone, Copy, Debug, Default, Eq, PartialEq)]
+pub struct Location {
+    pub row: usize,
+    pub column: usize,
+    pub file: FileName,
+}
+
+impl fmt::Display for Location {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        write!(f, "{}:{}:{}", self.file.0, self.row, self.column)
+    }
+}
+
+impl Ord for Location {
+    fn cmp(&self, other: &Self) -> Ordering {
+        let file_cmp = self.file.0.to_string().cmp(&other.file.0.to_string());
+        if file_cmp != Ordering::Equal {
+            return file_cmp;
+        }
+
+        let row_cmp = self.row.cmp(&other.row);
+        if row_cmp != Ordering::Equal {
+            return row_cmp;
+        }
+
+        self.column.cmp(&other.column)
+    }
+}
+
+impl PartialOrd for Location {
+    fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
+        Some(self.cmp(other))
+    }
+}
+
+impl Location {
+    pub fn visualize<'a>(
+        &self,
+        line: &'a str,
+        desc: impl fmt::Display + 'a,
+    ) -> impl fmt::Display + 'a {
+        struct Visualize<'a, D: fmt::Display> {
+            loc: Location,
+            line: &'a str,
+            desc: D,
+        }
+        impl<D: fmt::Display> fmt::Display for Visualize<'_, D> {
+            fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+                write!(
+                    f,
+                    "{}\n{}\n{arrow:>pad$}",
+                    self.desc,
+                    self.line,
+                    pad = self.loc.column,
+                    arrow = "↑",
+                )
+            }
+        }
+        Visualize { loc: *self, line, desc }
+    }
+}
+
+impl Location {
+    #[must_use]
+    pub fn new(row: usize, column: usize, file: FileName) -> Self {
+        Location { row, column, file }
+    }
+
+    #[must_use]
+    pub fn row(&self) -> usize {
+        self.row
+    }
+
+    #[must_use]
+    pub fn column(&self) -> usize {
+        self.column
+    }
+
+    pub fn reset(&mut self) {
+        self.row = 1;
+        self.column = 1;
+    }
+
+    pub fn go_right(&mut self) {
+        self.column += 1;
+    }
+
+    pub fn go_left(&mut self) {
+        self.column -= 1;
+    }
+
+    pub fn newline(&mut self) {
+        self.row += 1;
+        self.column = 1;
+    }
+}

nac3core/Cargo.toml

@@ -2,15 +2,30 @@
 name = "nac3core"
 version = "0.1.0"
 authors = ["M-Labs"]
-edition = "2018"
+edition = "2021"
+
+[features]
+no-escape-analysis = []
 
 [dependencies]
-num-bigint = "0.3"
-num-traits = "0.2"
-inkwell = { git = "https://github.com/TheDan64/inkwell", branch = "master", features = ["llvm10-0"] }
-rustpython-parser = { git = "https://github.com/RustPython/RustPython", branch = "master" }
-indoc = "1.0"
+itertools = "0.13"
+crossbeam = "0.8"
+indexmap = "2.6"
+parking_lot = "0.12"
+rayon = "1.10"
+nac3parser = { path = "../nac3parser" }
+strum = "0.26"
+strum_macros = "0.26"
+
+[dependencies.inkwell]
+version = "0.5"
+default-features = false
+features = ["llvm14-0-prefer-dynamic", "target-x86", "target-arm", "target-riscv", "no-libffi-linking"]
 
 [dev-dependencies]
 test-case = "1.2.0"
-indoc = "1.0"
+indoc = "2.0"
+insta = "=1.11.0"
+
+[build-dependencies]
+regex = "1.10"

nac3core/build.rs

@@ -0,0 +1,110 @@
+use std::{
+    env,
+    fs::File,
+    io::Write,
+    path::Path,
+    process::{Command, Stdio},
+};
+
+use regex::Regex;
+
+fn main() {
+    let out_dir = env::var("OUT_DIR").unwrap();
+    let out_dir = Path::new(&out_dir);
+    let irrt_dir = Path::new("irrt");
+
+    let irrt_cpp_path = irrt_dir.join("irrt.cpp");
+
+    /*
+     * HACK: Sadly, clang doesn't let us emit generic LLVM bitcode.
+     * Compiling for WASM32 and filtering the output with regex is the closest we can get.
+     */
+    let mut flags: Vec<&str> = vec![
+        "--target=wasm32",
+        "-x",
+        "c++",
+        "-std=c++20",
+        "-fno-discard-value-names",
+        "-fno-exceptions",
+        "-fno-rtti",
+        "-emit-llvm",
+        "-S",
+        "-Wall",
+        "-Wextra",
+        "-o",
+        "-",
+        "-I",
+        irrt_dir.to_str().unwrap(),
+        irrt_cpp_path.to_str().unwrap(),
+    ];
+
+    match env::var("PROFILE").as_deref() {
+        Ok("debug") => {
+            flags.push("-O0");
+            flags.push("-DIRRT_DEBUG_ASSERT");
+        }
+        Ok("release") => {
+            flags.push("-O3");
+        }
+        flavor => panic!("Unknown or missing build flavor {flavor:?}"),
+    }
+
+    // Tell Cargo to rerun if any file under `irrt_dir` (recursive) changes
+    println!("cargo:rerun-if-changed={}", irrt_dir.to_str().unwrap());
+
+    // Compile IRRT and capture the LLVM IR output
+    let output = Command::new("clang-irrt")
+        .args(flags)
+        .output()
+        .map(|o| {
+            assert!(o.status.success(), "{}", std::str::from_utf8(&o.stderr).unwrap());
+            o
+        })
+        .unwrap();
+
+    // https://github.com/rust-lang/regex/issues/244
+    let output = std::str::from_utf8(&output.stdout).unwrap().replace("\r\n", "\n");
+    let mut filtered_output = String::with_capacity(output.len());
+
+    // Filter out irrelevant IR
+    //
+    // Regex:
+    // - `(?ms:^define.*?\}$)` captures LLVM `define` blocks
+    // - `(?m:^declare.*?$)` captures LLVM `declare` lines
+    // - `(?m:^%.+?=\s*type\s*\{.+?\}$)` captures LLVM `type` declarations
+    // - `(?m:^@.+?=.+$)` captures global constants
+    let regex_filter = Regex::new(
+        r"(?ms:^define.*?\}$)|(?m:^declare.*?$)|(?m:^%.+?=\s*type\s*\{.+?\}$)|(?m:^@.+?=.+$)",
+    )
+    .unwrap();
+    for f in regex_filter.captures_iter(&output) {
+        assert_eq!(f.len(), 1);
+        filtered_output.push_str(&f[0]);
+        filtered_output.push('\n');
+    }
+
+    let filtered_output = Regex::new("(#\\d+)|(, *![0-9A-Za-z.]+)|(![0-9A-Za-z.]+)|(!\".*?\")")
+        .unwrap()
+        .replace_all(&filtered_output, "");
+
+    // For debugging
+    // Doing `DEBUG_DUMP_IRRT=1 cargo build -p nac3core` dumps the LLVM IR generated
+    const DEBUG_DUMP_IRRT: &str = "DEBUG_DUMP_IRRT";
+    println!("cargo:rerun-if-env-changed={DEBUG_DUMP_IRRT}");
+    if env::var(DEBUG_DUMP_IRRT).is_ok() {
+        let mut file = File::create(out_dir.join("irrt.ll")).unwrap();
+        file.write_all(output.as_bytes()).unwrap();
+
+        let mut file = File::create(out_dir.join("irrt-filtered.ll")).unwrap();
+        file.write_all(filtered_output.as_bytes()).unwrap();
+    }
+
+    let mut llvm_as = Command::new("llvm-as-irrt")
+        .stdin(Stdio::piped())
+        .arg("-o")
+        .arg(out_dir.join("irrt.bc"))
+        .spawn()
+        .unwrap();
+    llvm_as.stdin.as_mut().unwrap().write_all(filtered_output.as_bytes()).unwrap();
+    assert!(llvm_as.wait().unwrap().success());
+}

nac3core/irrt/irrt.cpp

@@ -0,0 +1,6 @@
+#include "irrt/exception.hpp"
+#include "irrt/int_types.hpp"
+#include "irrt/list.hpp"
+#include "irrt/math.hpp"
+#include "irrt/ndarray.hpp"
+#include "irrt/slice.hpp"

nac3core/irrt/irrt/cslice.hpp

@@ -0,0 +1,9 @@
+#pragma once
+
+#include "irrt/int_types.hpp"
+
+template<typename SizeT>
+struct CSlice {
+    uint8_t* base;
+    SizeT len;
+};

nac3core/irrt/irrt/debug.hpp

@@ -0,0 +1,25 @@
+#pragma once
+
+// Set in nac3core/build.rs
+#ifdef IRRT_DEBUG_ASSERT
+#define IRRT_DEBUG_ASSERT_BOOL true
+#else
+#define IRRT_DEBUG_ASSERT_BOOL false
+#endif
+
+#define raise_debug_assert(SizeT, msg, param1, param2, param3) \
+    raise_exception(SizeT, EXN_ASSERTION_ERROR, "IRRT debug assert failed: " msg, param1, param2, param3)
+
+#define debug_assert_eq(SizeT, lhs, rhs)                                           \
+    if constexpr (IRRT_DEBUG_ASSERT_BOOL) {                                        \
+        if ((lhs) != (rhs)) {                                                      \
+            raise_debug_assert(SizeT, "LHS = {0}. RHS = {1}", lhs, rhs, NO_PARAM); \
+        }                                                                          \
+    }
+
+#define debug_assert(SizeT, expr)                                                  \
+    if constexpr (IRRT_DEBUG_ASSERT_BOOL) {                                        \
+        if (!(expr)) {                                                             \
+            raise_debug_assert(SizeT, "Got false.", NO_PARAM, NO_PARAM, NO_PARAM); \
+        }                                                                          \
+    }

nac3core/irrt/irrt/exception.hpp

@@ -0,0 +1,82 @@
+#pragma once
+
+#include "irrt/cslice.hpp"
+#include "irrt/int_types.hpp"
+
+/**
+ * @brief The int type of ARTIQ exception IDs.
+ */
+typedef int32_t ExceptionId;
+
+/*
+ * Set of exceptions C++ IRRT can use.
+ * Must be synchronized with `setup_irrt_exceptions` in `nac3core/src/codegen/irrt/mod.rs`.
+ */
+extern "C" {
+ExceptionId EXN_INDEX_ERROR;
+ExceptionId EXN_VALUE_ERROR;
+ExceptionId EXN_ASSERTION_ERROR;
+ExceptionId EXN_TYPE_ERROR;
+}
+
+/**
+ * @brief Extern function to `__nac3_raise`
+ *
+ * The parameter `err` could be `Exception<int32_t>` or `Exception<int64_t>`. The caller
+ * must make sure to pass `Exception`s with the correct `SizeT` depending on the `size_t` of the runtime.
+ */
+extern "C" void __nac3_raise(void* err);
+
+namespace {
+/**
+ * @brief NAC3's Exception struct
+ */
+template<typename SizeT>
+struct Exception {
+    ExceptionId id;
+    CSlice<SizeT> filename;
+    int32_t line;
+    int32_t column;
+    CSlice<SizeT> function;
+    CSlice<SizeT> msg;
+    int64_t params[3];
+};
+
+constexpr int64_t NO_PARAM = 0;
+
+template<typename SizeT>
+void _raise_exception_helper(ExceptionId id,
+                             const char* filename,
+                             int32_t line,
+                             const char* function,
+                             const char* msg,
+                             int64_t param0,
+                             int64_t param1,
+                             int64_t param2) {
+    Exception<SizeT> e = {
+        .id = id,
+        .filename = {.base = reinterpret_cast<const uint8_t*>(filename), .len = __builtin_strlen(filename)},
+        .line = line,
+        .column = 0,
+        .function = {.base = reinterpret_cast<const uint8_t*>(function), .len = __builtin_strlen(function)},
+        .msg = {.base = reinterpret_cast<const uint8_t*>(msg), .len = __builtin_strlen(msg)},
+    };
+    e.params[0] = param0;
+    e.params[1] = param1;
+    e.params[2] = param2;
+    __nac3_raise(reinterpret_cast<void*>(&e));
+    __builtin_unreachable();
+}
+
+/**
+ * @brief Raise an exception with location details (location in the IRRT source files).
+ * @param SizeT The runtime `size_t` type.
+ * @param id The ID of the exception to raise.
+ * @param msg A global constant C-string of the error message.
+ *
+ * `param0` to `param2` are optional format arguments of `msg`. They should be set to
+ * `NO_PARAM` to indicate they are unused.
+ */
+#define raise_exception(SizeT, id, msg, param0, param1, param2) \
+    _raise_exception_helper<SizeT>(id, __FILE__, __LINE__, __FUNCTION__, msg, param0, param1, param2)
+}  // namespace

nac3core/irrt/irrt/int_types.hpp

@@ -0,0 +1,22 @@
+#pragma once
+
+#if __STDC_VERSION__ >= 202000
+using int8_t = _BitInt(8);
+using uint8_t = unsigned _BitInt(8);
+using int32_t = _BitInt(32);
+using uint32_t = unsigned _BitInt(32);
+using int64_t = _BitInt(64);
+using uint64_t = unsigned _BitInt(64);
+#else
+using int8_t = _ExtInt(8);
+using uint8_t = unsigned _ExtInt(8);
+using int32_t = _ExtInt(32);
+using uint32_t = unsigned _ExtInt(32);
+using int64_t = _ExtInt(64);
+using uint64_t = unsigned _ExtInt(64);
+#endif
+
+// NDArray indices are always `uint32_t`.
+using NDIndex = uint32_t;
+// The type of an index or a value describing the length of a range/slice is always `int32_t`.
+using SliceIndex = int32_t;

nac3core/irrt/irrt/list.hpp

@@ -0,0 +1,75 @@
+#pragma once
+
+#include "irrt/int_types.hpp"
+#include "irrt/math_util.hpp"
+
+extern "C" {
+// Handle list assignment and dropping part of the list when
+// both dest_step and src_step are +1.
+// - All the index must *not* be out-of-bound or negative,
+// - The end index is *inclusive*,
+// - The length of src and dest slice size should already
+// be checked: if dest.step == 1 then len(src) <= len(dest) else len(src) == len(dest)
+SliceIndex __nac3_list_slice_assign_var_size(SliceIndex dest_start,
+                                             SliceIndex dest_end,
+                                             SliceIndex dest_step,
+                                             uint8_t* dest_arr,
+                                             SliceIndex dest_arr_len,
+                                             SliceIndex src_start,
+                                             SliceIndex src_end,
+                                             SliceIndex src_step,
+                                             uint8_t* src_arr,
+                                             SliceIndex src_arr_len,
+                                             const SliceIndex size) {
+    /* if dest_arr_len == 0, do nothing since we do not support extending list */
+    if (dest_arr_len == 0)
+        return dest_arr_len;
+    /* if both step is 1, memmove directly, handle the dropping of the list, and shrink size */
+    if (src_step == dest_step && dest_step == 1) {
+        const SliceIndex src_len = (src_end >= src_start) ? (src_end - src_start + 1) : 0;
+        const SliceIndex dest_len = (dest_end >= dest_start) ? (dest_end - dest_start + 1) : 0;
+        if (src_len > 0) {
+            __builtin_memmove(dest_arr + dest_start * size, src_arr + src_start * size, src_len * size);
+        }
+        if (dest_len > 0) {
+            /* dropping */
+            __builtin_memmove(dest_arr + (dest_start + src_len) * size, dest_arr + (dest_end + 1) * size,
+                              (dest_arr_len - dest_end - 1) * size);
+        }
+        /* shrink size */
+        return dest_arr_len - (dest_len - src_len);
+    }
+    /* if two range overlaps, need alloca */
+    uint8_t need_alloca = (dest_arr == src_arr)
+                          && !(max(dest_start, dest_end) < min(src_start, src_end)
+                               || max(src_start, src_end) < min(dest_start, dest_end));
+    if (need_alloca) {
+        uint8_t* tmp = reinterpret_cast<uint8_t*>(__builtin_alloca(src_arr_len * size));
+        __builtin_memcpy(tmp, src_arr, src_arr_len * size);
+        src_arr = tmp;
+    }
+    SliceIndex src_ind = src_start;
+    SliceIndex dest_ind = dest_start;
+    for (; (src_step > 0) ? (src_ind <= src_end) : (src_ind >= src_end); src_ind += src_step, dest_ind += dest_step) {
+        /* for constant optimization */
+        if (size == 1) {
+            __builtin_memcpy(dest_arr + dest_ind, src_arr + src_ind, 1);
+        } else if (size == 4) {
+            __builtin_memcpy(dest_arr + dest_ind * 4, src_arr + src_ind * 4, 4);
+        } else if (size == 8) {
+            __builtin_memcpy(dest_arr + dest_ind * 8, src_arr + src_ind * 8, 8);
+        } else {
+            /* memcpy for var size, cannot overlap after previous alloca */
+            __builtin_memcpy(dest_arr + dest_ind * size, src_arr + src_ind * size, size);
+        }
+    }
+    /* only dest_step == 1 can we shrink the dest list. */
+    /* size should be ensured prior to calling this function */
+    if (dest_step == 1 && dest_end >= dest_start) {
+        __builtin_memmove(dest_arr + dest_ind * size, dest_arr + (dest_end + 1) * size,
+                          (dest_arr_len - dest_end - 1) * size);
+        return dest_arr_len - (dest_end - dest_ind) - 1;
+    }
+    return dest_arr_len;
+}
+}  // extern "C"

nac3core/irrt/irrt/math.hpp

@@ -0,0 +1,93 @@
+#pragma once
+
+namespace {
+// adapted from GNU Scientific Library: https://git.savannah.gnu.org/cgit/gsl.git/tree/sys/pow_int.c
+// need to make sure `exp >= 0` before calling this function
+template<typename T>
+T __nac3_int_exp_impl(T base, T exp) {
+    T res = 1;
+    /* repeated squaring method */
+    do {
+        if (exp & 1) {
+            res *= base; /* for n odd */
+        }
+        exp >>= 1;
+        base *= base;
+    } while (exp);
+    return res;
+}
+}  // namespace
+
+#define DEF_nac3_int_exp_(T)                   \
+    T __nac3_int_exp_##T(T base, T exp) {      \
+        return __nac3_int_exp_impl(base, exp); \
+    }
+
+extern "C" {
+
+// Putting semicolons here to make clang-format not reformat this into
+// a stair shape.
+DEF_nac3_int_exp_(int32_t);
+DEF_nac3_int_exp_(int64_t);
+DEF_nac3_int_exp_(uint32_t);
+DEF_nac3_int_exp_(uint64_t);
+
+int32_t __nac3_isinf(double x) {
+    return __builtin_isinf(x);
+}
+
+int32_t __nac3_isnan(double x) {
+    return __builtin_isnan(x);
+}
+
+double tgamma(double arg);
+
+double __nac3_gamma(double z) {
+    // Handling for denormals
+    //     | x                 | Python gamma(x) | C tgamma(x) |
+    // --- | ----------------- | --------------- | ----------- |
+    // (1) | nan               | nan             | nan         |
+    // (2) | -inf              | -inf            | inf         |
+    // (3) | inf               | inf             | inf         |
+    // (4) | 0.0               | inf             | inf         |
+    // (5) | {-1.0, -2.0, ...} | inf             | nan         |
+
+    // (1)-(3)
+    if (__builtin_isinf(z) || __builtin_isnan(z)) {
+        return z;
+    }
+
+    double v = tgamma(z);
+
+    // (4)-(5)
+    return __builtin_isinf(v) || __builtin_isnan(v) ? __builtin_inf() : v;
+}
+
+double lgamma(double arg);
+
+double __nac3_gammaln(double x) {
+    // libm's handling of value overflows differs from scipy:
+    // - scipy: gammaln(-inf) -> -inf
+    // - libm : lgamma(-inf) -> inf
+
+    if (__builtin_isinf(x)) {
+        return x;
+    }
+
+    return lgamma(x);
+}
+
+double j0(double x);
+
+double __nac3_j0(double x) {
+    // libm's handling of value overflows differs from scipy:
+    // - scipy: j0(inf) -> nan
+    // - libm : j0(inf) -> 0.0
+
+    if (__builtin_isinf(x)) {
+        return __builtin_nan("");
+    }
+
+    return j0(x);
+}
+}

nac3core/irrt/irrt/math_util.hpp

@@ -0,0 +1,13 @@
+#pragma once
+
+namespace {
+template<typename T>
+const T& max(const T& a, const T& b) {
+    return a > b ? a : b;
+}
+
+template<typename T>
+const T& min(const T& a, const T& b) {
+    return a > b ? b : a;
+}
+}  // namespace

nac3core/irrt/irrt/ndarray.hpp

@@ -0,0 +1,144 @@
+#pragma once
+
+#include "irrt/int_types.hpp"
+
+namespace {
+template<typename SizeT>
+SizeT __nac3_ndarray_calc_size_impl(const SizeT* list_data, SizeT list_len, SizeT begin_idx, SizeT end_idx) {
+    __builtin_assume(end_idx <= list_len);
+
+    SizeT num_elems = 1;
+    for (SizeT i = begin_idx; i < end_idx; ++i) {
+        SizeT val = list_data[i];
+        __builtin_assume(val > 0);
+        num_elems *= val;
+    }
+    return num_elems;
+}
+
+template<typename SizeT>
+void __nac3_ndarray_calc_nd_indices_impl(SizeT index, const SizeT* dims, SizeT num_dims, NDIndex* idxs) {
+    SizeT stride = 1;
+    for (SizeT dim = 0; dim < num_dims; dim++) {
+        SizeT i = num_dims - dim - 1;
+        __builtin_assume(dims[i] > 0);
+        idxs[i] = (index / stride) % dims[i];
+        stride *= dims[i];
+    }
+}
+
+template<typename SizeT>
+SizeT __nac3_ndarray_flatten_index_impl(const SizeT* dims, SizeT num_dims, const NDIndex* indices, SizeT num_indices) {
+    SizeT idx = 0;
+    SizeT stride = 1;
+    for (SizeT i = 0; i < num_dims; ++i) {
+        SizeT ri = num_dims - i - 1;
+        if (ri < num_indices) {
+            idx += stride * indices[ri];
+        }
+
+        __builtin_assume(dims[i] > 0);
+        stride *= dims[ri];
+    }
+    return idx;
+}
+
+template<typename SizeT>
+void __nac3_ndarray_calc_broadcast_impl(const SizeT* lhs_dims,
+                                        SizeT lhs_ndims,
+                                        const SizeT* rhs_dims,
+                                        SizeT rhs_ndims,
+                                        SizeT* out_dims) {
+    SizeT max_ndims = lhs_ndims > rhs_ndims ? lhs_ndims : rhs_ndims;
+
+    for (SizeT i = 0; i < max_ndims; ++i) {
+        const SizeT* lhs_dim_sz = i < lhs_ndims ? &lhs_dims[lhs_ndims - i - 1] : nullptr;
+        const SizeT* rhs_dim_sz = i < rhs_ndims ? &rhs_dims[rhs_ndims - i - 1] : nullptr;
+        SizeT* out_dim = &out_dims[max_ndims - i - 1];
+
+        if (lhs_dim_sz == nullptr) {
+            *out_dim = *rhs_dim_sz;
+        } else if (rhs_dim_sz == nullptr) {
+            *out_dim = *lhs_dim_sz;
+        } else if (*lhs_dim_sz == 1) {
+            *out_dim = *rhs_dim_sz;
+        } else if (*rhs_dim_sz == 1) {
+            *out_dim = *lhs_dim_sz;
+        } else if (*lhs_dim_sz == *rhs_dim_sz) {
+            *out_dim = *lhs_dim_sz;
+        } else {
+            __builtin_unreachable();
+        }
+    }
+}
+
+template<typename SizeT>
+void __nac3_ndarray_calc_broadcast_idx_impl(const SizeT* src_dims,
+                                            SizeT src_ndims,
+                                            const NDIndex* in_idx,
+                                            NDIndex* out_idx) {
+    for (SizeT i = 0; i < src_ndims; ++i) {
+        SizeT src_i = src_ndims - i - 1;
+        out_idx[src_i] = src_dims[src_i] == 1 ? 0 : in_idx[src_i];
+    }
+}
+}  // namespace
+
+extern "C" {
+uint32_t __nac3_ndarray_calc_size(const uint32_t* list_data, uint32_t list_len, uint32_t begin_idx, uint32_t end_idx) {
+    return __nac3_ndarray_calc_size_impl(list_data, list_len, begin_idx, end_idx);
+}
+
+uint64_t
+__nac3_ndarray_calc_size64(const uint64_t* list_data, uint64_t list_len, uint64_t begin_idx, uint64_t end_idx) {
+    return __nac3_ndarray_calc_size_impl(list_data, list_len, begin_idx, end_idx);
+}
+
+void __nac3_ndarray_calc_nd_indices(uint32_t index, const uint32_t* dims, uint32_t num_dims, NDIndex* idxs) {
+    __nac3_ndarray_calc_nd_indices_impl(index, dims, num_dims, idxs);
+}
+
+void __nac3_ndarray_calc_nd_indices64(uint64_t index, const uint64_t* dims, uint64_t num_dims, NDIndex* idxs) {
+    __nac3_ndarray_calc_nd_indices_impl(index, dims, num_dims, idxs);
+}
+
+uint32_t
+__nac3_ndarray_flatten_index(const uint32_t* dims, uint32_t num_dims, const NDIndex* indices, uint32_t num_indices) {
+    return __nac3_ndarray_flatten_index_impl(dims, num_dims, indices, num_indices);
+}
+
+uint64_t
+__nac3_ndarray_flatten_index64(const uint64_t* dims, uint64_t num_dims, const NDIndex* indices, uint64_t num_indices) {
+    return __nac3_ndarray_flatten_index_impl(dims, num_dims, indices, num_indices);
+}
+
+void __nac3_ndarray_calc_broadcast(const uint32_t* lhs_dims,
+                                   uint32_t lhs_ndims,
+                                   const uint32_t* rhs_dims,
+                                   uint32_t rhs_ndims,
+                                   uint32_t* out_dims) {
+    return __nac3_ndarray_calc_broadcast_impl(lhs_dims, lhs_ndims, rhs_dims, rhs_ndims, out_dims);
+}
+
+void __nac3_ndarray_calc_broadcast64(const uint64_t* lhs_dims,
+                                     uint64_t lhs_ndims,
+                                     const uint64_t* rhs_dims,
+                                     uint64_t rhs_ndims,
+                                     uint64_t* out_dims) {
+    return __nac3_ndarray_calc_broadcast_impl(lhs_dims, lhs_ndims, rhs_dims, rhs_ndims, out_dims);
+}
+
+void __nac3_ndarray_calc_broadcast_idx(const uint32_t* src_dims,
+                                       uint32_t src_ndims,
+                                       const NDIndex* in_idx,
+                                       NDIndex* out_idx) {
+    __nac3_ndarray_calc_broadcast_idx_impl(src_dims, src_ndims, in_idx, out_idx);
+}
+
+void __nac3_ndarray_calc_broadcast_idx64(const uint64_t* src_dims,
+                                         uint64_t src_ndims,
+                                         const NDIndex* in_idx,
+                                         NDIndex* out_idx) {
+    __nac3_ndarray_calc_broadcast_idx_impl(src_dims, src_ndims, in_idx, out_idx);
+}
+}

nac3core/irrt/irrt/slice.hpp

@@ -0,0 +1,28 @@
+#pragma once
+
+#include "irrt/int_types.hpp"
+
+extern "C" {
+SliceIndex __nac3_slice_index_bound(SliceIndex i, const SliceIndex len) {
+    if (i < 0) {
+        i = len + i;
+    }
+    if (i < 0) {
+        return 0;
+    } else if (i > len) {
+        return len;
+    }
+    return i;
+}
+
+SliceIndex __nac3_range_slice_len(const SliceIndex start, const SliceIndex end, const SliceIndex step) {
+    SliceIndex diff = end - start;
+    if (diff > 0 && step > 0) {
+        return ((diff - 1) / step) + 1;
+    } else if (diff < 0 && step < 0) {
+        return ((diff + 1) / step) + 1;
+    } else {
+        return 0;
+    }
+}
+}

nac3core/src/codegen/concrete_type.rs

@@ -0,0 +1,325 @@
+use std::collections::HashMap;
+
+use indexmap::IndexMap;
+
+use nac3parser::ast::StrRef;
+
+use crate::{
+    symbol_resolver::SymbolValue,
+    toplevel::DefinitionId,
+    typecheck::{
+        type_inferencer::PrimitiveStore,
+        typedef::{
+            into_var_map, FunSignature, FuncArg, Type, TypeEnum, TypeVar, TypeVarId, Unifier,
+        },
+    },
+};
+
+pub struct ConcreteTypeStore {
+    store: Vec<ConcreteTypeEnum>,
+}
+
+#[derive(Copy, Clone, Eq, PartialEq, Hash, Debug)]
+pub struct ConcreteType(usize);
+
+#[derive(Clone, Debug)]
+pub struct ConcreteFuncArg {
+    pub name: StrRef,
+    pub ty: ConcreteType,
+    pub default_value: Option<SymbolValue>,
+    pub is_vararg: bool,
+}
+
+#[derive(Clone, Debug)]
+pub enum Primitive {
+    Int32,
+    Int64,
+    UInt32,
+    UInt64,
+    Float,
+    Bool,
+    None,
+    Range,
+    Str,
+    Exception,
+}
+
+#[derive(Debug)]
+pub enum ConcreteTypeEnum {
+    TPrimitive(Primitive),
+    TTuple {
+        ty: Vec<ConcreteType>,
+        is_vararg_ctx: bool,
+    },
+    TObj {
+        obj_id: DefinitionId,
+        fields: HashMap<StrRef, (ConcreteType, bool)>,
+        params: IndexMap<TypeVarId, ConcreteType>,
+    },
+    TVirtual {
+        ty: ConcreteType,
+    },
+    TFunc {
+        args: Vec<ConcreteFuncArg>,
+        ret: ConcreteType,
+        vars: HashMap<TypeVarId, ConcreteType>,
+    },
+    TLiteral {
+        values: Vec<SymbolValue>,
+    },
+}
+
+impl ConcreteTypeStore {
+    #[must_use]
+    pub fn new() -> ConcreteTypeStore {
+        ConcreteTypeStore {
+            store: vec![
+                ConcreteTypeEnum::TPrimitive(Primitive::Int32),
+                ConcreteTypeEnum::TPrimitive(Primitive::Int64),
+                ConcreteTypeEnum::TPrimitive(Primitive::Float),
+                ConcreteTypeEnum::TPrimitive(Primitive::Bool),
+                ConcreteTypeEnum::TPrimitive(Primitive::None),
+                ConcreteTypeEnum::TPrimitive(Primitive::Range),
+                ConcreteTypeEnum::TPrimitive(Primitive::Str),
+                ConcreteTypeEnum::TPrimitive(Primitive::Exception),
+                ConcreteTypeEnum::TPrimitive(Primitive::UInt32),
+                ConcreteTypeEnum::TPrimitive(Primitive::UInt64),
+            ],
+        }
+    }
+
+    #[must_use]
+    pub fn get(&self, cty: ConcreteType) -> &ConcreteTypeEnum {
+        &self.store[cty.0]
+    }
+
+    pub fn from_signature(
+        &mut self,
+        unifier: &mut Unifier,
+        primitives: &PrimitiveStore,
+        signature: &FunSignature,
+        cache: &mut HashMap<Type, Option<ConcreteType>>,
+    ) -> ConcreteTypeEnum {
+        ConcreteTypeEnum::TFunc {
+            args: signature
+                .args
+                .iter()
+                .map(|arg| ConcreteFuncArg {
+                    name: arg.name,
+                    ty: if arg.is_vararg {
+                        let tuple_ty = unifier
+                            .add_ty(TypeEnum::TTuple { ty: vec![arg.ty], is_vararg_ctx: true });
+
+                        self.from_unifier_type(unifier, primitives, tuple_ty, cache)
+                    } else {
+                        self.from_unifier_type(unifier, primitives, arg.ty, cache)
+                    },
+                    default_value: arg.default_value.clone(),
+                    is_vararg: arg.is_vararg,
+                })
+                .collect(),
+            ret: self.from_unifier_type(unifier, primitives, signature.ret, cache),
+            vars: signature
+                .vars
+                .iter()
+                .map(|(id, ty)| (*id, self.from_unifier_type(unifier, primitives, *ty, cache)))
+                .collect(),
+        }
+    }
+
+    pub fn from_unifier_type(
+        &mut self,
+        unifier: &mut Unifier,
+        primitives: &PrimitiveStore,
+        ty: Type,
+        cache: &mut HashMap<Type, Option<ConcreteType>>,
+    ) -> ConcreteType {
+        let ty = unifier.get_representative(ty);
+        if unifier.unioned(ty, primitives.int32) {
+            ConcreteType(0)
+        } else if unifier.unioned(ty, primitives.int64) {
+            ConcreteType(1)
+        } else if unifier.unioned(ty, primitives.float) {
+            ConcreteType(2)
+        } else if unifier.unioned(ty, primitives.bool) {
+            ConcreteType(3)
+        } else if unifier.unioned(ty, primitives.none) {
+            ConcreteType(4)
+        } else if unifier.unioned(ty, primitives.range) {
+            ConcreteType(5)
+        } else if unifier.unioned(ty, primitives.str) {
+            ConcreteType(6)
+        } else if unifier.unioned(ty, primitives.exception) {
+            ConcreteType(7)
+        } else if unifier.unioned(ty, primitives.uint32) {
+            ConcreteType(8)
+        } else if unifier.unioned(ty, primitives.uint64) {
+            ConcreteType(9)
+        } else if let Some(cty) = cache.get(&ty) {
+            if let Some(cty) = cty {
+                *cty
+            } else {
+                let index = self.store.len();
+                // placeholder
+                self.store.push(ConcreteTypeEnum::TPrimitive(Primitive::Int32));
+                let result = ConcreteType(index);
+                cache.insert(ty, Some(result));
+                result
+            }
+        } else {
+            cache.insert(ty, None);
+            let ty_enum = unifier.get_ty(ty);
+            let result = match &*ty_enum {
+                TypeEnum::TTuple { ty, is_vararg_ctx } => ConcreteTypeEnum::TTuple {
+                    ty: ty
+                        .iter()
+                        .map(|t| self.from_unifier_type(unifier, primitives, *t, cache))
+                        .collect(),
+                    is_vararg_ctx: *is_vararg_ctx,
+                },
+                TypeEnum::TObj { obj_id, fields, params } => ConcreteTypeEnum::TObj {
+                    obj_id: *obj_id,
+                    fields: fields
+                        .iter()
+                        .filter_map(|(name, ty)| {
+                            // here we should not have type vars, but some partial instantiated
+                            // class methods can still have uninstantiated type vars, so
+                            // filter out all the methods, as this will not affect codegen
+                            if let TypeEnum::TFunc(..) = &*unifier.get_ty(ty.0) {
+                                None
+                            } else {
+                                Some((
+                                    *name,
+                                    (
+                                        self.from_unifier_type(unifier, primitives, ty.0, cache),
+                                        ty.1,
+                                    ),
+                                ))
+                            }
+                        })
+                        .collect(),
+                    params: params
+                        .iter()
+                        .map(|(id, ty)| {
+                            (*id, self.from_unifier_type(unifier, primitives, *ty, cache))
+                        })
+                        .collect(),
+                },
+                TypeEnum::TVirtual { ty } => ConcreteTypeEnum::TVirtual {
+                    ty: self.from_unifier_type(unifier, primitives, *ty, cache),
+                },
+                TypeEnum::TFunc(signature) => {
+                    self.from_signature(unifier, primitives, signature, cache)
+                }
+                TypeEnum::TLiteral { values, .. } => {
+                    ConcreteTypeEnum::TLiteral { values: values.clone() }
+                }
+                _ => unreachable!("{:?}", ty_enum.get_type_name()),
+            };
+            let index = if let Some(ConcreteType(index)) = cache.get(&ty).unwrap() {
+                self.store[*index] = result;
+                *index
+            } else {
+                self.store.push(result);
+                self.store.len() - 1
+            };
+            cache.insert(ty, Some(ConcreteType(index)));
+            ConcreteType(index)
+        }
+    }
+
+    pub fn to_unifier_type(
+        &self,
+        unifier: &mut Unifier,
+        primitives: &PrimitiveStore,
+        cty: ConcreteType,
+        cache: &mut HashMap<ConcreteType, Option<Type>>,
+    ) -> Type {
+        if let Some(ty) = cache.get_mut(&cty) {
+            return if let Some(ty) = ty {
+                *ty
+            } else {
+                *ty = Some(unifier.get_dummy_var().ty);
+                ty.unwrap()
+            };
+        }
+        cache.insert(cty, None);
+        let result = match &self.store[cty.0] {
+            ConcreteTypeEnum::TPrimitive(primitive) => {
+                let ty = match primitive {
+                    Primitive::Int32 => primitives.int32,
+                    Primitive::Int64 => primitives.int64,
+                    Primitive::UInt32 => primitives.uint32,
+                    Primitive::UInt64 => primitives.uint64,
+                    Primitive::Float => primitives.float,
+                    Primitive::Bool => primitives.bool,
+                    Primitive::None => primitives.none,
+                    Primitive::Range => primitives.range,
+                    Primitive::Str => primitives.str,
+                    Primitive::Exception => primitives.exception,
+                };
+                *cache.get_mut(&cty).unwrap() = Some(ty);
+                return ty;
+            }
+            ConcreteTypeEnum::TTuple { ty, is_vararg_ctx } => TypeEnum::TTuple {
+                ty: ty
+                    .iter()
+                    .map(|cty| self.to_unifier_type(unifier, primitives, *cty, cache))
+                    .collect(),
+                is_vararg_ctx: *is_vararg_ctx,
+            },
+            ConcreteTypeEnum::TVirtual { ty } => {
+                TypeEnum::TVirtual { ty: self.to_unifier_type(unifier, primitives, *ty, cache) }
+            }
+            ConcreteTypeEnum::TObj { obj_id, fields, params } => TypeEnum::TObj {
+                obj_id: *obj_id,
+                fields: fields
+                    .iter()
+                    .map(|(name, cty)| {
+                        (*name, (self.to_unifier_type(unifier, primitives, cty.0, cache), cty.1))
+                    })
+                    .collect::<HashMap<_, _>>(),
+                params: into_var_map(params.iter().map(|(&id, cty)| {
+                    let ty = self.to_unifier_type(unifier, primitives, *cty, cache);
+                    TypeVar { id, ty }
+                })),
+            },
+            ConcreteTypeEnum::TFunc { args, ret, vars } => TypeEnum::TFunc(FunSignature {
+                args: args
+                    .iter()
+                    .map(|arg| FuncArg {
+                        name: arg.name,
+                        ty: self.to_unifier_type(unifier, primitives, arg.ty, cache),
+                        default_value: arg.default_value.clone(),
+                        is_vararg: false,
+                    })
+                    .collect(),
+                ret: self.to_unifier_type(unifier, primitives, *ret, cache),
+                vars: into_var_map(vars.iter().map(|(&id, cty)| {
+                    let ty = self.to_unifier_type(unifier, primitives, *cty, cache);
+                    TypeVar { id, ty }
+                })),
+            }),
+            ConcreteTypeEnum::TLiteral { values, .. } => {
+                TypeEnum::TLiteral { values: values.clone(), loc: None }
+            }
+        };
+        let result = unifier.add_ty(result);
+        if let Some(ty) = cache.get(&cty).unwrap() {
+            unifier.unify(*ty, result).unwrap();
+        }
+        cache.insert(cty, Some(result));
+        result
+    }
+
+    pub fn add_cty(&mut self, cty: ConcreteTypeEnum) -> ConcreteType {
+        self.store.push(cty);
+        ConcreteType(self.store.len() - 1)
+    }
+}
+
+impl Default for ConcreteTypeStore {
+    fn default() -> Self {
+        Self::new()
+    }
+}

nac3core/src/codegen/extern_fns.rs

@@ -0,0 +1,193 @@
+use inkwell::{
+    attributes::{Attribute, AttributeLoc},
+    values::{BasicValueEnum, CallSiteValue, FloatValue, IntValue},
+};
+use itertools::Either;
+
+use crate::codegen::CodeGenContext;
+
+/// Macro to generate extern function
+/// Both function return type and function parameter type are `FloatValue`
+///
+/// Arguments:
+/// * `unary/binary`: Whether the extern function requires one (unary) or two (binary) operands
+/// * `$fn_name:ident`: The identifier of the rust function to be generated
+/// * `$extern_fn:literal`: Name of underlying extern function
+///
+/// Optional Arguments:
+/// * `$(,$attributes:literal)*)`: Attributes linked with the extern function.
+///   The default attributes are "mustprogress", "nofree", "nounwind", "willreturn", and "writeonly".
+///   These will be used unless other attributes are specified
+/// * `$(,$args:ident)*`: Operands of the extern function
+///   The data type of these operands will be set to `FloatValue`
+///  
+macro_rules! generate_extern_fn {
+    ("unary", $fn_name:ident, $extern_fn:literal) => {
+        generate_extern_fn!($fn_name, $extern_fn, arg, "mustprogress", "nofree", "nounwind", "willreturn", "writeonly");
+    };
+    ("unary", $fn_name:ident, $extern_fn:literal $(,$attributes:literal)*) => {
+        generate_extern_fn!($fn_name, $extern_fn, arg $(,$attributes)*);
+    };
+    ("binary", $fn_name:ident, $extern_fn:literal) => {
+        generate_extern_fn!($fn_name, $extern_fn, arg1, arg2, "mustprogress", "nofree", "nounwind", "willreturn", "writeonly");
+    };
+    ("binary", $fn_name:ident, $extern_fn:literal $(,$attributes:literal)*) => {
+        generate_extern_fn!($fn_name, $extern_fn, arg1, arg2 $(,$attributes)*);
+    };
+    ($fn_name:ident, $extern_fn:literal $(,$args:ident)* $(,$attributes:literal)*) => {
+        #[doc = concat!("Invokes the [`", stringify!($extern_fn), "`](https://en.cppreference.com/w/c/numeric/math/", stringify!($llvm_name), ") function." )]
+        pub fn $fn_name<'ctx>(
+            ctx: &CodeGenContext<'ctx, '_>
+            $(,$args: FloatValue<'ctx>)*,
+            name: Option<&str>,
+        ) -> FloatValue<'ctx> {
+            const FN_NAME: &str = $extern_fn;
+
+            let llvm_f64 = ctx.ctx.f64_type();
+            $(debug_assert_eq!($args.get_type(), llvm_f64);)*
+
+            let extern_fn = ctx.module.get_function(FN_NAME).unwrap_or_else(|| {
+                let fn_type = llvm_f64.fn_type(&[$($args.get_type().into()),*], false);
+                let func = ctx.module.add_function(FN_NAME, fn_type, None);
+                for attr in [$($attributes),*] {
+                    func.add_attribute(
+                        AttributeLoc::Function,
+                        ctx.ctx.create_enum_attribute(Attribute::get_named_enum_kind_id(attr), 0),
+                    );
+                }
+                func
+            });
+
+            ctx.builder
+                .build_call(extern_fn, &[$($args.into()),*], name.unwrap_or_default())
+                .map(CallSiteValue::try_as_basic_value)
+                .map(|v| v.map_left(BasicValueEnum::into_float_value))
+                .map(Either::unwrap_left)
+                .unwrap()
+        }
+    };
+}
+
+generate_extern_fn!("unary", call_tan, "tan");
+generate_extern_fn!("unary", call_asin, "asin");
+generate_extern_fn!("unary", call_acos, "acos");
+generate_extern_fn!("unary", call_atan, "atan");
+generate_extern_fn!("unary", call_sinh, "sinh");
+generate_extern_fn!("unary", call_cosh, "cosh");
+generate_extern_fn!("unary", call_tanh, "tanh");
+generate_extern_fn!("unary", call_asinh, "asinh");
+generate_extern_fn!("unary", call_acosh, "acosh");
+generate_extern_fn!("unary", call_atanh, "atanh");
+generate_extern_fn!("unary", call_expm1, "expm1");
+generate_extern_fn!(
+    "unary",
+    call_cbrt,
+    "cbrt",
+    "mustprogress",
+    "nofree",
+    "nosync",
+    "nounwind",
+    "readonly",
+    "willreturn"
+);
+generate_extern_fn!("unary", call_erf, "erf", "nounwind");
+generate_extern_fn!("unary", call_erfc, "erfc", "nounwind");
+generate_extern_fn!("unary", call_j1, "j1", "nounwind");
+
+generate_extern_fn!("binary", call_atan2, "atan2");
+generate_extern_fn!("binary", call_hypot, "hypot", "nounwind");
+generate_extern_fn!("binary", call_nextafter, "nextafter", "nounwind");
+
+/// Invokes the [`ldexp`](https://en.cppreference.com/w/c/numeric/math/ldexp) function.
+pub fn call_ldexp<'ctx>(
+    ctx: &CodeGenContext<'ctx, '_>,
+    arg: FloatValue<'ctx>,
+    exp: IntValue<'ctx>,
+    name: Option<&str>,
+) -> FloatValue<'ctx> {
+    const FN_NAME: &str = "ldexp";
+
+    let llvm_f64 = ctx.ctx.f64_type();
+    let llvm_i32 = ctx.ctx.i32_type();
+    debug_assert_eq!(arg.get_type(), llvm_f64);
+    debug_assert_eq!(exp.get_type(), llvm_i32);
+
+    let extern_fn = ctx.module.get_function(FN_NAME).unwrap_or_else(|| {
+        let fn_type = llvm_f64.fn_type(&[llvm_f64.into(), llvm_i32.into()], false);
+        let func = ctx.module.add_function(FN_NAME, fn_type, None);
+        for attr in ["mustprogress", "nofree", "nounwind", "willreturn"] {
+            func.add_attribute(
+                AttributeLoc::Function,
+                ctx.ctx.create_enum_attribute(Attribute::get_named_enum_kind_id(attr), 0),
+            );
+        }
+
+        func
+    });
+
+    ctx.builder
+        .build_call(extern_fn, &[arg.into(), exp.into()], name.unwrap_or_default())
+        .map(CallSiteValue::try_as_basic_value)
+        .map(|v| v.map_left(BasicValueEnum::into_float_value))
+        .map(Either::unwrap_left)
+        .unwrap()
+}
+
+/// Macro to generate `np_linalg` and `sp_linalg` functions
+/// The function takes as input `NDArray` and returns ()
+///
+/// Arguments:
+/// * `$fn_name:ident`: The identifier of the rust function to be generated
+/// * `$extern_fn:literal`: Name of underlying extern function
+/// * (2/3/4): Number of `NDArray` that function takes as input
+///
+/// Note:
+/// The operands and resulting `NDArray` are both passed as input to the funcion
+/// It is the responsibility of caller to ensure that output `NDArray` is properly allocated on stack
+/// The function changes the content of the output `NDArray` in-place
+macro_rules! generate_linalg_extern_fn {
+    ($fn_name:ident, $extern_fn:literal, 2) => {
+        generate_linalg_extern_fn!($fn_name, $extern_fn, mat1, mat2);
+    };
+    ($fn_name:ident, $extern_fn:literal, 3) => {
+        generate_linalg_extern_fn!($fn_name, $extern_fn, mat1, mat2, mat3);
+    };
+    ($fn_name:ident, $extern_fn:literal, 4) => {
+        generate_linalg_extern_fn!($fn_name, $extern_fn, mat1, mat2, mat3, mat4);
+    };
+    ($fn_name:ident, $extern_fn:literal $(,$input_matrix:ident)*) => {
+        #[doc = concat!("Invokes the linalg `", stringify!($extern_fn), " function." )]
+        pub fn $fn_name<'ctx>(
+            ctx: &mut CodeGenContext<'ctx, '_>
+            $(,$input_matrix: BasicValueEnum<'ctx>)*,
+            name: Option<&str>,
+        ){
+            const FN_NAME: &str = $extern_fn;
+            let extern_fn = ctx.module.get_function(FN_NAME).unwrap_or_else(|| {
+                let fn_type = ctx.ctx.void_type().fn_type(&[$($input_matrix.get_type().into()),*], false);
+
+                let func = ctx.module.add_function(FN_NAME, fn_type, None);
+                for attr in ["mustprogress", "nofree", "nounwind", "willreturn", "writeonly"] {
+                    func.add_attribute(
+                        AttributeLoc::Function,
+                        ctx.ctx.create_enum_attribute(Attribute::get_named_enum_kind_id(attr), 0),
+                    );
+                }
+                func
+            });
+
+            ctx.builder.build_call(extern_fn, &[$($input_matrix.into(),)*], name.unwrap_or_default()).unwrap();
+        }
+    };
+}
+
+generate_linalg_extern_fn!(call_np_linalg_cholesky, "np_linalg_cholesky", 2);
+generate_linalg_extern_fn!(call_np_linalg_qr, "np_linalg_qr", 3);
+generate_linalg_extern_fn!(call_np_linalg_svd, "np_linalg_svd", 4);
+generate_linalg_extern_fn!(call_np_linalg_inv, "np_linalg_inv", 2);
+generate_linalg_extern_fn!(call_np_linalg_pinv, "np_linalg_pinv", 2);
+generate_linalg_extern_fn!(call_np_linalg_matrix_power, "np_linalg_matrix_power", 3);
+generate_linalg_extern_fn!(call_np_linalg_det, "np_linalg_det", 2);
+generate_linalg_extern_fn!(call_sp_linalg_lu, "sp_linalg_lu", 3);
+generate_linalg_extern_fn!(call_sp_linalg_schur, "sp_linalg_schur", 3);
+generate_linalg_extern_fn!(call_sp_linalg_hessenberg, "sp_linalg_hessenberg", 3);

nac3core/src/codegen/generator.rs

@@ -0,0 +1,294 @@
+use inkwell::{
+    context::Context,
+    types::{BasicTypeEnum, IntType},
+    values::{BasicValueEnum, IntValue, PointerValue},
+};
+
+use nac3parser::ast::{Expr, Stmt, StrRef};
+
+use crate::{
+    codegen::{bool_to_i1, bool_to_i8, classes::ArraySliceValue, expr::*, stmt::*, CodeGenContext},
+    symbol_resolver::ValueEnum,
+    toplevel::{DefinitionId, TopLevelDef},
+    typecheck::typedef::{FunSignature, Type},
+};
+
+pub trait CodeGenerator {
+    /// Return the module name for the code generator.
+    fn get_name(&self) -> &str;
+
+    fn get_size_type<'ctx>(&self, ctx: &'ctx Context) -> IntType<'ctx>;
+
+    /// Generate function call and returns the function return value.
+    /// - obj: Optional object for method call.
+    /// - fun: Function signature and definition ID.
+    /// - params: Function parameters. Note that this does not include the object even if the
+    ///   function is a class method.
+    fn gen_call<'ctx>(
+        &mut self,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        obj: Option<(Type, ValueEnum<'ctx>)>,
+        fun: (&FunSignature, DefinitionId),
+        params: Vec<(Option<StrRef>, ValueEnum<'ctx>)>,
+    ) -> Result<Option<BasicValueEnum<'ctx>>, String>
+    where
+        Self: Sized,
+    {
+        gen_call(self, ctx, obj, fun, params)
+    }
+
+    /// Generate object constructor and returns the constructed object.
+    /// - signature: Function signature of the constructor.
+    /// - def: Class definition for the constructor class.
+    /// - params: Function parameters.
+    fn gen_constructor<'ctx>(
+        &mut self,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        signature: &FunSignature,
+        def: &TopLevelDef,
+        params: Vec<(Option<StrRef>, ValueEnum<'ctx>)>,
+    ) -> Result<BasicValueEnum<'ctx>, String>
+    where
+        Self: Sized,
+    {
+        gen_constructor(self, ctx, signature, def, params)
+    }
+
+    /// Generate a function instance.
+    /// - obj: Optional object for method call.
+    /// - fun: Function signature, definition ID and the substitution key.
+    /// - params: Function parameters. Note that this does not include the object even if the
+    ///   function is a class method.
+    ///
+    /// Note that this function should check if the function is generated in another thread (due to
+    /// possible race condition), see the default implementation for an example.
+    fn gen_func_instance<'ctx>(
+        &mut self,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        obj: Option<(Type, ValueEnum<'ctx>)>,
+        fun: (&FunSignature, &mut TopLevelDef, String),
+        id: usize,
+    ) -> Result<String, String> {
+        gen_func_instance(ctx, &obj, fun, id)
+    }
+
+    /// Generate the code for an expression.
+    fn gen_expr<'ctx>(
+        &mut self,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        expr: &Expr<Option<Type>>,
+    ) -> Result<Option<ValueEnum<'ctx>>, String>
+    where
+        Self: Sized,
+    {
+        gen_expr(self, ctx, expr)
+    }
+
+    /// Allocate memory for a variable and return a pointer pointing to it.
+    /// The default implementation places the allocations at the start of the function.
+    fn gen_var_alloc<'ctx>(
+        &mut self,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        ty: BasicTypeEnum<'ctx>,
+        name: Option<&str>,
+    ) -> Result<PointerValue<'ctx>, String> {
+        gen_var(ctx, ty, name)
+    }
+
+    /// Allocate memory for a variable and return a pointer pointing to it.
+    /// The default implementation places the allocations at the start of the function.
+    fn gen_array_var_alloc<'ctx>(
+        &mut self,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        ty: BasicTypeEnum<'ctx>,
+        size: IntValue<'ctx>,
+        name: Option<&'ctx str>,
+    ) -> Result<ArraySliceValue<'ctx>, String> {
+        gen_array_var(ctx, ty, size, name)
+    }
+
+    /// Return a pointer pointing to the target of the expression.
+    fn gen_store_target<'ctx>(
+        &mut self,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        pattern: &Expr<Option<Type>>,
+        name: Option<&str>,
+    ) -> Result<Option<PointerValue<'ctx>>, String>
+    where
+        Self: Sized,
+    {
+        gen_store_target(self, ctx, pattern, name)
+    }
+
+    /// Generate code for an assignment expression.
+    fn gen_assign<'ctx>(
+        &mut self,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        target: &Expr<Option<Type>>,
+        value: ValueEnum<'ctx>,
+        value_ty: Type,
+    ) -> Result<(), String>
+    where
+        Self: Sized,
+    {
+        gen_assign(self, ctx, target, value, value_ty)
+    }
+
+    /// Generate code for an assignment expression where LHS is a `"target_list"`.
+    ///
+    /// See <https://docs.python.org/3/reference/simple_stmts.html#assignment-statements>.
+    fn gen_assign_target_list<'ctx>(
+        &mut self,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        targets: &Vec<Expr<Option<Type>>>,
+        value: ValueEnum<'ctx>,
+        value_ty: Type,
+    ) -> Result<(), String>
+    where
+        Self: Sized,
+    {
+        gen_assign_target_list(self, ctx, targets, value, value_ty)
+    }
+
+    /// Generate code for an item assignment.
+    ///
+    /// i.e., `target[key] = value`
+    fn gen_setitem<'ctx>(
+        &mut self,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        target: &Expr<Option<Type>>,
+        key: &Expr<Option<Type>>,
+        value: ValueEnum<'ctx>,
+        value_ty: Type,
+    ) -> Result<(), String>
+    where
+        Self: Sized,
+    {
+        gen_setitem(self, ctx, target, key, value, value_ty)
+    }
+
+    /// Generate code for a while expression.
+    /// Return true if the while loop must early return
+    fn gen_while(
+        &mut self,
+        ctx: &mut CodeGenContext<'_, '_>,
+        stmt: &Stmt<Option<Type>>,
+    ) -> Result<(), String>
+    where
+        Self: Sized,
+    {
+        gen_while(self, ctx, stmt)
+    }
+
+    /// Generate code for a for expression.
+    /// Return true if the for loop must early return
+    fn gen_for(
+        &mut self,
+        ctx: &mut CodeGenContext<'_, '_>,
+        stmt: &Stmt<Option<Type>>,
+    ) -> Result<(), String>
+    where
+        Self: Sized,
+    {
+        gen_for(self, ctx, stmt)
+    }
+
+    /// Generate code for an if expression.
+    /// Return true if the statement must early return
+    fn gen_if(
+        &mut self,
+        ctx: &mut CodeGenContext<'_, '_>,
+        stmt: &Stmt<Option<Type>>,
+    ) -> Result<(), String>
+    where
+        Self: Sized,
+    {
+        gen_if(self, ctx, stmt)
+    }
+
+    fn gen_with(
+        &mut self,
+        ctx: &mut CodeGenContext<'_, '_>,
+        stmt: &Stmt<Option<Type>>,
+    ) -> Result<(), String>
+    where
+        Self: Sized,
+    {
+        gen_with(self, ctx, stmt)
+    }
+
+    /// Generate code for a statement
+    ///
+    /// Return true if the statement must early return
+    fn gen_stmt(
+        &mut self,
+        ctx: &mut CodeGenContext<'_, '_>,
+        stmt: &Stmt<Option<Type>>,
+    ) -> Result<(), String>
+    where
+        Self: Sized,
+    {
+        gen_stmt(self, ctx, stmt)
+    }
+
+    /// Generates code for a block statement.
+    fn gen_block<'a, I: Iterator<Item = &'a Stmt<Option<Type>>>>(
+        &mut self,
+        ctx: &mut CodeGenContext<'_, '_>,
+        stmts: I,
+    ) -> Result<(), String>
+    where
+        Self: Sized,
+    {
+        gen_block(self, ctx, stmts)
+    }
+
+    /// See [`bool_to_i1`].
+    fn bool_to_i1<'ctx>(
+        &self,
+        ctx: &CodeGenContext<'ctx, '_>,
+        bool_value: IntValue<'ctx>,
+    ) -> IntValue<'ctx> {
+        bool_to_i1(&ctx.builder, bool_value)
+    }
+
+    /// See [`bool_to_i8`].
+    fn bool_to_i8<'ctx>(
+        &self,
+        ctx: &CodeGenContext<'ctx, '_>,
+        bool_value: IntValue<'ctx>,
+    ) -> IntValue<'ctx> {
+        bool_to_i8(&ctx.builder, ctx.ctx, bool_value)
+    }
+}
+
+pub struct DefaultCodeGenerator {
+    name: String,
+    size_t: u32,
+}
+
+impl DefaultCodeGenerator {
+    #[must_use]
+    pub fn new(name: String, size_t: u32) -> DefaultCodeGenerator {
+        assert!(matches!(size_t, 32 | 64));
+        DefaultCodeGenerator { name, size_t }
+    }
+}
+
+impl CodeGenerator for DefaultCodeGenerator {
+    /// Returns the name for this [`CodeGenerator`].
+    fn get_name(&self) -> &str {
+        &self.name
+    }
+
+    /// Returns an LLVM integer type representing `size_t`.
+    fn get_size_type<'ctx>(&self, ctx: &'ctx Context) -> IntType<'ctx> {
+        // it should be unsigned, but we don't really need unsigned and this could save us from
+        // having to do a bit cast...
+        if self.size_t == 32 {
+            ctx.i32_type()
+        } else {
+            ctx.i64_type()
+        }
+    }
+}

nac3core/src/codegen/irrt/mod.rs

@@ -0,0 +1,952 @@
+use inkwell::{
+    attributes::{Attribute, AttributeLoc},
+    context::Context,
+    memory_buffer::MemoryBuffer,
+    module::Module,
+    types::{BasicTypeEnum, IntType},
+    values::{BasicValue, BasicValueEnum, CallSiteValue, FloatValue, IntValue},
+    AddressSpace, IntPredicate,
+};
+use itertools::Either;
+
+use nac3parser::ast::Expr;
+
+use super::{
+    classes::{
+        ArrayLikeIndexer, ArrayLikeValue, ArraySliceValue, ListValue, NDArrayValue,
+        TypedArrayLikeAccessor, TypedArrayLikeAdapter, UntypedArrayLikeAccessor,
+    },
+    llvm_intrinsics,
+    macros::codegen_unreachable,
+    stmt::gen_for_callback_incrementing,
+    CodeGenContext, CodeGenerator,
+};
+use crate::{symbol_resolver::SymbolResolver, typecheck::typedef::Type};
+
+#[must_use]
+pub fn load_irrt<'ctx>(ctx: &'ctx Context, symbol_resolver: &dyn SymbolResolver) -> Module<'ctx> {
+    let bitcode_buf = MemoryBuffer::create_from_memory_range(
+        include_bytes!(concat!(env!("OUT_DIR"), "/irrt.bc")),
+        "irrt_bitcode_buffer",
+    );
+    let irrt_mod = Module::parse_bitcode_from_buffer(&bitcode_buf, ctx).unwrap();
+    let inline_attr = Attribute::get_named_enum_kind_id("alwaysinline");
+    for symbol in &[
+        "__nac3_int_exp_int32_t",
+        "__nac3_int_exp_int64_t",
+        "__nac3_range_slice_len",
+        "__nac3_slice_index_bound",
+    ] {
+        let function = irrt_mod.get_function(symbol).unwrap();
+        function.add_attribute(AttributeLoc::Function, ctx.create_enum_attribute(inline_attr, 0));
+    }
+
+    // Initialize all global `EXN_*` exception IDs in IRRT with the [`SymbolResolver`].
+    let exn_id_type = ctx.i32_type();
+    let errors = &[
+        ("EXN_INDEX_ERROR", "0:IndexError"),
+        ("EXN_VALUE_ERROR", "0:ValueError"),
+        ("EXN_ASSERTION_ERROR", "0:AssertionError"),
+        ("EXN_TYPE_ERROR", "0:TypeError"),
+    ];
+    for (irrt_name, symbol_name) in errors {
+        let exn_id = symbol_resolver.get_string_id(symbol_name);
+        let exn_id = exn_id_type.const_int(exn_id as u64, false).as_basic_value_enum();
+
+        let global = irrt_mod.get_global(irrt_name).unwrap_or_else(|| {
+            panic!("Exception symbol name '{irrt_name}' should exist in the IRRT LLVM module")
+        });
+        global.set_initializer(&exn_id);
+    }
+
+    irrt_mod
+}
+
+// repeated squaring method adapted from GNU Scientific Library:
+// https://git.savannah.gnu.org/cgit/gsl.git/tree/sys/pow_int.c
+pub fn integer_power<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    base: IntValue<'ctx>,
+    exp: IntValue<'ctx>,
+    signed: bool,
+) -> IntValue<'ctx> {
+    let symbol = match (base.get_type().get_bit_width(), exp.get_type().get_bit_width(), signed) {
+        (32, 32, true) => "__nac3_int_exp_int32_t",
+        (64, 64, true) => "__nac3_int_exp_int64_t",
+        (32, 32, false) => "__nac3_int_exp_uint32_t",
+        (64, 64, false) => "__nac3_int_exp_uint64_t",
+        _ => codegen_unreachable!(ctx),
+    };
+    let base_type = base.get_type();
+    let pow_fun = ctx.module.get_function(symbol).unwrap_or_else(|| {
+        let fn_type = base_type.fn_type(&[base_type.into(), base_type.into()], false);
+        ctx.module.add_function(symbol, fn_type, None)
+    });
+    // throw exception when exp < 0
+    let ge_zero = ctx
+        .builder
+        .build_int_compare(
+            IntPredicate::SGE,
+            exp,
+            exp.get_type().const_zero(),
+            "assert_int_pow_ge_0",
+        )
+        .unwrap();
+    ctx.make_assert(
+        generator,
+        ge_zero,
+        "0:ValueError",
+        "integer power must be positive or zero",
+        [None, None, None],
+        ctx.current_loc,
+    );
+    ctx.builder
+        .build_call(pow_fun, &[base.into(), exp.into()], "call_int_pow")
+        .map(CallSiteValue::try_as_basic_value)
+        .map(|v| v.map_left(BasicValueEnum::into_int_value))
+        .map(Either::unwrap_left)
+        .unwrap()
+}
+
+pub fn calculate_len_for_slice_range<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    start: IntValue<'ctx>,
+    end: IntValue<'ctx>,
+    step: IntValue<'ctx>,
+) -> IntValue<'ctx> {
+    const SYMBOL: &str = "__nac3_range_slice_len";
+    let len_func = ctx.module.get_function(SYMBOL).unwrap_or_else(|| {
+        let i32_t = ctx.ctx.i32_type();
+        let fn_t = i32_t.fn_type(&[i32_t.into(), i32_t.into(), i32_t.into()], false);
+        ctx.module.add_function(SYMBOL, fn_t, None)
+    });
+
+    // assert step != 0, throw exception if not
+    let not_zero = ctx
+        .builder
+        .build_int_compare(IntPredicate::NE, step, step.get_type().const_zero(), "range_step_ne")
+        .unwrap();
+    ctx.make_assert(
+        generator,
+        not_zero,
+        "0:ValueError",
+        "step must not be zero",
+        [None, None, None],
+        ctx.current_loc,
+    );
+    ctx.builder
+        .build_call(len_func, &[start.into(), end.into(), step.into()], "calc_len")
+        .map(CallSiteValue::try_as_basic_value)
+        .map(|v| v.map_left(BasicValueEnum::into_int_value))
+        .map(Either::unwrap_left)
+        .unwrap()
+}
+
+/// NOTE: the output value of the end index of this function should be compared ***inclusively***,
+/// because python allows `a[2::-1]`, whose semantic is `[a[2], a[1], a[0]]`, which is equivalent to
+/// NO numeric slice in python.
+///
+/// equivalent code:
+/// ```pseudo_code
+/// match (start, end, step):
+///     case (s, e, None | Some(step)) if step > 0:
+///         return (
+///             match s:
+///                 case None:
+///                     0
+///                 case Some(s):
+///                     handle_in_bound(s)
+///             ,match e:
+///                 case None:
+///                     length - 1
+///                 case Some(e):
+///                     handle_in_bound(e) - 1
+///             ,step == None ? 1 : step
+///         )
+///     case (s, e, Some(step)) if step < 0:
+///         return (
+///             match s:
+///                 case None:
+///                     length - 1
+///                 case Some(s):
+///                     s = handle_in_bound(s)
+///                     if s == length:
+///                         s - 1
+///                     else:
+///                         s
+///             ,match e:
+///                 case None:
+///                     0
+///                 case Some(e):
+///                     handle_in_bound(e) + 1
+///             ,step
+///         )
+/// ```
+pub fn handle_slice_indices<'ctx, G: CodeGenerator>(
+    start: &Option<Box<Expr<Option<Type>>>>,
+    end: &Option<Box<Expr<Option<Type>>>>,
+    step: &Option<Box<Expr<Option<Type>>>>,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    generator: &mut G,
+    length: IntValue<'ctx>,
+) -> Result<Option<(IntValue<'ctx>, IntValue<'ctx>, IntValue<'ctx>)>, String> {
+    let int32 = ctx.ctx.i32_type();
+    let zero = int32.const_zero();
+    let one = int32.const_int(1, false);
+    let length = ctx.builder.build_int_truncate_or_bit_cast(length, int32, "leni32").unwrap();
+    Ok(Some(match (start, end, step) {
+        (s, e, None) => (
+            if let Some(s) = s.as_ref() {
+                match handle_slice_index_bound(s, ctx, generator, length)? {
+                    Some(v) => v,
+                    None => return Ok(None),
+                }
+            } else {
+                int32.const_zero()
+            },
+            {
+                let e = if let Some(s) = e.as_ref() {
+                    match handle_slice_index_bound(s, ctx, generator, length)? {
+                        Some(v) => v,
+                        None => return Ok(None),
+                    }
+                } else {
+                    length
+                };
+                ctx.builder.build_int_sub(e, one, "final_end").unwrap()
+            },
+            one,
+        ),
+        (s, e, Some(step)) => {
+            let step = if let Some(v) = generator.gen_expr(ctx, step)? {
+                v.to_basic_value_enum(ctx, generator, ctx.primitives.int32)?.into_int_value()
+            } else {
+                return Ok(None);
+            };
+            // assert step != 0, throw exception if not
+            let not_zero = ctx
+                .builder
+                .build_int_compare(
+                    IntPredicate::NE,
+                    step,
+                    step.get_type().const_zero(),
+                    "range_step_ne",
+                )
+                .unwrap();
+            ctx.make_assert(
+                generator,
+                not_zero,
+                "0:ValueError",
+                "slice step cannot be zero",
+                [None, None, None],
+                ctx.current_loc,
+            );
+            let len_id = ctx.builder.build_int_sub(length, one, "lenmin1").unwrap();
+            let neg = ctx
+                .builder
+                .build_int_compare(IntPredicate::SLT, step, zero, "step_is_neg")
+                .unwrap();
+            (
+                match s {
+                    Some(s) => {
+                        let Some(s) = handle_slice_index_bound(s, ctx, generator, length)? else {
+                            return Ok(None);
+                        };
+                        ctx.builder
+                            .build_select(
+                                ctx.builder
+                                    .build_and(
+                                        ctx.builder
+                                            .build_int_compare(
+                                                IntPredicate::EQ,
+                                                s,
+                                                length,
+                                                "s_eq_len",
+                                            )
+                                            .unwrap(),
+                                        neg,
+                                        "should_minus_one",
+                                    )
+                                    .unwrap(),
+                                ctx.builder.build_int_sub(s, one, "s_min").unwrap(),
+                                s,
+                                "final_start",
+                            )
+                            .map(BasicValueEnum::into_int_value)
+                            .unwrap()
+                    }
+                    None => ctx
+                        .builder
+                        .build_select(neg, len_id, zero, "stt")
+                        .map(BasicValueEnum::into_int_value)
+                        .unwrap(),
+                },
+                match e {
+                    Some(e) => {
+                        let Some(e) = handle_slice_index_bound(e, ctx, generator, length)? else {
+                            return Ok(None);
+                        };
+                        ctx.builder
+                            .build_select(
+                                neg,
+                                ctx.builder.build_int_add(e, one, "end_add_one").unwrap(),
+                                ctx.builder.build_int_sub(e, one, "end_sub_one").unwrap(),
+                                "final_end",
+                            )
+                            .map(BasicValueEnum::into_int_value)
+                            .unwrap()
+                    }
+                    None => ctx
+                        .builder
+                        .build_select(neg, zero, len_id, "end")
+                        .map(BasicValueEnum::into_int_value)
+                        .unwrap(),
+                },
+                step,
+            )
+        }
+    }))
+}
+
+/// this function allows index out of range, since python
+/// allows index out of range in slice (`a = [1,2,3]; a[1:10] == [2,3]`).
+pub fn handle_slice_index_bound<'ctx, G: CodeGenerator>(
+    i: &Expr<Option<Type>>,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    generator: &mut G,
+    length: IntValue<'ctx>,
+) -> Result<Option<IntValue<'ctx>>, String> {
+    const SYMBOL: &str = "__nac3_slice_index_bound";
+    let func = ctx.module.get_function(SYMBOL).unwrap_or_else(|| {
+        let i32_t = ctx.ctx.i32_type();
+        let fn_t = i32_t.fn_type(&[i32_t.into(), i32_t.into()], false);
+        ctx.module.add_function(SYMBOL, fn_t, None)
+    });
+
+    let i = if let Some(v) = generator.gen_expr(ctx, i)? {
+        v.to_basic_value_enum(ctx, generator, i.custom.unwrap())?
+    } else {
+        return Ok(None);
+    };
+    Ok(Some(
+        ctx.builder
+            .build_call(func, &[i.into(), length.into()], "bounded_ind")
+            .map(CallSiteValue::try_as_basic_value)
+            .map(|v| v.map_left(BasicValueEnum::into_int_value))
+            .map(Either::unwrap_left)
+            .unwrap(),
+    ))
+}
+
+/// This function handles 'end' **inclusively**.
+/// Order of tuples `assign_idx` and `value_idx` is ('start', 'end', 'step').
+/// Negative index should be handled before entering this function
+pub fn list_slice_assignment<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    ty: BasicTypeEnum<'ctx>,
+    dest_arr: ListValue<'ctx>,
+    dest_idx: (IntValue<'ctx>, IntValue<'ctx>, IntValue<'ctx>),
+    src_arr: ListValue<'ctx>,
+    src_idx: (IntValue<'ctx>, IntValue<'ctx>, IntValue<'ctx>),
+) {
+    let size_ty = generator.get_size_type(ctx.ctx);
+    let int8_ptr = ctx.ctx.i8_type().ptr_type(AddressSpace::default());
+    let int32 = ctx.ctx.i32_type();
+    let (fun_symbol, elem_ptr_type) = ("__nac3_list_slice_assign_var_size", int8_ptr);
+    let slice_assign_fun = {
+        let ty_vec = vec![
+            int32.into(),         // dest start idx
+            int32.into(),         // dest end idx
+            int32.into(),         // dest step
+            elem_ptr_type.into(), // dest arr ptr
+            int32.into(),         // dest arr len
+            int32.into(),         // src start idx
+            int32.into(),         // src end idx
+            int32.into(),         // src step
+            elem_ptr_type.into(), // src arr ptr
+            int32.into(),         // src arr len
+            int32.into(),         // size
+        ];
+        ctx.module.get_function(fun_symbol).unwrap_or_else(|| {
+            let fn_t = int32.fn_type(ty_vec.as_slice(), false);
+            ctx.module.add_function(fun_symbol, fn_t, None)
+        })
+    };
+
+    let zero = int32.const_zero();
+    let one = int32.const_int(1, false);
+    let dest_arr_ptr = dest_arr.data().base_ptr(ctx, generator);
+    let dest_arr_ptr =
+        ctx.builder.build_pointer_cast(dest_arr_ptr, elem_ptr_type, "dest_arr_ptr_cast").unwrap();
+    let dest_len = dest_arr.load_size(ctx, Some("dest.len"));
+    let dest_len = ctx.builder.build_int_truncate_or_bit_cast(dest_len, int32, "srclen32").unwrap();
+    let src_arr_ptr = src_arr.data().base_ptr(ctx, generator);
+    let src_arr_ptr =
+        ctx.builder.build_pointer_cast(src_arr_ptr, elem_ptr_type, "src_arr_ptr_cast").unwrap();
+    let src_len = src_arr.load_size(ctx, Some("src.len"));
+    let src_len = ctx.builder.build_int_truncate_or_bit_cast(src_len, int32, "srclen32").unwrap();
+
+    // index in bound and positive should be done
+    // assert if dest.step == 1 then len(src) <= len(dest) else len(src) == len(dest), and
+    // throw exception if not satisfied
+    let src_end = ctx
+        .builder
+        .build_select(
+            ctx.builder.build_int_compare(IntPredicate::SLT, src_idx.2, zero, "is_neg").unwrap(),
+            ctx.builder.build_int_sub(src_idx.1, one, "e_min_one").unwrap(),
+            ctx.builder.build_int_add(src_idx.1, one, "e_add_one").unwrap(),
+            "final_e",
+        )
+        .map(BasicValueEnum::into_int_value)
+        .unwrap();
+    let dest_end = ctx
+        .builder
+        .build_select(
+            ctx.builder.build_int_compare(IntPredicate::SLT, dest_idx.2, zero, "is_neg").unwrap(),
+            ctx.builder.build_int_sub(dest_idx.1, one, "e_min_one").unwrap(),
+            ctx.builder.build_int_add(dest_idx.1, one, "e_add_one").unwrap(),
+            "final_e",
+        )
+        .map(BasicValueEnum::into_int_value)
+        .unwrap();
+    let src_slice_len =
+        calculate_len_for_slice_range(generator, ctx, src_idx.0, src_end, src_idx.2);
+    let dest_slice_len =
+        calculate_len_for_slice_range(generator, ctx, dest_idx.0, dest_end, dest_idx.2);
+    let src_eq_dest = ctx
+        .builder
+        .build_int_compare(IntPredicate::EQ, src_slice_len, dest_slice_len, "slice_src_eq_dest")
+        .unwrap();
+    let src_slt_dest = ctx
+        .builder
+        .build_int_compare(IntPredicate::SLT, src_slice_len, dest_slice_len, "slice_src_slt_dest")
+        .unwrap();
+    let dest_step_eq_one = ctx
+        .builder
+        .build_int_compare(
+            IntPredicate::EQ,
+            dest_idx.2,
+            dest_idx.2.get_type().const_int(1, false),
+            "slice_dest_step_eq_one",
+        )
+        .unwrap();
+    let cond_1 = ctx.builder.build_and(dest_step_eq_one, src_slt_dest, "slice_cond_1").unwrap();
+    let cond = ctx.builder.build_or(src_eq_dest, cond_1, "slice_cond").unwrap();
+    ctx.make_assert(
+        generator,
+        cond,
+        "0:ValueError",
+        "attempt to assign sequence of size {0} to slice of size {1} with step size {2}",
+        [Some(src_slice_len), Some(dest_slice_len), Some(dest_idx.2)],
+        ctx.current_loc,
+    );
+
+    let new_len = {
+        let args = vec![
+            dest_idx.0.into(),   // dest start idx
+            dest_idx.1.into(),   // dest end idx
+            dest_idx.2.into(),   // dest step
+            dest_arr_ptr.into(), // dest arr ptr
+            dest_len.into(),     // dest arr len
+            src_idx.0.into(),    // src start idx
+            src_idx.1.into(),    // src end idx
+            src_idx.2.into(),    // src step
+            src_arr_ptr.into(),  // src arr ptr
+            src_len.into(),      // src arr len
+            {
+                let s = match ty {
+                    BasicTypeEnum::FloatType(t) => t.size_of(),
+                    BasicTypeEnum::IntType(t) => t.size_of(),
+                    BasicTypeEnum::PointerType(t) => t.size_of(),
+                    BasicTypeEnum::StructType(t) => t.size_of().unwrap(),
+                    _ => codegen_unreachable!(ctx),
+                };
+                ctx.builder.build_int_truncate_or_bit_cast(s, int32, "size").unwrap()
+            }
+            .into(),
+        ];
+        ctx.builder
+            .build_call(slice_assign_fun, args.as_slice(), "slice_assign")
+            .map(CallSiteValue::try_as_basic_value)
+            .map(|v| v.map_left(BasicValueEnum::into_int_value))
+            .map(Either::unwrap_left)
+            .unwrap()
+    };
+    // update length
+    let need_update =
+        ctx.builder.build_int_compare(IntPredicate::NE, new_len, dest_len, "need_update").unwrap();
+    let current = ctx.builder.get_insert_block().unwrap().get_parent().unwrap();
+    let update_bb = ctx.ctx.append_basic_block(current, "update");
+    let cont_bb = ctx.ctx.append_basic_block(current, "cont");
+    ctx.builder.build_conditional_branch(need_update, update_bb, cont_bb).unwrap();
+    ctx.builder.position_at_end(update_bb);
+    let new_len = ctx.builder.build_int_z_extend_or_bit_cast(new_len, size_ty, "new_len").unwrap();
+    dest_arr.store_size(ctx, generator, new_len);
+    ctx.builder.build_unconditional_branch(cont_bb).unwrap();
+    ctx.builder.position_at_end(cont_bb);
+}
+
+/// Generates a call to `isinf` in IR. Returns an `i1` representing the result.
+pub fn call_isinf<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &CodeGenContext<'ctx, '_>,
+    v: FloatValue<'ctx>,
+) -> IntValue<'ctx> {
+    let intrinsic_fn = ctx.module.get_function("__nac3_isinf").unwrap_or_else(|| {
+        let fn_type = ctx.ctx.i32_type().fn_type(&[ctx.ctx.f64_type().into()], false);
+        ctx.module.add_function("__nac3_isinf", fn_type, None)
+    });
+
+    let ret = ctx
+        .builder
+        .build_call(intrinsic_fn, &[v.into()], "isinf")
+        .map(CallSiteValue::try_as_basic_value)
+        .map(|v| v.map_left(BasicValueEnum::into_int_value))
+        .map(Either::unwrap_left)
+        .unwrap();
+
+    generator.bool_to_i1(ctx, ret)
+}
+
+/// Generates a call to `isnan` in IR. Returns an `i1` representing the result.
+pub fn call_isnan<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &CodeGenContext<'ctx, '_>,
+    v: FloatValue<'ctx>,
+) -> IntValue<'ctx> {
+    let intrinsic_fn = ctx.module.get_function("__nac3_isnan").unwrap_or_else(|| {
+        let fn_type = ctx.ctx.i32_type().fn_type(&[ctx.ctx.f64_type().into()], false);
+        ctx.module.add_function("__nac3_isnan", fn_type, None)
+    });
+
+    let ret = ctx
+        .builder
+        .build_call(intrinsic_fn, &[v.into()], "isnan")
+        .map(CallSiteValue::try_as_basic_value)
+        .map(|v| v.map_left(BasicValueEnum::into_int_value))
+        .map(Either::unwrap_left)
+        .unwrap();
+
+    generator.bool_to_i1(ctx, ret)
+}
+
+/// Generates a call to `gamma` in IR. Returns an `f64` representing the result.
+pub fn call_gamma<'ctx>(ctx: &CodeGenContext<'ctx, '_>, v: FloatValue<'ctx>) -> FloatValue<'ctx> {
+    let llvm_f64 = ctx.ctx.f64_type();
+
+    let intrinsic_fn = ctx.module.get_function("__nac3_gamma").unwrap_or_else(|| {
+        let fn_type = llvm_f64.fn_type(&[llvm_f64.into()], false);
+        ctx.module.add_function("__nac3_gamma", fn_type, None)
+    });
+
+    ctx.builder
+        .build_call(intrinsic_fn, &[v.into()], "gamma")
+        .map(CallSiteValue::try_as_basic_value)
+        .map(|v| v.map_left(BasicValueEnum::into_float_value))
+        .map(Either::unwrap_left)
+        .unwrap()
+}
+
+/// Generates a call to `gammaln` in IR. Returns an `f64` representing the result.
+pub fn call_gammaln<'ctx>(ctx: &CodeGenContext<'ctx, '_>, v: FloatValue<'ctx>) -> FloatValue<'ctx> {
+    let llvm_f64 = ctx.ctx.f64_type();
+
+    let intrinsic_fn = ctx.module.get_function("__nac3_gammaln").unwrap_or_else(|| {
+        let fn_type = llvm_f64.fn_type(&[llvm_f64.into()], false);
+        ctx.module.add_function("__nac3_gammaln", fn_type, None)
+    });
+
+    ctx.builder
+        .build_call(intrinsic_fn, &[v.into()], "gammaln")
+        .map(CallSiteValue::try_as_basic_value)
+        .map(|v| v.map_left(BasicValueEnum::into_float_value))
+        .map(Either::unwrap_left)
+        .unwrap()
+}
+
+/// Generates a call to `j0` in IR. Returns an `f64` representing the result.
+pub fn call_j0<'ctx>(ctx: &CodeGenContext<'ctx, '_>, v: FloatValue<'ctx>) -> FloatValue<'ctx> {
+    let llvm_f64 = ctx.ctx.f64_type();
+
+    let intrinsic_fn = ctx.module.get_function("__nac3_j0").unwrap_or_else(|| {
+        let fn_type = llvm_f64.fn_type(&[llvm_f64.into()], false);
+        ctx.module.add_function("__nac3_j0", fn_type, None)
+    });
+
+    ctx.builder
+        .build_call(intrinsic_fn, &[v.into()], "j0")
+        .map(CallSiteValue::try_as_basic_value)
+        .map(|v| v.map_left(BasicValueEnum::into_float_value))
+        .map(Either::unwrap_left)
+        .unwrap()
+}
+
+/// Generates a call to `__nac3_ndarray_calc_size`. Returns an [`IntValue`] representing the
+/// calculated total size.
+///
+/// * `dims` - An [`ArrayLikeIndexer`] containing the size of each dimension.
+/// * `range` - The dimension index to begin and end (exclusively) calculating the dimensions for,
+///   or [`None`] if starting from the first dimension and ending at the last dimension
+///   respectively.
+pub fn call_ndarray_calc_size<'ctx, G, Dims>(
+    generator: &G,
+    ctx: &CodeGenContext<'ctx, '_>,
+    dims: &Dims,
+    (begin, end): (Option<IntValue<'ctx>>, Option<IntValue<'ctx>>),
+) -> IntValue<'ctx>
+where
+    G: CodeGenerator + ?Sized,
+    Dims: ArrayLikeIndexer<'ctx>,
+{
+    let llvm_usize = generator.get_size_type(ctx.ctx);
+    let llvm_pusize = llvm_usize.ptr_type(AddressSpace::default());
+
+    let ndarray_calc_size_fn_name = match llvm_usize.get_bit_width() {
+        32 => "__nac3_ndarray_calc_size",
+        64 => "__nac3_ndarray_calc_size64",
+        bw => codegen_unreachable!(ctx, "Unsupported size type bit width: {}", bw),
+    };
+    let ndarray_calc_size_fn_t = llvm_usize.fn_type(
+        &[llvm_pusize.into(), llvm_usize.into(), llvm_usize.into(), llvm_usize.into()],
+        false,
+    );
+    let ndarray_calc_size_fn =
+        ctx.module.get_function(ndarray_calc_size_fn_name).unwrap_or_else(|| {
+            ctx.module.add_function(ndarray_calc_size_fn_name, ndarray_calc_size_fn_t, None)
+        });
+
+    let begin = begin.unwrap_or_else(|| llvm_usize.const_zero());
+    let end = end.unwrap_or_else(|| dims.size(ctx, generator));
+    ctx.builder
+        .build_call(
+            ndarray_calc_size_fn,
+            &[
+                dims.base_ptr(ctx, generator).into(),
+                dims.size(ctx, generator).into(),
+                begin.into(),
+                end.into(),
+            ],
+            "",
+        )
+        .map(CallSiteValue::try_as_basic_value)
+        .map(|v| v.map_left(BasicValueEnum::into_int_value))
+        .map(Either::unwrap_left)
+        .unwrap()
+}
+
+/// Generates a call to `__nac3_ndarray_calc_nd_indices`. Returns a [`TypeArrayLikeAdpater`]
+/// containing `i32` indices of the flattened index.
+///
+/// * `index` - The index to compute the multidimensional index for.
+/// * `ndarray` - LLVM pointer to the `NDArray`. This value must be the LLVM representation of an
+///   `NDArray`.
+pub fn call_ndarray_calc_nd_indices<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    index: IntValue<'ctx>,
+    ndarray: NDArrayValue<'ctx>,
+) -> TypedArrayLikeAdapter<'ctx, IntValue<'ctx>> {
+    let llvm_void = ctx.ctx.void_type();
+    let llvm_i32 = ctx.ctx.i32_type();
+    let llvm_usize = generator.get_size_type(ctx.ctx);
+    let llvm_pi32 = llvm_i32.ptr_type(AddressSpace::default());
+    let llvm_pusize = llvm_usize.ptr_type(AddressSpace::default());
+
+    let ndarray_calc_nd_indices_fn_name = match llvm_usize.get_bit_width() {
+        32 => "__nac3_ndarray_calc_nd_indices",
+        64 => "__nac3_ndarray_calc_nd_indices64",
+        bw => codegen_unreachable!(ctx, "Unsupported size type bit width: {}", bw),
+    };
+    let ndarray_calc_nd_indices_fn =
+        ctx.module.get_function(ndarray_calc_nd_indices_fn_name).unwrap_or_else(|| {
+            let fn_type = llvm_void.fn_type(
+                &[llvm_usize.into(), llvm_pusize.into(), llvm_usize.into(), llvm_pi32.into()],
+                false,
+            );
+
+            ctx.module.add_function(ndarray_calc_nd_indices_fn_name, fn_type, None)
+        });
+
+    let ndarray_num_dims = ndarray.load_ndims(ctx);
+    let ndarray_dims = ndarray.dim_sizes();
+
+    let indices = ctx.builder.build_array_alloca(llvm_i32, ndarray_num_dims, "").unwrap();
+
+    ctx.builder
+        .build_call(
+            ndarray_calc_nd_indices_fn,
+            &[
+                index.into(),
+                ndarray_dims.base_ptr(ctx, generator).into(),
+                ndarray_num_dims.into(),
+                indices.into(),
+            ],
+            "",
+        )
+        .unwrap();
+
+    TypedArrayLikeAdapter::from(
+        ArraySliceValue::from_ptr_val(indices, ndarray_num_dims, None),
+        Box::new(|_, v| v.into_int_value()),
+        Box::new(|_, v| v.into()),
+    )
+}
+
+fn call_ndarray_flatten_index_impl<'ctx, G, Indices>(
+    generator: &G,
+    ctx: &CodeGenContext<'ctx, '_>,
+    ndarray: NDArrayValue<'ctx>,
+    indices: &Indices,
+) -> IntValue<'ctx>
+where
+    G: CodeGenerator + ?Sized,
+    Indices: ArrayLikeIndexer<'ctx>,
+{
+    let llvm_i32 = ctx.ctx.i32_type();
+    let llvm_usize = generator.get_size_type(ctx.ctx);
+
+    let llvm_pi32 = llvm_i32.ptr_type(AddressSpace::default());
+    let llvm_pusize = llvm_usize.ptr_type(AddressSpace::default());
+
+    debug_assert_eq!(
+        IntType::try_from(indices.element_type(ctx, generator))
+            .map(IntType::get_bit_width)
+            .unwrap_or_default(),
+        llvm_i32.get_bit_width(),
+        "Expected i32 value for argument `indices` to `call_ndarray_flatten_index_impl`"
+    );
+    debug_assert_eq!(
+        indices.size(ctx, generator).get_type().get_bit_width(),
+        llvm_usize.get_bit_width(),
+        "Expected usize integer value for argument `indices_size` to `call_ndarray_flatten_index_impl`"
+    );
+
+    let ndarray_flatten_index_fn_name = match llvm_usize.get_bit_width() {
+        32 => "__nac3_ndarray_flatten_index",
+        64 => "__nac3_ndarray_flatten_index64",
+        bw => codegen_unreachable!(ctx, "Unsupported size type bit width: {}", bw),
+    };
+    let ndarray_flatten_index_fn =
+        ctx.module.get_function(ndarray_flatten_index_fn_name).unwrap_or_else(|| {
+            let fn_type = llvm_usize.fn_type(
+                &[llvm_pusize.into(), llvm_usize.into(), llvm_pi32.into(), llvm_usize.into()],
+                false,
+            );
+
+            ctx.module.add_function(ndarray_flatten_index_fn_name, fn_type, None)
+        });
+
+    let ndarray_num_dims = ndarray.load_ndims(ctx);
+    let ndarray_dims = ndarray.dim_sizes();
+
+    let index = ctx
+        .builder
+        .build_call(
+            ndarray_flatten_index_fn,
+            &[
+                ndarray_dims.base_ptr(ctx, generator).into(),
+                ndarray_num_dims.into(),
+                indices.base_ptr(ctx, generator).into(),
+                indices.size(ctx, generator).into(),
+            ],
+            "",
+        )
+        .map(CallSiteValue::try_as_basic_value)
+        .map(|v| v.map_left(BasicValueEnum::into_int_value))
+        .map(Either::unwrap_left)
+        .unwrap();
+
+    index
+}
+
+/// Generates a call to `__nac3_ndarray_flatten_index`. Returns the flattened index for the
+/// multidimensional index.
+///
+/// * `ndarray` - LLVM pointer to the `NDArray`. This value must be the LLVM representation of an
+///   `NDArray`.
+/// * `indices` - The multidimensional index to compute the flattened index for.
+pub fn call_ndarray_flatten_index<'ctx, G, Index>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    ndarray: NDArrayValue<'ctx>,
+    indices: &Index,
+) -> IntValue<'ctx>
+where
+    G: CodeGenerator + ?Sized,
+    Index: ArrayLikeIndexer<'ctx>,
+{
+    call_ndarray_flatten_index_impl(generator, ctx, ndarray, indices)
+}
+
+/// Generates a call to `__nac3_ndarray_calc_broadcast`. Returns a tuple containing the number of
+/// dimension and size of each dimension of the resultant `ndarray`.
+pub fn call_ndarray_calc_broadcast<'ctx, G: CodeGenerator + ?Sized>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    lhs: NDArrayValue<'ctx>,
+    rhs: NDArrayValue<'ctx>,
+) -> TypedArrayLikeAdapter<'ctx, IntValue<'ctx>> {
+    let llvm_usize = generator.get_size_type(ctx.ctx);
+    let llvm_pusize = llvm_usize.ptr_type(AddressSpace::default());
+
+    let ndarray_calc_broadcast_fn_name = match llvm_usize.get_bit_width() {
+        32 => "__nac3_ndarray_calc_broadcast",
+        64 => "__nac3_ndarray_calc_broadcast64",
+        bw => codegen_unreachable!(ctx, "Unsupported size type bit width: {}", bw),
+    };
+    let ndarray_calc_broadcast_fn =
+        ctx.module.get_function(ndarray_calc_broadcast_fn_name).unwrap_or_else(|| {
+            let fn_type = llvm_usize.fn_type(
+                &[
+                    llvm_pusize.into(),
+                    llvm_usize.into(),
+                    llvm_pusize.into(),
+                    llvm_usize.into(),
+                    llvm_pusize.into(),
+                ],
+                false,
+            );
+
+            ctx.module.add_function(ndarray_calc_broadcast_fn_name, fn_type, None)
+        });
+
+    let lhs_ndims = lhs.load_ndims(ctx);
+    let rhs_ndims = rhs.load_ndims(ctx);
+    let min_ndims = llvm_intrinsics::call_int_umin(ctx, lhs_ndims, rhs_ndims, None);
+
+    gen_for_callback_incrementing(
+        generator,
+        ctx,
+        None,
+        llvm_usize.const_zero(),
+        (min_ndims, false),
+        |generator, ctx, _, idx| {
+            let idx = ctx.builder.build_int_sub(min_ndims, idx, "").unwrap();
+            let (lhs_dim_sz, rhs_dim_sz) = unsafe {
+                (
+                    lhs.dim_sizes().get_typed_unchecked(ctx, generator, &idx, None),
+                    rhs.dim_sizes().get_typed_unchecked(ctx, generator, &idx, None),
+                )
+            };
+
+            let llvm_usize_const_one = llvm_usize.const_int(1, false);
+            let lhs_eqz = ctx
+                .builder
+                .build_int_compare(IntPredicate::EQ, lhs_dim_sz, llvm_usize_const_one, "")
+                .unwrap();
+            let rhs_eqz = ctx
+                .builder
+                .build_int_compare(IntPredicate::EQ, rhs_dim_sz, llvm_usize_const_one, "")
+                .unwrap();
+            let lhs_or_rhs_eqz = ctx.builder.build_or(lhs_eqz, rhs_eqz, "").unwrap();
+
+            let lhs_eq_rhs = ctx
+                .builder
+                .build_int_compare(IntPredicate::EQ, lhs_dim_sz, rhs_dim_sz, "")
+                .unwrap();
+
+            let is_compatible = ctx.builder.build_or(lhs_or_rhs_eqz, lhs_eq_rhs, "").unwrap();
+
+            ctx.make_assert(
+                generator,
+                is_compatible,
+                "0:ValueError",
+                "operands could not be broadcast together",
+                [None, None, None],
+                ctx.current_loc,
+            );
+
+            Ok(())
+        },
+        llvm_usize.const_int(1, false),
+    )
+    .unwrap();
+
+    let max_ndims = llvm_intrinsics::call_int_umax(ctx, lhs_ndims, rhs_ndims, None);
+    let lhs_dims = lhs.dim_sizes().base_ptr(ctx, generator);
+    let lhs_ndims = lhs.load_ndims(ctx);
+    let rhs_dims = rhs.dim_sizes().base_ptr(ctx, generator);
+    let rhs_ndims = rhs.load_ndims(ctx);
+    let out_dims = ctx.builder.build_array_alloca(llvm_usize, max_ndims, "").unwrap();
+    let out_dims = ArraySliceValue::from_ptr_val(out_dims, max_ndims, None);
+
+    ctx.builder
+        .build_call(
+            ndarray_calc_broadcast_fn,
+            &[
+                lhs_dims.into(),
+                lhs_ndims.into(),
+                rhs_dims.into(),
+                rhs_ndims.into(),
+                out_dims.base_ptr(ctx, generator).into(),
+            ],
+            "",
+        )
+        .unwrap();
+
+    TypedArrayLikeAdapter::from(
+        out_dims,
+        Box::new(|_, v| v.into_int_value()),
+        Box::new(|_, v| v.into()),
+    )
+}
+
+/// Generates a call to `__nac3_ndarray_calc_broadcast_idx`. Returns an [`ArrayAllocaValue`]
+/// containing the indices used for accessing `array` corresponding to the index of the broadcasted
+/// array `broadcast_idx`.
+pub fn call_ndarray_calc_broadcast_index<
+    'ctx,
+    G: CodeGenerator + ?Sized,
+    BroadcastIdx: UntypedArrayLikeAccessor<'ctx>,
+>(
+    generator: &mut G,
+    ctx: &mut CodeGenContext<'ctx, '_>,
+    array: NDArrayValue<'ctx>,
+    broadcast_idx: &BroadcastIdx,
+) -> TypedArrayLikeAdapter<'ctx, IntValue<'ctx>> {
+    let llvm_i32 = ctx.ctx.i32_type();
+    let llvm_usize = generator.get_size_type(ctx.ctx);
+    let llvm_pi32 = llvm_i32.ptr_type(AddressSpace::default());
+    let llvm_pusize = llvm_usize.ptr_type(AddressSpace::default());
+
+    let ndarray_calc_broadcast_fn_name = match llvm_usize.get_bit_width() {
+        32 => "__nac3_ndarray_calc_broadcast_idx",
+        64 => "__nac3_ndarray_calc_broadcast_idx64",
+        bw => codegen_unreachable!(ctx, "Unsupported size type bit width: {}", bw),
+    };
+    let ndarray_calc_broadcast_fn =
+        ctx.module.get_function(ndarray_calc_broadcast_fn_name).unwrap_or_else(|| {
+            let fn_type = llvm_usize.fn_type(
+                &[llvm_pusize.into(), llvm_usize.into(), llvm_pi32.into(), llvm_pi32.into()],
+                false,
+            );
+
+            ctx.module.add_function(ndarray_calc_broadcast_fn_name, fn_type, None)
+        });
+
+    let broadcast_size = broadcast_idx.size(ctx, generator);
+    let out_idx = ctx.builder.build_array_alloca(llvm_i32, broadcast_size, "").unwrap();
+
+    let array_dims = array.dim_sizes().base_ptr(ctx, generator);
+    let array_ndims = array.load_ndims(ctx);
+    let broadcast_idx_ptr = unsafe {
+        broadcast_idx.ptr_offset_unchecked(ctx, generator, &llvm_usize.const_zero(), None)
+    };
+
+    ctx.builder
+        .build_call(
+            ndarray_calc_broadcast_fn,
+            &[array_dims.into(), array_ndims.into(), broadcast_idx_ptr.into(), out_idx.into()],
+            "",
+        )
+        .unwrap();
+
+    TypedArrayLikeAdapter::from(
+        ArraySliceValue::from_ptr_val(out_idx, broadcast_size, None),
+        Box::new(|_, v| v.into_int_value()),
+        Box::new(|_, v| v.into()),
+    )
+}

nac3core/src/codegen/llvm_intrinsics.rs

@@ -0,0 +1,345 @@
+use inkwell::{
+    context::Context,
+    intrinsics::Intrinsic,
+    types::{AnyTypeEnum::IntType, FloatType},
+    values::{BasicValueEnum, CallSiteValue, FloatValue, IntValue, PointerValue},
+    AddressSpace,
+};
+use itertools::Either;
+
+use crate::codegen::CodeGenContext;
+
+/// Returns the string representation for the floating-point type `ft` when used in intrinsic
+/// functions.
+fn get_float_intrinsic_repr(ctx: &Context, ft: FloatType) -> &'static str {
+    // Standard LLVM floating-point types
+    if ft == ctx.f16_type() {
+        return "f16";
+    }
+    if ft == ctx.f32_type() {
+        return "f32";
+    }
+    if ft == ctx.f64_type() {
+        return "f64";
+    }
+    if ft == ctx.f128_type() {
+        return "f128";
+    }
+
+    // Non-standard floating-point types
+    if ft == ctx.x86_f80_type() {
+        return "f80";
+    }
+    if ft == ctx.ppc_f128_type() {
+        return "ppcf128";
+    }
+
+    unreachable!()
+}
+
+/// Invokes the [`llvm.va_start`](https://llvm.org/docs/LangRef.html#llvm-va-start-intrinsic)
+/// intrinsic.
+pub fn call_va_start<'ctx>(ctx: &CodeGenContext<'ctx, '_>, arglist: PointerValue<'ctx>) {
+    const FN_NAME: &str = "llvm.va_start";
+
+    let intrinsic_fn = ctx.module.get_function(FN_NAME).unwrap_or_else(|| {
+        let llvm_void = ctx.ctx.void_type();
+        let llvm_i8 = ctx.ctx.i8_type();
+        let llvm_p0i8 = llvm_i8.ptr_type(AddressSpace::default());
+        let fn_type = llvm_void.fn_type(&[llvm_p0i8.into()], false);
+
+        ctx.module.add_function(FN_NAME, fn_type, None)
+    });
+
+    ctx.builder.build_call(intrinsic_fn, &[arglist.into()], "").unwrap();
+}
+
+/// Invokes the [`llvm.va_start`](https://llvm.org/docs/LangRef.html#llvm-va-start-intrinsic)
+/// intrinsic.
+pub fn call_va_end<'ctx>(ctx: &CodeGenContext<'ctx, '_>, arglist: PointerValue<'ctx>) {
+    const FN_NAME: &str = "llvm.va_end";
+
+    let intrinsic_fn = ctx.module.get_function(FN_NAME).unwrap_or_else(|| {
+        let llvm_void = ctx.ctx.void_type();
+        let llvm_i8 = ctx.ctx.i8_type();
+        let llvm_p0i8 = llvm_i8.ptr_type(AddressSpace::default());
+        let fn_type = llvm_void.fn_type(&[llvm_p0i8.into()], false);
+
+        ctx.module.add_function(FN_NAME, fn_type, None)
+    });
+
+    ctx.builder.build_call(intrinsic_fn, &[arglist.into()], "").unwrap();
+}
+
+/// Invokes the [`llvm.stacksave`](https://llvm.org/docs/LangRef.html#llvm-stacksave-intrinsic)
+/// intrinsic.
+pub fn call_stacksave<'ctx>(
+    ctx: &CodeGenContext<'ctx, '_>,
+    name: Option<&str>,
+) -> PointerValue<'ctx> {
+    const FN_NAME: &str = "llvm.stacksave";
+
+    let intrinsic_fn = Intrinsic::find(FN_NAME)
+        .and_then(|intrinsic| intrinsic.get_declaration(&ctx.module, &[]))
+        .unwrap();
+
+    ctx.builder
+        .build_call(intrinsic_fn, &[], name.unwrap_or_default())
+        .map(CallSiteValue::try_as_basic_value)
+        .map(|v| v.map_left(BasicValueEnum::into_pointer_value))
+        .map(Either::unwrap_left)
+        .unwrap()
+}
+
+/// Invokes the
+/// [`llvm.stackrestore`](https://llvm.org/docs/LangRef.html#llvm-stackrestore-intrinsic) intrinsic.
+///
+/// - `ptr`: The pointer storing the address to restore the stack to.
+pub fn call_stackrestore<'ctx>(ctx: &CodeGenContext<'ctx, '_>, ptr: PointerValue<'ctx>) {
+    const FN_NAME: &str = "llvm.stackrestore";
+
+    /*
+        SEE https://github.com/TheDan64/inkwell/issues/496
+
+        We want `llvm.stackrestore`, but the following would generate `llvm.stackrestore.p0i8`.
+        ```ignore
+        let intrinsic_fn = Intrinsic::find(FN_NAME)
+            .and_then(|intrinsic| intrinsic.get_declaration(&ctx.module, &[llvm_p0i8.into()]))
+            .unwrap();
+        ```
+
+        Temp workaround by manually declaring the intrinsic with the correct function name instead.
+    */
+    let intrinsic_fn = ctx.module.get_function(FN_NAME).unwrap_or_else(|| {
+        let llvm_void = ctx.ctx.void_type();
+        let llvm_i8 = ctx.ctx.i8_type();
+        let llvm_p0i8 = llvm_i8.ptr_type(AddressSpace::default());
+        let fn_type = llvm_void.fn_type(&[llvm_p0i8.into()], false);
+
+        ctx.module.add_function(FN_NAME, fn_type, None)
+    });
+
+    ctx.builder.build_call(intrinsic_fn, &[ptr.into()], "").unwrap();
+}
+
+/// Invokes the [`llvm.memcpy`](https://llvm.org/docs/LangRef.html#llvm-memcpy-intrinsic) intrinsic.
+///
+/// * `dest` - The pointer to the destination. Must be a pointer to an integer type.
+/// * `src` - The pointer to the source. Must be a pointer to an integer type.
+/// * `len` - The number of bytes to copy.
+/// * `is_volatile` - Whether the `memcpy` operation should be `volatile`.
+pub fn call_memcpy<'ctx>(
+    ctx: &CodeGenContext<'ctx, '_>,
+    dest: PointerValue<'ctx>,
+    src: PointerValue<'ctx>,
+    len: IntValue<'ctx>,
+    is_volatile: IntValue<'ctx>,
+) {
+    const FN_NAME: &str = "llvm.memcpy";
+
+    debug_assert!(dest.get_type().get_element_type().is_int_type());
+    debug_assert!(src.get_type().get_element_type().is_int_type());
+    debug_assert_eq!(
+        dest.get_type().get_element_type().into_int_type().get_bit_width(),
+        src.get_type().get_element_type().into_int_type().get_bit_width(),
+    );
+    debug_assert!(matches!(len.get_type().get_bit_width(), 32 | 64));
+    debug_assert_eq!(is_volatile.get_type().get_bit_width(), 1);
+
+    let llvm_dest_t = dest.get_type();
+    let llvm_src_t = src.get_type();
+    let llvm_len_t = len.get_type();
+
+    let intrinsic_fn = Intrinsic::find(FN_NAME)
+        .and_then(|intrinsic| {
+            intrinsic.get_declaration(
+                &ctx.module,
+                &[llvm_dest_t.into(), llvm_src_t.into(), llvm_len_t.into()],
+            )
+        })
+        .unwrap();
+
+    ctx.builder
+        .build_call(intrinsic_fn, &[dest.into(), src.into(), len.into(), is_volatile.into()], "")
+        .unwrap();
+}
+
+/// Invokes the `llvm.memcpy` intrinsic.
+///
+/// Unlike [`call_memcpy`], this function accepts any type of pointer value. If `dest` or `src` is
+/// not a pointer to an integer, the pointer(s) will be cast to `i8*` before invoking `memcpy`.
+pub fn call_memcpy_generic<'ctx>(
+    ctx: &CodeGenContext<'ctx, '_>,
+    dest: PointerValue<'ctx>,
+    src: PointerValue<'ctx>,
+    len: IntValue<'ctx>,
+    is_volatile: IntValue<'ctx>,
+) {
+    let llvm_i8 = ctx.ctx.i8_type();
+    let llvm_p0i8 = llvm_i8.ptr_type(AddressSpace::default());
+
+    let dest_elem_t = dest.get_type().get_element_type();
+    let src_elem_t = src.get_type().get_element_type();
+
+    let dest = if matches!(dest_elem_t, IntType(t) if t.get_bit_width() == 8) {
+        dest
+    } else {
+        ctx.builder
+            .build_bit_cast(dest, llvm_p0i8, "")
+            .map(BasicValueEnum::into_pointer_value)
+            .unwrap()
+    };
+    let src = if matches!(src_elem_t, IntType(t) if t.get_bit_width() == 8) {
+        src
+    } else {
+        ctx.builder
+            .build_bit_cast(src, llvm_p0i8, "")
+            .map(BasicValueEnum::into_pointer_value)
+            .unwrap()
+    };
+
+    call_memcpy(ctx, dest, src, len, is_volatile);
+}
+
+/// Macro to find and generate build call for llvm intrinsic (body of llvm intrinsic function)
+///
+/// Arguments:
+/// * `$ctx:ident`: Reference to the current Code Generation Context
+/// * `$name:ident`: Optional name to be assigned to the llvm build call (Option<&str>)
+/// * `$llvm_name:literal`: Name of underlying llvm intrinsic function
+/// * `$map_fn:ident`: Mapping function to be applied on `BasicValue` (`BasicValue` -> Function Return Type).
+///   Use `BasicValueEnum::into_int_value` for Integer return type and
+///   `BasicValueEnum::into_float_value` for Float return type
+/// * `$llvm_ty:ident`: Type of first operand
+/// * `,($val:ident)*`: Comma separated list of operands
+macro_rules! generate_llvm_intrinsic_fn_body {
+    ($ctx:ident, $name:ident, $llvm_name:literal, $map_fn:expr, $llvm_ty:ident $(,$val:ident)*) => {{
+        const FN_NAME: &str = concat!("llvm.", $llvm_name);
+        let intrinsic_fn = Intrinsic::find(FN_NAME).and_then(|intrinsic| intrinsic.get_declaration(&$ctx.module, &[$llvm_ty.into()])).unwrap();
+        $ctx.builder.build_call(intrinsic_fn, &[$($val.into()),*],  $name.unwrap_or_default()).map(CallSiteValue::try_as_basic_value).map(|v| v.map_left($map_fn)).map(Either::unwrap_left).unwrap()
+    }};
+}
+
+/// Macro to generate the llvm intrinsic function using [`generate_llvm_intrinsic_fn_body`].
+///
+/// Arguments:
+/// * `float/int`: Indicates the return and argument type of the function
+/// * `$fn_name:ident`: The identifier of the rust function to be generated
+/// * `$llvm_name:literal`: Name of underlying llvm intrinsic function.
+///   Omit "llvm." prefix from the function name i.e. use "ceil" instead of "llvm.ceil"
+/// * `$val:ident`: The operand for unary operations
+/// * `$val1:ident`, `$val2:ident`: The operands for binary operations
+macro_rules! generate_llvm_intrinsic_fn {
+    ("float", $fn_name:ident, $llvm_name:literal, $val:ident) => {
+        #[doc = concat!("Invokes the [`", stringify!($llvm_name), "`](https://llvm.org/docs/LangRef.html#llvm-", stringify!($llvm_name), "-intrinsic) intrinsic." )]
+        pub fn $fn_name<'ctx> (
+            ctx: &CodeGenContext<'ctx, '_>,
+            $val: FloatValue<'ctx>,
+            name: Option<&str>,
+        ) -> FloatValue<'ctx> {
+            let llvm_ty = $val.get_type();
+            generate_llvm_intrinsic_fn_body!(ctx, name, $llvm_name, BasicValueEnum::into_float_value, llvm_ty, $val)
+        }
+    };
+    ("float", $fn_name:ident, $llvm_name:literal, $val1:ident, $val2:ident) => {
+        #[doc = concat!("Invokes the [`", stringify!($llvm_name), "`](https://llvm.org/docs/LangRef.html#llvm-", stringify!($llvm_name), "-intrinsic) intrinsic." )]
+        pub fn $fn_name<'ctx> (
+            ctx: &CodeGenContext<'ctx, '_>,
+            $val1: FloatValue<'ctx>,
+            $val2: FloatValue<'ctx>,
+            name: Option<&str>,
+        ) -> FloatValue<'ctx> {
+            debug_assert_eq!($val1.get_type(), $val2.get_type());
+            let llvm_ty = $val1.get_type();
+            generate_llvm_intrinsic_fn_body!(ctx, name, $llvm_name, BasicValueEnum::into_float_value, llvm_ty, $val1, $val2)
+        }
+    };
+    ("int", $fn_name:ident, $llvm_name:literal, $val1:ident, $val2:ident) => {
+        #[doc = concat!("Invokes the [`", stringify!($llvm_name), "`](https://llvm.org/docs/LangRef.html#llvm-", stringify!($llvm_name), "-intrinsic) intrinsic." )]
+        pub fn $fn_name<'ctx> (
+            ctx: &CodeGenContext<'ctx, '_>,
+            $val1: IntValue<'ctx>,
+            $val2: IntValue<'ctx>,
+            name: Option<&str>,
+        ) -> IntValue<'ctx> {
+            debug_assert_eq!($val1.get_type().get_bit_width(), $val2.get_type().get_bit_width());
+            let llvm_ty = $val1.get_type();
+            generate_llvm_intrinsic_fn_body!(ctx, name, $llvm_name, BasicValueEnum::into_int_value, llvm_ty, $val1, $val2)
+        }
+    };
+}
+
+/// Invokes the [`llvm.abs`](https://llvm.org/docs/LangRef.html#llvm-abs-intrinsic) intrinsic.
+///
+/// * `src` - The value for which the absolute value is to be returned.
+/// * `is_int_min_poison` - Whether `poison` is to be returned if `src` is `INT_MIN`.
+pub fn call_int_abs<'ctx>(
+    ctx: &CodeGenContext<'ctx, '_>,
+    src: IntValue<'ctx>,
+    is_int_min_poison: IntValue<'ctx>,
+    name: Option<&str>,
+) -> IntValue<'ctx> {
+    debug_assert_eq!(is_int_min_poison.get_type().get_bit_width(), 1);
+    debug_assert!(is_int_min_poison.is_const());
+
+    let src_type = src.get_type();
+    generate_llvm_intrinsic_fn_body!(
+        ctx,
+        name,
+        "abs",
+        BasicValueEnum::into_int_value,
+        src_type,
+        src,
+        is_int_min_poison
+    )
+}
+
+generate_llvm_intrinsic_fn!("int", call_int_smax, "smax", a, b);
+generate_llvm_intrinsic_fn!("int", call_int_smin, "smin", a, b);
+generate_llvm_intrinsic_fn!("int", call_int_umax, "umax", a, b);
+generate_llvm_intrinsic_fn!("int", call_int_umin, "umin", a, b);
+generate_llvm_intrinsic_fn!("int", call_expect, "expect", val, expected_val);
+
+generate_llvm_intrinsic_fn!("float", call_float_sqrt, "sqrt", val);
+generate_llvm_intrinsic_fn!("float", call_float_sin, "sin", val);
+generate_llvm_intrinsic_fn!("float", call_float_cos, "cos", val);
+generate_llvm_intrinsic_fn!("float", call_float_pow, "pow", val, power);
+generate_llvm_intrinsic_fn!("float", call_float_exp, "exp", val);
+generate_llvm_intrinsic_fn!("float", call_float_exp2, "exp2", val);
+generate_llvm_intrinsic_fn!("float", call_float_log, "log", val);
+generate_llvm_intrinsic_fn!("float", call_float_log10, "log10", val);
+generate_llvm_intrinsic_fn!("float", call_float_log2, "log2", val);
+generate_llvm_intrinsic_fn!("float", call_float_fabs, "fabs", src);
+generate_llvm_intrinsic_fn!("float", call_float_minnum, "minnum", val, power);
+generate_llvm_intrinsic_fn!("float", call_float_maxnum, "maxnum", val, power);
+generate_llvm_intrinsic_fn!("float", call_float_copysign, "copysign", mag, sgn);
+generate_llvm_intrinsic_fn!("float", call_float_floor, "floor", val);
+generate_llvm_intrinsic_fn!("float", call_float_ceil, "ceil", val);
+generate_llvm_intrinsic_fn!("float", call_float_round, "round", val);
+generate_llvm_intrinsic_fn!("float", call_float_rint, "rint", val);
+
+/// Invokes the [`llvm.powi`](https://llvm.org/docs/LangRef.html#llvm-powi-intrinsic) intrinsic.
+pub fn call_float_powi<'ctx>(
+    ctx: &CodeGenContext<'ctx, '_>,
+    val: FloatValue<'ctx>,
+    power: IntValue<'ctx>,
+    name: Option<&str>,
+) -> FloatValue<'ctx> {
+    const FN_NAME: &str = "llvm.powi";
+
+    let llvm_val_t = val.get_type();
+    let llvm_power_t = power.get_type();
+
+    let intrinsic_fn = Intrinsic::find(FN_NAME)
+        .and_then(|intrinsic| {
+            intrinsic.get_declaration(&ctx.module, &[llvm_val_t.into(), llvm_power_t.into()])
+        })
+        .unwrap();
+
+    ctx.builder
+        .build_call(intrinsic_fn, &[val.into(), power.into()], name.unwrap_or_default())
+        .map(CallSiteValue::try_as_basic_value)
+        .map(|v| v.map_left(BasicValueEnum::into_float_value))
+        .map(Either::unwrap_left)
+        .unwrap()
+}

nac3core/src/codegen/test.rs

@@ -0,0 +1,476 @@
+use std::{
+    collections::{HashMap, HashSet},
+    sync::Arc,
+};
+
+use indexmap::IndexMap;
+use indoc::indoc;
+use inkwell::{
+    targets::{InitializationConfig, Target},
+    OptimizationLevel,
+};
+use nac3parser::{
+    ast::{fold::Fold, FileName, StrRef},
+    parser::parse_program,
+};
+use parking_lot::RwLock;
+
+use crate::{
+    codegen::{
+        classes::{ListType, NDArrayType, ProxyType, RangeType},
+        concrete_type::ConcreteTypeStore,
+        CodeGenContext, CodeGenLLVMOptions, CodeGenTargetMachineOptions, CodeGenTask,
+        CodeGenerator, DefaultCodeGenerator, WithCall, WorkerRegistry,
+    },
+    symbol_resolver::{SymbolResolver, ValueEnum},
+    toplevel::{
+        composer::{ComposerConfig, TopLevelComposer},
+        DefinitionId, FunInstance, TopLevelContext, TopLevelDef,
+    },
+    typecheck::{
+        type_inferencer::{FunctionData, IdentifierInfo, Inferencer, PrimitiveStore},
+        typedef::{FunSignature, FuncArg, Type, TypeEnum, Unifier, VarMap},
+    },
+};
+
+struct Resolver {
+    id_to_type: HashMap<StrRef, Type>,
+    id_to_def: RwLock<HashMap<StrRef, DefinitionId>>,
+    class_names: HashMap<StrRef, Type>,
+}
+
+impl Resolver {
+    pub fn add_id_def(&self, id: StrRef, def: DefinitionId) {
+        self.id_to_def.write().insert(id, def);
+    }
+}
+
+impl SymbolResolver for Resolver {
+    fn get_default_param_value(
+        &self,
+        _: &nac3parser::ast::Expr,
+    ) -> Option<crate::symbol_resolver::SymbolValue> {
+        unimplemented!()
+    }
+
+    fn get_symbol_type(
+        &self,
+        _: &mut Unifier,
+        _: &[Arc<RwLock<TopLevelDef>>],
+        _: &PrimitiveStore,
+        str: StrRef,
+    ) -> Result<Type, String> {
+        self.id_to_type.get(&str).copied().ok_or_else(|| format!("cannot find symbol `{str}`"))
+    }
+
+    fn get_symbol_value<'ctx>(
+        &self,
+        _: StrRef,
+        _: &mut CodeGenContext<'ctx, '_>,
+        _: &mut dyn CodeGenerator,
+    ) -> Option<ValueEnum<'ctx>> {
+        unimplemented!()
+    }
+
+    fn get_identifier_def(&self, id: StrRef) -> Result<DefinitionId, HashSet<String>> {
+        self.id_to_def
+            .read()
+            .get(&id)
+            .copied()
+            .ok_or_else(|| HashSet::from([format!("cannot find symbol `{id}`")]))
+    }
+
+    fn get_string_id(&self, _: &str) -> i32 {
+        unimplemented!()
+    }
+
+    fn get_exception_id(&self, _tyid: usize) -> usize {
+        unimplemented!()
+    }
+}
+
+#[test]
+fn test_primitives() {
+    let source = indoc! { "
+        c = a + b
+        d = a if c == 1 else 0
+        return d
+        "};
+    let statements = parse_program(source, FileName::default()).unwrap();
+
+    let composer = TopLevelComposer::new(Vec::new(), Vec::new(), ComposerConfig::default(), 32).0;
+    let mut unifier = composer.unifier.clone();
+    let primitives = composer.primitives_ty;
+    let top_level = Arc::new(composer.make_top_level_context());
+    unifier.top_level = Some(top_level.clone());
+
+    let resolver = Arc::new(Resolver {
+        id_to_type: HashMap::new(),
+        id_to_def: RwLock::new(HashMap::new()),
+        class_names: HashMap::default(),
+    }) as Arc<dyn SymbolResolver + Send + Sync>;
+
+    let threads = vec![DefaultCodeGenerator::new("test".into(), 32).into()];
+    let signature = FunSignature {
+        args: vec![
+            FuncArg {
+                name: "a".into(),
+                ty: primitives.int32,
+                default_value: None,
+                is_vararg: false,
+            },
+            FuncArg {
+                name: "b".into(),
+                ty: primitives.int32,
+                default_value: None,
+                is_vararg: false,
+            },
+        ],
+        ret: primitives.int32,
+        vars: VarMap::new(),
+    };
+
+    let mut store = ConcreteTypeStore::new();
+    let mut cache = HashMap::new();
+    let signature = store.from_signature(&mut unifier, &primitives, &signature, &mut cache);
+    let signature = store.add_cty(signature);
+
+    let mut function_data = FunctionData {
+        resolver: resolver.clone(),
+        bound_variables: Vec::new(),
+        return_type: Some(primitives.int32),
+    };
+    let mut virtual_checks = Vec::new();
+    let mut calls = HashMap::new();
+    let mut identifiers: HashMap<_, _> =
+        ["a".into(), "b".into()].map(|id| (id, IdentifierInfo::default())).into();
+    let mut inferencer = Inferencer {
+        top_level: &top_level,
+        function_data: &mut function_data,
+        unifier: &mut unifier,
+        variable_mapping: HashMap::default(),
+        primitives: &primitives,
+        virtual_checks: &mut virtual_checks,
+        calls: &mut calls,
+        defined_identifiers: identifiers.clone(),
+        in_handler: false,
+    };
+    inferencer.variable_mapping.insert("a".into(), inferencer.primitives.int32);
+    inferencer.variable_mapping.insert("b".into(), inferencer.primitives.int32);
+
+    let statements = statements
+        .into_iter()
+        .map(|v| inferencer.fold_stmt(v))
+        .collect::<Result<Vec<_>, _>>()
+        .unwrap();
+
+    inferencer.check_block(&statements, &mut identifiers).unwrap();
+    let top_level = Arc::new(TopLevelContext {
+        definitions: Arc::new(RwLock::new(std::mem::take(&mut *top_level.definitions.write()))),
+        unifiers: Arc::new(RwLock::new(vec![(unifier.get_shared_unifier(), primitives)])),
+        personality_symbol: None,
+    });
+
+    let task = CodeGenTask {
+        subst: Vec::default(),
+        symbol_name: "testing".into(),
+        body: Arc::new(statements),
+        unifier_index: 0,
+        calls: Arc::new(calls),
+        resolver,
+        store,
+        signature,
+        id: 0,
+    };
+    let f = Arc::new(WithCall::new(Box::new(|module| {
+        // the following IR is equivalent to
+        // ```
+        // ; ModuleID = 'test.ll'
+        // source_filename = "test"
+        //
+        // ; Function Attrs: norecurse nounwind readnone
+        // define i32 @testing(i32 %0, i32 %1) local_unnamed_addr #0 {
+        // init:
+        //   %add = add i32 %1, %0
+        //   %cmp = icmp eq i32 %add, 1
+        //   %ifexpr = select i1 %cmp, i32 %0, i32 0
+        //   ret i32 %ifexpr
+        // }
+        //
+        // attributes #0 = { norecurse nounwind readnone }
+        // ```
+        // after O2 optimization
+
+        let expected = indoc! {"
+            ; ModuleID = 'test'
+            source_filename = \"test\"
+            target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128\"
+            target triple = \"x86_64-unknown-linux-gnu\"
+
+            ; Function Attrs: mustprogress nofree norecurse nosync nounwind readnone willreturn
+            define i32 @testing(i32 %0, i32 %1) local_unnamed_addr #0 !dbg !4 {
+            init:
+              %add = add i32 %1, %0, !dbg !9
+              %cmp = icmp eq i32 %add, 1, !dbg !10
+              %. = select i1 %cmp, i32 %0, i32 0, !dbg !11
+              ret i32 %., !dbg !12
+            }
+
+            attributes #0 = { mustprogress nofree norecurse nosync nounwind readnone willreturn }
+
+            !llvm.module.flags = !{!0, !1}
+            !llvm.dbg.cu = !{!2}
+
+            !0 = !{i32 2, !\"Debug Info Version\", i32 3}
+            !1 = !{i32 2, !\"Dwarf Version\", i32 4}
+            !2 = distinct !DICompileUnit(language: DW_LANG_Python, file: !3, producer: \"NAC3\", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug)
+            !3 = !DIFile(filename: \"unknown\", directory: \"\")
+            !4 = distinct !DISubprogram(name: \"testing\", linkageName: \"testing\", scope: null, file: !3, line: 1, type: !5, scopeLine: 1, flags: DIFlagPublic, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !2, retainedNodes: !8)
+            !5 = !DISubroutineType(flags: DIFlagPublic, types: !6)
+            !6 = !{!7}
+            !7 = !DIBasicType(name: \"_\", flags: DIFlagPublic)
+            !8 = !{}
+            !9 = !DILocation(line: 1, column: 9, scope: !4)
+            !10 = !DILocation(line: 2, column: 15, scope: !4)
+            !11 = !DILocation(line: 0, scope: !4)
+            !12 = !DILocation(line: 3, column: 8, scope: !4)
+        "}
+        .trim();
+        assert_eq!(expected, module.print_to_string().to_str().unwrap().trim());
+    })));
+
+    Target::initialize_all(&InitializationConfig::default());
+
+    let llvm_options = CodeGenLLVMOptions {
+        opt_level: OptimizationLevel::Default,
+        target: CodeGenTargetMachineOptions::from_host_triple(),
+    };
+    let (registry, handles) = WorkerRegistry::create_workers(threads, top_level, &llvm_options, &f);
+    registry.add_task(task);
+    registry.wait_tasks_complete(handles);
+}
+
+#[test]
+fn test_simple_call() {
+    let source_1 = indoc! { "
+        a = foo(a)
+        return a * 2
+        "};
+    let statements_1 = parse_program(source_1, FileName::default()).unwrap();
+
+    let source_2 = indoc! { "
+        return a + 1
+        "};
+    let statements_2 = parse_program(source_2, FileName::default()).unwrap();
+
+    let composer = TopLevelComposer::new(Vec::new(), Vec::new(), ComposerConfig::default(), 32).0;
+    let mut unifier = composer.unifier.clone();
+    let primitives = composer.primitives_ty;
+    let top_level = Arc::new(composer.make_top_level_context());
+    unifier.top_level = Some(top_level.clone());
+
+    let signature = FunSignature {
+        args: vec![FuncArg {
+            name: "a".into(),
+            ty: primitives.int32,
+            default_value: None,
+            is_vararg: false,
+        }],
+        ret: primitives.int32,
+        vars: VarMap::new(),
+    };
+    let fun_ty = unifier.add_ty(TypeEnum::TFunc(signature.clone()));
+    let mut store = ConcreteTypeStore::new();
+    let mut cache = HashMap::new();
+    let signature = store.from_signature(&mut unifier, &primitives, &signature, &mut cache);
+    let signature = store.add_cty(signature);
+
+    let foo_id = top_level.definitions.read().len();
+    top_level.definitions.write().push(Arc::new(RwLock::new(TopLevelDef::Function {
+        name: "foo".to_string(),
+        simple_name: "foo".into(),
+        signature: fun_ty,
+        var_id: vec![],
+        instance_to_stmt: HashMap::new(),
+        instance_to_symbol: HashMap::new(),
+        resolver: None,
+        codegen_callback: None,
+        loc: None,
+    })));
+
+    let resolver = Resolver {
+        id_to_type: HashMap::new(),
+        id_to_def: RwLock::new(HashMap::new()),
+        class_names: HashMap::default(),
+    };
+    resolver.add_id_def("foo".into(), DefinitionId(foo_id));
+    let resolver = Arc::new(resolver) as Arc<dyn SymbolResolver + Send + Sync>;
+
+    if let TopLevelDef::Function { resolver: r, .. } =
+        &mut *top_level.definitions.read()[foo_id].write()
+    {
+        *r = Some(resolver.clone());
+    } else {
+        unreachable!()
+    }
+
+    let threads = vec![DefaultCodeGenerator::new("test".into(), 32).into()];
+    let mut function_data = FunctionData {
+        resolver: resolver.clone(),
+        bound_variables: Vec::new(),
+        return_type: Some(primitives.int32),
+    };
+    let mut virtual_checks = Vec::new();
+    let mut calls = HashMap::new();
+    let mut identifiers: HashMap<_, _> =
+        ["a".into(), "foo".into()].map(|id| (id, IdentifierInfo::default())).into();
+    let mut inferencer = Inferencer {
+        top_level: &top_level,
+        function_data: &mut function_data,
+        unifier: &mut unifier,
+        variable_mapping: HashMap::default(),
+        primitives: &primitives,
+        virtual_checks: &mut virtual_checks,
+        calls: &mut calls,
+        defined_identifiers: identifiers.clone(),
+        in_handler: false,
+    };
+    inferencer.variable_mapping.insert("a".into(), inferencer.primitives.int32);
+    inferencer.variable_mapping.insert("foo".into(), fun_ty);
+
+    let statements_1 = statements_1
+        .into_iter()
+        .map(|v| inferencer.fold_stmt(v))
+        .collect::<Result<Vec<_>, _>>()
+        .unwrap();
+
+    let calls1 = inferencer.calls.clone();
+    inferencer.calls.clear();
+
+    let statements_2 = statements_2
+        .into_iter()
+        .map(|v| inferencer.fold_stmt(v))
+        .collect::<Result<Vec<_>, _>>()
+        .unwrap();
+
+    if let TopLevelDef::Function { instance_to_stmt, .. } =
+        &mut *top_level.definitions.read()[foo_id].write()
+    {
+        instance_to_stmt.insert(
+            String::new(),
+            FunInstance {
+                body: Arc::new(statements_2),
+                calls: Arc::new(inferencer.calls.clone()),
+                subst: IndexMap::default(),
+                unifier_id: 0,
+            },
+        );
+    } else {
+        unreachable!()
+    }
+
+    inferencer.check_block(&statements_1, &mut identifiers).unwrap();
+    let top_level = Arc::new(TopLevelContext {
+        definitions: Arc::new(RwLock::new(std::mem::take(&mut *top_level.definitions.write()))),
+        unifiers: Arc::new(RwLock::new(vec![(unifier.get_shared_unifier(), primitives)])),
+        personality_symbol: None,
+    });
+
+    let task = CodeGenTask {
+        subst: Vec::default(),
+        symbol_name: "testing".to_string(),
+        body: Arc::new(statements_1),
+        calls: Arc::new(calls1),
+        unifier_index: 0,
+        resolver,
+        signature,
+        store,
+        id: 0,
+    };
+    let f = Arc::new(WithCall::new(Box::new(|module| {
+        let expected = indoc! {"
+            ; ModuleID = 'test'
+            source_filename = \"test\"
+            target datalayout = \"e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128\"
+            target triple = \"x86_64-unknown-linux-gnu\"
+
+            ; Function Attrs: mustprogress nofree norecurse nosync nounwind readnone willreturn
+            define i32 @testing(i32 %0) local_unnamed_addr #0 !dbg !5 {
+            init:
+              %add.i = shl i32 %0, 1, !dbg !10
+              %mul = add i32 %add.i, 2, !dbg !10
+              ret i32 %mul, !dbg !10
+            }
+
+            ; Function Attrs: mustprogress nofree norecurse nosync nounwind readnone willreturn
+            define i32 @foo.0(i32 %0) local_unnamed_addr #0 !dbg !11 {
+            init:
+              %add = add i32 %0, 1, !dbg !12
+              ret i32 %add, !dbg !12
+            }
+
+            attributes #0 = { mustprogress nofree norecurse nosync nounwind readnone willreturn }
+
+            !llvm.module.flags = !{!0, !1}
+            !llvm.dbg.cu = !{!2, !4}
+
+            !0 = !{i32 2, !\"Debug Info Version\", i32 3}
+            !1 = !{i32 2, !\"Dwarf Version\", i32 4}
+            !2 = distinct !DICompileUnit(language: DW_LANG_Python, file: !3, producer: \"NAC3\", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug)
+            !3 = !DIFile(filename: \"unknown\", directory: \"\")
+            !4 = distinct !DICompileUnit(language: DW_LANG_Python, file: !3, producer: \"NAC3\", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug)
+            !5 = distinct !DISubprogram(name: \"testing\", linkageName: \"testing\", scope: null, file: !3, line: 1, type: !6, scopeLine: 1, flags: DIFlagPublic, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !2, retainedNodes: !9)
+            !6 = !DISubroutineType(flags: DIFlagPublic, types: !7)
+            !7 = !{!8}
+            !8 = !DIBasicType(name: \"_\", flags: DIFlagPublic)
+            !9 = !{}
+            !10 = !DILocation(line: 2, column: 12, scope: !5)
+            !11 = distinct !DISubprogram(name: \"foo.0\", linkageName: \"foo.0\", scope: null, file: !3, line: 1, type: !6, scopeLine: 1, flags: DIFlagPublic, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !9)
+            !12 = !DILocation(line: 1, column: 12, scope: !11)
+        "}
+        .trim();
+        assert_eq!(expected, module.print_to_string().to_str().unwrap().trim());
+    })));
+
+    Target::initialize_all(&InitializationConfig::default());
+
+    let llvm_options = CodeGenLLVMOptions {
+        opt_level: OptimizationLevel::Default,
+        target: CodeGenTargetMachineOptions::from_host_triple(),
+    };
+    let (registry, handles) = WorkerRegistry::create_workers(threads, top_level, &llvm_options, &f);
+    registry.add_task(task);
+    registry.wait_tasks_complete(handles);
+}
+
+#[test]
+fn test_classes_list_type_new() {
+    let ctx = inkwell::context::Context::create();
+    let generator = DefaultCodeGenerator::new(String::new(), 64);
+
+    let llvm_i32 = ctx.i32_type();
+    let llvm_usize = generator.get_size_type(&ctx);
+
+    let llvm_list = ListType::new(&generator, &ctx, llvm_i32.into());
+    assert!(ListType::is_type(llvm_list.as_base_type(), llvm_usize).is_ok());
+}
+
+#[test]
+fn test_classes_range_type_new() {
+    let ctx = inkwell::context::Context::create();
+
+    let llvm_range = RangeType::new(&ctx);
+    assert!(RangeType::is_type(llvm_range.as_base_type()).is_ok());
+}
+
+#[test]
+fn test_classes_ndarray_type_new() {
+    let ctx = inkwell::context::Context::create();
+    let generator = DefaultCodeGenerator::new(String::new(), 64);
+
+    let llvm_i32 = ctx.i32_type();
+    let llvm_usize = generator.get_size_type(&ctx);
+
+    let llvm_ndarray = NDArrayType::new(&generator, &ctx, llvm_i32.into());
+    assert!(NDArrayType::is_type(llvm_ndarray.as_base_type(), llvm_usize).is_ok());
+}

nac3core/src/lib.rs

@@ -1,13 +1,29 @@
-#![warn(clippy::all)]
-#![allow(clippy::clone_double_ref)]
+#![deny(
+    future_incompatible,
+    let_underscore,
+    nonstandard_style,
+    rust_2024_compatibility,
+    clippy::all
+)]
+#![warn(clippy::pedantic)]
+#![allow(
+    dead_code,
+    clippy::cast_possible_truncation,
+    clippy::cast_sign_loss,
+    clippy::enum_glob_use,
+    clippy::missing_errors_doc,
+    clippy::missing_panics_doc,
+    clippy::module_name_repetitions,
+    clippy::similar_names,
+    clippy::too_many_lines,
+    clippy::wildcard_imports
+)]
 
-#[cfg(test)]
-extern crate test_case;
-
-extern crate num_bigint;
-extern crate inkwell;
-extern crate rustpython_parser;
-extern crate indoc;
-
-mod typecheck;
+// users of nac3core need to use the same version of these dependencies, so expose them as nac3core::*
+pub use inkwell;
+pub use nac3parser;
 
+pub mod codegen;
+pub mod symbol_resolver;
+pub mod toplevel;
+pub mod typecheck;

nac3core/src/symbol_resolver.rs

@@ -0,0 +1,617 @@
+use std::{
+    collections::{HashMap, HashSet},
+    fmt::{Debug, Display},
+    rc::Rc,
+    sync::Arc,
+};
+
+use inkwell::values::{BasicValueEnum, FloatValue, IntValue, PointerValue, StructValue};
+use itertools::{chain, izip, Itertools};
+use parking_lot::RwLock;
+
+use nac3parser::ast::{Constant, Expr, Location, StrRef};
+
+use crate::{
+    codegen::{CodeGenContext, CodeGenerator},
+    toplevel::{type_annotation::TypeAnnotation, DefinitionId, TopLevelDef},
+    typecheck::{
+        type_inferencer::PrimitiveStore,
+        typedef::{Type, TypeEnum, Unifier, VarMap},
+    },
+};
+
+#[derive(Clone, PartialEq, Debug)]
+pub enum SymbolValue {
+    I32(i32),
+    I64(i64),
+    U32(u32),
+    U64(u64),
+    Str(String),
+    Double(f64),
+    Bool(bool),
+    Tuple(Vec<SymbolValue>),
+    OptionSome(Box<SymbolValue>),
+    OptionNone,
+}
+
+impl SymbolValue {
+    /// Creates a [`SymbolValue`] from a [`Constant`].
+    ///
+    /// * `constant` - The constant to create the value from.
+    /// * `expected_ty` - The expected type of the [`SymbolValue`].
+    pub fn from_constant(
+        constant: &Constant,
+        expected_ty: Type,
+        primitives: &PrimitiveStore,
+        unifier: &mut Unifier,
+    ) -> Result<Self, String> {
+        match constant {
+            Constant::None => {
+                if unifier.unioned(expected_ty, primitives.option) {
+                    Ok(SymbolValue::OptionNone)
+                } else {
+                    Err(format!("Expected {expected_ty:?}, but got Option"))
+                }
+            }
+            Constant::Bool(b) => {
+                if unifier.unioned(expected_ty, primitives.bool) {
+                    Ok(SymbolValue::Bool(*b))
+                } else {
+                    Err(format!("Expected {expected_ty:?}, but got bool"))
+                }
+            }
+            Constant::Str(s) => {
+                if unifier.unioned(expected_ty, primitives.str) {
+                    Ok(SymbolValue::Str(s.to_string()))
+                } else {
+                    Err(format!("Expected {expected_ty:?}, but got str"))
+                }
+            }
+            Constant::Int(i) => {
+                if unifier.unioned(expected_ty, primitives.int32) {
+                    i32::try_from(*i).map(SymbolValue::I32).map_err(|e| e.to_string())
+                } else if unifier.unioned(expected_ty, primitives.int64) {
+                    i64::try_from(*i).map(SymbolValue::I64).map_err(|e| e.to_string())
+                } else if unifier.unioned(expected_ty, primitives.uint32) {
+                    u32::try_from(*i).map(SymbolValue::U32).map_err(|e| e.to_string())
+                } else if unifier.unioned(expected_ty, primitives.uint64) {
+                    u64::try_from(*i).map(SymbolValue::U64).map_err(|e| e.to_string())
+                } else {
+                    Err(format!("Expected {}, but got int", unifier.stringify(expected_ty)))
+                }
+            }
+            Constant::Tuple(t) => {
+                let expected_ty = unifier.get_ty(expected_ty);
+                let TypeEnum::TTuple { ty, is_vararg_ctx } = expected_ty.as_ref() else {
+                    return Err(format!(
+                        "Expected {:?}, but got Tuple",
+                        expected_ty.get_type_name()
+                    ));
+                };
+
+                assert!(*is_vararg_ctx || ty.len() == t.len());
+
+                let elems = t
+                    .iter()
+                    .zip(ty)
+                    .map(|(constant, ty)| Self::from_constant(constant, *ty, primitives, unifier))
+                    .collect::<Result<Vec<SymbolValue>, _>>()?;
+                Ok(SymbolValue::Tuple(elems))
+            }
+            Constant::Float(f) => {
+                if unifier.unioned(expected_ty, primitives.float) {
+                    Ok(SymbolValue::Double(*f))
+                } else {
+                    Err(format!("Expected {expected_ty:?}, but got float"))
+                }
+            }
+            _ => Err(format!("Unsupported value type {constant:?}")),
+        }
+    }
+
+    /// Creates a [`SymbolValue`] from a [`Constant`], with its type being inferred from the constant value.
+    ///
+    /// * `constant` - The constant to create the value from.
+    pub fn from_constant_inferred(constant: &Constant) -> Result<Self, String> {
+        match constant {
+            Constant::None => Ok(SymbolValue::OptionNone),
+            Constant::Bool(b) => Ok(SymbolValue::Bool(*b)),
+            Constant::Str(s) => Ok(SymbolValue::Str(s.to_string())),
+            Constant::Int(i) => {
+                let i = *i;
+                if i >= 0 {
+                    i32::try_from(i)
+                        .map(SymbolValue::I32)
+                        .or_else(|_| i64::try_from(i).map(SymbolValue::I64))
+                        .map_err(|_| {
+                            format!("Literal cannot be expressed as any integral type: {i}")
+                        })
+                } else {
+                    u32::try_from(i)
+                        .map(SymbolValue::U32)
+                        .or_else(|_| u64::try_from(i).map(SymbolValue::U64))
+                        .map_err(|_| {
+                            format!("Literal cannot be expressed as any integral type: {i}")
+                        })
+                }
+            }
+            Constant::Tuple(t) => {
+                let elems = t
+                    .iter()
+                    .map(Self::from_constant_inferred)
+                    .collect::<Result<Vec<SymbolValue>, _>>()?;
+                Ok(SymbolValue::Tuple(elems))
+            }
+            Constant::Float(f) => Ok(SymbolValue::Double(*f)),
+            _ => Err(format!("Unsupported value type {constant:?}")),
+        }
+    }
+
+    /// Returns the [`Type`] representing the data type of this value.
+    pub fn get_type(&self, primitives: &PrimitiveStore, unifier: &mut Unifier) -> Type {
+        match self {
+            SymbolValue::I32(_) => primitives.int32,
+            SymbolValue::I64(_) => primitives.int64,
+            SymbolValue::U32(_) => primitives.uint32,
+            SymbolValue::U64(_) => primitives.uint64,
+            SymbolValue::Str(_) => primitives.str,
+            SymbolValue::Double(_) => primitives.float,
+            SymbolValue::Bool(_) => primitives.bool,
+            SymbolValue::Tuple(vs) => {
+                let vs_tys = vs.iter().map(|v| v.get_type(primitives, unifier)).collect::<Vec<_>>();
+                unifier.add_ty(TypeEnum::TTuple { ty: vs_tys, is_vararg_ctx: false })
+            }
+            SymbolValue::OptionSome(_) | SymbolValue::OptionNone => primitives.option,
+        }
+    }
+
+    /// Returns the [`TypeAnnotation`] representing the data type of this value.
+    pub fn get_type_annotation(
+        &self,
+        primitives: &PrimitiveStore,
+        unifier: &mut Unifier,
+    ) -> TypeAnnotation {
+        match self {
+            SymbolValue::Bool(..)
+            | SymbolValue::Double(..)
+            | SymbolValue::I32(..)
+            | SymbolValue::I64(..)
+            | SymbolValue::U32(..)
+            | SymbolValue::U64(..)
+            | SymbolValue::Str(..) => TypeAnnotation::Primitive(self.get_type(primitives, unifier)),
+            SymbolValue::Tuple(vs) => {
+                let vs_tys = vs
+                    .iter()
+                    .map(|v| v.get_type_annotation(primitives, unifier))
+                    .collect::<Vec<_>>();
+                TypeAnnotation::Tuple(vs_tys)
+            }
+            SymbolValue::OptionNone => TypeAnnotation::CustomClass {
+                id: primitives.option.obj_id(unifier).unwrap(),
+                params: Vec::default(),
+            },
+            SymbolValue::OptionSome(v) => {
+                let ty = v.get_type_annotation(primitives, unifier);
+                TypeAnnotation::CustomClass {
+                    id: primitives.option.obj_id(unifier).unwrap(),
+                    params: vec![ty],
+                }
+            }
+        }
+    }
+
+    /// Returns the [`TypeEnum`] representing the data type of this value.
+    pub fn get_type_enum(
+        &self,
+        primitives: &PrimitiveStore,
+        unifier: &mut Unifier,
+    ) -> Rc<TypeEnum> {
+        let ty = self.get_type(primitives, unifier);
+        unifier.get_ty(ty)
+    }
+}
+
+impl Display for SymbolValue {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match self {
+            SymbolValue::I32(i) => write!(f, "{i}"),
+            SymbolValue::I64(i) => write!(f, "int64({i})"),
+            SymbolValue::U32(i) => write!(f, "uint32({i})"),
+            SymbolValue::U64(i) => write!(f, "uint64({i})"),
+            SymbolValue::Str(s) => write!(f, "\"{s}\""),
+            SymbolValue::Double(d) => write!(f, "{d}"),
+            SymbolValue::Bool(b) => {
+                if *b {
+                    write!(f, "True")
+                } else {
+                    write!(f, "False")
+                }
+            }
+            SymbolValue::Tuple(t) => {
+                write!(f, "({})", t.iter().map(|v| format!("{v}")).collect::<Vec<_>>().join(", "))
+            }
+            SymbolValue::OptionSome(v) => write!(f, "Some({v})"),
+            SymbolValue::OptionNone => write!(f, "none"),
+        }
+    }
+}
+
+impl TryFrom<SymbolValue> for u64 {
+    type Error = ();
+
+    /// Tries to convert a [`SymbolValue`] into a [`u64`], returning [`Err`] if the value is not
+    /// numeric or if the value cannot be converted into a `u64` without overflow.
+    fn try_from(value: SymbolValue) -> Result<Self, Self::Error> {
+        match value {
+            SymbolValue::I32(v) => u64::try_from(v).map_err(|_| ()),
+            SymbolValue::I64(v) => u64::try_from(v).map_err(|_| ()),
+            SymbolValue::U32(v) => Ok(u64::from(v)),
+            SymbolValue::U64(v) => Ok(v),
+            _ => Err(()),
+        }
+    }
+}
+
+impl TryFrom<SymbolValue> for i128 {
+    type Error = ();
+
+    /// Tries to convert a [`SymbolValue`] into a [`i128`], returning [`Err`] if the value is not
+    /// numeric.
+    fn try_from(value: SymbolValue) -> Result<Self, Self::Error> {
+        match value {
+            SymbolValue::I32(v) => Ok(i128::from(v)),
+            SymbolValue::I64(v) => Ok(i128::from(v)),
+            SymbolValue::U32(v) => Ok(i128::from(v)),
+            SymbolValue::U64(v) => Ok(i128::from(v)),
+            _ => Err(()),
+        }
+    }
+}
+
+pub trait StaticValue {
+    /// Returns a unique identifier for this value.
+    fn get_unique_identifier(&self) -> u64;
+
+    /// Returns the constant object represented by this unique identifier.
+    fn get_const_obj<'ctx>(
+        &self,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        generator: &mut dyn CodeGenerator,
+    ) -> BasicValueEnum<'ctx>;
+
+    /// Converts this value to a LLVM [`BasicValueEnum`].
+    fn to_basic_value_enum<'ctx>(
+        &self,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        generator: &mut dyn CodeGenerator,
+        expected_ty: Type,
+    ) -> Result<BasicValueEnum<'ctx>, String>;
+
+    /// Returns a field within this value.
+    fn get_field<'ctx>(
+        &self,
+        name: StrRef,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+    ) -> Option<ValueEnum<'ctx>>;
+
+    /// Returns a single element of this tuple.
+    fn get_tuple_element<'ctx>(&self, index: u32) -> Option<ValueEnum<'ctx>>;
+}
+
+#[derive(Clone)]
+pub enum ValueEnum<'ctx> {
+    /// [`ValueEnum`] representing a static value.
+    Static(Arc<dyn StaticValue + Send + Sync>),
+
+    /// [`ValueEnum`] representing a dynamic value.
+    Dynamic(BasicValueEnum<'ctx>),
+}
+
+impl<'ctx> From<BasicValueEnum<'ctx>> for ValueEnum<'ctx> {
+    fn from(v: BasicValueEnum<'ctx>) -> Self {
+        ValueEnum::Dynamic(v)
+    }
+}
+
+impl<'ctx> From<PointerValue<'ctx>> for ValueEnum<'ctx> {
+    fn from(v: PointerValue<'ctx>) -> Self {
+        ValueEnum::Dynamic(v.into())
+    }
+}
+
+impl<'ctx> From<IntValue<'ctx>> for ValueEnum<'ctx> {
+    fn from(v: IntValue<'ctx>) -> Self {
+        ValueEnum::Dynamic(v.into())
+    }
+}
+
+impl<'ctx> From<FloatValue<'ctx>> for ValueEnum<'ctx> {
+    fn from(v: FloatValue<'ctx>) -> Self {
+        ValueEnum::Dynamic(v.into())
+    }
+}
+
+impl<'ctx> From<StructValue<'ctx>> for ValueEnum<'ctx> {
+    fn from(v: StructValue<'ctx>) -> Self {
+        ValueEnum::Dynamic(v.into())
+    }
+}
+
+impl<'ctx> ValueEnum<'ctx> {
+    /// Converts this [`ValueEnum`] to a [`BasicValueEnum`].
+    pub fn to_basic_value_enum<'a>(
+        self,
+        ctx: &mut CodeGenContext<'ctx, 'a>,
+        generator: &mut dyn CodeGenerator,
+        expected_ty: Type,
+    ) -> Result<BasicValueEnum<'ctx>, String> {
+        match self {
+            ValueEnum::Static(v) => v.to_basic_value_enum(ctx, generator, expected_ty),
+            ValueEnum::Dynamic(v) => Ok(v),
+        }
+    }
+}
+
+pub trait SymbolResolver {
+    /// Get type of type variable identifier or top-level function type,
+    fn get_symbol_type(
+        &self,
+        unifier: &mut Unifier,
+        top_level_defs: &[Arc<RwLock<TopLevelDef>>],
+        primitives: &PrimitiveStore,
+        str: StrRef,
+    ) -> Result<Type, String>;
+
+    /// Get the top-level definition of identifiers.
+    fn get_identifier_def(&self, str: StrRef) -> Result<DefinitionId, HashSet<String>>;
+
+    fn get_symbol_value<'ctx>(
+        &self,
+        str: StrRef,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        generator: &mut dyn CodeGenerator,
+    ) -> Option<ValueEnum<'ctx>>;
+
+    fn get_default_param_value(&self, expr: &Expr) -> Option<SymbolValue>;
+    fn get_string_id(&self, s: &str) -> i32;
+    fn get_exception_id(&self, tyid: usize) -> usize;
+
+    fn handle_deferred_eval(
+        &self,
+        _unifier: &mut Unifier,
+        _top_level_defs: &[Arc<RwLock<TopLevelDef>>],
+        _primitives: &PrimitiveStore,
+    ) -> Result<(), String> {
+        Ok(())
+    }
+}
+
+thread_local! {
+    static IDENTIFIER_ID: [StrRef; 11] = [
+        "int32".into(),
+        "int64".into(),
+        "float".into(),
+        "bool".into(),
+        "virtual".into(),
+        "tuple".into(),
+        "str".into(),
+        "Exception".into(),
+        "uint32".into(),
+        "uint64".into(),
+        "Literal".into(),
+    ];
+}
+
+/// Converts a type annotation into a [Type].
+pub fn parse_type_annotation<T>(
+    resolver: &dyn SymbolResolver,
+    top_level_defs: &[Arc<RwLock<TopLevelDef>>],
+    unifier: &mut Unifier,
+    primitives: &PrimitiveStore,
+    expr: &Expr<T>,
+) -> Result<Type, HashSet<String>> {
+    use nac3parser::ast::ExprKind::*;
+    let ids = IDENTIFIER_ID.with(|ids| *ids);
+    let int32_id = ids[0];
+    let int64_id = ids[1];
+    let float_id = ids[2];
+    let bool_id = ids[3];
+    let virtual_id = ids[4];
+    let tuple_id = ids[5];
+    let str_id = ids[6];
+    let exn_id = ids[7];
+    let uint32_id = ids[8];
+    let uint64_id = ids[9];
+    let literal_id = ids[10];
+
+    let name_handling = |id: &StrRef, loc: Location, unifier: &mut Unifier| {
+        if *id == int32_id {
+            Ok(primitives.int32)
+        } else if *id == int64_id {
+            Ok(primitives.int64)
+        } else if *id == uint32_id {
+            Ok(primitives.uint32)
+        } else if *id == uint64_id {
+            Ok(primitives.uint64)
+        } else if *id == float_id {
+            Ok(primitives.float)
+        } else if *id == bool_id {
+            Ok(primitives.bool)
+        } else if *id == str_id {
+            Ok(primitives.str)
+        } else if *id == exn_id {
+            Ok(primitives.exception)
+        } else {
+            let obj_id = resolver.get_identifier_def(*id);
+            if let Ok(obj_id) = obj_id {
+                let def = top_level_defs[obj_id.0].read();
+                if let TopLevelDef::Class { fields, methods, type_vars, .. } = &*def {
+                    if !type_vars.is_empty() {
+                        return Err(HashSet::from([format!(
+                            "Unexpected number of type parameters: expected {} but got 0",
+                            type_vars.len()
+                        )]));
+                    }
+                    let fields = chain(
+                        fields.iter().map(|(k, v, m)| (*k, (*v, *m))),
+                        methods.iter().map(|(k, v, _)| (*k, (*v, false))),
+                    )
+                    .collect();
+                    Ok(unifier.add_ty(TypeEnum::TObj { obj_id, fields, params: VarMap::default() }))
+                } else {
+                    Err(HashSet::from([format!("Cannot use function name as type at {loc}")]))
+                }
+            } else {
+                let ty =
+                    resolver.get_symbol_type(unifier, top_level_defs, primitives, *id).map_err(
+                        |e| HashSet::from([format!("Unknown type annotation at {loc}: {e}")]),
+                    )?;
+                if let TypeEnum::TVar { .. } = &*unifier.get_ty(ty) {
+                    Ok(ty)
+                } else {
+                    Err(HashSet::from([format!("Unknown type annotation {id} at {loc}")]))
+                }
+            }
+        }
+    };
+
+    let subscript_name_handle = |id: &StrRef, slice: &Expr<T>, unifier: &mut Unifier| {
+        if *id == virtual_id {
+            let ty = parse_type_annotation(resolver, top_level_defs, unifier, primitives, slice)?;
+            Ok(unifier.add_ty(TypeEnum::TVirtual { ty }))
+        } else if *id == tuple_id {
+            if let Tuple { elts, .. } = &slice.node {
+                let ty = elts
+                    .iter()
+                    .map(|elt| {
+                        parse_type_annotation(resolver, top_level_defs, unifier, primitives, elt)
+                    })
+                    .collect::<Result<Vec<_>, _>>()?;
+                Ok(unifier.add_ty(TypeEnum::TTuple { ty, is_vararg_ctx: false }))
+            } else {
+                Err(HashSet::from(["Expected multiple elements for tuple".into()]))
+            }
+        } else if *id == literal_id {
+            let mut parse_literal = |elt: &Expr<T>| {
+                let ty = parse_type_annotation(resolver, top_level_defs, unifier, primitives, elt)?;
+                let ty_enum = &*unifier.get_ty_immutable(ty);
+                match ty_enum {
+                    TypeEnum::TLiteral { values, .. } => Ok(values.clone()),
+                    _ => Err(HashSet::from([format!(
+                        "Expected literal in type argument for Literal at {}",
+                        elt.location
+                    )])),
+                }
+            };
+
+            let values = if let Tuple { elts, .. } = &slice.node {
+                elts.iter().map(&mut parse_literal).collect::<Result<Vec<_>, _>>()?
+            } else {
+                vec![parse_literal(slice)?]
+            }
+            .into_iter()
+            .flatten()
+            .collect_vec();
+
+            Ok(unifier.get_fresh_literal(values, Some(slice.location)))
+        } else {
+            let types = if let Tuple { elts, .. } = &slice.node {
+                elts.iter()
+                    .map(|v| {
+                        parse_type_annotation(resolver, top_level_defs, unifier, primitives, v)
+                    })
+                    .collect::<Result<Vec<_>, _>>()?
+            } else {
+                vec![parse_type_annotation(resolver, top_level_defs, unifier, primitives, slice)?]
+            };
+
+            let obj_id = resolver.get_identifier_def(*id)?;
+            let def = top_level_defs[obj_id.0].read();
+            if let TopLevelDef::Class { fields, methods, type_vars, .. } = &*def {
+                if types.len() != type_vars.len() {
+                    return Err(HashSet::from([format!(
+                        "Unexpected number of type parameters: expected {} but got {}",
+                        type_vars.len(),
+                        types.len()
+                    )]));
+                }
+                let mut subst = VarMap::new();
+                for (var, ty) in izip!(type_vars.iter(), types.iter()) {
+                    let id = if let TypeEnum::TVar { id, .. } = &*unifier.get_ty(*var) {
+                        *id
+                    } else {
+                        unreachable!()
+                    };
+                    subst.insert(id, *ty);
+                }
+                let mut fields = fields
+                    .iter()
+                    .map(|(attr, ty, is_mutable)| {
+                        let ty = unifier.subst(*ty, &subst).unwrap_or(*ty);
+                        (*attr, (ty, *is_mutable))
+                    })
+                    .collect::<HashMap<_, _>>();
+                fields.extend(methods.iter().map(|(attr, ty, _)| {
+                    let ty = unifier.subst(*ty, &subst).unwrap_or(*ty);
+                    (*attr, (ty, false))
+                }));
+                Ok(unifier.add_ty(TypeEnum::TObj { obj_id, fields, params: subst }))
+            } else {
+                Err(HashSet::from(["Cannot use function name as type".into()]))
+            }
+        }
+    };
+
+    match &expr.node {
+        Name { id, .. } => name_handling(id, expr.location, unifier),
+        Subscript { value, slice, .. } => {
+            if let Name { id, .. } = &value.node {
+                subscript_name_handle(id, slice, unifier)
+            } else {
+                Err(HashSet::from([format!("unsupported type expression at {}", expr.location)]))
+            }
+        }
+        Constant { value, .. } => SymbolValue::from_constant_inferred(value)
+            .map(|v| unifier.get_fresh_literal(vec![v], Some(expr.location)))
+            .map_err(|err| HashSet::from([err])),
+        _ => Err(HashSet::from([format!("unsupported type expression at {}", expr.location)])),
+    }
+}
+
+impl dyn SymbolResolver + Send + Sync {
+    pub fn parse_type_annotation<T>(
+        &self,
+        top_level_defs: &[Arc<RwLock<TopLevelDef>>],
+        unifier: &mut Unifier,
+        primitives: &PrimitiveStore,
+        expr: &Expr<T>,
+    ) -> Result<Type, HashSet<String>> {
+        parse_type_annotation(self, top_level_defs, unifier, primitives, expr)
+    }
+
+    pub fn get_type_name(
+        &self,
+        top_level_defs: &[Arc<RwLock<TopLevelDef>>],
+        unifier: &mut Unifier,
+        ty: Type,
+    ) -> String {
+        unifier.internal_stringify(
+            ty,
+            &mut |id| {
+                let TopLevelDef::Class { name, .. } = &*top_level_defs[id].read() else {
+                    unreachable!("expected class definition")
+                };
+
+                name.to_string()
+            },
+            &mut |id| format!("typevar{id}"),
+            &mut None,
+        )
+    }
+}
+
+impl Debug for dyn SymbolResolver + Send + Sync {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        write!(f, "")
+    }
+}

nac3core/src/toplevel/mod.rs

@@ -0,0 +1,176 @@
+use std::{
+    borrow::BorrowMut,
+    collections::{HashMap, HashSet},
+    fmt::Debug,
+    iter::FromIterator,
+    sync::Arc,
+};
+
+use inkwell::values::BasicValueEnum;
+use itertools::Itertools;
+use parking_lot::RwLock;
+
+use nac3parser::ast::{self, Expr, Location, Stmt, StrRef};
+
+use crate::{
+    codegen::{CodeGenContext, CodeGenerator},
+    symbol_resolver::{SymbolResolver, ValueEnum},
+    typecheck::{
+        type_inferencer::{CodeLocation, PrimitiveStore},
+        typedef::{
+            CallId, FunSignature, FuncArg, SharedUnifier, Type, TypeEnum, TypeVarId, Unifier,
+            VarMap,
+        },
+    },
+};
+
+#[derive(PartialEq, Eq, PartialOrd, Ord, Clone, Copy, Hash, Debug)]
+pub struct DefinitionId(pub usize);
+
+pub mod builtins;
+pub mod composer;
+pub mod helper;
+pub mod numpy;
+pub mod type_annotation;
+use composer::*;
+use type_annotation::*;
+#[cfg(test)]
+mod test;
+
+type GenCallCallback = dyn for<'ctx, 'a> Fn(
+        &mut CodeGenContext<'ctx, 'a>,
+        Option<(Type, ValueEnum<'ctx>)>,
+        (&FunSignature, DefinitionId),
+        Vec<(Option<StrRef>, ValueEnum<'ctx>)>,
+        &mut dyn CodeGenerator,
+    ) -> Result<Option<BasicValueEnum<'ctx>>, String>
+    + Send
+    + Sync;
+
+pub struct GenCall {
+    fp: Box<GenCallCallback>,
+}
+
+impl GenCall {
+    #[must_use]
+    pub fn new(fp: Box<GenCallCallback>) -> GenCall {
+        GenCall { fp }
+    }
+
+    /// Creates a dummy instance of [`GenCall`], which invokes [`unreachable!()`] with the given
+    /// `reason`.
+    #[must_use]
+    pub fn create_dummy(reason: String) -> GenCall {
+        Self::new(Box::new(move |_, _, _, _, _| unreachable!("{reason}")))
+    }
+
+    pub fn run<'ctx>(
+        &self,
+        ctx: &mut CodeGenContext<'ctx, '_>,
+        obj: Option<(Type, ValueEnum<'ctx>)>,
+        fun: (&FunSignature, DefinitionId),
+        args: Vec<(Option<StrRef>, ValueEnum<'ctx>)>,
+        generator: &mut dyn CodeGenerator,
+    ) -> Result<Option<BasicValueEnum<'ctx>>, String> {
+        (self.fp)(ctx, obj, fun, args, generator)
+    }
+}
+
+impl Debug for GenCall {
+    fn fmt(&self, _: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        Ok(())
+    }
+}
+
+#[derive(Clone, Debug)]
+pub struct FunInstance {
+    pub body: Arc<Vec<Stmt<Option<Type>>>>,
+    pub calls: Arc<HashMap<CodeLocation, CallId>>,
+    pub subst: VarMap,
+    pub unifier_id: usize,
+}
+
+#[derive(Debug, Clone)]
+pub enum TopLevelDef {
+    Class {
+        /// Name for error messages and symbols.
+        name: StrRef,
+        /// Object ID used for [`TypeEnum`].
+        object_id: DefinitionId,
+        /// type variables bounded to the class.
+        type_vars: Vec<Type>,
+        /// Class fields.
+        ///
+        /// Name and type is mutable.
+        fields: Vec<(StrRef, Type, bool)>,
+        /// Class Attributes.
+        ///
+        /// Name, type, value.
+        attributes: Vec<(StrRef, Type, ast::Constant)>,
+        /// Class methods, pointing to the corresponding function definition.
+        methods: Vec<(StrRef, Type, DefinitionId)>,
+        /// Ancestor classes, including itself.
+        ancestors: Vec<TypeAnnotation>,
+        /// Symbol resolver of the module defined the class; [None] if it is built-in type.
+        resolver: Option<Arc<dyn SymbolResolver + Send + Sync>>,
+        /// Constructor type.
+        constructor: Option<Type>,
+        /// Definition location.
+        loc: Option<Location>,
+    },
+    Function {
+        /// Prefix for symbol, should be unique globally.
+        name: String,
+        /// Simple name, the same as in method/function definition.
+        simple_name: StrRef,
+        /// Function signature.
+        signature: Type,
+        /// Instantiated type variable IDs.
+        var_id: Vec<TypeVarId>,
+        /// Function instance to symbol mapping
+        ///
+        /// * Key: String representation of type variable values, sorted by variable ID in ascending
+        ///   order, including type variables associated with the class.
+        /// * Value: Function symbol name.
+        instance_to_symbol: HashMap<String, String>,
+        /// Function instances to annotated AST mapping
+        ///
+        /// * Key: String representation of type variable values, sorted by variable ID in ascending
+        ///   order, including type variables associated with the class. Excluding rigid type
+        ///   variables.
+        ///
+        /// Rigid type variables that would be substituted when the function is instantiated.
+        instance_to_stmt: HashMap<String, FunInstance>,
+        /// Symbol resolver of the module defined the class.
+        resolver: Option<Arc<dyn SymbolResolver + Send + Sync>>,
+        /// Custom code generation callback.
+        codegen_callback: Option<Arc<GenCall>>,
+        /// Definition location.
+        loc: Option<Location>,
+    },
+    Variable {
+        /// Qualified name of the global variable, should be unique globally.
+        name: String,
+
+        /// Simple name, the same as in method/function definition.
+        simple_name: StrRef,
+
+        /// Type of the global variable.
+        ty: Type,
+
+        /// The declared type of the global variable.
+        ty_decl: Expr,
+
+        /// Symbol resolver of the module defined the class.
+        resolver: Option<Arc<dyn SymbolResolver + Send + Sync>>,
+
+        /// Definition location.
+        loc: Option<Location>,
+    },
+}
+
+pub struct TopLevelContext {
+    pub definitions: Arc<RwLock<Vec<Arc<RwLock<TopLevelDef>>>>>,
+    pub unifiers: Arc<RwLock<Vec<(SharedUnifier, PrimitiveStore)>>>,
+    pub personality_symbol: Option<String>,
+}

nac3core/src/toplevel/numpy.rs

@@ -0,0 +1,86 @@
+use itertools::Itertools;
+
+use crate::{
+    toplevel::helper::PrimDef,
+    typecheck::{
+        type_inferencer::PrimitiveStore,
+        typedef::{Type, TypeEnum, TypeVarId, Unifier, VarMap},
+    },
+};
+
+/// Creates a `ndarray` [`Type`] with the given type arguments.
+///
+/// * `dtype` - The element type of the `ndarray`, or [`None`] if the type variable is not
+///   specialized.
+/// * `ndims` - The number of dimensions of the `ndarray`, or [`None`] if the type variable is not
+///   specialized.
+pub fn make_ndarray_ty(
+    unifier: &mut Unifier,
+    primitives: &PrimitiveStore,
+    dtype: Option<Type>,
+    ndims: Option<Type>,
+) -> Type {
+    subst_ndarray_tvars(unifier, primitives.ndarray, dtype, ndims)
+}
+
+/// Substitutes type variables in `ndarray`.
+///
+/// * `dtype` - The element type of the `ndarray`, or [`None`] if the type variable is not
+///   specialized.
+/// * `ndims` - The number of dimensions of the `ndarray`, or [`None`] if the type variable is not
+///   specialized.
+pub fn subst_ndarray_tvars(
+    unifier: &mut Unifier,
+    ndarray: Type,
+    dtype: Option<Type>,
+    ndims: Option<Type>,
+) -> Type {
+    let TypeEnum::TObj { obj_id, params, .. } = &*unifier.get_ty_immutable(ndarray) else {
+        panic!("Expected `ndarray` to be TObj, but got {}", unifier.stringify(ndarray))
+    };
+    debug_assert_eq!(*obj_id, PrimDef::NDArray.id());
+
+    if dtype.is_none() && ndims.is_none() {
+        return ndarray;
+    }
+
+    let tvar_ids = params.iter().map(|(obj_id, _)| *obj_id).collect_vec();
+    debug_assert_eq!(tvar_ids.len(), 2);
+
+    let mut tvar_subst = VarMap::new();
+    if let Some(dtype) = dtype {
+        tvar_subst.insert(tvar_ids[0], dtype);
+    }
+    if let Some(ndims) = ndims {
+        tvar_subst.insert(tvar_ids[1], ndims);
+    }
+
+    unifier.subst(ndarray, &tvar_subst).unwrap_or(ndarray)
+}
+
+fn unpack_ndarray_tvars(unifier: &mut Unifier, ndarray: Type) -> Vec<(TypeVarId, Type)> {
+    let TypeEnum::TObj { obj_id, params, .. } = &*unifier.get_ty_immutable(ndarray) else {
+        panic!("Expected `ndarray` to be TObj, but got {}", unifier.stringify(ndarray))
+    };
+    debug_assert_eq!(*obj_id, PrimDef::NDArray.id());
+    debug_assert_eq!(params.len(), 2);
+
+    params
+        .iter()
+        .sorted_by_key(|(obj_id, _)| *obj_id)
+        .map(|(var_id, ty)| (*var_id, *ty))
+        .collect_vec()
+}
+
+/// Unpacks the type variable IDs of `ndarray` into a tuple. The elements of the tuple corresponds
+/// to `dtype` (the element type) and `ndims` (the number of dimensions) of the `ndarray`
+/// respectively.
+pub fn unpack_ndarray_var_ids(unifier: &mut Unifier, ndarray: Type) -> (TypeVarId, TypeVarId) {
+    unpack_ndarray_tvars(unifier, ndarray).into_iter().map(|v| v.0).collect_tuple().unwrap()
+}
+
+/// Unpacks the type variables of `ndarray` into a tuple. The elements of the tuple corresponds to
+/// `dtype` (the element type) and `ndims` (the number of dimensions) of the `ndarray` respectively.
+pub fn unpack_ndarray_var_tys(unifier: &mut Unifier, ndarray: Type) -> (Type, Type) {
+    unpack_ndarray_tvars(unifier, ndarray).into_iter().map(|v| v.1).collect_tuple().unwrap()
+}

nac3core/src/toplevel/snapshots/nac3core__toplevel__test__test_analyze__generic_class.snap

@@ -0,0 +1,12 @@
+---
+source: nac3core/src/toplevel/test.rs
+expression: res_vec
+---
+[
+    "Class {\nname: \"Generic_A\",\nancestors: [\"Generic_A[V]\", \"B\"],\nfields: [\"aa\", \"a\"],\nmethods: [(\"__init__\", \"fn[[], none]\"), (\"foo\", \"fn[[b:T], none]\"), (\"fun\", \"fn[[a:int32], V]\")],\ntype_vars: [\"V\"]\n}\n",
+    "Function {\nname: \"Generic_A.__init__\",\nsig: \"fn[[], none]\",\nvar_id: []\n}\n",
+    "Function {\nname: \"Generic_A.fun\",\nsig: \"fn[[a:int32], V]\",\nvar_id: [TypeVarId(241)]\n}\n",
+    "Class {\nname: \"B\",\nancestors: [\"B\"],\nfields: [\"aa\"],\nmethods: [(\"__init__\", \"fn[[], none]\"), (\"foo\", \"fn[[b:T], none]\")],\ntype_vars: []\n}\n",
+    "Function {\nname: \"B.__init__\",\nsig: \"fn[[], none]\",\nvar_id: []\n}\n",
+    "Function {\nname: \"B.foo\",\nsig: \"fn[[b:T], none]\",\nvar_id: []\n}\n",
+]

nac3core/src/toplevel/snapshots/nac3core__toplevel__test__test_analyze__inheritance_override.snap

@@ -0,0 +1,15 @@
+---
+source: nac3core/src/toplevel/test.rs
+expression: res_vec
+---
+[
+    "Class {\nname: \"A\",\nancestors: [\"A[T]\"],\nfields: [\"a\", \"b\", \"c\"],\nmethods: [(\"__init__\", \"fn[[t:T], none]\"), (\"fun\", \"fn[[a:int32, b:T], list[virtual[B[bool]]]]\"), (\"foo\", \"fn[[c:C], none]\")],\ntype_vars: [\"T\"]\n}\n",
+    "Function {\nname: \"A.__init__\",\nsig: \"fn[[t:T], none]\",\nvar_id: []\n}\n",
+    "Function {\nname: \"A.fun\",\nsig: \"fn[[a:int32, b:T], list[virtual[B[bool]]]]\",\nvar_id: []\n}\n",
+    "Function {\nname: \"A.foo\",\nsig: \"fn[[c:C], none]\",\nvar_id: []\n}\n",
+    "Class {\nname: \"B\",\nancestors: [\"B[typevar230]\", \"A[float]\"],\nfields: [\"a\", \"b\", \"c\", \"d\"],\nmethods: [(\"__init__\", \"fn[[], none]\"), (\"fun\", \"fn[[a:int32, b:T], list[virtual[B[bool]]]]\"), (\"foo\", \"fn[[c:C], none]\")],\ntype_vars: [\"typevar230\"]\n}\n",
+    "Function {\nname: \"B.__init__\",\nsig: \"fn[[], none]\",\nvar_id: []\n}\n",
+    "Function {\nname: \"B.fun\",\nsig: \"fn[[a:int32, b:T], list[virtual[B[bool]]]]\",\nvar_id: []\n}\n",
+    "Class {\nname: \"C\",\nancestors: [\"C\", \"B[bool]\", \"A[float]\"],\nfields: [\"a\", \"b\", \"c\", \"d\", \"e\"],\nmethods: [(\"__init__\", \"fn[[], none]\"), (\"fun\", \"fn[[a:int32, b:T], list[virtual[B[bool]]]]\"), (\"foo\", \"fn[[c:C], none]\")],\ntype_vars: []\n}\n",
+    "Function {\nname: \"C.__init__\",\nsig: \"fn[[], none]\",\nvar_id: []\n}\n",
+]

nac3core/src/toplevel/snapshots/nac3core__toplevel__test__test_analyze__list_tuple_generic.snap

@@ -0,0 +1,13 @@
+---
+source: nac3core/src/toplevel/test.rs
+expression: res_vec
+---
+[
+    "Function {\nname: \"foo\",\nsig: \"fn[[a:list[int32], b:tuple[T, float]], A[B, bool]]\",\nvar_id: []\n}\n",
+    "Class {\nname: \"A\",\nancestors: [\"A[T, V]\"],\nfields: [\"a\", \"b\"],\nmethods: [(\"__init__\", \"fn[[v:V], none]\"), (\"fun\", \"fn[[a:T], V]\")],\ntype_vars: [\"T\", \"V\"]\n}\n",
+    "Function {\nname: \"A.__init__\",\nsig: \"fn[[v:V], none]\",\nvar_id: [TypeVarId(243)]\n}\n",
+    "Function {\nname: \"A.fun\",\nsig: \"fn[[a:T], V]\",\nvar_id: [TypeVarId(248)]\n}\n",
+    "Function {\nname: \"gfun\",\nsig: \"fn[[a:A[list[float], int32]], none]\",\nvar_id: []\n}\n",
+    "Class {\nname: \"B\",\nancestors: [\"B\"],\nfields: [],\nmethods: [(\"__init__\", \"fn[[], none]\")],\ntype_vars: []\n}\n",
+    "Function {\nname: \"B.__init__\",\nsig: \"fn[[], none]\",\nvar_id: []\n}\n",
+]

nac3core/src/toplevel/snapshots/nac3core__toplevel__test__test_analyze__self1.snap

@@ -0,0 +1,13 @@
+---
+source: nac3core/src/toplevel/test.rs
+expression: res_vec
+---
+[
+    "Class {\nname: \"A\",\nancestors: [\"A[typevar229, typevar230]\"],\nfields: [\"a\", \"b\"],\nmethods: [(\"__init__\", \"fn[[a:A[float, bool], b:B], none]\"), (\"fun\", \"fn[[a:A[float, bool]], A[bool, int32]]\")],\ntype_vars: [\"typevar229\", \"typevar230\"]\n}\n",
+    "Function {\nname: \"A.__init__\",\nsig: \"fn[[a:A[float, bool], b:B], none]\",\nvar_id: []\n}\n",
+    "Function {\nname: \"A.fun\",\nsig: \"fn[[a:A[float, bool]], A[bool, int32]]\",\nvar_id: []\n}\n",
+    "Class {\nname: \"B\",\nancestors: [\"B\", \"A[int64, bool]\"],\nfields: [\"a\", \"b\"],\nmethods: [(\"__init__\", \"fn[[], none]\"), (\"fun\", \"fn[[a:A[float, bool]], A[bool, int32]]\"), (\"foo\", \"fn[[b:B], B]\"), (\"bar\", \"fn[[a:A[list[B], int32]], tuple[A[virtual[A[B, int32]], bool], B]]\")],\ntype_vars: []\n}\n",
+    "Function {\nname: \"B.__init__\",\nsig: \"fn[[], none]\",\nvar_id: []\n}\n",
+    "Function {\nname: \"B.foo\",\nsig: \"fn[[b:B], B]\",\nvar_id: []\n}\n",
+    "Function {\nname: \"B.bar\",\nsig: \"fn[[a:A[list[B], int32]], tuple[A[virtual[A[B, int32]], bool], B]]\",\nvar_id: []\n}\n",
+]

nac3core/src/toplevel/snapshots/nac3core__toplevel__test__test_analyze__simple_class_compose.snap

@@ -0,0 +1,17 @@
+---
+source: nac3core/src/toplevel/test.rs
+expression: res_vec
+---
+[
+    "Class {\nname: \"A\",\nancestors: [\"A\"],\nfields: [\"a\"],\nmethods: [(\"__init__\", \"fn[[], none]\"), (\"fun\", \"fn[[b:B], none]\"), (\"foo\", \"fn[[a:T, b:V], none]\")],\ntype_vars: []\n}\n",
+    "Function {\nname: \"A.__init__\",\nsig: \"fn[[], none]\",\nvar_id: []\n}\n",
+    "Function {\nname: \"A.fun\",\nsig: \"fn[[b:B], none]\",\nvar_id: []\n}\n",
+    "Function {\nname: \"A.foo\",\nsig: \"fn[[a:T, b:V], none]\",\nvar_id: [TypeVarId(249)]\n}\n",
+    "Class {\nname: \"B\",\nancestors: [\"B\", \"C\", \"A\"],\nfields: [\"a\"],\nmethods: [(\"__init__\", \"fn[[], none]\"), (\"fun\", \"fn[[b:B], none]\"), (\"foo\", \"fn[[a:T, b:V], none]\")],\ntype_vars: []\n}\n",
+    "Function {\nname: \"B.__init__\",\nsig: \"fn[[], none]\",\nvar_id: []\n}\n",
+    "Class {\nname: \"C\",\nancestors: [\"C\", \"A\"],\nfields: [\"a\"],\nmethods: [(\"__init__\", \"fn[[], none]\"), (\"fun\", \"fn[[b:B], none]\"), (\"foo\", \"fn[[a:T, b:V], none]\")],\ntype_vars: []\n}\n",
+    "Function {\nname: \"C.__init__\",\nsig: \"fn[[], none]\",\nvar_id: []\n}\n",
+    "Function {\nname: \"C.fun\",\nsig: \"fn[[b:B], none]\",\nvar_id: []\n}\n",
+    "Function {\nname: \"foo\",\nsig: \"fn[[a:A], none]\",\nvar_id: []\n}\n",
+    "Function {\nname: \"ff\",\nsig: \"fn[[a:T], V]\",\nvar_id: [TypeVarId(257)]\n}\n",
+]

nac3core/src/toplevel/snapshots/nac3core__toplevel__test__test_analyze__simple_pass_in_class.snap

@@ -0,0 +1,9 @@
+---
+source: nac3core/src/toplevel/test.rs
+assertion_line: 549
+expression: res_vec
+
+---
+[
+    "Class {\nname: \"A\",\nancestors: [\"A\"],\nfields: [],\nmethods: [],\ntype_vars: []\n}\n",
+]

nac3core/src/toplevel/test.rs

@@ -0,0 +1,838 @@
+use std::{collections::HashMap, sync::Arc};
+
+use indoc::indoc;
+use parking_lot::Mutex;
+use test_case::test_case;
+
+use nac3parser::{
+    ast::{fold::Fold, FileName},
+    parser::parse_program,
+};
+
+use super::*;
+use crate::{
+    codegen::CodeGenContext,
+    symbol_resolver::{SymbolResolver, ValueEnum},
+    toplevel::{helper::PrimDef, DefinitionId},
+    typecheck::{
+        type_inferencer::PrimitiveStore,
+        typedef::{into_var_map, Type, Unifier},
+    },
+};
+
+struct ResolverInternal {
+    id_to_type: Mutex<HashMap<StrRef, Type>>,
+    id_to_def: Mutex<HashMap<StrRef, DefinitionId>>,
+    class_names: Mutex<HashMap<StrRef, Type>>,
+}
+
+impl ResolverInternal {
+    fn add_id_def(&self, id: StrRef, def: DefinitionId) {
+        self.id_to_def.lock().insert(id, def);
+    }
+
+    fn add_id_type(&self, id: StrRef, ty: Type) {
+        self.id_to_type.lock().insert(id, ty);
+    }
+}
+
+struct Resolver(Arc<ResolverInternal>);
+
+impl SymbolResolver for Resolver {
+    fn get_default_param_value(
+        &self,
+        _: &ast::Expr,
+    ) -> Option<crate::symbol_resolver::SymbolValue> {
+        unimplemented!()
+    }
+
+    fn get_symbol_type(
+        &self,
+        _: &mut Unifier,
+        _: &[Arc<RwLock<TopLevelDef>>],
+        _: &PrimitiveStore,
+        str: StrRef,
+    ) -> Result<Type, String> {
+        self.0
+            .id_to_type
+            .lock()
+            .get(&str)
+            .copied()
+            .ok_or_else(|| format!("cannot find symbol `{str}`"))
+    }
+
+    fn get_symbol_value<'ctx>(
+        &self,
+        _: StrRef,
+        _: &mut CodeGenContext<'ctx, '_>,
+        _: &mut dyn CodeGenerator,
+    ) -> Option<ValueEnum<'ctx>> {
+        unimplemented!()
+    }
+
+    fn get_identifier_def(&self, id: StrRef) -> Result<DefinitionId, HashSet<String>> {
+        self.0
+            .id_to_def
+            .lock()
+            .get(&id)
+            .copied()
+            .ok_or_else(|| HashSet::from(["Unknown identifier".to_string()]))
+    }
+
+    fn get_string_id(&self, _: &str) -> i32 {
+        unimplemented!()
+    }
+
+    fn get_exception_id(&self, _tyid: usize) -> usize {
+        unimplemented!()
+    }
+}
+
+#[test_case(
+    vec![
+        indoc! {"
+            def fun(a: int32) -> int32:
+                return a
+        "},
+        indoc! {"
+            class A:
+                def __init__(self):
+                    self.a: int32 = 3
+        "},
+        indoc! {"
+            class B:
+                def __init__(self):
+                    self.b: float = 4.3
+
+                def fun(self):
+                    self.b = self.b + 3.0
+        "},
+        indoc! {"
+            def foo(a: float):
+                a + 1.0
+        "},
+        indoc! {"
+            class C(B):
+                def __init__(self):
+                    self.c: int32 = 4
+                    self.a: bool = True
+        "},
+    ];
+    "register"
+)]
+fn test_simple_register(source: Vec<&str>) {
+    let mut composer =
+        TopLevelComposer::new(Vec::new(), Vec::new(), ComposerConfig::default(), 64).0;
+
+    for s in source {
+        let ast = parse_program(s, FileName::default()).unwrap();
+        let ast = ast[0].clone();
+
+        composer.register_top_level(ast, None, "", false).unwrap();
+    }
+}
+
+#[test_case(
+    indoc! {"
+        class A:
+            def foo(self):
+                pass
+        a = A()
+    "};
+    "register"
+)]
+fn test_simple_register_without_constructor(source: &str) {
+    let mut composer =
+        TopLevelComposer::new(Vec::new(), Vec::new(), ComposerConfig::default(), 64).0;
+    let ast = parse_program(source, FileName::default()).unwrap();
+    let ast = ast[0].clone();
+    composer.register_top_level(ast, None, "", true).unwrap();
+}
+
+#[test_case(
+    &[
+        indoc! {"
+            def fun(a: int32) -> int32:
+                return a
+        "},
+        indoc! {"
+            def foo(a: float):
+                a + 1.0
+        "},
+        indoc! {"
+            def f(b: int64) -> int32:
+                return 3
+        "},
+    ],
+    &[
+        "fn[[a:0], 0]",
+        "fn[[a:2], 4]",
+        "fn[[b:1], 0]",
+    ],
+    &[
+        "fun",
+        "foo",
+        "f"
+    ];
+    "function compose"
+)]
+fn test_simple_function_analyze(source: &[&str], tys: &[&str], names: &[&str]) {
+    let mut composer =
+        TopLevelComposer::new(Vec::new(), Vec::new(), ComposerConfig::default(), 64).0;
+
+    let internal_resolver = Arc::new(ResolverInternal {
+        id_to_def: Mutex::default(),
+        id_to_type: Mutex::default(),
+        class_names: Mutex::default(),
+    });
+    let resolver =
+        Arc::new(Resolver(internal_resolver.clone())) as Arc<dyn SymbolResolver + Send + Sync>;
+
+    for s in source {
+        let ast = parse_program(s, FileName::default()).unwrap();
+        let ast = ast[0].clone();
+
+        let (id, def_id, ty) =
+            composer.register_top_level(ast, Some(resolver.clone()), "", false).unwrap();
+        internal_resolver.add_id_def(id, def_id);
+        if let Some(ty) = ty {
+            internal_resolver.add_id_type(id, ty);
+        }
+    }
+
+    composer.start_analysis(true).unwrap();
+
+    for (i, (def, _)) in composer.definition_ast_list.iter().skip(composer.builtin_num).enumerate()
+    {
+        let def = &*def.read();
+        if let TopLevelDef::Function { signature, name, .. } = def {
+            let ty_str = composer.unifier.internal_stringify(
+                *signature,
+                &mut |id| id.to_string(),
+                &mut |id| id.to_string(),
+                &mut None,
+            );
+            assert_eq!(ty_str, tys[i]);
+            assert_eq!(name, names[i]);
+        }
+    }
+}
+
+#[test_case(
+    &[
+        indoc! {"
+            class A():
+                a: int32
+                def __init__(self):
+                    self.a = 3
+                def fun(self, b: B):
+                    pass
+                def foo(self, a: T, b: V):
+                    pass
+        "},
+        indoc! {"
+            class B(C):
+                def __init__(self):
+                    pass
+        "},
+        indoc! {"
+            class C(A):
+                def __init__(self):
+                    pass
+                def fun(self, b: B):
+                    a = 1
+                    pass
+        "},
+        indoc! {"
+            def foo(a: A):
+                pass
+        "},
+        indoc! {"
+            def ff(a: T) -> V:
+                pass
+        "}
+    ],
+    &[];
+    "simple class compose"
+)]
+#[test_case(
+    &[
+        indoc! {"
+            class Generic_A(Generic[V], B):
+                a: int64
+                def __init__(self):
+                    self.a = 123123123123
+                def fun(self, a: int32) -> V:
+                    pass
+        "},
+        indoc! {"
+            class B:
+                aa: bool
+                def __init__(self):
+                    self.aa = False
+                def foo(self, b: T):
+                    pass
+        "}
+    ],
+    &[];
+    "generic class"
+)]
+#[test_case(
+    &[
+        indoc! {"
+            def foo(a: list[int32], b: tuple[T, float]) -> A[B, bool]:
+                pass
+        "},
+        indoc! {"
+            class A(Generic[T, V]):
+                a: T
+                b: V
+                def __init__(self, v: V):
+                    self.a = 1
+                    self.b = v
+                def fun(self, a: T) -> V:
+                    pass
+        "},
+        indoc! {"
+            def gfun(a: A[list[float], int32]):
+                pass
+        "},
+        indoc! {"
+            class B:
+                def __init__(self):
+                    pass
+        "}
+    ],
+    &[];
+    "list tuple generic"
+)]
+#[test_case(
+    &[
+        indoc! {"
+            class A(Generic[T, V]):
+                a: A[float, bool]
+                b: B
+                def __init__(self, a: A[float, bool], b: B):
+                    self.a = a
+                    self.b = b
+                def fun(self, a: A[float, bool]) -> A[bool, int32]:
+                    pass
+        "},
+        indoc! {"
+            class B(A[int64, bool]):
+                def __init__(self):
+                    pass
+                def foo(self, b: B) -> B:
+                    pass
+                def bar(self, a: A[list[B], int32]) -> tuple[A[virtual[A[B, int32]], bool], B]:
+                    pass
+        "}
+    ],
+    &[];
+    "self1"
+)]
+#[test_case(
+    &[
+        indoc! {"
+            class A(Generic[T]):
+                a: int32
+                b: T
+                c: A[int64]
+                def __init__(self, t: T):
+                    self.a = 3
+                    self.b = T
+                def fun(self, a: int32, b: T) -> list[virtual[B[bool]]]:
+                    pass
+                def foo(self, c: C):
+                    pass
+        "},
+        indoc! {"
+            class B(Generic[V], A[float]):
+                d: C
+                def __init__(self):
+                    pass
+                def fun(self, a: int32, b: T) -> list[virtual[B[bool]]]:
+                    # override
+                    pass
+        "},
+        indoc! {"
+            class C(B[bool]):
+                e: int64
+                def __init__(self):
+                    pass
+        "}
+    ],
+    &[];
+    "inheritance_override"
+)]
+#[test_case(
+    &[
+        indoc! {"
+            class A(Generic[T]):
+                def __init__(self):
+                    pass
+                def fun(self, a: A[T]) -> A[T]:
+                    pass
+        "}
+    ],
+    &["application of type vars to generic class is not currently supported (at unknown:4:24)"];
+    "err no type var in generic app"
+)]
+#[test_case(
+    &[
+        indoc! {"
+            class A(B):
+                def __init__(self):
+                    pass
+        "},
+        indoc! {"
+            class B(A):
+                def __init__(self):
+                    pass
+        "}
+    ],
+    &["cyclic inheritance detected"];
+    "cyclic1"
+)]
+#[test_case(
+    &[
+        indoc! {"
+            class A(B[bool, int64]):
+                def __init__(self):
+                    pass
+        "},
+        indoc! {"
+            class B(Generic[V, T], C[int32]):
+                def __init__(self):
+                    pass
+        "},
+        indoc! {"
+            class C(Generic[T], A):
+                def __init__(self):
+                    pass
+        "},
+    ],
+    &["cyclic inheritance detected"];
+    "cyclic2"
+)]
+#[test_case(
+    &[
+        indoc! {"
+            class A:
+                pass
+        "}
+    ],
+    &["5: Class {\nname: \"A\",\ndef_id: DefinitionId(5),\nancestors: [CustomClassKind { id: DefinitionId(5), params: [] }],\nfields: [],\nmethods: [],\ntype_vars: []\n}"];
+    "simple pass in class"
+)]
+#[test_case(
+    &[indoc! {"
+        class A:
+            def __init__():
+                pass
+    "}],
+    &["__init__ method must have a `self` parameter (at unknown:2:5)"];
+    "err no self_1"
+)]
+#[test_case(
+    &[
+        indoc! {"
+            class A(B, Generic[T], C):
+                def __init__(self):
+                    pass
+        "},
+        indoc! {"
+            class B:
+                def __init__(self):
+                    pass
+        "},
+        indoc! {"
+            class C:
+                def __init__(self):
+                    pass
+        "}
+
+    ],
+    &["a class definition can only have at most one base class declaration and one generic declaration (at unknown:1:24)"];
+    "err multiple inheritance"
+)]
+#[test_case(
+    &[
+        indoc! {"
+            class A(Generic[T]):
+                a: int32
+                b: T
+                c: A[int64]
+                def __init__(self, t: T):
+                    self.a = 3
+                    self.b = T
+                def fun(self, a: int32, b: T) -> list[virtual[B[bool]]]:
+                    pass
+        "},
+        indoc! {"
+            class B(Generic[V], A[float]):
+                def __init__(self):
+                    pass
+                def fun(self, a: int32, b: T) -> list[virtual[B[int32]]]:
+                    # override
+                    pass
+        "}
+    ],
+    &["method fun has same name as ancestors' method, but incompatible type"];
+    "err_incompatible_inheritance_method"
+)]
+#[test_case(
+    &[
+        indoc! {"
+            class A(Generic[T]):
+                a: int32
+                b: T
+                c: A[int64]
+                def __init__(self, t: T):
+                    self.a = 3
+                    self.b = T
+                def fun(self, a: int32, b: T) -> list[virtual[B[bool]]]:
+                    pass
+        "},
+        indoc! {"
+            class B(Generic[V], A[float]):
+                a: int32
+                def __init__(self):
+                    pass
+                def fun(self, a: int32, b: T) -> list[virtual[B[bool]]]:
+                    # override
+                    pass
+        "}
+    ],
+    &["field `a` has already declared in the ancestor classes"];
+    "err_incompatible_inheritance_field"
+)]
+#[test_case(
+    &[
+        indoc! {"
+            class A:
+                def __init__(self):
+                    pass
+        "},
+        indoc! {"
+            class A:
+                a: int32
+                def __init__(self):
+                    pass
+        "}
+    ],
+    &["duplicate definition of class `A` (at unknown:1:1)"];
+    "class same name"
+)]
+fn test_analyze(source: &[&str], res: &[&str]) {
+    let print = false;
+    let mut composer =
+        TopLevelComposer::new(Vec::new(), Vec::new(), ComposerConfig::default(), 64).0;
+
+    let internal_resolver = make_internal_resolver_with_tvar(
+        vec![
+            ("T".into(), vec![]),
+            ("V".into(), vec![composer.primitives_ty.bool, composer.primitives_ty.int32]),
+            ("G".into(), vec![composer.primitives_ty.bool, composer.primitives_ty.int64]),
+        ],
+        &mut composer.unifier,
+        print,
+    );
+    let resolver =
+        Arc::new(Resolver(internal_resolver.clone())) as Arc<dyn SymbolResolver + Send + Sync>;
+
+    for s in source {
+        let ast = parse_program(s, FileName::default()).unwrap();
+        let ast = ast[0].clone();
+
+        let (id, def_id, ty) = {
+            match composer.register_top_level(ast, Some(resolver.clone()), "", false) {
+                Ok(x) => x,
+                Err(msg) => {
+                    if print {
+                        println!("{msg}");
+                    } else {
+                        assert_eq!(res[0], msg);
+                    }
+                    return;
+                }
+            }
+        };
+        internal_resolver.add_id_def(id, def_id);
+        if let Some(ty) = ty {
+            internal_resolver.add_id_type(id, ty);
+        }
+    }
+
+    if let Err(msg) = composer.start_analysis(false) {
+        if print {
+            println!("{}", msg.iter().sorted().join("\n----------\n"));
+        } else {
+            assert_eq!(res[0], msg.iter().next().unwrap());
+        }
+    } else {
+        // skip 5 to skip primitives
+        let mut res_vec: Vec<String> = Vec::new();
+        for (def, _) in composer.definition_ast_list.iter().skip(composer.builtin_num) {
+            let def = &*def.read();
+            res_vec.push(format!("{}\n", def.to_string(composer.unifier.borrow_mut())));
+        }
+        insta::assert_debug_snapshot!(res_vec);
+    }
+}
+
+#[test_case(
+    vec![
+        indoc! {"
+            def fun(a: int32, b: int32) -> int32:
+                return a + b
+        "},
+        indoc! {"
+            def fib(n: int32) -> int32:
+                if n <= 2:
+                    return 1
+                a = fib(n - 1)
+                b = fib(n - 2)
+                return fib(n - 1)
+        "}
+    ],
+    &[];
+    "simple function"
+)]
+#[test_case(
+    vec![
+        indoc! {"
+            class A:
+                a: int32
+                def __init__(self):
+                    self.a = 3
+                def fun(self) -> int32:
+                    b = self.a + 3
+                    return b * self.a
+                def clone(self) -> A:
+                    SELF = self
+                    return SELF
+                def sum(self) -> int32:
+                    if self.a == 0:
+                        return self.a
+                    else:
+                        a = self.a
+                        self.a = self.a - 1
+                        return a + self.sum()
+                def fib(self, a: int32) -> int32:
+                    if a <= 2:
+                        return 1
+                    return self.fib(a - 1) + self.fib(a - 2)
+        "},
+        indoc! {"
+            def fun(a: A) -> int32:
+                return a.fun() + 2
+        "}
+    ],
+    &[];
+    "simple class body"
+)]
+#[test_case(
+    vec![
+        indoc! {"
+            def fun(a: V, c: G, t: T) -> V:
+                b = a
+                cc = c
+                ret = fun(b, cc, t)
+                return ret * ret
+        "},
+        indoc! {"
+            def sum_three(l: list[V]) -> V:
+                return l[0] + l[1] + l[2]
+        "},
+        indoc! {"
+            def sum_sq_pair(p: tuple[V, V]) -> list[V]:
+                a = p[0]
+                b = p[1]
+                a = a**a
+                b = b**b
+                return [a, b]
+        "}
+    ],
+    &[];
+    "type var fun"
+)]
+#[test_case(
+    vec![
+        indoc! {"
+            class A(Generic[G]):
+                a: G
+                b: bool
+                def __init__(self, aa: G):
+                    self.a = aa
+                    if 2 > 1:
+                        self.b = True
+                    else:
+                        # self.b = False
+                        pass
+                def fun(self, a: G) -> list[G]:
+                    ret = [a, self.a]
+                    return ret if self.b else self.fun(self.a)
+        "}
+    ],
+    &[];
+    "type var class"
+)]
+#[test_case(
+    vec![
+        indoc! {"
+            class A:
+                def fun(self):
+                    pass
+        "},
+        indoc!{"
+            class B:
+                a: int32
+                b: bool
+                def __init__(self):
+                    # self.b = False
+                    if 3 > 2:
+                        self.a = 3
+                        self.b = False
+                    else:
+                        self.a = 4
+                        self.b = True
+        "}
+    ],
+    &[];
+    "no_init_inst_check"
+)]
+fn test_inference(source: Vec<&str>, res: &[&str]) {
+    let print = true;
+    let mut composer =
+        TopLevelComposer::new(Vec::new(), Vec::new(), ComposerConfig::default(), 64).0;
+
+    let internal_resolver = make_internal_resolver_with_tvar(
+        vec![
+            ("T".into(), vec![]),
+            (
+                "V".into(),
+                vec![
+                    composer.primitives_ty.float,
+                    composer.primitives_ty.int32,
+                    composer.primitives_ty.int64,
+                ],
+            ),
+            ("G".into(), vec![composer.primitives_ty.bool, composer.primitives_ty.int64]),
+        ],
+        &mut composer.unifier,
+        print,
+    );
+    let resolver =
+        Arc::new(Resolver(internal_resolver.clone())) as Arc<dyn SymbolResolver + Send + Sync>;
+
+    for s in source {
+        let ast = parse_program(s, FileName::default()).unwrap();
+        let ast = ast[0].clone();
+
+        let (id, def_id, ty) = {
+            match composer.register_top_level(ast, Some(resolver.clone()), "", false) {
+                Ok(x) => x,
+                Err(msg) => {
+                    if print {
+                        println!("{msg}");
+                    } else {
+                        assert_eq!(res[0], msg);
+                    }
+                    return;
+                }
+            }
+        };
+        internal_resolver.add_id_def(id, def_id);
+        if let Some(ty) = ty {
+            internal_resolver.add_id_type(id, ty);
+        }
+    }
+
+    if let Err(msg) = composer.start_analysis(true) {
+        if print {
+            println!("{}", msg.iter().sorted().join("\n----------\n"));
+        } else {
+            assert_eq!(res[0], msg.iter().next().unwrap());
+        }
+    } else {
+        // skip 5 to skip primitives
+        let mut stringify_folder = TypeToStringFolder { unifier: &mut composer.unifier };
+        for (def, _) in composer.definition_ast_list.iter().skip(composer.builtin_num) {
+            let def = &*def.read();
+
+            if let TopLevelDef::Function { instance_to_stmt, name, .. } = def {
+                println!(
+                    "=========`{}`: number of instances: {}===========",
+                    name,
+                    instance_to_stmt.len()
+                );
+                for inst in instance_to_stmt {
+                    let ast = &inst.1.body;
+                    for b in ast.iter() {
+                        println!("{:?}", stringify_folder.fold_stmt(b.clone()).unwrap());
+                        println!("--------------------");
+                    }
+                    println!("\n");
+                }
+            }
+        }
+    }
+}
+
+fn make_internal_resolver_with_tvar(
+    tvars: Vec<(StrRef, Vec<Type>)>,
+    unifier: &mut Unifier,
+    print: bool,
+) -> Arc<ResolverInternal> {
+    let list_elem_tvar = unifier.get_fresh_var(Some("list_elem".into()), None);
+    let list = unifier.add_ty(TypeEnum::TObj {
+        obj_id: PrimDef::List.id(),
+        fields: HashMap::new(),
+        params: into_var_map([list_elem_tvar]),
+    });
+
+    let res: Arc<ResolverInternal> = ResolverInternal {
+        id_to_def: Mutex::new(HashMap::from([("list".into(), PrimDef::List.id())])),
+        id_to_type: tvars
+            .into_iter()
+            .map(|(name, range)| {
+                (name, {
+                    let tvar = unifier.get_fresh_var_with_range(range.as_slice(), None, None);
+                    if print {
+                        println!("{}: {:?}, typevar{}", name, tvar.ty, tvar.id);
+                    }
+                    tvar.ty
+                })
+            })
+            .collect::<HashMap<_, _>>()
+            .into(),
+        class_names: Mutex::new(HashMap::from([("list".into(), list)])),
+    }
+    .into();
+    if print {
+        println!();
+    }
+    res
+}
+
+struct TypeToStringFolder<'a> {
+    unifier: &'a mut Unifier,
+}
+
+impl<'a> Fold<Option<Type>> for TypeToStringFolder<'a> {
+    type TargetU = String;
+    type Error = String;
+    fn map_user(&mut self, user: Option<Type>) -> Result<Self::TargetU, Self::Error> {
+        Ok(if let Some(ty) = user {
+            self.unifier.internal_stringify(
+                ty,
+                &mut |id| format!("class{id}"),
+                &mut |id| format!("typevar{id}"),
+                &mut None,
+            )
+        } else {
+            "None".into()
+        })
+    }
+}

nac3core/src/toplevel/type_annotation.rs

@@ -0,0 +1,660 @@
+use strum::IntoEnumIterator;
+
+use nac3parser::ast::Constant;
+
+use super::*;
+use crate::{
+    symbol_resolver::SymbolValue,
+    toplevel::helper::{PrimDef, PrimDefDetails},
+    typecheck::typedef::VarMap,
+};
+
+#[derive(Clone, Debug)]
+pub enum TypeAnnotation {
+    Primitive(Type),
+    // we use type vars kind at params to represent self type
+    CustomClass {
+        id: DefinitionId,
+        // params can also be type var
+        params: Vec<TypeAnnotation>,
+    },
+    // can only be CustomClassKind
+    Virtual(Box<TypeAnnotation>),
+    TypeVar(Type),
+    /// A `Literal` allowing a subset of literals.
+    Literal(Vec<Constant>),
+    Tuple(Vec<TypeAnnotation>),
+}
+
+impl TypeAnnotation {
+    pub fn stringify(&self, unifier: &mut Unifier) -> String {
+        use TypeAnnotation::*;
+        match self {
+            Primitive(ty) | TypeVar(ty) => unifier.stringify(*ty),
+            CustomClass { id, params } => {
+                let class_name = if let Some(ref top) = unifier.top_level {
+                    if let TopLevelDef::Class { name, .. } = &*top.definitions.read()[id.0].read() {
+                        (*name).into()
+                    } else {
+                        unreachable!()
+                    }
+                } else {
+                    format!("class_def_{}", id.0)
+                };
+                format!("{}{}", class_name, {
+                    let param_list =
+                        params.iter().map(|p| p.stringify(unifier)).collect_vec().join(", ");
+                    if param_list.is_empty() {
+                        String::new()
+                    } else {
+                        format!("[{param_list}]")
+                    }
+                })
+            }
+            Literal(values) => {
+                format!("Literal({})", values.iter().map(|v| format!("{v:?}")).join(", "))
+            }
+            Virtual(ty) => format!("virtual[{}]", ty.stringify(unifier)),
+            Tuple(types) => {
+                format!(
+                    "tuple[{}]",
+                    types.iter().map(|p| p.stringify(unifier)).collect_vec().join(", ")
+                )
+            }
+        }
+    }
+}
+
+/// Parses an AST expression `expr` into a [`TypeAnnotation`].
+///
+/// * `locked` - A [`HashMap`] containing the IDs of known definitions, mapped to a [`Vec`] of all
+///   generic variables associated with the definition.
+/// * `type_var` - The type variable associated with the type argument currently being parsed. Pass
+///   [`None`] when this function is invoked externally.
+pub fn parse_ast_to_type_annotation_kinds<T, S: std::hash::BuildHasher + Clone>(
+    resolver: &(dyn SymbolResolver + Send + Sync),
+    top_level_defs: &[Arc<RwLock<TopLevelDef>>],
+    unifier: &mut Unifier,
+    primitives: &PrimitiveStore,
+    expr: &ast::Expr<T>,
+    // the key stores the type_var of this topleveldef::class, we only need this field here
+    locked: HashMap<DefinitionId, Vec<Type>, S>,
+) -> Result<TypeAnnotation, HashSet<String>> {
+    let name_handle = |id: &StrRef,
+                       unifier: &mut Unifier,
+                       locked: HashMap<DefinitionId, Vec<Type>, S>| {
+        if id == &"int32".into() {
+            Ok(TypeAnnotation::Primitive(primitives.int32))
+        } else if id == &"int64".into() {
+            Ok(TypeAnnotation::Primitive(primitives.int64))
+        } else if id == &"uint32".into() {
+            Ok(TypeAnnotation::Primitive(primitives.uint32))
+        } else if id == &"uint64".into() {
+            Ok(TypeAnnotation::Primitive(primitives.uint64))
+        } else if id == &"float".into() {
+            Ok(TypeAnnotation::Primitive(primitives.float))
+        } else if id == &"bool".into() {
+            Ok(TypeAnnotation::Primitive(primitives.bool))
+        } else if id == &"str".into() {
+            Ok(TypeAnnotation::Primitive(primitives.str))
+        } else if id == &"Exception".into() {
+            Ok(TypeAnnotation::CustomClass { id: PrimDef::Exception.id(), params: Vec::default() })
+        } else if let Ok(obj_id) = resolver.get_identifier_def(*id) {
+            let type_vars = {
+                let def_read = top_level_defs[obj_id.0].try_read();
+                if let Some(def_read) = def_read {
+                    if let TopLevelDef::Class { type_vars, .. } = &*def_read {
+                        type_vars.clone()
+                    } else {
+                        return Err(HashSet::from([format!(
+                            "function cannot be used as a type (at {})",
+                            expr.location
+                        )]));
+                    }
+                } else {
+                    locked.get(&obj_id).unwrap().clone()
+                }
+            };
+            // check param number here
+            if !type_vars.is_empty() {
+                return Err(HashSet::from([format!(
+                    "expect {} type variable parameter but got 0 (at {})",
+                    type_vars.len(),
+                    expr.location,
+                )]));
+            }
+            Ok(TypeAnnotation::CustomClass { id: obj_id, params: vec![] })
+        } else if let Ok(ty) = resolver.get_symbol_type(unifier, top_level_defs, primitives, *id) {
+            if let TypeEnum::TVar { .. } = unifier.get_ty(ty).as_ref() {
+                let var = unifier.get_fresh_var(Some(*id), Some(expr.location)).ty;
+                unifier.unify(var, ty).unwrap();
+                Ok(TypeAnnotation::TypeVar(ty))
+            } else {
+                Err(HashSet::from([format!(
+                    "`{}` is not a valid type annotation (at {})",
+                    id, expr.location
+                )]))
+            }
+        } else {
+            Err(HashSet::from([format!(
+                "`{}` is not a valid type annotation (at {})",
+                id, expr.location
+            )]))
+        }
+    };
+
+    let class_name_handle =
+        |id: &StrRef,
+         slice: &ast::Expr<T>,
+         unifier: &mut Unifier,
+         mut locked: HashMap<DefinitionId, Vec<Type>, S>| {
+            if ["virtual".into(), "Generic".into(), "tuple".into(), "Option".into()].contains(id) {
+                return Err(HashSet::from([format!(
+                    "keywords cannot be class name (at {})",
+                    expr.location
+                )]));
+            }
+            let obj_id = resolver.get_identifier_def(*id)?;
+            let type_vars = {
+                let def_read = top_level_defs[obj_id.0].try_read();
+                if let Some(def_read) = def_read {
+                    let TopLevelDef::Class { type_vars, .. } = &*def_read else {
+                        unreachable!("must be class here")
+                    };
+
+                    type_vars.clone()
+                } else {
+                    locked.get(&obj_id).unwrap().clone()
+                }
+            };
+            // we do not check whether the application of type variables are compatible here
+            let param_type_infos = {
+                let params_ast = if let ast::ExprKind::Tuple { elts, .. } = &slice.node {
+                    elts.iter().collect_vec()
+                } else {
+                    vec![slice]
+                };
+                if type_vars.len() != params_ast.len() {
+                    return Err(HashSet::from([format!(
+                        "expect {} type parameters but got {} (at {})",
+                        type_vars.len(),
+                        params_ast.len(),
+                        params_ast[0].location,
+                    )]));
+                }
+                let result = params_ast
+                    .iter()
+                    .map(|x| {
+                        parse_ast_to_type_annotation_kinds(
+                            resolver,
+                            top_level_defs,
+                            unifier,
+                            primitives,
+                            x,
+                            {
+                                locked.insert(obj_id, type_vars.clone());
+                                locked.clone()
+                            },
+                        )
+                    })
+                    .collect::<Result<Vec<_>, _>>()?;
+                // make sure the result do not contain any type vars
+                let no_type_var =
+                    result.iter().all(|x| get_type_var_contained_in_type_annotation(x).is_empty());
+                if no_type_var {
+                    result
+                } else {
+                    return Err(HashSet::from([
+                        format!(
+                            "application of type vars to generic class is not currently supported (at {})",
+                            params_ast[0].location
+                        ),
+                    ]));
+                }
+            };
+            Ok(TypeAnnotation::CustomClass { id: obj_id, params: param_type_infos })
+        };
+
+    match &expr.node {
+        ast::ExprKind::Name { id, .. } => name_handle(id, unifier, locked),
+        // virtual
+        ast::ExprKind::Subscript { value, slice, .. }
+            if {
+                matches!(&value.node, ast::ExprKind::Name { id, .. } if id == &"virtual".into())
+            } =>
+        {
+            let def = parse_ast_to_type_annotation_kinds(
+                resolver,
+                top_level_defs,
+                unifier,
+                primitives,
+                slice.as_ref(),
+                locked,
+            )?;
+            if !matches!(def, TypeAnnotation::CustomClass { .. }) {
+                unreachable!("must be concretized custom class kind in the virtual")
+            }
+            Ok(TypeAnnotation::Virtual(def.into()))
+        }
+
+        // option
+        ast::ExprKind::Subscript { value, slice, .. }
+            if {
+                matches!(&value.node, ast::ExprKind::Name { id, .. } if id == &"Option".into())
+            } =>
+        {
+            let def_ann = parse_ast_to_type_annotation_kinds(
+                resolver,
+                top_level_defs,
+                unifier,
+                primitives,
+                slice.as_ref(),
+                locked,
+            )?;
+            let id =
+                if let TypeEnum::TObj { obj_id, .. } = unifier.get_ty(primitives.option).as_ref() {
+                    *obj_id
+                } else {
+                    unreachable!()
+                };
+            Ok(TypeAnnotation::CustomClass { id, params: vec![def_ann] })
+        }
+
+        // tuple
+        ast::ExprKind::Subscript { value, slice, .. }
+            if {
+                matches!(&value.node, ast::ExprKind::Name { id, .. } if id == &"tuple".into())
+            } =>
+        {
+            let tup_elts = {
+                if let ast::ExprKind::Tuple { elts, .. } = &slice.node {
+                    elts.as_slice()
+                } else {
+                    std::slice::from_ref(slice.as_ref())
+                }
+            };
+            let type_annotations = tup_elts
+                .iter()
+                .map(|e| {
+                    parse_ast_to_type_annotation_kinds(
+                        resolver,
+                        top_level_defs,
+                        unifier,
+                        primitives,
+                        e,
+                        locked.clone(),
+                    )
+                })
+                .collect::<Result<Vec<_>, _>>()?;
+            Ok(TypeAnnotation::Tuple(type_annotations))
+        }
+
+        // Literal
+        ast::ExprKind::Subscript { value, slice, .. }
+            if {
+                matches!(&value.node, ast::ExprKind::Name { id, .. } if id == &"Literal".into())
+            } =>
+        {
+            let tup_elts = {
+                if let ast::ExprKind::Tuple { elts, .. } = &slice.node {
+                    elts.as_slice()
+                } else {
+                    std::slice::from_ref(slice.as_ref())
+                }
+            };
+            let type_annotations = tup_elts
+                .iter()
+                .map(|e| match &e.node {
+                    ast::ExprKind::Constant { value, .. } => {
+                        Ok(TypeAnnotation::Literal(vec![value.clone()]))
+                    }
+                    _ => parse_ast_to_type_annotation_kinds(
+                        resolver,
+                        top_level_defs,
+                        unifier,
+                        primitives,
+                        e,
+                        locked.clone(),
+                    ),
+                })
+                .collect::<Result<Vec<_>, _>>()?
+                .into_iter()
+                .flat_map(|type_ann| match type_ann {
+                    TypeAnnotation::Literal(values) => values,
+                    _ => unreachable!(),
+                })
+                .collect_vec();
+
+            if type_annotations.len() == 1 {
+                Ok(TypeAnnotation::Literal(type_annotations))
+            } else {
+                Err(HashSet::from([format!(
+                    "multiple literal bounds are currently unsupported (at {})",
+                    value.location
+                )]))
+            }
+        }
+
+        // custom class
+        ast::ExprKind::Subscript { value, slice, .. } => {
+            if let ast::ExprKind::Name { id, .. } = &value.node {
+                class_name_handle(id, slice, unifier, locked)
+            } else {
+                Err(HashSet::from([format!(
+                    "unsupported expression type for class name (at {})",
+                    value.location
+                )]))
+            }
+        }
+
+        ast::ExprKind::Constant { value, .. } => Ok(TypeAnnotation::Literal(vec![value.clone()])),
+
+        _ => Err(HashSet::from([format!(
+            "unsupported expression for type annotation (at {})",
+            expr.location
+        )])),
+    }
+}
+
+// no need to have the `locked` parameter, unlike the `parse_ast_to_type_annotation_kinds`, since
+// when calling this function, there should be no topleveldef::class being write, and this function
+// also only read the toplevedefs
+pub fn get_type_from_type_annotation_kinds(
+    top_level_defs: &[Arc<RwLock<TopLevelDef>>],
+    unifier: &mut Unifier,
+    primitives: &PrimitiveStore,
+    ann: &TypeAnnotation,
+    subst_list: &mut Option<Vec<Type>>,
+) -> Result<Type, HashSet<String>> {
+    match ann {
+        TypeAnnotation::CustomClass { id: obj_id, params } => {
+            let def_read = top_level_defs[obj_id.0].read();
+            let class_def: &TopLevelDef = &def_read;
+            let TopLevelDef::Class { fields, methods, type_vars, .. } = class_def else {
+                unreachable!("should be class def here")
+            };
+
+            if type_vars.len() != params.len() {
+                return Err(HashSet::from([format!(
+                    "unexpected number of type parameters: expected {} but got {}",
+                    type_vars.len(),
+                    params.len()
+                )]));
+            }
+
+            let param_ty = params
+                .iter()
+                .map(|x| {
+                    get_type_from_type_annotation_kinds(
+                        top_level_defs,
+                        unifier,
+                        primitives,
+                        x,
+                        subst_list,
+                    )
+                })
+                .collect::<Result<Vec<_>, _>>()?;
+
+            let ty = if let Some(prim_def) = PrimDef::iter().find(|prim| prim.id() == *obj_id) {
+                // Primitive TopLevelDefs do not contain all fields that are present in their Type
+                // counterparts, so directly perform subst on the Type instead.
+
+                let PrimDefDetails::PrimClass { get_ty_fn, .. } = prim_def.details() else {
+                    unreachable!()
+                };
+
+                let base_ty = get_ty_fn(primitives);
+                let params =
+                    if let TypeEnum::TObj { params, .. } = &*unifier.get_ty_immutable(base_ty) {
+                        params.clone()
+                    } else {
+                        unreachable!()
+                    };
+
+                unifier
+                    .subst(
+                        get_ty_fn(primitives),
+                        &params
+                            .iter()
+                            .zip(param_ty)
+                            .map(|(obj_tv, param)| (*obj_tv.0, param))
+                            .collect(),
+                    )
+                    .unwrap_or(base_ty)
+            } else {
+                let subst = {
+                    // check for compatible range
+                    // TODO: if allow type var to be applied(now this disallowed in the parse_to_type_annotation), need more check
+                    let mut result = VarMap::new();
+                    for (tvar, p) in type_vars.iter().zip(param_ty) {
+                        match unifier.get_ty(*tvar).as_ref() {
+                            TypeEnum::TVar {
+                                id,
+                                range,
+                                fields: None,
+                                name,
+                                loc,
+                                is_const_generic: false,
+                            } => {
+                                let ok: bool = {
+                                    // create a temp type var and unify to check compatibility
+                                    p == *tvar || {
+                                        let temp = unifier.get_fresh_var_with_range(
+                                            range.as_slice(),
+                                            *name,
+                                            *loc,
+                                        );
+                                        unifier.unify(temp.ty, p).is_ok()
+                                    }
+                                };
+                                if ok {
+                                    result.insert(*id, p);
+                                } else {
+                                    return Err(HashSet::from([format!(
+                                        "cannot apply type {} to type variable with id {:?}",
+                                        unifier.internal_stringify(
+                                            p,
+                                            &mut |id| format!("class{id}"),
+                                            &mut |id| format!("typevar{id}"),
+                                            &mut None
+                                        ),
+                                        *id
+                                    )]));
+                                }
+                            }
+
+                            TypeEnum::TVar {
+                                id, range, name, loc, is_const_generic: true, ..
+                            } => {
+                                let ty = range[0];
+                                let ok: bool = {
+                                    // create a temp type var and unify to check compatibility
+                                    p == *tvar || {
+                                        let temp =
+                                            unifier.get_fresh_const_generic_var(ty, *name, *loc);
+                                        unifier.unify(temp.ty, p).is_ok()
+                                    }
+                                };
+                                if ok {
+                                    result.insert(*id, p);
+                                } else {
+                                    return Err(HashSet::from([format!(
+                                        "cannot apply type {} to type variable {}",
+                                        unifier.stringify(p),
+                                        name.unwrap_or_else(|| format!("typevar{id}").into()),
+                                    )]));
+                                }
+                            }
+
+                            _ => unreachable!("must be generic type var"),
+                        }
+                    }
+                    result
+                };
+                // Class Attributes keep a copy with Class Definition and are not added to objects
+                let mut tobj_fields = methods
+                    .iter()
+                    .map(|(name, ty, _)| {
+                        let subst_ty = unifier.subst(*ty, &subst).unwrap_or(*ty);
+                        // methods are immutable
+                        (*name, (subst_ty, false))
+                    })
+                    .collect::<HashMap<_, _>>();
+                tobj_fields.extend(fields.iter().map(|(name, ty, mutability)| {
+                    let subst_ty = unifier.subst(*ty, &subst).unwrap_or(*ty);
+                    (*name, (subst_ty, *mutability))
+                }));
+                let need_subst = !subst.is_empty();
+                let ty = unifier.add_ty(TypeEnum::TObj {
+                    obj_id: *obj_id,
+                    fields: tobj_fields,
+                    params: subst,
+                });
+
+                if need_subst {
+                    if let Some(wl) = subst_list.as_mut() {
+                        wl.push(ty);
+                    }
+                }
+
+                ty
+            };
+
+            Ok(ty)
+        }
+        TypeAnnotation::Primitive(ty) | TypeAnnotation::TypeVar(ty) => Ok(*ty),
+        TypeAnnotation::Literal(values) => {
+            let values = values
+                .iter()
+                .map(SymbolValue::from_constant_inferred)
+                .collect::<Result<Vec<_>, _>>()
+                .map_err(|err| HashSet::from([err]))?;
+
+            let var = unifier.get_fresh_literal(values, None);
+            Ok(var)
+        }
+        TypeAnnotation::Virtual(ty) => {
+            let ty = get_type_from_type_annotation_kinds(
+                top_level_defs,
+                unifier,
+                primitives,
+                ty.as_ref(),
+                subst_list,
+            )?;
+            Ok(unifier.add_ty(TypeEnum::TVirtual { ty }))
+        }
+        TypeAnnotation::Tuple(tys) => {
+            let tys = tys
+                .iter()
+                .map(|x| {
+                    get_type_from_type_annotation_kinds(
+                        top_level_defs,
+                        unifier,
+                        primitives,
+                        x,
+                        subst_list,
+                    )
+                })
+                .collect::<Result<Vec<_>, _>>()?;
+            Ok(unifier.add_ty(TypeEnum::TTuple { ty: tys, is_vararg_ctx: false }))
+        }
+    }
+}
+
+/// given an def id, return a type annotation of self \
+/// ```python
+/// class A(Generic[T, V]):
+///     def fun(self):
+/// ```
+/// the type of `self` should be similar to `A[T, V]`, where `T`, `V`
+/// considered to be type variables associated with the class \
+/// \
+/// But note that here we do not make a duplication of `T`, `V`, we directly
+/// use them as they are in the [`TopLevelDef::Class`] since those in the
+/// `TopLevelDef::Class.type_vars` will be substitute later when seeing applications/instantiations
+/// the Type of their fields and methods will also be subst when application/instantiation
+#[must_use]
+pub fn make_self_type_annotation(type_vars: &[Type], object_id: DefinitionId) -> TypeAnnotation {
+    TypeAnnotation::CustomClass {
+        id: object_id,
+        params: type_vars.iter().map(|ty| TypeAnnotation::TypeVar(*ty)).collect_vec(),
+    }
+}
+
+/// get all the occurences of type vars contained in a type annotation
+/// e.g. `A[int, B[T], V, virtual[C[G]]]` => [T, V, G]
+/// this function will not make a duplicate of type var
+#[must_use]
+pub fn get_type_var_contained_in_type_annotation(ann: &TypeAnnotation) -> Vec<TypeAnnotation> {
+    let mut result: Vec<TypeAnnotation> = Vec::new();
+    match ann {
+        TypeAnnotation::TypeVar(..) => result.push(ann.clone()),
+        TypeAnnotation::Virtual(ann) => {
+            result.extend(get_type_var_contained_in_type_annotation(ann.as_ref()));
+        }
+        TypeAnnotation::CustomClass { params, .. } => {
+            for p in params {
+                result.extend(get_type_var_contained_in_type_annotation(p));
+            }
+        }
+        TypeAnnotation::Tuple(anns) => {
+            for a in anns {
+                result.extend(get_type_var_contained_in_type_annotation(a));
+            }
+        }
+        TypeAnnotation::Primitive(..) | TypeAnnotation::Literal { .. } => {}
+    }
+    result
+}
+
+/// check the type compatibility for overload
+pub fn check_overload_type_annotation_compatible(
+    this: &TypeAnnotation,
+    other: &TypeAnnotation,
+    unifier: &mut Unifier,
+) -> bool {
+    match (this, other) {
+        (TypeAnnotation::Primitive(a), TypeAnnotation::Primitive(b)) => a == b,
+        (TypeAnnotation::TypeVar(a), TypeAnnotation::TypeVar(b)) => {
+            let a = unifier.get_ty(*a);
+            let a = &*a;
+            let b = unifier.get_ty(*b);
+            let b = &*b;
+            let (
+                TypeEnum::TVar { id: a, fields: None, .. },
+                TypeEnum::TVar { id: b, fields: None, .. },
+            ) = (a, b)
+            else {
+                unreachable!("must be type var")
+            };
+
+            a == b
+        }
+        (TypeAnnotation::Virtual(a), TypeAnnotation::Virtual(b)) => {
+            check_overload_type_annotation_compatible(a.as_ref(), b.as_ref(), unifier)
+        }
+
+        (TypeAnnotation::Tuple(a), TypeAnnotation::Tuple(b)) => {
+            a.len() == b.len() && {
+                a.iter()
+                    .zip(b)
+                    .all(|(a, b)| check_overload_type_annotation_compatible(a, b, unifier))
+            }
+        }
+
+        (
+            TypeAnnotation::CustomClass { id: a, params: a_p },
+            TypeAnnotation::CustomClass { id: b, params: b_p },
+        ) => {
+            a.0 == b.0 && {
+                a_p.len() == b_p.len() && {
+                    a_p.iter()
+                        .zip(b_p)
+                        .all(|(a, b)| check_overload_type_annotation_compatible(a, b, unifier))
+                }
+            }
+        }
+
+        _ => false,
+    }
+}

nac3core/src/typecheck/context/global_context.rs

@@ -1,109 +0,0 @@
-use super::super::typedef::*;
-use std::collections::HashMap;
-use std::rc::Rc;
-
-/// Structure for storing top-level type definitions.
-/// Used for collecting type signature from source code.
-/// Can be converted to `InferenceContext` for type inference in functions.
-pub struct GlobalContext<'a> {
-    /// List of primitive definitions.
-    pub(super) primitive_defs: Vec<TypeDef<'a>>,
-    /// List of class definitions.
-    pub(super) class_defs: Vec<ClassDef<'a>>,
-    /// List of parametric type definitions.
-    pub(super) parametric_defs: Vec<ParametricDef<'a>>,
-    /// List of type variable definitions.
-    pub(super) var_defs: Vec<VarDef<'a>>,
-    /// Function name to signature mapping.
-    pub(super) fn_table: HashMap<&'a str, FnDef>,
-
-    primitives: Vec<Type>,
-    variables: Vec<Type>,
-}
-
-impl<'a> GlobalContext<'a> {
-    pub fn new(primitive_defs: Vec<TypeDef<'a>>) -> GlobalContext {
-        let mut primitives = Vec::new();
-        for (i, t) in primitive_defs.iter().enumerate() {
-            primitives.push(TypeEnum::PrimitiveType(PrimitiveId(i)).into());
-        }
-        GlobalContext {
-            primitive_defs,
-            class_defs: Vec::new(),
-            parametric_defs: Vec::new(),
-            var_defs: Vec::new(),
-            fn_table: HashMap::new(),
-            primitives,
-            variables: Vec::new(),
-        }
-    }
-
-    pub fn add_class(&mut self, def: ClassDef<'a>) -> ClassId {
-        self.class_defs.push(def);
-        ClassId(self.class_defs.len() - 1)
-    }
-
-    pub fn add_parametric(&mut self, def: ParametricDef<'a>) -> ParamId {
-        self.parametric_defs.push(def);
-        ParamId(self.parametric_defs.len() - 1)
-    }
-
-    pub fn add_variable(&mut self, def: VarDef<'a>) -> VariableId {
-        self.add_variable_private(def)
-    }
-
-    pub fn add_variable_private(&mut self, def: VarDef<'a>) -> VariableId {
-        self.var_defs.push(def);
-        self.variables
-            .push(TypeEnum::TypeVariable(VariableId(self.var_defs.len() - 1)).into());
-        VariableId(self.var_defs.len() - 1)
-    }
-
-    pub fn add_fn(&mut self, name: &'a str, def: FnDef) {
-        self.fn_table.insert(name, def);
-    }
-
-    pub fn get_fn_def(&self, name: &str) -> Option<&FnDef> {
-        self.fn_table.get(name)
-    }
-
-    pub fn get_primitive_def_mut(&mut self, id: PrimitiveId) -> &mut TypeDef<'a> {
-        self.primitive_defs.get_mut(id.0).unwrap()
-    }
-
-    pub fn get_primitive_def(&self, id: PrimitiveId) -> &TypeDef {
-        self.primitive_defs.get(id.0).unwrap()
-    }
-
-    pub fn get_class_def_mut(&mut self, id: ClassId) -> &mut ClassDef<'a> {
-        self.class_defs.get_mut(id.0).unwrap()
-    }
-
-    pub fn get_class_def(&self, id: ClassId) -> &ClassDef {
-        self.class_defs.get(id.0).unwrap()
-    }
-
-    pub fn get_parametric_def_mut(&mut self, id: ParamId) -> &mut ParametricDef<'a> {
-        self.parametric_defs.get_mut(id.0).unwrap()
-    }
-
-    pub fn get_parametric_def(&self, id: ParamId) -> &ParametricDef {
-        self.parametric_defs.get(id.0).unwrap()
-    }
-
-    pub fn get_variable_def_mut(&mut self, id: VariableId) -> &mut VarDef<'a> {
-        self.var_defs.get_mut(id.0).unwrap()
-    }
-
-    pub fn get_variable_def(&self, id: VariableId) -> &VarDef {
-        self.var_defs.get(id.0).unwrap()
-    }
-
-    pub fn get_primitive(&self, id: PrimitiveId) -> Type {
-        self.primitives.get(id.0).unwrap().clone()
-    }
-
-    pub fn get_variable(&self, id: VariableId) -> Type {
-        self.variables.get(id.0).unwrap().clone()
-    }
-}

nac3core/src/typecheck/context/inference_context.rs

@@ -1,216 +0,0 @@
-use super::super::location::{FileID, Location};
-use super::super::symbol_resolver::*;
-use super::super::typedef::*;
-use super::GlobalContext;
-use rustpython_parser::ast;
-use std::boxed::Box;
-use std::collections::HashMap;
-
-pub struct ContextStack {
-    /// stack level, starts from 0
-    level: u32,
-    /// stack of symbol definitions containing (name, level) where `level` is the smallest level
-    /// where the name is assigned a value
-    sym_def: Vec<(String, u32)>,
-}
-
-pub struct InferenceContext<'a> {
-    /// global context
-    global: GlobalContext<'a>,
-    /// per source symbol resolver
-    resolver: Box<dyn SymbolResolver>,
-    /// File ID
-    file: FileID,
-
-    /// identifier to (type, readable, location) mapping.
-    /// an identifier might be defined earlier but has no value (for some code path), thus not
-    /// readable.
-    sym_table: HashMap<String, (Type, bool, Location)>,
-    /// stack
-    stack: ContextStack,
-}
-
-// non-trivial implementations here
-impl<'a> InferenceContext<'a> {
-    pub fn new(
-        global: GlobalContext,
-        resolver: Box<dyn SymbolResolver>,
-        file: FileID,
-    ) -> InferenceContext {
-        InferenceContext {
-            global,
-            resolver,
-            file,
-            sym_table: HashMap::new(),
-            stack: ContextStack {
-                level: 0,
-                sym_def: Vec::new(),
-            },
-        }
-    }
-
-    /// execute the function with new scope.
-    /// variable assignment would be limited within the scope (not readable outside), and type
-    /// returns the list of variables assigned within the scope, and the result of the function
-    pub fn with_scope<F, R>(&mut self, f: F) -> (Vec<(String, Type, Location)>, R)
-    where
-        F: FnOnce(&mut Self) -> R,
-    {
-        self.start_scope();
-        let result = f(self);
-        let poped_names = self.end_scope();
-        (poped_names, result)
-    }
-
-    pub fn start_scope(&mut self) {
-        self.stack.level += 1;
-    }
-
-    pub fn end_scope(&mut self) -> Vec<(String, Type, Location)> {
-        self.stack.level -= 1;
-        let mut poped_names = Vec::new();
-        while !self.stack.sym_def.is_empty() {
-            let (_, level) = self.stack.sym_def.last().unwrap();
-            if *level > self.stack.level {
-                let (name, _) = self.stack.sym_def.pop().unwrap();
-                let (t, b, l) = self.sym_table.get_mut(&name).unwrap();
-                // set it to be unreadable
-                *b = false;
-                poped_names.push((name, t.clone(), *l));
-            } else {
-                break;
-            }
-        }
-        poped_names
-    }
-
-    /// assign a type to an identifier.
-    /// may return error if the identifier was defined but with different type
-    pub fn assign(&mut self, name: String, ty: Type, loc: ast::Location) -> Result<Type, String> {
-        if let Some((t, x, _)) = self.sym_table.get_mut(&name) {
-            if t == &ty {
-                if !*x {
-                    self.stack.sym_def.push((name, self.stack.level));
-                }
-                *x = true;
-                Ok(ty)
-            } else {
-                Err("different types".into())
-            }
-        } else {
-            self.stack.sym_def.push((name.clone(), self.stack.level));
-            self.sym_table.insert(
-                name,
-                (ty.clone(), true, Location::CodeRange(self.file, loc)),
-            );
-            Ok(ty)
-        }
-    }
-
-    /// get the type of an identifier
-    /// may return error if the identifier is not defined, and cannot be resolved with the
-    /// resolution function.
-    pub fn resolve(&self, name: &str) -> Result<Type, String> {
-        if let Some((t, x, _)) = self.sym_table.get(name) {
-            if *x {
-                Ok(t.clone())
-            } else {
-                Err("may not be defined".into())
-            }
-        } else {
-            match self.resolver.get_symbol_type(name) {
-                Some(SymbolType::Identifier(t)) => Ok(t),
-                Some(SymbolType::TypeName(_)) => Err("is not a value".into()),
-                _ => Err("unbounded identifier".into()),
-            }
-        }
-    }
-
-    pub fn get_location(&self, name: &str) -> Option<Location> {
-        if let Some((_, _, l)) = self.sym_table.get(name) {
-            Some(*l)
-        } else {
-            self.resolver.get_symbol_location(name)
-        }
-    }
-
-    /// check if an identifier is already defined
-    pub fn defined(&self, name: &String) -> bool {
-        self.sym_table.get(name).is_some()
-    }
-}
-
-// trivial getters:
-impl<'a> InferenceContext<'a> {
-    pub fn get_primitive(&self, id: PrimitiveId) -> Type {
-        TypeEnum::PrimitiveType(id).into()
-    }
-
-    pub fn get_variable(&self, id: VariableId) -> Type {
-        TypeEnum::TypeVariable(id).into()
-    }
-
-    pub fn get_fn_def(&self, name: &str) -> Option<&FnDef> {
-        self.global.fn_table.get(name)
-    }
-    pub fn get_primitive_def(&self, id: PrimitiveId) -> &TypeDef {
-        self.global.primitive_defs.get(id.0).unwrap()
-    }
-    pub fn get_class_def(&self, id: ClassId) -> &ClassDef {
-        self.global.class_defs.get(id.0).unwrap()
-    }
-    pub fn get_parametric_def(&self, id: ParamId) -> &ParametricDef {
-        self.global.parametric_defs.get(id.0).unwrap()
-    }
-    pub fn get_variable_def(&self, id: VariableId) -> &VarDef {
-        self.global.var_defs.get(id.0).unwrap()
-    }
-    pub fn get_type(&self, name: &str) -> Result<Type, String> {
-        match self.resolver.get_symbol_type(name) {
-            Some(SymbolType::TypeName(t)) => Ok(t),
-            Some(SymbolType::Identifier(_)) => Err("not a type".into()),
-            _ => Err("unbounded identifier".into()),
-        }
-    }
-}
-
-impl TypeEnum {
-    pub fn subst(&self, map: &HashMap<VariableId, Type>) -> TypeEnum {
-        match self {
-            TypeEnum::TypeVariable(id) => map.get(id).map(|v| v.as_ref()).unwrap_or(self).clone(),
-            TypeEnum::ParametricType(id, params) => TypeEnum::ParametricType(
-                *id,
-                params
-                    .iter()
-                    .map(|v| v.as_ref().subst(map).into())
-                    .collect(),
-            ),
-            _ => self.clone(),
-        }
-    }
-
-    pub fn get_subst(&self, ctx: &InferenceContext) -> HashMap<VariableId, Type> {
-        match self {
-            TypeEnum::ParametricType(id, params) => {
-                let vars = &ctx.get_parametric_def(*id).params;
-                vars.iter()
-                    .zip(params)
-                    .map(|(v, p)| (*v, p.as_ref().clone().into()))
-                    .collect()
-            }
-            // if this proves to be slow, we can use option type
-            _ => HashMap::new(),
-        }
-    }
-
-    pub fn get_base<'b: 'a, 'a>(&'a self, ctx: &'b InferenceContext) -> Option<&'b TypeDef> {
-        match self {
-            TypeEnum::PrimitiveType(id) => Some(ctx.get_primitive_def(*id)),
-            TypeEnum::ClassType(id) | TypeEnum::VirtualClassType(id) => {
-                Some(&ctx.get_class_def(*id).base)
-            }
-            TypeEnum::ParametricType(id, _) => Some(&ctx.get_parametric_def(*id).base),
-            _ => None,
-        }
-    }
-}

nac3core/src/typecheck/context/mod.rs

@@ -1,4 +0,0 @@
-mod inference_context;
-mod global_context;
-pub use inference_context::InferenceContext;
-pub use global_context::GlobalContext;

nac3core/src/typecheck/function_check.rs

@@ -0,0 +1,423 @@
+use std::{
+    collections::{HashMap, HashSet},
+    iter::once,
+};
+
+use nac3parser::ast::{
+    self, Constant, Expr, ExprKind,
+    Operator::{LShift, RShift},
+    Stmt, StmtKind, StrRef,
+};
+
+use super::{
+    type_inferencer::{IdentifierInfo, Inferencer},
+    typedef::{Type, TypeEnum},
+};
+use crate::toplevel::helper::PrimDef;
+
+impl<'a> Inferencer<'a> {
+    fn should_have_value(&mut self, expr: &Expr<Option<Type>>) -> Result<(), HashSet<String>> {
+        if matches!(expr.custom, Some(ty) if self.unifier.unioned(ty, self.primitives.none)) {
+            Err(HashSet::from([format!("Error at {}: cannot have value none", expr.location)]))
+        } else {
+            Ok(())
+        }
+    }
+
+    fn check_pattern(
+        &mut self,
+        pattern: &Expr<Option<Type>>,
+        defined_identifiers: &mut HashMap<StrRef, IdentifierInfo>,
+    ) -> Result<(), HashSet<String>> {
+        match &pattern.node {
+            ExprKind::Name { id, .. } if id == &"none".into() => {
+                Err(HashSet::from([format!("cannot assign to a `none` (at {})", pattern.location)]))
+            }
+            ExprKind::Name { id, .. } => {
+                if !defined_identifiers.contains_key(id) {
+                    defined_identifiers.insert(*id, IdentifierInfo::default());
+                }
+                self.should_have_value(pattern)?;
+                Ok(())
+            }
+            ExprKind::List { elts, .. } | ExprKind::Tuple { elts, .. } => {
+                for elt in elts {
+                    self.check_pattern(elt, defined_identifiers)?;
+                    self.should_have_value(elt)?;
+                }
+                Ok(())
+            }
+            ExprKind::Starred { value, .. } => {
+                self.check_pattern(value, defined_identifiers)?;
+                self.should_have_value(value)?;
+                Ok(())
+            }
+            ExprKind::Subscript { value, slice, .. } => {
+                self.check_expr(value, defined_identifiers)?;
+                self.should_have_value(value)?;
+                self.check_expr(slice, defined_identifiers)?;
+                if let TypeEnum::TTuple { .. } = &*self.unifier.get_ty(value.custom.unwrap()) {
+                    return Err(HashSet::from([format!(
+                        "Error at {}: cannot assign to tuple element",
+                        value.location
+                    )]));
+                }
+                Ok(())
+            }
+            ExprKind::Constant { .. } => Err(HashSet::from([format!(
+                "cannot assign to a constant (at {})",
+                pattern.location
+            )])),
+            _ => self.check_expr(pattern, defined_identifiers),
+        }
+    }
+
+    fn check_expr(
+        &mut self,
+        expr: &Expr<Option<Type>>,
+        defined_identifiers: &mut HashMap<StrRef, IdentifierInfo>,
+    ) -> Result<(), HashSet<String>> {
+        // there are some cases where the custom field is None
+        if let Some(ty) = &expr.custom {
+            if !matches!(&expr.node, ExprKind::Constant { value: Constant::Ellipsis, .. })
+                && !ty.obj_id(self.unifier).is_some_and(|id| id == PrimDef::List.id())
+                && !self.unifier.is_concrete(*ty, &self.function_data.bound_variables)
+            {
+                return Err(HashSet::from([format!(
+                    "expected concrete type at {} but got {}",
+                    expr.location,
+                    self.unifier.get_ty(*ty).get_type_name()
+                )]));
+            }
+        }
+        match &expr.node {
+            ExprKind::Name { id, .. } => {
+                if id == &"none".into() {
+                    return Ok(());
+                }
+                self.should_have_value(expr)?;
+                if !defined_identifiers.contains_key(id) {
+                    match self.function_data.resolver.get_symbol_type(
+                        self.unifier,
+                        &self.top_level.definitions.read(),
+                        self.primitives,
+                        *id,
+                    ) {
+                        Ok(_) => {
+                            self.defined_identifiers.insert(*id, IdentifierInfo::default());
+                        }
+                        Err(e) => {
+                            return Err(HashSet::from([format!(
+                                "type error at identifier `{}` ({}) at {}",
+                                id, e, expr.location
+                            )]))
+                        }
+                    }
+                }
+            }
+            ExprKind::List { elts, .. }
+            | ExprKind::Tuple { elts, .. }
+            | ExprKind::BoolOp { values: elts, .. } => {
+                for elt in elts {
+                    self.check_expr(elt, defined_identifiers)?;
+                    self.should_have_value(elt)?;
+                }
+            }
+            ExprKind::Attribute { value, .. } => {
+                self.check_expr(value, defined_identifiers)?;
+                self.should_have_value(value)?;
+            }
+            ExprKind::BinOp { left, op, right } => {
+                self.check_expr(left, defined_identifiers)?;
+                self.check_expr(right, defined_identifiers)?;
+                self.should_have_value(left)?;
+                self.should_have_value(right)?;
+
+                // Check whether a bitwise shift has a negative RHS constant value
+                if *op == LShift || *op == RShift {
+                    if let ExprKind::Constant { value, .. } = &right.node {
+                        let Constant::Int(rhs_val) = value else { unreachable!() };
+
+                        if *rhs_val < 0 {
+                            return Err(HashSet::from([format!(
+                                "shift count is negative at {}",
+                                right.location
+                            )]));
+                        }
+                    }
+                }
+            }
+            ExprKind::UnaryOp { operand, .. } => {
+                self.check_expr(operand, defined_identifiers)?;
+                self.should_have_value(operand)?;
+            }
+            ExprKind::Compare { left, comparators, .. } => {
+                for elt in once(left.as_ref()).chain(comparators.iter()) {
+                    self.check_expr(elt, defined_identifiers)?;
+                    self.should_have_value(elt)?;
+                }
+            }
+            ExprKind::Subscript { value, slice, .. } => {
+                self.should_have_value(value)?;
+                self.check_expr(value, defined_identifiers)?;
+                self.check_expr(slice, defined_identifiers)?;
+            }
+            ExprKind::IfExp { test, body, orelse } => {
+                self.check_expr(test, defined_identifiers)?;
+                self.check_expr(body, defined_identifiers)?;
+                self.check_expr(orelse, defined_identifiers)?;
+            }
+            ExprKind::Slice { lower, upper, step } => {
+                for elt in [lower.as_ref(), upper.as_ref(), step.as_ref()].iter().flatten() {
+                    self.should_have_value(elt)?;
+                    self.check_expr(elt, defined_identifiers)?;
+                }
+            }
+            ExprKind::Lambda { args, body } => {
+                let mut defined_identifiers = defined_identifiers.clone();
+                for arg in &args.args {
+                    // TODO: should we check the types here?
+                    defined_identifiers.entry(arg.node.arg).or_default();
+                }
+                self.check_expr(body, &mut defined_identifiers)?;
+            }
+            ExprKind::ListComp { elt, generators, .. } => {
+                // in our type inference stage, we already make sure that there is only 1 generator
+                let ast::Comprehension { target, iter, ifs, .. } = &generators[0];
+                self.check_expr(iter, defined_identifiers)?;
+                self.should_have_value(iter)?;
+                let mut defined_identifiers = defined_identifiers.clone();
+                self.check_pattern(target, &mut defined_identifiers)?;
+                self.should_have_value(target)?;
+                for term in once(elt.as_ref()).chain(ifs.iter()) {
+                    self.check_expr(term, &mut defined_identifiers)?;
+                    self.should_have_value(term)?;
+                }
+            }
+            ExprKind::Call { func, args, keywords } => {
+                for expr in once(func.as_ref())
+                    .chain(args.iter())
+                    .chain(keywords.iter().map(|v| v.node.value.as_ref()))
+                {
+                    self.check_expr(expr, defined_identifiers)?;
+                    self.should_have_value(expr)?;
+                }
+            }
+            ExprKind::Constant { .. } => {}
+            _ => {
+                unimplemented!()
+            }
+        }
+        Ok(())
+    }
+
+    /// Check that the return value is a non-`alloca` type, effectively only allowing primitive types.
+    ///
+    /// This is a workaround preventing the caller from using a variable `alloca`-ed in the body, which
+    /// is freed when the function returns.
+    fn check_return_value_ty(&mut self, ret_ty: Type) -> bool {
+        if cfg!(feature = "no-escape-analysis") {
+            true
+        } else {
+            match &*self.unifier.get_ty_immutable(ret_ty) {
+                TypeEnum::TObj { .. } => [
+                    self.primitives.int32,
+                    self.primitives.int64,
+                    self.primitives.uint32,
+                    self.primitives.uint64,
+                    self.primitives.float,
+                    self.primitives.bool,
+                ]
+                .iter()
+                .any(|allowed_ty| self.unifier.unioned(ret_ty, *allowed_ty)),
+                TypeEnum::TTuple { ty, .. } => ty.iter().all(|t| self.check_return_value_ty(*t)),
+                _ => false,
+            }
+        }
+    }
+
+    // check statements for proper identifier def-use and return on all paths
+    fn check_stmt(
+        &mut self,
+        stmt: &Stmt<Option<Type>>,
+        defined_identifiers: &mut HashMap<StrRef, IdentifierInfo>,
+    ) -> Result<bool, HashSet<String>> {
+        match &stmt.node {
+            StmtKind::For { target, iter, body, orelse, .. } => {
+                self.check_expr(iter, defined_identifiers)?;
+                self.should_have_value(iter)?;
+                let mut local_defined_identifiers = defined_identifiers.clone();
+                for stmt in orelse {
+                    self.check_stmt(stmt, &mut local_defined_identifiers)?;
+                }
+                let mut local_defined_identifiers = defined_identifiers.clone();
+                self.check_pattern(target, &mut local_defined_identifiers)?;
+                self.should_have_value(target)?;
+                for stmt in body {
+                    self.check_stmt(stmt, &mut local_defined_identifiers)?;
+                }
+                Ok(false)
+            }
+            StmtKind::If { test, body, orelse, .. } => {
+                self.check_expr(test, defined_identifiers)?;
+                self.should_have_value(test)?;
+                let mut body_identifiers = defined_identifiers.clone();
+                let mut orelse_identifiers = defined_identifiers.clone();
+                let body_returned = self.check_block(body, &mut body_identifiers)?;
+                let orelse_returned = self.check_block(orelse, &mut orelse_identifiers)?;
+
+                for ident in body_identifiers.keys() {
+                    if !defined_identifiers.contains_key(ident)
+                        && orelse_identifiers.contains_key(ident)
+                    {
+                        defined_identifiers.insert(*ident, IdentifierInfo::default());
+                    }
+                }
+                Ok(body_returned && orelse_returned)
+            }
+            StmtKind::While { test, body, orelse, .. } => {
+                self.check_expr(test, defined_identifiers)?;
+                self.should_have_value(test)?;
+                let mut defined_identifiers = defined_identifiers.clone();
+                self.check_block(body, &mut defined_identifiers)?;
+                self.check_block(orelse, &mut defined_identifiers)?;
+                Ok(false)
+            }
+            StmtKind::With { items, body, .. } => {
+                let mut new_defined_identifiers = defined_identifiers.clone();
+                for item in items {
+                    self.check_expr(&item.context_expr, defined_identifiers)?;
+                    if let Some(var) = item.optional_vars.as_ref() {
+                        self.check_pattern(var, &mut new_defined_identifiers)?;
+                    }
+                }
+                self.check_block(body, &mut new_defined_identifiers)?;
+                Ok(false)
+            }
+            StmtKind::Try { body, handlers, orelse, finalbody, .. } => {
+                self.check_block(body, &mut defined_identifiers.clone())?;
+                self.check_block(orelse, &mut defined_identifiers.clone())?;
+                for handler in handlers {
+                    let mut defined_identifiers = defined_identifiers.clone();
+                    let ast::ExcepthandlerKind::ExceptHandler { name, body, .. } = &handler.node;
+                    if let Some(name) = name {
+                        defined_identifiers.insert(*name, IdentifierInfo::default());
+                    }
+                    self.check_block(body, &mut defined_identifiers)?;
+                }
+                self.check_block(finalbody, defined_identifiers)?;
+                Ok(false)
+            }
+            StmtKind::Expr { value, .. } => {
+                self.check_expr(value, defined_identifiers)?;
+                Ok(false)
+            }
+            StmtKind::Assign { targets, value, .. } => {
+                self.check_expr(value, defined_identifiers)?;
+                self.should_have_value(value)?;
+                for target in targets {
+                    self.check_pattern(target, defined_identifiers)?;
+                }
+                Ok(false)
+            }
+            StmtKind::AnnAssign { target, value, .. } => {
+                if let Some(value) = value {
+                    self.check_expr(value, defined_identifiers)?;
+                    self.should_have_value(value)?;
+                    self.check_pattern(target, defined_identifiers)?;
+                }
+                Ok(false)
+            }
+            StmtKind::Return { value, .. } => {
+                if let Some(value) = value {
+                    self.check_expr(value, defined_identifiers)?;
+                    self.should_have_value(value)?;
+
+                    // Check that the return value is a non-`alloca` type, effectively only allowing primitive types.
+                    // This is a workaround preventing the caller from using a variable `alloca`-ed in the body, which
+                    // is freed when the function returns.
+                    if let Some(ret_ty) = value.custom {
+                        // Explicitly allow ellipsis as a return value, as the type of the ellipsis is contextually
+                        // inferred and just generates an unconditional assertion
+                        if matches!(
+                            value.node,
+                            ExprKind::Constant { value: Constant::Ellipsis, .. }
+                        ) {
+                            return Ok(true);
+                        }
+
+                        if !self.check_return_value_ty(ret_ty) {
+                            return Err(HashSet::from([
+                                format!(
+                                    "return value of type {} must be a primitive or a tuple of primitives at {}",
+                                    self.unifier.stringify(ret_ty),
+                                    value.location,
+                                ),
+                            ]));
+                        }
+                    }
+                }
+                Ok(true)
+            }
+            StmtKind::Raise { exc, .. } => {
+                if let Some(value) = exc {
+                    self.check_expr(value, defined_identifiers)?;
+                }
+                Ok(true)
+            }
+            StmtKind::Global { names, .. } => {
+                for id in names {
+                    if let Some(id_info) = defined_identifiers.get(id) {
+                        if !id_info.is_global {
+                            return Err(HashSet::from([format!(
+                                "name '{id}' is assigned to before global declaration at {}",
+                                stmt.location,
+                            )]));
+                        }
+
+                        continue;
+                    }
+
+                    match self.function_data.resolver.get_symbol_type(
+                        self.unifier,
+                        &self.top_level.definitions.read(),
+                        self.primitives,
+                        *id,
+                    ) {
+                        Ok(_) => {
+                            self.defined_identifiers
+                                .insert(*id, IdentifierInfo { is_global: true });
+                        }
+                        Err(e) => {
+                            return Err(HashSet::from([format!(
+                                "type error at identifier `{}` ({}) at {}",
+                                id, e, stmt.location
+                            )]))
+                        }
+                    }
+                }
+
+                Ok(false)
+            }
+            // break, raise, etc.
+            _ => Ok(false),
+        }
+    }
+
+    pub fn check_block(
+        &mut self,
+        block: &[Stmt<Option<Type>>],
+        defined_identifiers: &mut HashMap<StrRef, IdentifierInfo>,
+    ) -> Result<bool, HashSet<String>> {
+        let mut ret = false;
+        for stmt in block {
+            if ret {
+                eprintln!("warning: dead code at {}\n", stmt.location);
+            }
+            if self.check_stmt(stmt, defined_identifiers)? {
+                ret = true;
+            }
+        }
+        Ok(ret)
+    }
+}

nac3core/src/typecheck/inference_core.rs

@@ -1,526 +0,0 @@
-use super::context::InferenceContext;
-use super::typedef::{TypeEnum::*, *};
-use std::collections::HashMap;
-
-fn find_subst(
-    ctx: &InferenceContext,
-    valuation: &Option<(VariableId, Type)>,
-    sub: &mut HashMap<VariableId, Type>,
-    mut a: Type,
-    mut b: Type,
-) -> Result<(), String> {
-    // TODO: fix error messages later
-    if let TypeVariable(id) = a.as_ref() {
-        if let Some((assumption_id, t)) = valuation {
-            if assumption_id == id {
-                a = t.clone();
-            }
-        }
-    }
-
-    let mut substituted = false;
-    if let TypeVariable(id) = b.as_ref() {
-        if let Some(c) = sub.get(&id) {
-            b = c.clone();
-            substituted = true;
-        }
-    }
-
-    match (a.as_ref(), b.as_ref()) {
-        (BotType, _) => Ok(()),
-        (TypeVariable(id_a), TypeVariable(id_b)) => {
-            if substituted {
-                return if id_a == id_b {
-                    Ok(())
-                } else {
-                    Err("different variables".to_string())
-                };
-            }
-            let v_a = ctx.get_variable_def(*id_a);
-            let v_b = ctx.get_variable_def(*id_b);
-            if !v_b.bound.is_empty() {
-                if v_a.bound.is_empty() {
-                    return Err("unbounded a".to_string());
-                } else {
-                    let diff: Vec<_> = v_a
-                        .bound
-                        .iter()
-                        .filter(|x| !v_b.bound.contains(x))
-                        .collect();
-                    if !diff.is_empty() {
-                        return Err("different domain".to_string());
-                    }
-                }
-            }
-            sub.insert(*id_b, a.clone());
-            Ok(())
-        }
-        (TypeVariable(id_a), _) => {
-            let v_a = ctx.get_variable_def(*id_a);
-            if v_a.bound.len() == 1 && v_a.bound[0].as_ref() == b.as_ref() {
-                Ok(())
-            } else {
-                Err("different domain".to_string())
-            }
-        }
-        (_, TypeVariable(id_b)) => {
-            let v_b = ctx.get_variable_def(*id_b);
-            if v_b.bound.is_empty() || v_b.bound.contains(&a) {
-                sub.insert(*id_b, a.clone());
-                Ok(())
-            } else {
-                Err("different domain".to_string())
-            }
-        }
-        (_, VirtualClassType(id_b)) => {
-            let mut parents;
-            match a.as_ref() {
-                ClassType(id_a) => {
-                    parents = [*id_a].to_vec();
-                }
-                VirtualClassType(id_a) => {
-                    parents = [*id_a].to_vec();
-                }
-                _ => {
-                    return Err("cannot substitute non-class type into virtual class".to_string());
-                }
-            };
-            while !parents.is_empty() {
-                if *id_b == parents[0] {
-                    return Ok(());
-                }
-                let c = ctx.get_class_def(parents.remove(0));
-                parents.extend_from_slice(&c.parents);
-            }
-            Err("not subtype".to_string())
-        }
-        (ParametricType(id_a, param_a), ParametricType(id_b, param_b)) => {
-            if id_a != id_b || param_a.len() != param_b.len() {
-                Err("different parametric types".to_string())
-            } else {
-                for (x, y) in param_a.iter().zip(param_b.iter()) {
-                    find_subst(ctx, valuation, sub, x.clone(), y.clone())?;
-                }
-                Ok(())
-            }
-        }
-        (_, _) => {
-            if a == b {
-                Ok(())
-            } else {
-                Err("not equal".to_string())
-            }
-        }
-    }
-}
-
-fn resolve_call_rec(
-    ctx: &InferenceContext,
-    valuation: &Option<(VariableId, Type)>,
-    obj: Option<Type>,
-    func: &str,
-    args: &[Type],
-) -> Result<Option<Type>, String> {
-    let mut subst = obj
-        .as_ref()
-        .map(|v| v.get_subst(ctx))
-        .unwrap_or_else(HashMap::new);
-
-    let fun = match &obj {
-        Some(obj) => {
-            let base = match obj.as_ref() {
-                PrimitiveType(id) => &ctx.get_primitive_def(*id),
-                ClassType(id) | VirtualClassType(id) => &ctx.get_class_def(*id).base,
-                ParametricType(id, _) => &ctx.get_parametric_def(*id).base,
-                _ => return Err("not supported".to_string()),
-            };
-            base.methods.get(func)
-        }
-        None => ctx.get_fn_def(func),
-    }
-    .ok_or_else(|| "no such function".to_string())?;
-
-    if args.len() != fun.args.len() {
-        return Err("incorrect parameter number".to_string());
-    }
-    for (a, b) in args.iter().zip(fun.args.iter()) {
-        find_subst(ctx, valuation, &mut subst, a.clone(), b.clone())?;
-    }
-    let result = fun.result.as_ref().map(|v| v.subst(&subst));
-    Ok(result.map(|result| {
-        if let SelfType = result {
-            obj.unwrap()
-        } else {
-            result.into()
-        }
-    }))
-}
-
-pub fn resolve_call(
-    ctx: &InferenceContext,
-    obj: Option<Type>,
-    func: &str,
-    args: &[Type],
-) -> Result<Option<Type>, String> {
-    resolve_call_rec(ctx, &None, obj, func, args)
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-    use super::super::context::GlobalContext;
-    use super::super::primitives::*;
-    use std::rc::Rc;
-
-    fn get_inference_context(ctx: GlobalContext) -> InferenceContext {
-        // InferenceContext::new(ctx, Box::new(|_| Err("unbounded identifier".into())))
-        crate::typecheck::type_check::test::new_ctx().ctx
-    }
-
-    #[test]
-    fn test_simple_generic() {
-        let mut ctx = basic_ctx();
-        let v1 = ctx.add_variable(VarDef {
-            name: "V1",
-            bound: vec![ctx.get_primitive(INT32_TYPE), ctx.get_primitive(FLOAT_TYPE)],
-        });
-        let v1 = ctx.get_variable(v1);
-        let v2 = ctx.add_variable(VarDef {
-            name: "V2",
-            bound: vec![
-                ctx.get_primitive(BOOL_TYPE),
-                ctx.get_primitive(INT32_TYPE),
-                ctx.get_primitive(FLOAT_TYPE),
-            ],
-        });
-        let v2 = ctx.get_variable(v2);
-        let ctx = get_inference_context(ctx);
-
-        assert_eq!(
-            resolve_call(&ctx, None, "int32", &[ctx.get_primitive(FLOAT_TYPE)]),
-            Ok(Some(ctx.get_primitive(INT32_TYPE)))
-        );
-
-        assert_eq!(
-            resolve_call(&ctx, None, "int32", &[ctx.get_primitive(INT32_TYPE)],),
-            Ok(Some(ctx.get_primitive(INT32_TYPE)))
-        );
-
-        assert_eq!(
-            resolve_call(&ctx, None, "float", &[ctx.get_primitive(INT32_TYPE)]),
-            Ok(Some(ctx.get_primitive(FLOAT_TYPE)))
-        );
-
-        assert_eq!(
-            resolve_call(&ctx, None, "float", &[ctx.get_primitive(BOOL_TYPE)]),
-            Err("different domain".to_string())
-        );
-
-        assert_eq!(
-            resolve_call(&ctx, None, "float", &[]),
-            Err("incorrect parameter number".to_string())
-        );
-
-        assert_eq!(
-            resolve_call(&ctx, None, "float", &[v1]),
-            Ok(Some(ctx.get_primitive(FLOAT_TYPE)))
-        );
-
-        assert_eq!(
-            resolve_call(&ctx, None, "float", &[v2]),
-            Err("different domain".to_string())
-        );
-    }
-
-    #[test]
-    fn test_methods() {
-        let mut ctx = basic_ctx();
-
-        let v0 = ctx.add_variable(VarDef {
-            name: "V0",
-            bound: vec![],
-        });
-        let v0 = ctx.get_variable(v0);
-
-        let int32 = ctx.get_primitive(INT32_TYPE);
-        let int64 = ctx.get_primitive(INT64_TYPE);
-        let ctx = get_inference_context(ctx);
-
-        // simple cases
-        assert_eq!(
-            resolve_call(&ctx, Some(int32.clone()), "__add__", &[int32.clone()]),
-            Ok(Some(int32.clone()))
-        );
-
-        assert_ne!(
-            resolve_call(&ctx, Some(int32.clone()), "__add__", &[int32.clone()]),
-            Ok(Some(int64.clone()))
-        );
-
-        assert_eq!(
-            resolve_call(&ctx, Some(int32), "__add__", &[int64]),
-            Err("not equal".to_string())
-        );
-
-        // with type variables
-        assert_eq!(
-            resolve_call(&ctx, Some(v0.clone()), "__add__", &[v0.clone()]),
-            Err("not supported".into())
-        );
-    }
-
-    #[test]
-    fn test_multi_generic() {
-        let mut ctx = basic_ctx();
-        let v0 = ctx.add_variable(VarDef {
-            name: "V0",
-            bound: vec![],
-        });
-        let v0 = ctx.get_variable(v0);
-        let v1 = ctx.add_variable(VarDef {
-            name: "V1",
-            bound: vec![],
-        });
-        let v1 = ctx.get_variable(v1);
-        let v2 = ctx.add_variable(VarDef {
-            name: "V2",
-            bound: vec![],
-        });
-        let v2 = ctx.get_variable(v2);
-        let v3 = ctx.add_variable(VarDef {
-            name: "V3",
-            bound: vec![],
-        });
-        let v3 = ctx.get_variable(v3);
-
-        ctx.add_fn(
-            "foo",
-            FnDef {
-                args: vec![v0.clone(), v0.clone(), v1.clone()],
-                result: Some(v0.clone()),
-            },
-        );
-
-        ctx.add_fn(
-            "foo1",
-            FnDef {
-                args: vec![ParametricType(TUPLE_TYPE, vec![v0.clone(), v0.clone(), v1]).into()],
-                result: Some(v0),
-            },
-        );
-        let ctx = get_inference_context(ctx);
-
-        assert_eq!(
-            resolve_call(&ctx, None, "foo", &[v2.clone(), v2.clone(), v2.clone()]),
-            Ok(Some(v2.clone()))
-        );
-        assert_eq!(
-            resolve_call(&ctx, None, "foo", &[v2.clone(), v2.clone(), v3.clone()]),
-            Ok(Some(v2.clone()))
-        );
-        assert_eq!(
-            resolve_call(&ctx, None, "foo", &[v2.clone(), v3.clone(), v3.clone()]),
-            Err("different variables".to_string())
-        );
-
-        assert_eq!(
-            resolve_call(
-                &ctx,
-                None,
-                "foo1",
-                &[ParametricType(TUPLE_TYPE, vec![v2.clone(), v2.clone(), v2.clone()]).into()]
-            ),
-            Ok(Some(v2.clone()))
-        );
-        assert_eq!(
-            resolve_call(
-                &ctx,
-                None,
-                "foo1",
-                &[ParametricType(TUPLE_TYPE, vec![v2.clone(), v2.clone(), v3.clone()]).into()]
-            ),
-            Ok(Some(v2.clone()))
-        );
-        assert_eq!(
-            resolve_call(
-                &ctx,
-                None,
-                "foo1",
-                &[ParametricType(TUPLE_TYPE, vec![v2, v3.clone(), v3]).into()]
-            ),
-            Err("different variables".to_string())
-        );
-    }
-
-    #[test]
-    fn test_class_generics() {
-        let mut ctx = basic_ctx();
-
-        let list = ctx.get_parametric_def_mut(LIST_TYPE);
-        let t = Rc::new(TypeVariable(list.params[0]));
-        list.base.methods.insert(
-            "head",
-            FnDef {
-                args: vec![],
-                result: Some(t.clone()),
-            },
-        );
-        list.base.methods.insert(
-            "append",
-            FnDef {
-                args: vec![t],
-                result: None,
-            },
-        );
-
-        let v0 = ctx.add_variable(VarDef {
-            name: "V0",
-            bound: vec![],
-        });
-        let v0 = ctx.get_variable(v0);
-        let v1 = ctx.add_variable(VarDef {
-            name: "V1",
-            bound: vec![],
-        });
-        let v1 = ctx.get_variable(v1);
-        let ctx = get_inference_context(ctx);
-
-        assert_eq!(
-            resolve_call(
-                &ctx,
-                Some(ParametricType(LIST_TYPE, vec![v0.clone()]).into()),
-                "head",
-                &[]
-            ),
-            Ok(Some(v0.clone()))
-        );
-        assert_eq!(
-            resolve_call(
-                &ctx,
-                Some(ParametricType(LIST_TYPE, vec![v0.clone()]).into()),
-                "append",
-                &[v0.clone()]
-            ),
-            Ok(None)
-        );
-        assert_eq!(
-            resolve_call(
-                &ctx,
-                Some(ParametricType(LIST_TYPE, vec![v0]).into()),
-                "append",
-                &[v1]
-            ),
-            Err("different variables".to_string())
-        );
-    }
-
-    #[test]
-    fn test_virtual_class() {
-        let mut ctx = basic_ctx();
-
-        let foo = ctx.add_class(ClassDef {
-            base: TypeDef {
-                name: "Foo",
-                methods: HashMap::new(),
-                fields: HashMap::new(),
-            },
-            parents: vec![],
-        });
-
-        let foo1 = ctx.add_class(ClassDef {
-            base: TypeDef {
-                name: "Foo1",
-                methods: HashMap::new(),
-                fields: HashMap::new(),
-            },
-            parents: vec![foo],
-        });
-
-        let foo2 = ctx.add_class(ClassDef {
-            base: TypeDef {
-                name: "Foo2",
-                methods: HashMap::new(),
-                fields: HashMap::new(),
-            },
-            parents: vec![foo1],
-        });
-
-        let bar = ctx.add_class(ClassDef {
-            base: TypeDef {
-                name: "bar",
-                methods: HashMap::new(),
-                fields: HashMap::new(),
-            },
-            parents: vec![],
-        });
-
-        ctx.add_fn(
-            "foo",
-            FnDef {
-                args: vec![VirtualClassType(foo).into()],
-                result: None,
-            },
-        );
-        ctx.add_fn(
-            "foo1",
-            FnDef {
-                args: vec![VirtualClassType(foo1).into()],
-                result: None,
-            },
-        );
-        let ctx = get_inference_context(ctx);
-
-        assert_eq!(
-            resolve_call(&ctx, None, "foo", &[ClassType(foo).into()]),
-            Ok(None)
-        );
-
-        assert_eq!(
-            resolve_call(&ctx, None, "foo", &[ClassType(foo1).into()]),
-            Ok(None)
-        );
-
-        assert_eq!(
-            resolve_call(&ctx, None, "foo", &[ClassType(foo2).into()]),
-            Ok(None)
-        );
-
-        assert_eq!(
-            resolve_call(&ctx, None, "foo", &[ClassType(bar).into()]),
-            Err("not subtype".to_string())
-        );
-
-        assert_eq!(
-            resolve_call(&ctx, None, "foo1", &[ClassType(foo1).into()]),
-            Ok(None)
-        );
-
-        assert_eq!(
-            resolve_call(&ctx, None, "foo1", &[ClassType(foo2).into()]),
-            Ok(None)
-        );
-
-        assert_eq!(
-            resolve_call(&ctx, None, "foo1", &[ClassType(foo).into()]),
-            Err("not subtype".to_string())
-        );
-
-        // virtual class substitution
-        assert_eq!(
-            resolve_call(&ctx, None, "foo", &[VirtualClassType(foo).into()]),
-            Ok(None)
-        );
-        assert_eq!(
-            resolve_call(&ctx, None, "foo", &[VirtualClassType(foo1).into()]),
-            Ok(None)
-        );
-        assert_eq!(
-            resolve_call(&ctx, None, "foo", &[VirtualClassType(foo2).into()]),
-            Ok(None)
-        );
-        assert_eq!(
-            resolve_call(&ctx, None, "foo", &[VirtualClassType(bar).into()]),
-            Err("not subtype".to_string())
-        );
-    }
-}

nac3core/src/typecheck/location.rs

@@ -1,31 +0,0 @@
-use rustpython_parser::ast;
-use std::vec::Vec;
-
-#[derive(Clone, Copy, PartialEq)]
-pub struct FileID(pub u32);
-
-#[derive(Clone, Copy, PartialEq)]
-pub enum Location {
-    CodeRange(FileID, ast::Location),
-    Builtin
-}
-
-pub struct FileRegistry {
-    files: Vec<String>,
-}
-
-impl FileRegistry {
-    pub fn new() -> FileRegistry {
-        FileRegistry { files: Vec::new() }
-    }
-
-    pub fn add_file(&mut self, path: &str) -> FileID {
-        let index = self.files.len() as u32;
-        self.files.push(path.to_owned());
-        FileID(index)
-    }
-
-    pub fn query_file(&self, id: FileID) -> &str {
-        &self.files[id.0 as usize]
-    }
-}

nac3core/src/typecheck/magic_methods.rs

@@ -1,58 +1,781 @@
-use rustpython_parser::ast::{Cmpop, Operator, Unaryop};
+use std::{cmp::max, collections::HashMap, rc::Rc};
 
-pub fn binop_name(op: &Operator) -> &'static str {
-    match op {
-        Operator::Add => "__add__",
-        Operator::Sub => "__sub__",
-        Operator::Div => "__truediv__",
-        Operator::Mod => "__mod__",
-        Operator::Mult => "__mul__",
-        Operator::Pow => "__pow__",
-        Operator::BitOr => "__or__",
-        Operator::BitXor => "__xor__",
-        Operator::BitAnd => "__and__",
-        Operator::LShift => "__lshift__",
-        Operator::RShift => "__rshift__",
-        Operator::FloorDiv => "__floordiv__",
-        Operator::MatMult => "__matmul__",
+use itertools::{iproduct, Itertools};
+use strum::IntoEnumIterator;
+
+use nac3parser::ast::{Cmpop, Operator, StrRef, Unaryop};
+
+use crate::{
+    symbol_resolver::SymbolValue,
+    toplevel::{
+        helper::PrimDef,
+        numpy::{make_ndarray_ty, unpack_ndarray_var_tys},
+    },
+    typecheck::{
+        type_inferencer::*,
+        typedef::{FunSignature, FuncArg, Type, TypeEnum, Unifier, VarMap},
+    },
+};
+
+/// The variant of a binary operator.
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum BinopVariant {
+    /// The normal variant.
+    /// For addition, it would be `+`.
+    Normal,
+    /// The "Augmented Assigning Operator" variant.
+    /// For addition, it would be `+=`.
+    AugAssign,
+}
+
+/// A binary operator with its variant.
+#[derive(Debug, Clone, Copy)]
+pub struct Binop {
+    /// The base [`Operator`] of this binary operator.
+    pub base: Operator,
+    /// The variant of this binary operator.
+    pub variant: BinopVariant,
+}
+
+impl Binop {
+    /// Make a [`Binop`] of the normal variant from an [`Operator`].
+    #[must_use]
+    pub fn normal(base: Operator) -> Self {
+        Binop { base, variant: BinopVariant::Normal }
+    }
+
+    /// Make a [`Binop`] of the aug assign variant from an [`Operator`].
+    #[must_use]
+    pub fn aug_assign(base: Operator) -> Self {
+        Binop { base, variant: BinopVariant::AugAssign }
     }
 }
 
-pub fn binop_assign_name(op: &Operator) -> &'static str {
-    match op {
-        Operator::Add => "__iadd__",
-        Operator::Sub => "__isub__",
-        Operator::Div => "__itruediv__",
-        Operator::Mod => "__imod__",
-        Operator::Mult => "__imul__",
-        Operator::Pow => "__ipow__",
-        Operator::BitOr => "__ior__",
-        Operator::BitXor => "__ixor__",
-        Operator::BitAnd => "__iand__",
-        Operator::LShift => "__ilshift__",
-        Operator::RShift => "__irshift__",
-        Operator::FloorDiv => "__ifloordiv__",
-        Operator::MatMult => "__imatmul__",
-    }
+/// Details about an operator (unary, binary, etc...) in Python
+#[derive(Debug, Clone, Copy)]
+pub struct OpInfo {
+    /// The method name of the binary operator.
+    /// For addition, this would be `__add__`, and `__iadd__` if
+    /// it is the augmented assigning variant.
+    pub method_name: &'static str,
+    /// The symbol of the binary operator.
+    /// For addition, this would be `+`, and `+=` if
+    /// it is the augmented assigning variant.
+    pub symbol: &'static str,
 }
 
-pub fn unaryop_name(op: &Unaryop) -> &'static str {
-    match op {
-        Unaryop::UAdd   => "__pos__",
-        Unaryop::USub   => "__neg__",
-        Unaryop::Not    => "__not__",
-        Unaryop::Invert => "__inv__",
-    }
+/// Helper macro to conveniently build an [`OpInfo`].
+///
+/// Example usage: `make_info("add", "+")` generates `OpInfo { name: "__add__", symbol: "+" }`
+macro_rules! make_op_info {
+    ($name:expr, $symbol:expr) => {
+        OpInfo { method_name: concat!("__", $name, "__"), symbol: $symbol }
+    };
 }
 
-pub fn comparison_name(op: &Cmpop) -> Option<&'static str> {
+pub trait HasOpInfo {
+    fn op_info(&self) -> OpInfo;
+}
+
+fn try_get_cmpop_info(op: Cmpop) -> Option<OpInfo> {
     match op {
-        Cmpop::Lt    => Some("__lt__"),
-        Cmpop::LtE   => Some("__le__"),
-        Cmpop::Gt    => Some("__gt__"),
-        Cmpop::GtE   => Some("__ge__"),
-        Cmpop::Eq    => Some("__eq__"),
-        Cmpop::NotEq => Some("__ne__"),
+        Cmpop::Lt => Some(make_op_info!("lt", "<")),
+        Cmpop::LtE => Some(make_op_info!("le", "<=")),
+        Cmpop::Gt => Some(make_op_info!("gt", ">")),
+        Cmpop::GtE => Some(make_op_info!("ge", ">=")),
+        Cmpop::Eq => Some(make_op_info!("eq", "==")),
+        Cmpop::NotEq => Some(make_op_info!("ne", "!=")),
         _ => None,
     }
 }
+
+impl OpInfo {
+    #[must_use]
+    pub fn supports_cmpop(op: Cmpop) -> bool {
+        try_get_cmpop_info(op).is_some()
+    }
+}
+
+impl HasOpInfo for Cmpop {
+    fn op_info(&self) -> OpInfo {
+        try_get_cmpop_info(*self).expect("{self:?} is not supported")
+    }
+}
+
+impl HasOpInfo for Binop {
+    fn op_info(&self) -> OpInfo {
+        // Helper macro to generate both the normal variant [`OpInfo`] and the
+        // augmented assigning variant [`OpInfo`] for a binary operator conveniently.
+        macro_rules! info {
+            ($name:literal, $symbol:literal) => {
+                (
+                    make_op_info!($name, $symbol),
+                    make_op_info!(concat!("i", $name), concat!($symbol, "=")),
+                )
+            };
+        }
+
+        let (normal_variant, aug_assign_variant) = match self.base {
+            Operator::Add => info!("add", "+"),
+            Operator::Sub => info!("sub", "-"),
+            Operator::Div => info!("truediv", "/"),
+            Operator::Mod => info!("mod", "%"),
+            Operator::Mult => info!("mul", "*"),
+            Operator::Pow => info!("pow", "**"),
+            Operator::BitOr => info!("or", "|"),
+            Operator::BitXor => info!("xor", "^"),
+            Operator::BitAnd => info!("and", "&"),
+            Operator::LShift => info!("lshift", "<<"),
+            Operator::RShift => info!("rshift", ">>"),
+            Operator::FloorDiv => info!("floordiv", "//"),
+            Operator::MatMult => info!("matmul", "@"),
+        };
+
+        match self.variant {
+            BinopVariant::Normal => normal_variant,
+            BinopVariant::AugAssign => aug_assign_variant,
+        }
+    }
+}
+
+impl HasOpInfo for Unaryop {
+    fn op_info(&self) -> OpInfo {
+        match self {
+            Unaryop::UAdd => make_op_info!("pos", "+"),
+            Unaryop::USub => make_op_info!("neg", "-"),
+            Unaryop::Not => make_op_info!("not", "not"), // i.e., `not False`, so the symbol is just `not`.
+            Unaryop::Invert => make_op_info!("inv", "~"),
+        }
+    }
+}
+
+pub(super) fn with_fields<F>(unifier: &mut Unifier, ty: Type, f: F)
+where
+    F: FnOnce(&mut Unifier, &mut HashMap<StrRef, (Type, bool)>),
+{
+    let (id, mut fields, params) =
+        if let TypeEnum::TObj { obj_id, fields, params } = &*unifier.get_ty(ty) {
+            (*obj_id, fields.clone(), params.clone())
+        } else {
+            unreachable!()
+        };
+    f(unifier, &mut fields);
+    unsafe {
+        let unification_table = unifier.get_unification_table();
+        unification_table.set_value(ty, Rc::new(TypeEnum::TObj { obj_id: id, fields, params }));
+    }
+}
+
+pub fn impl_binop(
+    unifier: &mut Unifier,
+    _store: &PrimitiveStore,
+    ty: Type,
+    other_ty: &[Type],
+    ret_ty: Option<Type>,
+    ops: &[Operator],
+) {
+    with_fields(unifier, ty, |unifier, fields| {
+        let (other_ty, other_var_id) = if other_ty.len() == 1 {
+            (other_ty[0], None)
+        } else {
+            let tvar = unifier.get_fresh_var_with_range(other_ty, Some("N".into()), None);
+            (tvar.ty, Some(tvar.id))
+        };
+
+        let function_vars = if let Some(var_id) = other_var_id {
+            vec![(var_id, other_ty)].into_iter().collect::<VarMap>()
+        } else {
+            VarMap::new()
+        };
+
+        let ret_ty = ret_ty.unwrap_or_else(|| unifier.get_fresh_var(None, None).ty);
+
+        for (base_op, variant) in iproduct!(ops, [BinopVariant::Normal, BinopVariant::AugAssign]) {
+            let op = Binop { base: *base_op, variant };
+            fields.insert(op.op_info().method_name.into(), {
+                (
+                    unifier.add_ty(TypeEnum::TFunc(FunSignature {
+                        ret: ret_ty,
+                        vars: function_vars.clone(),
+                        args: vec![FuncArg {
+                            ty: other_ty,
+                            default_value: None,
+                            name: "other".into(),
+                            is_vararg: false,
+                        }],
+                    })),
+                    false,
+                )
+            });
+        }
+    });
+}
+
+pub fn impl_unaryop(unifier: &mut Unifier, ty: Type, ret_ty: Option<Type>, ops: &[Unaryop]) {
+    with_fields(unifier, ty, |unifier, fields| {
+        let ret_ty = ret_ty.unwrap_or_else(|| unifier.get_fresh_var(None, None).ty);
+
+        for op in ops {
+            fields.insert(
+                op.op_info().method_name.into(),
+                (
+                    unifier.add_ty(TypeEnum::TFunc(FunSignature {
+                        ret: ret_ty,
+                        vars: VarMap::new(),
+                        args: vec![],
+                    })),
+                    false,
+                ),
+            );
+        }
+    });
+}
+
+pub fn impl_cmpop(
+    unifier: &mut Unifier,
+    _store: &PrimitiveStore,
+    ty: Type,
+    other_ty: &[Type],
+    ops: &[Cmpop],
+    ret_ty: Option<Type>,
+) {
+    with_fields(unifier, ty, |unifier, fields| {
+        let (other_ty, other_var_id) = if other_ty.len() == 1 {
+            (other_ty[0], None)
+        } else {
+            let tvar = unifier.get_fresh_var_with_range(other_ty, Some("N".into()), None);
+            (tvar.ty, Some(tvar.id))
+        };
+
+        let function_vars = if let Some(var_id) = other_var_id {
+            vec![(var_id, other_ty)].into_iter().collect::<VarMap>()
+        } else {
+            VarMap::new()
+        };
+
+        let ret_ty = ret_ty.unwrap_or_else(|| unifier.get_fresh_var(None, None).ty);
+
+        for op in ops {
+            fields.insert(
+                op.op_info().method_name.into(),
+                (
+                    unifier.add_ty(TypeEnum::TFunc(FunSignature {
+                        ret: ret_ty,
+                        vars: function_vars.clone(),
+                        args: vec![FuncArg {
+                            ty: other_ty,
+                            default_value: None,
+                            name: "other".into(),
+                            is_vararg: false,
+                        }],
+                    })),
+                    false,
+                ),
+            );
+        }
+    });
+}
+
+/// `Add`, `Sub`, `Mult`
+pub fn impl_basic_arithmetic(
+    unifier: &mut Unifier,
+    store: &PrimitiveStore,
+    ty: Type,
+    other_ty: &[Type],
+    ret_ty: Option<Type>,
+) {
+    impl_binop(
+        unifier,
+        store,
+        ty,
+        other_ty,
+        ret_ty,
+        &[Operator::Add, Operator::Sub, Operator::Mult],
+    );
+}
+
+/// `Pow`
+pub fn impl_pow(
+    unifier: &mut Unifier,
+    store: &PrimitiveStore,
+    ty: Type,
+    other_ty: &[Type],
+    ret_ty: Option<Type>,
+) {
+    impl_binop(unifier, store, ty, other_ty, ret_ty, &[Operator::Pow]);
+}
+
+/// `BitOr`, `BitXor`, `BitAnd`
+pub fn impl_bitwise_arithmetic(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type) {
+    impl_binop(
+        unifier,
+        store,
+        ty,
+        &[ty],
+        Some(ty),
+        &[Operator::BitAnd, Operator::BitOr, Operator::BitXor],
+    );
+}
+
+/// `LShift`, `RShift`
+pub fn impl_bitwise_shift(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type) {
+    impl_binop(
+        unifier,
+        store,
+        ty,
+        &[store.int32, store.uint32],
+        Some(ty),
+        &[Operator::LShift, Operator::RShift],
+    );
+}
+
+/// `Div`
+pub fn impl_div(
+    unifier: &mut Unifier,
+    store: &PrimitiveStore,
+    ty: Type,
+    other_ty: &[Type],
+    ret_ty: Option<Type>,
+) {
+    impl_binop(unifier, store, ty, other_ty, ret_ty, &[Operator::Div]);
+}
+
+/// `FloorDiv`
+pub fn impl_floordiv(
+    unifier: &mut Unifier,
+    store: &PrimitiveStore,
+    ty: Type,
+    other_ty: &[Type],
+    ret_ty: Option<Type>,
+) {
+    impl_binop(unifier, store, ty, other_ty, ret_ty, &[Operator::FloorDiv]);
+}
+
+/// `Mod`
+pub fn impl_mod(
+    unifier: &mut Unifier,
+    store: &PrimitiveStore,
+    ty: Type,
+    other_ty: &[Type],
+    ret_ty: Option<Type>,
+) {
+    impl_binop(unifier, store, ty, other_ty, ret_ty, &[Operator::Mod]);
+}
+
+/// [`Operator::MatMult`]
+pub fn impl_matmul(
+    unifier: &mut Unifier,
+    store: &PrimitiveStore,
+    ty: Type,
+    other_ty: &[Type],
+    ret_ty: Option<Type>,
+) {
+    impl_binop(unifier, store, ty, other_ty, ret_ty, &[Operator::MatMult]);
+}
+
+/// `UAdd`, `USub`
+pub fn impl_sign(unifier: &mut Unifier, _store: &PrimitiveStore, ty: Type, ret_ty: Option<Type>) {
+    impl_unaryop(unifier, ty, ret_ty, &[Unaryop::UAdd, Unaryop::USub]);
+}
+
+/// `Invert`
+pub fn impl_invert(unifier: &mut Unifier, _store: &PrimitiveStore, ty: Type, ret_ty: Option<Type>) {
+    impl_unaryop(unifier, ty, ret_ty, &[Unaryop::Invert]);
+}
+
+/// `Not`
+pub fn impl_not(unifier: &mut Unifier, _store: &PrimitiveStore, ty: Type, ret_ty: Option<Type>) {
+    impl_unaryop(unifier, ty, ret_ty, &[Unaryop::Not]);
+}
+
+/// `Lt`, `LtE`, `Gt`, `GtE`
+pub fn impl_comparison(
+    unifier: &mut Unifier,
+    store: &PrimitiveStore,
+    ty: Type,
+    other_ty: &[Type],
+    ret_ty: Option<Type>,
+) {
+    impl_cmpop(
+        unifier,
+        store,
+        ty,
+        other_ty,
+        &[Cmpop::Lt, Cmpop::Gt, Cmpop::LtE, Cmpop::GtE],
+        ret_ty,
+    );
+}
+
+/// `Eq`, `NotEq`
+pub fn impl_eq(
+    unifier: &mut Unifier,
+    store: &PrimitiveStore,
+    ty: Type,
+    other_ty: &[Type],
+    ret_ty: Option<Type>,
+) {
+    impl_cmpop(unifier, store, ty, other_ty, &[Cmpop::Eq, Cmpop::NotEq], ret_ty);
+}
+
+/// Returns the expected return type of binary operations with at least one `ndarray` operand.
+pub fn typeof_ndarray_broadcast(
+    unifier: &mut Unifier,
+    primitives: &PrimitiveStore,
+    left: Type,
+    right: Type,
+) -> Result<Type, String> {
+    let is_left_ndarray = left.obj_id(unifier).is_some_and(|id| id == PrimDef::NDArray.id());
+    let is_right_ndarray = right.obj_id(unifier).is_some_and(|id| id == PrimDef::NDArray.id());
+
+    assert!(is_left_ndarray || is_right_ndarray);
+
+    if is_left_ndarray && is_right_ndarray {
+        // Perform broadcasting on two ndarray operands.
+
+        let (left_ty_dtype, left_ty_ndims) = unpack_ndarray_var_tys(unifier, left);
+        let (right_ty_dtype, right_ty_ndims) = unpack_ndarray_var_tys(unifier, right);
+
+        assert!(unifier.unioned(left_ty_dtype, right_ty_dtype));
+
+        let left_ty_ndims = match &*unifier.get_ty_immutable(left_ty_ndims) {
+            TypeEnum::TLiteral { values, .. } => values.clone(),
+            _ => unreachable!(),
+        };
+        let right_ty_ndims = match &*unifier.get_ty_immutable(right_ty_ndims) {
+            TypeEnum::TLiteral { values, .. } => values.clone(),
+            _ => unreachable!(),
+        };
+
+        let res_ndims = left_ty_ndims
+            .into_iter()
+            .cartesian_product(right_ty_ndims)
+            .map(|(left, right)| {
+                let left_val = u64::try_from(left).unwrap();
+                let right_val = u64::try_from(right).unwrap();
+
+                max(left_val, right_val)
+            })
+            .unique()
+            .map(SymbolValue::U64)
+            .collect_vec();
+        let res_ndims = unifier.get_fresh_literal(res_ndims, None);
+
+        Ok(make_ndarray_ty(unifier, primitives, Some(left_ty_dtype), Some(res_ndims)))
+    } else {
+        let (ndarray_ty, scalar_ty) = if is_left_ndarray { (left, right) } else { (right, left) };
+
+        let (ndarray_ty_dtype, _) = unpack_ndarray_var_tys(unifier, ndarray_ty);
+
+        if unifier.unioned(ndarray_ty_dtype, scalar_ty) {
+            Ok(ndarray_ty)
+        } else {
+            let (expected_ty, actual_ty) = if is_left_ndarray {
+                (ndarray_ty_dtype, scalar_ty)
+            } else {
+                (scalar_ty, ndarray_ty_dtype)
+            };
+
+            Err(format!(
+                "Expected right-hand side operand to be {}, got {}",
+                unifier.stringify(expected_ty),
+                unifier.stringify(actual_ty),
+            ))
+        }
+    }
+}
+
+/// Returns the return type given a binary operator and its primitive operands.
+pub fn typeof_binop(
+    unifier: &mut Unifier,
+    primitives: &PrimitiveStore,
+    op: Operator,
+    lhs: Type,
+    rhs: Type,
+) -> Result<Option<Type>, String> {
+    let op = Binop { base: op, variant: BinopVariant::Normal };
+
+    let is_left_list = lhs.obj_id(unifier).is_some_and(|id| id == PrimDef::List.id());
+    let is_right_list = rhs.obj_id(unifier).is_some_and(|id| id == PrimDef::List.id());
+    let is_left_ndarray = lhs.obj_id(unifier).is_some_and(|id| id == PrimDef::NDArray.id());
+    let is_right_ndarray = rhs.obj_id(unifier).is_some_and(|id| id == PrimDef::NDArray.id());
+
+    Ok(Some(match op.base {
+        Operator::Add | Operator::Sub | Operator::Mult | Operator::Mod | Operator::FloorDiv => {
+            if is_left_list || is_right_list {
+                if ![Operator::Add, Operator::Mult].contains(&op.base) {
+                    return Err(format!(
+                        "Binary operator {} not supported for list",
+                        op.op_info().symbol
+                    ));
+                }
+
+                if is_left_list {
+                    lhs
+                } else {
+                    rhs
+                }
+            } else if is_left_ndarray || is_right_ndarray {
+                typeof_ndarray_broadcast(unifier, primitives, lhs, rhs)?
+            } else if unifier.unioned(lhs, rhs) {
+                lhs
+            } else {
+                return Ok(None);
+            }
+        }
+
+        Operator::MatMult => {
+            // NOTE: NumPy matmul's LHS and RHS must both be ndarrays. Scalars are not allowed.
+            match (&*unifier.get_ty(lhs), &*unifier.get_ty(rhs)) {
+                (
+                    TypeEnum::TObj { obj_id: lhs_obj_id, .. },
+                    TypeEnum::TObj { obj_id: rhs_obj_id, .. },
+                ) if *lhs_obj_id == primitives.ndarray.obj_id(unifier).unwrap()
+                    && *rhs_obj_id == primitives.ndarray.obj_id(unifier).unwrap() =>
+                {
+                    // LHS and RHS have valid types
+                }
+                _ => {
+                    let lhs_str = unifier.stringify(lhs);
+                    let rhs_str = unifier.stringify(rhs);
+                    return Err(format!("ndarray.__matmul__ only accepts ndarray operands, but left operand has type {lhs_str}, and right operand has type {rhs_str}"));
+                }
+            }
+
+            let (_, lhs_ndims) = unpack_ndarray_var_tys(unifier, lhs);
+            let lhs_ndims = match &*unifier.get_ty_immutable(lhs_ndims) {
+                TypeEnum::TLiteral { values, .. } => {
+                    assert_eq!(values.len(), 1);
+                    u64::try_from(values[0].clone()).unwrap()
+                }
+                _ => unreachable!(),
+            };
+            let (_, rhs_ndims) = unpack_ndarray_var_tys(unifier, rhs);
+            let rhs_ndims = match &*unifier.get_ty_immutable(rhs_ndims) {
+                TypeEnum::TLiteral { values, .. } => {
+                    assert_eq!(values.len(), 1);
+                    u64::try_from(values[0].clone()).unwrap()
+                }
+                _ => unreachable!(),
+            };
+
+            match (lhs_ndims, rhs_ndims) {
+                (2, 2) => typeof_ndarray_broadcast(unifier, primitives, lhs, rhs)?,
+                (lhs, rhs) if lhs == 0 || rhs == 0 => {
+                    return Err(format!(
+                    "Input operand {} does not have enough dimensions (has {lhs}, requires {rhs})",
+                    u8::from(rhs == 0)
+                ))
+                }
+                (lhs, rhs) => {
+                    return Err(format!(
+                        "ndarray.__matmul__ on {lhs}D and {rhs}D operands not supported"
+                    ))
+                }
+            }
+        }
+
+        Operator::Div => {
+            if is_left_ndarray || is_right_ndarray {
+                typeof_ndarray_broadcast(unifier, primitives, lhs, rhs)?
+            } else if unifier.unioned(lhs, rhs) {
+                primitives.float
+            } else {
+                return Ok(None);
+            }
+        }
+
+        Operator::Pow => {
+            if is_left_ndarray || is_right_ndarray {
+                typeof_ndarray_broadcast(unifier, primitives, lhs, rhs)?
+            } else if [
+                primitives.int32,
+                primitives.int64,
+                primitives.uint32,
+                primitives.uint64,
+                primitives.float,
+            ]
+            .into_iter()
+            .any(|ty| unifier.unioned(lhs, ty))
+            {
+                lhs
+            } else {
+                return Ok(None);
+            }
+        }
+
+        Operator::LShift | Operator::RShift => lhs,
+        Operator::BitOr | Operator::BitXor | Operator::BitAnd => {
+            if unifier.unioned(lhs, rhs) {
+                lhs
+            } else {
+                return Ok(None);
+            }
+        }
+    }))
+}
+
+pub fn typeof_unaryop(
+    unifier: &mut Unifier,
+    primitives: &PrimitiveStore,
+    op: Unaryop,
+    operand: Type,
+) -> Result<Option<Type>, String> {
+    let operand_obj_id = operand.obj_id(unifier);
+
+    if op == Unaryop::Not
+        && operand_obj_id.is_some_and(|id| id == primitives.ndarray.obj_id(unifier).unwrap())
+    {
+        return Err(
+            "The truth value of an array with more than one element is ambiguous".to_string()
+        );
+    }
+
+    Ok(match op {
+        Unaryop::Not => match operand_obj_id {
+            Some(v) if v == PrimDef::NDArray.id() => Some(operand),
+            Some(_) => Some(primitives.bool),
+            _ => None,
+        },
+
+        Unaryop::Invert => {
+            if operand_obj_id.is_some_and(|id| id == PrimDef::Bool.id()) {
+                Some(primitives.int32)
+            } else if operand_obj_id.is_some_and(|id| PrimDef::iter().any(|prim| id == prim.id())) {
+                Some(operand)
+            } else {
+                None
+            }
+        }
+
+        Unaryop::UAdd | Unaryop::USub => {
+            if operand_obj_id.is_some_and(|id| id == PrimDef::NDArray.id()) {
+                let (dtype, _) = unpack_ndarray_var_tys(unifier, operand);
+                if dtype.obj_id(unifier).is_some_and(|id| id == PrimDef::Bool.id()) {
+                    return Err(if op == Unaryop::UAdd {
+                        "The ufunc 'positive' cannot be applied to ndarray[bool, N]".to_string()
+                    } else {
+                        "The numpy boolean negative, the `-` operator, is not supported, use the `~` operator function instead.".to_string()
+                    });
+                }
+
+                Some(operand)
+            } else if operand_obj_id.is_some_and(|id| id == PrimDef::Bool.id()) {
+                Some(primitives.int32)
+            } else if operand_obj_id.is_some_and(|id| PrimDef::iter().any(|prim| id == prim.id())) {
+                Some(operand)
+            } else {
+                None
+            }
+        }
+    })
+}
+
+/// Returns the return type given a comparison operator and its primitive operands.
+pub fn typeof_cmpop(
+    unifier: &mut Unifier,
+    primitives: &PrimitiveStore,
+    _op: Cmpop,
+    lhs: Type,
+    rhs: Type,
+) -> Result<Option<Type>, String> {
+    let is_left_ndarray = lhs.obj_id(unifier).is_some_and(|id| id == PrimDef::NDArray.id());
+    let is_right_ndarray = rhs.obj_id(unifier).is_some_and(|id| id == PrimDef::NDArray.id());
+
+    Ok(Some(if is_left_ndarray || is_right_ndarray {
+        let brd = typeof_ndarray_broadcast(unifier, primitives, lhs, rhs)?;
+        let (_, ndims) = unpack_ndarray_var_tys(unifier, brd);
+
+        make_ndarray_ty(unifier, primitives, Some(primitives.bool), Some(ndims))
+    } else if unifier.unioned(lhs, rhs) {
+        primitives.bool
+    } else {
+        return Ok(None);
+    }))
+}
+
+pub fn set_primitives_magic_methods(store: &PrimitiveStore, unifier: &mut Unifier) {
+    let PrimitiveStore {
+        int32: int32_t,
+        int64: int64_t,
+        float: float_t,
+        bool: bool_t,
+        uint32: uint32_t,
+        uint64: uint64_t,
+        str: str_t,
+        list: list_t,
+        ndarray: ndarray_t,
+        ..
+    } = *store;
+    let size_t = store.usize();
+
+    /* int ======== */
+    for t in [int32_t, int64_t, uint32_t, uint64_t] {
+        let ndarray_int_t = make_ndarray_ty(unifier, store, Some(t), None);
+        impl_basic_arithmetic(unifier, store, t, &[t, ndarray_int_t], None);
+        impl_pow(unifier, store, t, &[t, ndarray_int_t], None);
+        impl_bitwise_arithmetic(unifier, store, t);
+        impl_bitwise_shift(unifier, store, t);
+        impl_div(unifier, store, t, &[t, ndarray_int_t], None);
+        impl_floordiv(unifier, store, t, &[t, ndarray_int_t], None);
+        impl_mod(unifier, store, t, &[t, ndarray_int_t], None);
+        impl_invert(unifier, store, t, Some(t));
+        impl_not(unifier, store, t, Some(bool_t));
+        impl_comparison(unifier, store, t, &[t, ndarray_int_t], None);
+        impl_eq(unifier, store, t, &[t, ndarray_int_t], None);
+    }
+    for t in [int32_t, int64_t] {
+        impl_sign(unifier, store, t, Some(t));
+    }
+
+    /* float ======== */
+    let ndarray_float_t = make_ndarray_ty(unifier, store, Some(float_t), None);
+    let ndarray_int32_t = make_ndarray_ty(unifier, store, Some(int32_t), None);
+    impl_basic_arithmetic(unifier, store, float_t, &[float_t, ndarray_float_t], None);
+    impl_pow(unifier, store, float_t, &[int32_t, float_t, ndarray_int32_t, ndarray_float_t], None);
+    impl_div(unifier, store, float_t, &[float_t, ndarray_float_t], None);
+    impl_floordiv(unifier, store, float_t, &[float_t, ndarray_float_t], None);
+    impl_mod(unifier, store, float_t, &[float_t, ndarray_float_t], None);
+    impl_sign(unifier, store, float_t, Some(float_t));
+    impl_not(unifier, store, float_t, Some(bool_t));
+    impl_comparison(unifier, store, float_t, &[float_t, ndarray_float_t], None);
+    impl_eq(unifier, store, float_t, &[float_t, ndarray_float_t], None);
+
+    /* bool ======== */
+    let ndarray_bool_t = make_ndarray_ty(unifier, store, Some(bool_t), None);
+    impl_invert(unifier, store, bool_t, Some(int32_t));
+    impl_not(unifier, store, bool_t, Some(bool_t));
+    impl_sign(unifier, store, bool_t, Some(int32_t));
+    impl_eq(unifier, store, bool_t, &[bool_t, ndarray_bool_t], None);
+
+    /* str ========= */
+    impl_cmpop(unifier, store, str_t, &[str_t], &[Cmpop::Eq, Cmpop::NotEq], Some(bool_t));
+
+    /* list ======== */
+    impl_binop(unifier, store, list_t, &[list_t], Some(list_t), &[Operator::Add]);
+    impl_binop(unifier, store, list_t, &[int32_t, int64_t], Some(list_t), &[Operator::Mult]);
+    impl_cmpop(unifier, store, list_t, &[list_t], &[Cmpop::Eq, Cmpop::NotEq], Some(bool_t));
+
+    /* ndarray ===== */
+    let ndarray_usized_ndims_tvar =
+        unifier.get_fresh_const_generic_var(size_t, Some("ndarray_ndims".into()), None);
+    let ndarray_unsized_t =
+        make_ndarray_ty(unifier, store, None, Some(ndarray_usized_ndims_tvar.ty));
+    let (ndarray_dtype_t, _) = unpack_ndarray_var_tys(unifier, ndarray_t);
+    let (ndarray_unsized_dtype_t, _) = unpack_ndarray_var_tys(unifier, ndarray_unsized_t);
+    impl_basic_arithmetic(
+        unifier,
+        store,
+        ndarray_t,
+        &[ndarray_unsized_t, ndarray_unsized_dtype_t],
+        None,
+    );
+    impl_pow(unifier, store, ndarray_t, &[ndarray_unsized_t, ndarray_unsized_dtype_t], None);
+    impl_div(unifier, store, ndarray_t, &[ndarray_t, ndarray_dtype_t], None);
+    impl_floordiv(unifier, store, ndarray_t, &[ndarray_unsized_t, ndarray_unsized_dtype_t], None);
+    impl_mod(unifier, store, ndarray_t, &[ndarray_unsized_t, ndarray_unsized_dtype_t], None);
+    impl_matmul(unifier, store, ndarray_t, &[ndarray_t], Some(ndarray_t));
+    impl_sign(unifier, store, ndarray_t, Some(ndarray_t));
+    impl_invert(unifier, store, ndarray_t, Some(ndarray_t));
+    impl_eq(unifier, store, ndarray_t, &[ndarray_unsized_t, ndarray_unsized_dtype_t], None);
+    impl_comparison(unifier, store, ndarray_t, &[ndarray_unsized_t, ndarray_unsized_dtype_t], None);
+}

nac3core/src/typecheck/mod.rs

@@ -1,8 +1,6 @@
-pub mod context;
-pub mod inference_core;
-pub mod location;
+mod function_check;
 pub mod magic_methods;
-pub mod primitives;
-pub mod symbol_resolver;
+pub mod type_error;
+pub mod type_inferencer;
 pub mod typedef;
-pub mod type_check;
+mod unification_table;

nac3core/src/typecheck/primitives.rs

@@ -1,184 +0,0 @@
-use super::typedef::{TypeEnum::*, *};
-use super::context::*;
-use std::collections::HashMap;
-
-pub const TUPLE_TYPE: ParamId = ParamId(0);
-pub const LIST_TYPE: ParamId = ParamId(1);
-
-pub const BOOL_TYPE: PrimitiveId = PrimitiveId(0);
-pub const INT32_TYPE: PrimitiveId = PrimitiveId(1);
-pub const INT64_TYPE: PrimitiveId = PrimitiveId(2);
-pub const FLOAT_TYPE: PrimitiveId = PrimitiveId(3);
-
-fn impl_math(def: &mut TypeDef, ty: &Type) {
-    let result = Some(ty.clone());
-    let fun = FnDef {
-        args: vec![ty.clone()],
-        result: result.clone(),
-    };
-    def.methods.insert("__add__", fun.clone());
-    def.methods.insert("__sub__", fun.clone());
-    def.methods.insert("__mul__", fun.clone());
-    def.methods.insert(
-        "__neg__",
-        FnDef {
-            args: vec![],
-            result,
-        },
-    );
-    def.methods.insert(
-        "__truediv__",
-        FnDef {
-            args: vec![ty.clone()],
-            result: Some(PrimitiveType(FLOAT_TYPE).into()),
-        },
-    );
-    def.methods.insert("__floordiv__", fun.clone());
-    def.methods.insert("__mod__", fun.clone());
-    def.methods.insert("__pow__", fun);
-}
-
-fn impl_bits(def: &mut TypeDef, ty: &Type) {
-    let result = Some(ty.clone());
-    let fun = FnDef {
-        args: vec![PrimitiveType(INT32_TYPE).into()],
-        result,
-    };
-
-    def.methods.insert("__lshift__", fun.clone());
-    def.methods.insert("__rshift__", fun);
-    def.methods.insert(
-        "__xor__",
-        FnDef {
-            args: vec![ty.clone()],
-            result: Some(ty.clone()),
-        },
-    );
-}
-
-fn impl_eq(def: &mut TypeDef, ty: &Type) {
-    let fun = FnDef {
-        args: vec![ty.clone()],
-        result: Some(PrimitiveType(BOOL_TYPE).into()),
-    };
-
-    def.methods.insert("__eq__", fun.clone());
-    def.methods.insert("__ne__", fun);
-}
-
-fn impl_order(def: &mut TypeDef, ty: &Type) {
-    let fun = FnDef {
-        args: vec![ty.clone()],
-        result: Some(PrimitiveType(BOOL_TYPE).into()),
-    };
-
-    def.methods.insert("__lt__", fun.clone());
-    def.methods.insert("__gt__", fun.clone());
-    def.methods.insert("__le__", fun.clone());
-    def.methods.insert("__ge__", fun);
-}
-
-pub fn basic_ctx() -> GlobalContext<'static> {
-    let primitives = [
-        TypeDef {
-            name: "bool",
-            fields: HashMap::new(),
-            methods: HashMap::new(),
-        },
-        TypeDef {
-            name: "int32",
-            fields: HashMap::new(),
-            methods: HashMap::new(),
-        },
-        TypeDef {
-            name: "int64",
-            fields: HashMap::new(),
-            methods: HashMap::new(),
-        },
-        TypeDef {
-            name: "float",
-            fields: HashMap::new(),
-            methods: HashMap::new(),
-        },
-    ]
-    .to_vec();
-    let mut ctx = GlobalContext::new(primitives);
-
-    let b = ctx.get_primitive(BOOL_TYPE);
-    let b_def = ctx.get_primitive_def_mut(BOOL_TYPE);
-    impl_eq(b_def, &b);
-    let int32 = ctx.get_primitive(INT32_TYPE);
-    let int32_def = ctx.get_primitive_def_mut(INT32_TYPE);
-    impl_math(int32_def, &int32);
-    impl_bits(int32_def, &int32);
-    impl_order(int32_def, &int32);
-    impl_eq(int32_def, &int32);
-    let int64 = ctx.get_primitive(INT64_TYPE);
-    let int64_def = ctx.get_primitive_def_mut(INT64_TYPE);
-    impl_math(int64_def, &int64);
-    impl_bits(int64_def, &int64);
-    impl_order(int64_def, &int64);
-    impl_eq(int64_def, &int64);
-    let float = ctx.get_primitive(FLOAT_TYPE);
-    let float_def = ctx.get_primitive_def_mut(FLOAT_TYPE);
-    impl_math(float_def, &float);
-    impl_order(float_def, &float);
-    impl_eq(float_def, &float);
-
-    let t = ctx.add_variable_private(VarDef {
-        name: "T",
-        bound: vec![],
-    });
-
-    ctx.add_parametric(ParametricDef {
-        base: TypeDef {
-            name: "tuple",
-            fields: HashMap::new(),
-            methods: HashMap::new(),
-        },
-        // we have nothing for tuple, so no param def
-        params: vec![],
-    });
-
-    ctx.add_parametric(ParametricDef {
-        base: TypeDef {
-            name: "list",
-            fields: HashMap::new(),
-            methods: HashMap::new(),
-        },
-        params: vec![t],
-    });
-
-    let i = ctx.add_variable_private(VarDef {
-        name: "I",
-        bound: vec![
-            PrimitiveType(INT32_TYPE).into(),
-            PrimitiveType(INT64_TYPE).into(),
-            PrimitiveType(FLOAT_TYPE).into(),
-        ],
-    });
-    let args = vec![TypeVariable(i).into()];
-    ctx.add_fn(
-        "int32",
-        FnDef {
-            args: args.clone(),
-            result: Some(PrimitiveType(INT32_TYPE).into()),
-        },
-    );
-    ctx.add_fn(
-        "int64",
-        FnDef {
-            args: args.clone(),
-            result: Some(PrimitiveType(INT64_TYPE).into()),
-        },
-    );
-    ctx.add_fn(
-        "float",
-        FnDef {
-            args,
-            result: Some(PrimitiveType(FLOAT_TYPE).into()),
-        },
-    );
-
-    ctx
-}

nac3core/src/typecheck/symbol_resolver.rs

@@ -1,23 +0,0 @@
-use super::typedef::Type;
-use super::location::Location;
-
-pub enum SymbolType {
-    TypeName(Type),
-    Identifier(Type),
-}
-
-pub enum SymbolValue<'a> {
-    I32(i32),
-    I64(i64),
-    Double(f64),
-    Bool(bool),
-    Tuple(&'a [SymbolValue<'a>]),
-    Bytes(&'a [u8]),
-}
-
-pub trait SymbolResolver {
-    fn get_symbol_type(&self, str: &str) -> Option<SymbolType>;
-    fn get_symbol_value(&self, str: &str) -> Option<SymbolValue>;
-    fn get_symbol_location(&self, str: &str) -> Option<Location>;
-    // handle function call etc.
-}

nac3core/src/typecheck/type_check.rs

@@ -1,752 +0,0 @@
-use std::convert::TryInto;
-
-use crate::typecheck::context::InferenceContext;
-use crate::typecheck::inference_core;
-use crate::typecheck::magic_methods;
-use crate::typecheck::typedef::{Type, TypeEnum};
-use crate::typecheck::primitives;
-use rustpython_parser::ast;
-
-struct NaiveFolder;
-impl ast::fold::Fold<()> for NaiveFolder {
-    type TargetU = Option<Type>;
-    type Error = String;
-    fn map_user(&mut self, _user: ()) -> Result<Self::TargetU, Self::Error> {
-        Ok(None)
-    }
-}
-
-pub struct TypeInferencer<'a> {
-    pub ctx: InferenceContext<'a>,
-    pub error_stack: Vec<(String, ast::Location)>
-}
-
-impl<'a> ast::fold::Fold<()> for TypeInferencer<'a> {
-    type TargetU = Option<Type>;
-    type Error = String;
-
-    fn map_user(&mut self, _user: ()) -> Result<Self::TargetU, Self::Error> {
-        Ok(None)
-    }
-
-    fn fold_expr(&mut self, node: ast::Expr<()>) -> Result<ast::Expr<Self::TargetU>, Self::Error> {
-        
-        self.error_stack.push(("Checking ".to_string() + node.node.name(), node.location));
-
-        let expr = match &node.node {
-            ast::ExprKind::ListComp { .. } => return self.fold_listcomp(node),
-            _ => rustpython_parser::ast::fold::fold_expr(self, node)?
-        };
-
-        let ret = Ok(ast::Expr {
-            // compute type info and store in the custom field
-            custom: match &expr.node {
-                ast::ExprKind::Constant {value, kind: _} => self.infer_constant(value),
-                ast::ExprKind::Name {id, ctx: _} => Ok(Some(self.ctx.resolve(id)?)),
-                ast::ExprKind::List {elts, ctx: _} => self.infer_list(elts),
-                ast::ExprKind::Tuple {elts, ctx: _} => self.infer_tuple(elts),
-                ast::ExprKind::Attribute {value, attr, ctx: _} => self.infer_attribute(value, attr),
-                ast::ExprKind::BoolOp {op: _, values} => self.infer_bool_ops(values),
-                ast::ExprKind::BinOp {left, op, right} => self.infer_bin_ops(left, op, right),
-                ast::ExprKind::UnaryOp {op, operand} => self.infer_unary_ops(op, operand),
-                ast::ExprKind::Compare {left, ops, comparators} => self.infer_compare(left, ops, comparators),
-                ast::ExprKind::Call {func, args, keywords} => self.infer_call(func, args, keywords),
-                ast::ExprKind::Subscript {value, slice, ctx: _} => self.infer_subscript(value, slice),
-                ast::ExprKind::IfExp {test, body, orelse} => self.infer_if_expr(test, body, orelse),
-                ast::ExprKind::ListComp {elt: _, generators: _} => unreachable!("should not earch here, the list comp should have been folded before"), // already folded
-                ast::ExprKind::Slice { .. } => Ok(None), // special handling for slice, which is supported
-                _ => Err("not supported yet".into())
-            }?,
-            location: expr.location,
-            node: expr.node
-        });
-        self.error_stack.pop();
-        ret
-    }
-
-    fn fold_stmt(&mut self, node: ast::Stmt<()>) -> Result<ast::Stmt<Self::TargetU>, Self::Error> {
-        let stmt = match node.node {
-            ast::StmtKind::AnnAssign {target, annotation, value, simple} => {
-                let target_folded = Box::new(self.fold_expr( *target)?);
-                let value = if let Some(v) = value {
-                    let value_folded = Box::new(self.fold_expr(*v)?);
-                    if target_folded.custom == value_folded.custom {
-                        Some(value_folded)
-                    } else {
-                        return Err("Assignment LHF does not have the same type as RHS".into())
-                    }
-                } else {
-                    None
-                };
-                // TODO check consistency with type annotation
-                ast::Located {
-                    location: node.location,
-                    custom: None,
-                    node: ast::StmtKind::AnnAssign {
-                        target: target_folded, 
-                        annotation: Box::new(NaiveFolder.fold_expr(*annotation)?), 
-                        value, 
-                        simple
-                    },
-                }
-            }
-            _ => ast::fold::fold_stmt(self, node)?
-        };
-
-        match &stmt.node {
-            ast::StmtKind::For { target, iter, .. } => {
-                if let Some(TypeEnum::ParametricType(primitives::LIST_TYPE, ls)) = iter.custom.as_deref() {
-                    unimplemented!()
-                    // TODO
-                } else {
-                    return Err("can only iterate over list".into())
-                }
-            }
-            ast::StmtKind::If { test, .. } | ast::StmtKind::While { test, .. } => {
-                if test.custom != Some(self.ctx.get_primitive(primitives::BOOL_TYPE)) {
-                    return Err("Test should be bool".into());
-                }
-            }
-            ast::StmtKind::Assign { targets, value, .. } => {
-                unimplemented!();
-                // TODO
-            }
-            ast::StmtKind::AnnAssign { .. } | ast::StmtKind::Expr { .. } => {}
-            ast::StmtKind::Break | ast::StmtKind::Continue => {}
-            ast::StmtKind::Return { value } => {
-                unimplemented!()
-                // TODO
-            }
-            _ => return Err("Unsupported statement type".to_string()),
-        }
-
-        Ok(stmt)
-    }
-}
-
-impl<'a> TypeInferencer<'a> {
-    fn infer_constant(&self, constant: &ast::Constant) -> Result<Option<Type>, String> {
-        match constant {
-            ast::Constant::Bool(_) => 
-                Ok(Some(self.ctx.get_primitive(primitives::BOOL_TYPE))),
-            
-            ast::Constant::Int(val) => {
-                let int32: Result<i32, _> = val.try_into();
-                let int64: Result<i64, _> = val.try_into();
-                
-                if int32.is_ok() {
-                    Ok(Some(self.ctx.get_primitive(primitives::INT32_TYPE)))
-                } else if int64.is_ok() {
-                    Ok(Some(self.ctx.get_primitive(primitives::INT64_TYPE)))
-                } else {
-                    Err("Integer out of bound".into())
-                }
-            },
-    
-            ast::Constant::Float(_) => 
-                Ok(Some(self.ctx.get_primitive(primitives::FLOAT_TYPE))),
-
-            ast::Constant::Tuple(vals) => {
-                let result = vals
-                    .iter()
-                    .map(|x| self.infer_constant(x))
-                    .collect::<Vec<_>>();
-                
-                if result.iter().all(|x| x.is_ok()) {
-                    Ok(Some(TypeEnum::ParametricType(
-                        primitives::TUPLE_TYPE, 
-                        result
-                            .into_iter()
-                            .map(|x| x.unwrap().unwrap())
-                            .collect::<Vec<_>>(),
-                    ).into()))
-                } else {
-                    Err("Some elements in tuple cannot be typed".into())
-                }
-            }
-
-            _ => Err("not supported".into())
-        }
-    }
-
-    fn infer_list(&self, elts: &[ast::Expr<Option<Type>>]) -> Result<Option<Type>, String> {
-        if elts.is_empty() {
-            Ok(Some(TypeEnum::ParametricType(primitives::LIST_TYPE, vec![TypeEnum::BotType.into()]).into()))
-        } else {
-            let types = elts
-                .iter()
-                .map(|x| &x.custom)
-                .collect::<Vec<_>>();
-            
-            if types.iter().all(|x| x.is_some()) {
-                let head = types.iter().next().unwrap(); // here unwrap alone should be fine after the previous check
-                if types.iter().all(|x| x.eq(head)) {
-                    Ok(Some(TypeEnum::ParametricType(primitives::LIST_TYPE, vec![(*head).clone().unwrap()]).into()))
-                } else {
-                    Err("inhomogeneous list is not allowed".into())
-                }
-            } else {
-                Err("list elements must have some type".into())
-            }
-        }
-    }
-
-    fn infer_tuple(&self, elts: &[ast::Expr<Option<Type>>]) -> Result<Option<Type>, String> {
-        let types = elts
-            .iter()
-            .map(|x| (x.custom).clone())
-            .collect::<Vec<_>>();
-
-        if types.iter().all(|x| x.is_some()) {
-            Ok(Some(TypeEnum::ParametricType(
-                primitives::TUPLE_TYPE, 
-                types.into_iter().map(|x| x.unwrap()).collect()).into())) // unwrap alone should be fine after the previous check
-        } else {
-            Err("tuple elements must have some type".into())
-        }
-    }
-
-    fn infer_attribute(&self, value: &ast::Expr<Option<Type>>, attr: &str) -> Result<Option<Type>, String> {
-        let ty = value.custom.clone().ok_or_else(|| "no value".to_string())?;
-        if let TypeEnum::TypeVariable(id) = ty.as_ref() {
-            let v = self.ctx.get_variable_def(*id);
-            if v.bound.is_empty() {
-                return Err("no fields on unbounded type variable".into());
-            }
-            let ty = v.bound[0].get_base(&self.ctx).and_then(|v| v.fields.get(attr));
-            if ty.is_none() {
-                return Err("unknown field".into());
-            }
-            for x in v.bound[1..].iter() {
-                let ty1 = x.get_base(&self.ctx).and_then(|v| v.fields.get(attr));
-                if ty1 != ty {
-                    return Err("unknown field (type mismatch between variants)".into());
-                }
-            }
-            return Ok(Some(ty.unwrap().clone()));
-        }
-        
-        match ty.get_base(&self.ctx) {
-            Some(b) => match b.fields.get(attr) {
-                Some(t) => Ok(Some(t.clone())),
-                None => Err("no such field".into()),
-            },
-            None => Err("this object has no fields".into()),
-        }
-    }
-
-    fn infer_bool_ops(&self, values: &[ast::Expr<Option<Type>>]) -> Result<Option<Type>, String> {
-        assert_eq!(values.len(), 2); 
-        let left = values[0].custom.clone().ok_or_else(|| "no value".to_string())?;
-        let right = values[1].custom.clone().ok_or_else(|| "no value".to_string())?;
-        let b = self.ctx.get_primitive(primitives::BOOL_TYPE);
-        if left == b && right == b {
-            Ok(Some(b))
-        } else {
-            Err("bool operands must be bool".to_string())
-        }
-    }
-
-    fn infer_bin_ops(&self, left: &ast::Expr<Option<Type>>, op: &ast::Operator, right: &ast::Expr<Option<Type>>) -> Result<Option<Type>, String> {
-        inference_core::resolve_call(
-            &self.ctx, 
-            Some(left.custom.clone().ok_or_else(|| "no value".to_string())?), 
-            magic_methods::binop_name(op), 
-            &[right.custom.clone().ok_or_else(|| "no value".to_string())?])
-            .map_err(|_| "unsupported binary operator between the oprands".to_string())
-    }
-
-    fn infer_unary_ops(&self, op: &ast::Unaryop, operand: &ast::Expr<Option<Type>>) -> Result<Option<Type>, String> {
-        if let ast::Unaryop::Not = op {
-            if operand.custom == Some(self.ctx.get_primitive(primitives::BOOL_TYPE)) {
-                Ok(Some(self.ctx.get_primitive(primitives::BOOL_TYPE)))
-            } else {
-                Err("logical not must be applied to bool".into())
-            }
-        } else {
-            inference_core::resolve_call(&self.ctx, operand.custom.clone(), magic_methods::unaryop_name(op), &[])
-            .map_err(|_| "unsupported unary operator".to_string())
-        }
-    }
-
-    fn infer_compare(&self, left: &ast::Expr<Option<Type>>, ops: &[ast::Cmpop], comparators: &[ast::Expr<Option<Type>>]) -> Result<Option<Type>, String> {
-        if left.custom.is_none() || (!comparators.iter().all(|x| x.custom.is_some())) {
-            Err("comparison operands must have type".into())
-        } else {
-            let bool_type = Some(self.ctx.get_primitive(primitives::BOOL_TYPE));
-            let ty_first = inference_core::resolve_call(
-                &self.ctx, 
-                Some(left.custom.clone().ok_or_else(|| "comparator must be able to be typed".to_string())?), 
-                magic_methods::comparison_name(&ops[0]).ok_or_else(|| "unsupported comparison".to_string())?, 
-                &[comparators[0].custom.clone().ok_or_else(|| "comparator must be able to be typed".to_string())?])
-                .map_err(|_| "Comparison between the comparators are not supportes".to_string())?;
-            if ty_first != bool_type {
-                return Err("comparison result must be boolean".into());
-            }
-
-            for ((a, b), op) 
-            in comparators[..(comparators.len() - 1)]
-                .iter()
-                .zip(comparators[1..].iter())
-                .zip(ops[1..].iter()) {
-                    let ty = inference_core::resolve_call(
-                        &self.ctx, 
-                        Some(a.custom.clone().ok_or_else(|| "comparator must be able to be typed".to_string())?.clone()), 
-                        magic_methods::comparison_name(op).ok_or_else(|| "unsupported comparison".to_string())?,
-                        &[b.custom.clone().ok_or_else(|| "comparator must be able to be typed".to_string())?.clone()])
-                        .map_err(|_| "Comparison between the comparators are not supportes".to_string())?;
-                    if ty != bool_type {
-                        return Err("comparison result must be boolean".into());
-                    }
-                }
-            Ok(bool_type)
-        }
-    }
-
-    fn infer_call(&self, func: &ast::Expr<Option<Type>>, args: &[ast::Expr<Option<Type>>], _keywords: &[ast::Keyword<Option<Type>>]) -> Result<Option<Type>, String> {
-        if args.iter().all(|x| x.custom.is_some()) {
-            match &func.node {
-                ast::ExprKind::Name {id, ctx: _} 
-                    => inference_core::resolve_call(
-                        &self.ctx, 
-                        None, 
-                        id, 
-                        &args.iter().map(|x| x.custom.clone().unwrap()).collect::<Vec<_>>()),
-                
-                ast::ExprKind::Attribute {value, attr, ctx: _}
-                    => inference_core::resolve_call(
-                        &self.ctx, 
-                        Some(value.custom.clone().ok_or_else(|| "no value".to_string())?), 
-                        &attr, 
-                        &args.iter().map(|x| x.custom.clone().unwrap()).collect::<Vec<_>>()),
-                
-                _ => Err("not supported".into())
-            }
-        } else {
-            Err("function params must have type".into())
-        }
-    }
-
-    fn infer_subscript(&self, value: &ast::Expr<Option<Type>>, slice: &ast::Expr<Option<Type>>) -> Result<Option<Type>, String> {
-        let val_type = value.custom.as_ref().ok_or_else(|| "no value".to_string())?.as_ref();
-        if let TypeEnum::ParametricType(primitives::LIST_TYPE, ls) = val_type {
-            if let ast::ExprKind::Slice {lower, upper, step} = &slice.node {
-                let int32_type = self.ctx.get_primitive(primitives::INT32_TYPE);
-                let l = lower.as_ref().map_or(
-                    Ok(&int32_type),
-                    |x| x.custom.as_ref().ok_or_else(|| "lower bound cannot be typped".to_string()))?;
-                let u = upper.as_ref().map_or(
-                    Ok(&int32_type),
-                    |x| x.custom.as_ref().ok_or_else(|| "upper bound cannot be typped".to_string()))?;
-                let s = step.as_ref().map_or(
-                    Ok(&int32_type),
-                    |x| x.custom.as_ref().ok_or_else(|| "step cannot be typped".to_string()))?;
-                
-                if l == &int32_type && u == &int32_type && s == &int32_type {
-                    Ok(value.custom.clone())
-                } else {
-                    Err("slice must be int32 type".into())
-                }
-            } else if slice.custom == Some(self.ctx.get_primitive(primitives::INT32_TYPE)) {
-                Ok(Some(ls[0].clone()))
-            } else {
-                Err("slice or index must be int32 type".into())
-            }
-        } else if let TypeEnum::ParametricType(primitives::TUPLE_TYPE, ls) = val_type {
-            if let ast::ExprKind::Constant {kind: _, value: ast::Constant::Int(val)} = &slice.node {
-                let ind: Result<usize, _> = val.try_into();
-                if ind.is_ok() && ind.unwrap() < ls.len() {
-                    Ok(Some(ls[ind.unwrap()].clone()))
-                } else {
-                    Err("tuple constant index out of range".into())
-                }
-            } else {
-                Err("tuple index can only be constant".into())
-            }
-        } else {
-            Err("subscript is not supported for types other than list or tuple".into())
-        }
-    }
-
-    fn infer_if_expr(&self, test: &ast::Expr<Option<Type>>, body: &ast::Expr<Option<Type>>, orelse: &ast::Expr<Option<Type>>) -> Result<Option<Type>, String> {
-        if test.custom != Some(self.ctx.get_primitive(primitives::BOOL_TYPE)) {
-            Err("test should be bool".into())
-        } else if body.custom == orelse.custom {
-            Ok(body.custom.clone())
-        } else {
-            Err("divergent type at if expression".into())
-        }
-    }
-
-    fn _infer_list_comprehesion(&self, elt: &ast::Expr<Option<Type>>, generators: &[ast::Comprehension<Option<Type>>]) -> Result<Option<Type>, String> {
-        if generators[0]
-            .ifs
-            .iter()
-            .all(|x| x.custom == Some(self.ctx.get_primitive(primitives::BOOL_TYPE))) {
-                Ok(Some(TypeEnum::ParametricType(
-                    primitives::LIST_TYPE, 
-                    vec![elt.custom.clone().ok_or_else(|| "elements should have value".to_string())?]).into()))
-            } else {
-                Err("test must be bool".into())
-            }
-    }
-
-    // some pre-folds need special handling
-    fn fold_listcomp(&mut self, expr: ast::Expr<()>) -> Result<ast::Expr<Option<Type>>, String> {
-        
-        self.error_stack.push(("list comprehension at ".into(), expr.location));
-
-        if let ast::Expr {
-            location, 
-            custom: _, 
-            node: ast::ExprKind::ListComp {
-                elt, 
-                mut generators}} = expr {
-            // if is list comprehension, need special pre-fold
-            if generators.len() != 1 {
-                return Err("only 1 generator statement is supported".into());
-            }
-            let gen = generators.remove(0);
-            if gen.is_async {
-                return Err("async is not supported".into());
-            }
-
-            let ast::Comprehension {iter, 
-                target, 
-                ifs, 
-                is_async} = gen;
-            let iter_folded = Box::new(self.fold_expr(*iter)?);
-
-            let ret = if let TypeEnum::ParametricType(
-                primitives::LIST_TYPE, 
-                ls) = iter_folded
-                    .custom
-                    .as_ref()
-                    .ok_or_else(|| "no value".to_string())?
-                    .as_ref()
-                    .clone() {
-
-                let result: Result<ast::Expr<Option<Type>>, String>;
-                self.ctx.start_scope();
-                {
-                    self.infer_simple_binding(&target, ls[0].clone())?;
-                    let elt_folded = Box::new(self.fold_expr(*elt)?);
-                    let target_folded = Box::new(self.fold_expr(*target)?);
-                    let ifs_folded = ifs
-                        .into_iter()
-                        .map(|x| self.fold_expr(x))
-                        .collect::<Result<Vec<ast::Expr<Option<Type>>>, _>>()?;
-                    
-                    result = 
-                        if ifs_folded
-                        .iter()
-                        .all(|x| x.custom == Some(self.ctx.get_primitive(primitives::BOOL_TYPE))) {
-                        // only pop the error stack when return Ok(..)
-                        self.error_stack.pop();
-                        
-                        Ok(ast::Expr {
-                            location,
-                            custom: Some(TypeEnum::ParametricType(
-                                primitives::LIST_TYPE, 
-                                vec![elt_folded
-                                    .custom
-                                    .clone()
-                                    .ok_or_else(|| "elements cannot be typped".to_string())?]).into()),
-                            node: ast::ExprKind::ListComp {
-                                elt: elt_folded,
-                                generators: vec![ast::Comprehension {
-                                    target: target_folded,
-                                    ifs: ifs_folded,
-                                    iter: iter_folded,
-                                    is_async
-                                }]
-                            }
-                        })
-                    } else {
-                        Err("test must be bool".into())
-                    };
-                }
-                self.ctx.end_scope();                
-                result
-            } else {
-                Err("iteration is supported for list only".into())
-            };
-            ret
-        } else {
-            panic!("this function is for list comprehensions only!");
-        }
-    }
-
-    fn infer_simple_binding<T>(&mut self, name: &ast::Expr<T>, ty: Type) -> Result<(), String> {
-        self.error_stack.push(("resolving list comprehension variables".into(), name.location));
-        let ret = match &name.node {
-            ast::ExprKind::Name {id, ctx: _} => {
-                if id == "_" {
-                    self.error_stack.pop();
-                    Ok(())
-                } else if self.ctx.defined(id) {
-                    Err("duplicated naming".into())
-                } else {
-                    self.ctx.assign(id.clone(), ty, name.location)?;
-                    self.error_stack.pop();
-                    Ok(())
-                }
-            }
-    
-            ast::ExprKind::Tuple {elts, ctx: _} => {
-                if let TypeEnum::ParametricType(primitives::TUPLE_TYPE, ls) = ty.as_ref() {
-                    if elts.len() == ls.len() {
-                        for (a, b) in elts.iter().zip(ls.iter()) {
-                            self.infer_simple_binding(a, b.clone())?;
-                        }
-                        self.error_stack.pop();
-                        Ok(())
-                    } else {
-                        Err("different length".into())
-                    }
-                } else {
-                    Err("not supported".into())
-                }
-            }
-            _ => Err("not supported".into())
-        };
-        
-        ret
-    }
-
-    fn fold_expr(&mut self, node: ast::Expr<()>) -> Result<ast::Expr<Option<Type>>, String> {
-        let result = <Self as ast::fold::Fold<()>>::fold_expr(self, node);
-        if result.is_err() {
-            println!("{:?}", result);
-            println!("{:?}", self.error_stack.pop().unwrap());
-        }
-        result
-    }
-}
-
-#[cfg(test)]
-pub mod test {
-    use crate::typecheck::{symbol_resolver::SymbolResolver, symbol_resolver::*, location::*};
-    use rustpython_parser::ast::Expr;
-    use super::*;
-    
-    #[cfg(test)]
-    use test_case::test_case;
-    
-    pub fn new_ctx<'a>() -> TypeInferencer<'a> {
-        struct S;
-        impl SymbolResolver for S {
-            fn get_symbol_location(&self, _str: &str) -> Option<Location> { None }
-            fn get_symbol_type(&self, _str: &str) -> Option<SymbolType> { None }
-            fn get_symbol_value(&self, _str: &str) -> Option<SymbolValue> { None }
-        }
-
-        TypeInferencer {
-            ctx: InferenceContext::new(primitives::basic_ctx(), Box::new(S{}), FileID(3)),
-            error_stack: Vec::new()
-        }
-    }
-
-    #[test]
-    fn test_i32() {
-        let mut inferencer = new_ctx();
-        let ast: Expr = Expr {
-            location: ast::Location::new(0, 0),
-            custom: (),
-            node: ast::ExprKind::Constant {
-                value: ast::Constant::Int(123.into()),
-                kind: None
-            }
-        };
-
-        let new_ast = inferencer.fold_expr(ast);
-        assert_eq!(
-            new_ast,
-            Ok(ast::Expr {
-                location: ast::Location::new(0, 0),
-                custom: Some(inferencer.ctx.get_primitive(primitives::INT32_TYPE)),
-                node: ast::ExprKind::Constant {
-                    value: ast::Constant::Int(123.into()),
-                    kind: None
-                }
-            })
-        );
-    }
-
-    #[test]
-    fn test_i64() {
-        let mut inferencer = new_ctx();
-
-        let location = ast::Location::new(0, 0);
-        let num: i64 = 99999999999;
-
-        let ast: Expr = Expr {
-            location,
-            custom: (),
-            node: ast::ExprKind::Constant {
-                value: ast::Constant::Int(num.into()),
-                kind: None,
-            }
-        };
-        
-        let new_ast = inferencer.fold_expr(ast).unwrap();
-
-        assert_eq!(
-            new_ast, 
-            Expr {
-                location,
-                custom: Some(inferencer.ctx.get_primitive(primitives::INT64_TYPE)),
-                node: ast::ExprKind::Constant {
-                    value: ast::Constant::Int(num.into()),
-                    kind: None,
-                }
-            }
-        );
-    }
-
-    #[test]
-    fn test_tuple() {
-        let mut inferencer = new_ctx();
-        let i32_t = inferencer.ctx.get_primitive(primitives::INT32_TYPE);
-        let float_t = inferencer.ctx.get_primitive(primitives::FLOAT_TYPE);
-        let ast = rustpython_parser::parser::parse_expression("(123, 123.123, 999999999)").unwrap();
-        let loc = ast.location;
-        let folded = inferencer.fold_expr(ast).unwrap();
-        assert_eq!(
-            folded,
-            ast::Expr {
-                location: loc,
-                custom: Some(TypeEnum::ParametricType(primitives::TUPLE_TYPE, vec![i32_t.clone(), float_t.clone(), i32_t.clone()]).into()),
-                node: ast::ExprKind::Tuple {
-                    ctx: ast::ExprContext::Load, 
-                    elts: vec![
-                        ast::Expr {
-                            location: ast::Location::new(1, 2),
-                            custom: Some(i32_t.clone()),
-                            node: ast::ExprKind::Constant {
-                                value: ast::Constant::Int(123.into()),
-                                kind: None
-                            }
-                        },
-                        ast::Expr {
-                            location: ast::Location::new(1, 7),
-                            custom: Some(float_t),
-                            node: ast::ExprKind::Constant {
-                                value: ast::Constant::Float(123.123),
-                                kind: None
-                            }
-                        },
-                        ast::Expr {
-                            location: ast::Location::new(1, 16),
-                            custom: Some(i32_t),
-                            node: ast::ExprKind::Constant {
-                                value: ast::Constant::Int(999999999.into()),
-                                kind: None
-                            }
-                        },
-                    ]
-                }
-            }
-        );
-
-    }
-
-    #[test]
-    fn test_list() {
-        let mut inferencer = new_ctx();
-        let location = ast::Location::new(0, 0);
-
-        let ast: Expr = Expr {
-            location,
-            custom: (),
-            node: ast::ExprKind::List {
-                ctx: ast::ExprContext::Load,
-                elts: vec![
-                    Expr {
-                        location,
-                        custom: (),
-                        node: ast::ExprKind::Constant {
-                            value: ast::Constant::Int(1.into()),
-                            kind: None,
-                        },
-                    },
-
-                    Expr {
-                        location,
-                        custom: (),
-                        node: ast::ExprKind::Constant {
-                            value: ast::Constant::Int(2.into()),
-                            kind: None,
-                        },
-                    },
-                ],
-            }
-        };
-
-        let new_ast = inferencer.fold_expr(ast).unwrap();
-        assert_eq!(
-            new_ast,
-            Expr {
-                location,
-                custom: Some(TypeEnum::ParametricType(primitives::LIST_TYPE, vec![inferencer.ctx.get_primitive(primitives::INT32_TYPE)]).into()),
-                node: ast::ExprKind::List {
-                    ctx: ast::ExprContext::Load,
-                    elts: vec![
-                        Expr {
-                            location,
-                            custom: Some(inferencer.ctx.get_primitive(primitives::INT32_TYPE)),
-                            node: ast::ExprKind::Constant {
-                                value: ast::Constant::Int(1.into()),
-                                kind: None,
-                            },
-                        },
-
-                        Expr {
-                            location,
-                            custom: Some(inferencer.ctx.get_primitive(primitives::INT32_TYPE)),
-                            // custom: None,
-                            node: ast::ExprKind::Constant {
-                                value: ast::Constant::Int(2.into()),
-                                kind: None,
-                            },
-                        },
-                    ],
-                }
-            }
-        );
-    }
-
-    #[test_case("False == [True or True, False][0]")]
-    #[test_case("1 < 2 < 3")]
-    #[test_case("1 + [123, 1232][0]")]
-    #[test_case("not True")]
-    #[test_case("[[1]][0][0]")]
-    #[test_case("[[1]][0]")]
-    #[test_case("[[(1, 2), (2, 3), (3, 4)], [(2, 4), (4, 6)]][0]")]
-    #[test_case("[1, 2, 3, 4, 5][1: 2]")]
-    #[test_case("4 if False and True else 8")]
-    #[test_case("(1, 2, 3, 4)[1]")]
-    #[test_case("(1, True, 3, False)[1]")]
-    fn test_mix(prog: &'static str) {
-        let mut inf = new_ctx();
-        let ast = rustpython_parser::parser::parse_expression(prog).unwrap();
-        let folded = inf.fold_expr(ast).unwrap();
-        // println!("{:?}\n", folded.custom);
-    }
-
-    #[test_case("[1, True, 2]")]
-    #[test_case("True if 1 else False")]
-    #[test_case("1 if True else False")]
-    #[test_case("1 and 2")]
-    #[test_case("False or 1")]
-    #[test_case("1 + False")]
-    #[test_case("1 < 2 > False")]
-    #[test_case("not 2")]
-    #[test_case("-True")]
-    fn test_err_msg(prog: &'static str) {
-        let mut inf = new_ctx();
-        let ast = rustpython_parser::parser::parse_expression(prog).unwrap();
-        let _folded = inf.fold_expr(ast);
-        println!("")
-    }
-}

nac3core/src/typecheck/type_error.rs

@@ -0,0 +1,242 @@
+use std::{collections::HashMap, fmt::Display};
+
+use itertools::Itertools;
+
+use nac3parser::ast::{Cmpop, Location, StrRef};
+
+use super::{
+    magic_methods::Binop,
+    typedef::{RecordKey, Type, Unifier},
+};
+use crate::typecheck::{magic_methods::HasOpInfo, typedef::TypeEnum};
+
+#[derive(Debug, Clone)]
+pub enum TypeErrorKind {
+    GotMultipleValues {
+        name: StrRef,
+    },
+    TooManyArguments {
+        expected_min_count: usize,
+        expected_max_count: usize,
+        got_count: usize,
+    },
+    MissingArgs {
+        missing_arg_names: Vec<StrRef>,
+    },
+    UnknownArgName(StrRef),
+    IncorrectArgType {
+        name: StrRef,
+        expected: Type,
+        got: Type,
+    },
+    UnsupportedBinaryOpTypes {
+        operator: Binop,
+        lhs_type: Type,
+        rhs_type: Type,
+        expected_rhs_type: Type,
+    },
+    UnsupportedComparsionOpTypes {
+        operator: Cmpop,
+        lhs_type: Type,
+        rhs_type: Type,
+        expected_rhs_type: Type,
+    },
+    FieldUnificationError {
+        field: RecordKey,
+        types: (Type, Type),
+        loc: (Option<Location>, Option<Location>),
+    },
+    IncompatibleRange(Type, Vec<Type>),
+    IncompatibleTypes(Type, Type),
+    MutationError(RecordKey, Type),
+    NoSuchField(RecordKey, Type),
+    TupleIndexOutOfBounds {
+        index: i32,
+        len: i32,
+    },
+    RequiresTypeAnn,
+    PolymorphicFunctionPointer,
+    NoSuchAttribute(RecordKey, Type),
+}
+
+#[derive(Debug, Clone)]
+pub struct TypeError {
+    pub kind: TypeErrorKind,
+    pub loc: Option<Location>,
+}
+
+impl TypeError {
+    #[must_use]
+    pub fn new(kind: TypeErrorKind, loc: Option<Location>) -> TypeError {
+        TypeError { kind, loc }
+    }
+
+    #[must_use]
+    pub fn at(mut self, loc: Option<Location>) -> TypeError {
+        self.loc = self.loc.or(loc);
+        self
+    }
+
+    #[must_use]
+    pub fn to_display(self, unifier: &Unifier) -> DisplayTypeError {
+        DisplayTypeError { err: self, unifier }
+    }
+}
+
+pub struct DisplayTypeError<'a> {
+    pub err: TypeError,
+    pub unifier: &'a Unifier,
+}
+
+fn loc_to_str(loc: Option<Location>) -> String {
+    match loc {
+        Some(loc) => format!("(in {loc})"),
+        None => String::new(),
+    }
+}
+
+impl<'a> Display for DisplayTypeError<'a> {
+    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
+        use TypeErrorKind::*;
+        let mut notes = Some(HashMap::new());
+        match &self.err.kind {
+            GotMultipleValues { name } => {
+                write!(f, "For multiple values for parameter {name}")
+            }
+            TooManyArguments { expected_min_count, expected_max_count, got_count } => {
+                debug_assert!(expected_min_count <= expected_max_count);
+                if expected_min_count == expected_max_count {
+                    let expected_count = expected_min_count; // or expected_max_count
+                    write!(f, "Too many arguments. Expected {expected_count} but got {got_count}")
+                } else {
+                    write!(f, "Too many arguments. Expected {expected_min_count} to {expected_max_count} arguments but got {got_count}")
+                }
+            }
+            MissingArgs { missing_arg_names } => {
+                let args = missing_arg_names.iter().join(", ");
+                write!(f, "Missing arguments: {args}")
+            }
+            UnsupportedBinaryOpTypes { operator, lhs_type, rhs_type, expected_rhs_type } => {
+                let op_symbol = operator.op_info().symbol;
+
+                let lhs_type_str = self.unifier.stringify_with_notes(*lhs_type, &mut notes);
+                let rhs_type_str = self.unifier.stringify_with_notes(*rhs_type, &mut notes);
+                let expected_rhs_type_str =
+                    self.unifier.stringify_with_notes(*expected_rhs_type, &mut notes);
+
+                write!(f, "Unsupported operand type(s) for {op_symbol}: '{lhs_type_str}' and '{rhs_type_str}' (right operand should have type {expected_rhs_type_str})")
+            }
+            UnsupportedComparsionOpTypes { operator, lhs_type, rhs_type, expected_rhs_type } => {
+                let op_symbol = operator.op_info().symbol;
+
+                let lhs_type_str = self.unifier.stringify_with_notes(*lhs_type, &mut notes);
+                let rhs_type_str = self.unifier.stringify_with_notes(*rhs_type, &mut notes);
+                let expected_rhs_type_str =
+                    self.unifier.stringify_with_notes(*expected_rhs_type, &mut notes);
+
+                write!(f, "'{op_symbol}' not supported between instances of '{lhs_type_str}' and '{rhs_type_str}' (right operand should have type {expected_rhs_type_str})")
+            }
+            UnknownArgName(name) => {
+                write!(f, "Unknown argument name: {name}")
+            }
+            IncorrectArgType { name, expected, got } => {
+                let expected = self.unifier.stringify_with_notes(*expected, &mut notes);
+                let got = self.unifier.stringify_with_notes(*got, &mut notes);
+                write!(f, "Incorrect argument type for parameter {name}. Expected {expected}, but got {got}")
+            }
+            FieldUnificationError { field, types, loc } => {
+                let lhs = self.unifier.stringify_with_notes(types.0, &mut notes);
+                let rhs = self.unifier.stringify_with_notes(types.1, &mut notes);
+                write!(
+                    f,
+                    "Unable to unify field {}: Got types {}{} and {}{}",
+                    field,
+                    lhs,
+                    loc_to_str(loc.0),
+                    rhs,
+                    loc_to_str(loc.1)
+                )
+            }
+            IncompatibleRange(t, ts) => {
+                let t = self.unifier.stringify_with_notes(*t, &mut notes);
+                let ts = ts
+                    .iter()
+                    .map(|t| self.unifier.stringify_with_notes(*t, &mut notes))
+                    .collect::<Vec<_>>();
+                write!(f, "Expected any one of these types: {}, but got {}", ts.join(", "), t)
+            }
+            IncompatibleTypes(t1, t2) => {
+                let type1 = self.unifier.get_ty_immutable(*t1);
+                let type2 = self.unifier.get_ty_immutable(*t2);
+                match (&*type1, &*type2) {
+                    (TypeEnum::TCall(calls), _) => {
+                        let loc = self.unifier.calls[calls[0].0].loc;
+                        let result = write!(
+                            f,
+                            "{} is not callable",
+                            self.unifier.stringify_with_notes(*t2, &mut notes)
+                        );
+                        if let Some(loc) = loc {
+                            result?;
+                            write!(f, " (in {loc})")?;
+                            return Ok(());
+                        }
+                        result
+                    }
+                    (
+                        TypeEnum::TTuple { ty: ty1, is_vararg_ctx: is_vararg1 },
+                        TypeEnum::TTuple { ty: ty2, is_vararg_ctx: is_vararg2 },
+                    ) if !is_vararg1 && !is_vararg2 && ty1.len() != ty2.len() => {
+                        let t1 = self.unifier.stringify_with_notes(*t1, &mut notes);
+                        let t2 = self.unifier.stringify_with_notes(*t2, &mut notes);
+                        write!(f, "Tuple length mismatch: got {t1} and {t2}")
+                    }
+                    _ => {
+                        let t1 = self.unifier.stringify_with_notes(*t1, &mut notes);
+                        let t2 = self.unifier.stringify_with_notes(*t2, &mut notes);
+                        write!(f, "Incompatible types: {t1} and {t2}")
+                    }
+                }
+            }
+            MutationError(name, t) => {
+                if let TypeEnum::TTuple { .. } = &*self.unifier.get_ty_immutable(*t) {
+                    write!(f, "Cannot assign to an element of a tuple")
+                } else {
+                    let t = self.unifier.stringify_with_notes(*t, &mut notes);
+                    write!(f, "Cannot assign to field {name} of {t}, which is immutable")
+                }
+            }
+            NoSuchField(name, t) => {
+                let t = self.unifier.stringify_with_notes(*t, &mut notes);
+                write!(f, "`{t}::{name}` field/method does not exist")
+            }
+            NoSuchAttribute(name, t) => {
+                let t = self.unifier.stringify_with_notes(*t, &mut notes);
+                write!(f, "`{t}::{name}` is not a class attribute")
+            }
+            TupleIndexOutOfBounds { index, len } => {
+                write!(
+                    f,
+                    "Tuple index out of bounds. Got {index} but tuple has only {len} elements"
+                )
+            }
+            RequiresTypeAnn => {
+                write!(f, "Unable to infer virtual object type: Type annotation required")
+            }
+            PolymorphicFunctionPointer => {
+                write!(f, "Polymorphic function pointers is not supported")
+            }
+        }?;
+        if let Some(loc) = self.err.loc {
+            write!(f, " at {loc}")?;
+        }
+        let notes = notes.unwrap();
+        if !notes.is_empty() {
+            write!(f, "\n\nNotes:")?;
+            for line in notes.values() {
+                write!(f, "\n    {line}")?;
+            }
+        }
+        Ok(())
+    }
+}

nac3core/src/typecheck/type_inferencer/test.rs

@@ -0,0 +1,779 @@
+use std::iter::zip;
+
+use indexmap::IndexMap;
+use indoc::indoc;
+use parking_lot::RwLock;
+use test_case::test_case;
+
+use nac3parser::{ast::FileName, parser::parse_program};
+
+use super::*;
+use crate::{
+    codegen::{CodeGenContext, CodeGenerator},
+    symbol_resolver::ValueEnum,
+    toplevel::{helper::PrimDef, DefinitionId, TopLevelDef},
+    typecheck::{magic_methods::with_fields, typedef::*},
+};
+
+struct Resolver {
+    id_to_type: HashMap<StrRef, Type>,
+    id_to_def: HashMap<StrRef, DefinitionId>,
+    class_names: HashMap<StrRef, Type>,
+}
+
+impl SymbolResolver for Resolver {
+    fn get_default_param_value(
+        &self,
+        _: &ast::Expr,
+    ) -> Option<crate::symbol_resolver::SymbolValue> {
+        unimplemented!()
+    }
+
+    fn get_symbol_type(
+        &self,
+        _: &mut Unifier,
+        _: &[Arc<RwLock<TopLevelDef>>],
+        _: &PrimitiveStore,
+        str: StrRef,
+    ) -> Result<Type, String> {
+        self.id_to_type.get(&str).copied().ok_or_else(|| format!("cannot find symbol `{str}`"))
+    }
+
+    fn get_symbol_value<'ctx>(
+        &self,
+        _: StrRef,
+        _: &mut CodeGenContext<'ctx, '_>,
+        _: &mut dyn CodeGenerator,
+    ) -> Option<ValueEnum<'ctx>> {
+        unimplemented!()
+    }
+
+    fn get_identifier_def(&self, id: StrRef) -> Result<DefinitionId, HashSet<String>> {
+        self.id_to_def
+            .get(&id)
+            .copied()
+            .ok_or_else(|| HashSet::from(["Unknown identifier".to_string()]))
+    }
+
+    fn get_string_id(&self, _: &str) -> i32 {
+        unimplemented!()
+    }
+
+    fn get_exception_id(&self, _tyid: usize) -> usize {
+        unimplemented!()
+    }
+}
+
+struct TestEnvironment {
+    pub unifier: Unifier,
+    pub function_data: FunctionData,
+    pub primitives: PrimitiveStore,
+    pub id_to_name: HashMap<usize, StrRef>,
+    pub identifier_mapping: HashMap<StrRef, Type>,
+    pub virtual_checks: Vec<(Type, Type, Location)>,
+    pub calls: HashMap<CodeLocation, CallId>,
+    pub top_level: TopLevelContext,
+}
+
+impl TestEnvironment {
+    pub fn basic_test_env() -> TestEnvironment {
+        let mut unifier = Unifier::new();
+
+        let int32 = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::Int32.id(),
+            fields: HashMap::new(),
+            params: VarMap::new(),
+        });
+        with_fields(&mut unifier, int32, |unifier, fields| {
+            let add_ty = unifier.add_ty(TypeEnum::TFunc(FunSignature {
+                args: vec![FuncArg {
+                    name: "other".into(),
+                    ty: int32,
+                    default_value: None,
+                    is_vararg: false,
+                }],
+                ret: int32,
+                vars: VarMap::new(),
+            }));
+            fields.insert("__add__".into(), (add_ty, false));
+        });
+        let int64 = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::Int64.id(),
+            fields: HashMap::new(),
+            params: VarMap::new(),
+        });
+        let float = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::Float.id(),
+            fields: HashMap::new(),
+            params: VarMap::new(),
+        });
+        let bool = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::Bool.id(),
+            fields: HashMap::new(),
+            params: VarMap::new(),
+        });
+        let none = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::None.id(),
+            fields: HashMap::new(),
+            params: VarMap::new(),
+        });
+        let range = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::Range.id(),
+            fields: HashMap::new(),
+            params: VarMap::new(),
+        });
+        let str = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::Str.id(),
+            fields: HashMap::new(),
+            params: VarMap::new(),
+        });
+        let exception = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::Exception.id(),
+            fields: HashMap::new(),
+            params: VarMap::new(),
+        });
+        let uint32 = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::UInt32.id(),
+            fields: HashMap::new(),
+            params: VarMap::new(),
+        });
+        let uint64 = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::UInt64.id(),
+            fields: HashMap::new(),
+            params: VarMap::new(),
+        });
+        let option = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::Option.id(),
+            fields: HashMap::new(),
+            params: VarMap::new(),
+        });
+        let list_elem_tvar = unifier.get_fresh_var(Some("list_elem".into()), None);
+        let list = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::List.id(),
+            fields: HashMap::new(),
+            params: into_var_map([list_elem_tvar]),
+        });
+        let ndarray_dtype_tvar = unifier.get_fresh_var(Some("ndarray_dtype".into()), None);
+        let ndarray_ndims_tvar =
+            unifier.get_fresh_const_generic_var(uint64, Some("ndarray_ndims".into()), None);
+        let ndarray = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::NDArray.id(),
+            fields: HashMap::new(),
+            params: into_var_map([ndarray_dtype_tvar, ndarray_ndims_tvar]),
+        });
+        let primitives = PrimitiveStore {
+            int32,
+            int64,
+            float,
+            bool,
+            none,
+            range,
+            str,
+            exception,
+            uint32,
+            uint64,
+            option,
+            list,
+            ndarray,
+            size_t: 64,
+        };
+        unifier.put_primitive_store(&primitives);
+        set_primitives_magic_methods(&primitives, &mut unifier);
+
+        let id_to_name: HashMap<_, _> = [
+            (0, "int32".into()),
+            (1, "int64".into()),
+            (2, "float".into()),
+            (3, "bool".into()),
+            (4, "none".into()),
+            (5, "range".into()),
+            (6, "str".into()),
+            (7, "exception".into()),
+        ]
+        .into();
+
+        let mut identifier_mapping = HashMap::new();
+        identifier_mapping.insert("None".into(), none);
+
+        let resolver = Arc::new(Resolver {
+            id_to_type: identifier_mapping.clone(),
+            id_to_def: HashMap::default(),
+            class_names: HashMap::default(),
+        }) as Arc<dyn SymbolResolver + Send + Sync>;
+
+        TestEnvironment {
+            top_level: TopLevelContext {
+                definitions: Arc::default(),
+                unifiers: Arc::default(),
+                personality_symbol: None,
+            },
+            unifier,
+            function_data: FunctionData {
+                resolver,
+                bound_variables: Vec::new(),
+                return_type: None,
+            },
+            primitives,
+            id_to_name,
+            identifier_mapping,
+            virtual_checks: Vec::new(),
+            calls: HashMap::new(),
+        }
+    }
+
+    fn new() -> TestEnvironment {
+        let mut unifier = Unifier::new();
+        let mut identifier_mapping = HashMap::new();
+        let mut top_level_defs: Vec<Arc<RwLock<TopLevelDef>>> = Vec::new();
+        let int32 = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::Int32.id(),
+            fields: HashMap::new(),
+            params: VarMap::new(),
+        });
+        with_fields(&mut unifier, int32, |unifier, fields| {
+            let add_ty = unifier.add_ty(TypeEnum::TFunc(FunSignature {
+                args: vec![FuncArg {
+                    name: "other".into(),
+                    ty: int32,
+                    default_value: None,
+                    is_vararg: false,
+                }],
+                ret: int32,
+                vars: VarMap::new(),
+            }));
+            fields.insert("__add__".into(), (add_ty, false));
+        });
+        let int64 = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::Int64.id(),
+            fields: HashMap::new(),
+            params: VarMap::new(),
+        });
+        let float = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::Float.id(),
+            fields: HashMap::new(),
+            params: VarMap::new(),
+        });
+        let bool = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::Bool.id(),
+            fields: HashMap::new(),
+            params: VarMap::new(),
+        });
+        let none = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::None.id(),
+            fields: HashMap::new(),
+            params: VarMap::new(),
+        });
+        let range = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::Range.id(),
+            fields: HashMap::new(),
+            params: VarMap::new(),
+        });
+        let str = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::Str.id(),
+            fields: HashMap::new(),
+            params: VarMap::new(),
+        });
+        let exception = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::Exception.id(),
+            fields: HashMap::new(),
+            params: VarMap::new(),
+        });
+        let uint32 = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::UInt32.id(),
+            fields: HashMap::new(),
+            params: VarMap::new(),
+        });
+        let uint64 = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::UInt64.id(),
+            fields: HashMap::new(),
+            params: VarMap::new(),
+        });
+        let option = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::Option.id(),
+            fields: HashMap::new(),
+            params: VarMap::new(),
+        });
+        let list_elem_tvar = unifier.get_fresh_var(Some("list_elem".into()), None);
+        let list = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::List.id(),
+            fields: HashMap::new(),
+            params: into_var_map([list_elem_tvar]),
+        });
+        let ndarray = unifier.add_ty(TypeEnum::TObj {
+            obj_id: PrimDef::NDArray.id(),
+            fields: HashMap::new(),
+            params: VarMap::new(),
+        });
+        identifier_mapping.insert("None".into(), none);
+        for (i, name) in [
+            "int32",
+            "int64",
+            "float",
+            "bool",
+            "none",
+            "range",
+            "str",
+            "Exception",
+            "uint32",
+            "uint64",
+            "Option",
+            "list",
+            "ndarray",
+        ]
+        .iter()
+        .enumerate()
+        {
+            top_level_defs.push(
+                RwLock::new(TopLevelDef::Class {
+                    name: (*name).into(),
+                    object_id: DefinitionId(i),
+                    type_vars: Vec::default(),
+                    fields: Vec::default(),
+                    attributes: Vec::default(),
+                    methods: Vec::default(),
+                    ancestors: Vec::default(),
+                    resolver: None,
+                    constructor: None,
+                    loc: None,
+                })
+                .into(),
+            );
+        }
+        let defs = 12;
+
+        let primitives = PrimitiveStore {
+            int32,
+            int64,
+            float,
+            bool,
+            none,
+            range,
+            str,
+            exception,
+            uint32,
+            uint64,
+            option,
+            list,
+            ndarray,
+            size_t: 64,
+        };
+
+        unifier.put_primitive_store(&primitives);
+
+        let tvar = unifier.get_dummy_var();
+
+        let foo_ty = unifier.add_ty(TypeEnum::TObj {
+            obj_id: DefinitionId(defs + 1),
+            fields: [("a".into(), (tvar.ty, true))].into(),
+            params: into_var_map([tvar]),
+        });
+        top_level_defs.push(
+            RwLock::new(TopLevelDef::Class {
+                name: "Foo".into(),
+                object_id: DefinitionId(defs + 1),
+                type_vars: vec![tvar.ty],
+                fields: [("a".into(), tvar.ty, true)].into(),
+                attributes: Vec::default(),
+                methods: Vec::default(),
+                ancestors: Vec::default(),
+                resolver: None,
+                constructor: None,
+                loc: None,
+            })
+            .into(),
+        );
+
+        identifier_mapping.insert(
+            "Foo".into(),
+            unifier.add_ty(TypeEnum::TFunc(FunSignature {
+                args: vec![],
+                ret: foo_ty,
+                vars: into_var_map([tvar]),
+            })),
+        );
+
+        let fun = unifier.add_ty(TypeEnum::TFunc(FunSignature {
+            args: vec![],
+            ret: int32,
+            vars: IndexMap::default(),
+        }));
+        let bar = unifier.add_ty(TypeEnum::TObj {
+            obj_id: DefinitionId(defs + 2),
+            fields: [("a".into(), (int32, true)), ("b".into(), (fun, true))].into(),
+            params: IndexMap::default(),
+        });
+        top_level_defs.push(
+            RwLock::new(TopLevelDef::Class {
+                name: "Bar".into(),
+                object_id: DefinitionId(defs + 2),
+                type_vars: Vec::default(),
+                fields: [("a".into(), int32, true), ("b".into(), fun, true)].into(),
+                attributes: Vec::default(),
+                methods: Vec::default(),
+                ancestors: Vec::default(),
+                resolver: None,
+                constructor: None,
+                loc: None,
+            })
+            .into(),
+        );
+        identifier_mapping.insert(
+            "Bar".into(),
+            unifier.add_ty(TypeEnum::TFunc(FunSignature {
+                args: vec![],
+                ret: bar,
+                vars: IndexMap::default(),
+            })),
+        );
+
+        let bar2 = unifier.add_ty(TypeEnum::TObj {
+            obj_id: DefinitionId(defs + 3),
+            fields: [("a".into(), (bool, true)), ("b".into(), (fun, false))].into(),
+            params: IndexMap::default(),
+        });
+        top_level_defs.push(
+            RwLock::new(TopLevelDef::Class {
+                name: "Bar2".into(),
+                object_id: DefinitionId(defs + 3),
+                type_vars: Vec::default(),
+                fields: [("a".into(), bool, true), ("b".into(), fun, false)].into(),
+                attributes: Vec::default(),
+                methods: Vec::default(),
+                ancestors: Vec::default(),
+                resolver: None,
+                constructor: None,
+                loc: None,
+            })
+            .into(),
+        );
+        identifier_mapping.insert(
+            "Bar2".into(),
+            unifier.add_ty(TypeEnum::TFunc(FunSignature {
+                args: vec![],
+                ret: bar2,
+                vars: IndexMap::default(),
+            })),
+        );
+        let class_names: HashMap<_, _> = [("Bar".into(), bar), ("Bar2".into(), bar2)].into();
+
+        let id_to_name = [
+            "int32".into(),
+            "int64".into(),
+            "float".into(),
+            "bool".into(),
+            "none".into(),
+            "range".into(),
+            "str".into(),
+            "exception".into(),
+            "uint32".into(),
+            "uint64".into(),
+            "option".into(),
+            "list".into(),
+            "ndarray".into(),
+            "Foo".into(),
+            "Bar".into(),
+            "Bar2".into(),
+        ]
+        .into_iter()
+        .enumerate()
+        .collect();
+
+        let top_level = TopLevelContext {
+            definitions: Arc::new(top_level_defs.into()),
+            unifiers: Arc::default(),
+            personality_symbol: None,
+        };
+
+        let resolver = Arc::new(Resolver {
+            id_to_type: identifier_mapping.clone(),
+            id_to_def: [
+                ("Foo".into(), DefinitionId(defs + 1)),
+                ("Bar".into(), DefinitionId(defs + 2)),
+                ("Bar2".into(), DefinitionId(defs + 3)),
+            ]
+            .into(),
+            class_names,
+        }) as Arc<dyn SymbolResolver + Send + Sync>;
+
+        TestEnvironment {
+            unifier,
+            top_level,
+            function_data: FunctionData {
+                resolver,
+                bound_variables: Vec::new(),
+                return_type: None,
+            },
+            primitives,
+            id_to_name,
+            identifier_mapping,
+            virtual_checks: Vec::new(),
+            calls: HashMap::new(),
+        }
+    }
+
+    fn get_inferencer(&mut self) -> Inferencer {
+        Inferencer {
+            top_level: &self.top_level,
+            function_data: &mut self.function_data,
+            unifier: &mut self.unifier,
+            variable_mapping: HashMap::default(),
+            primitives: &mut self.primitives,
+            virtual_checks: &mut self.virtual_checks,
+            calls: &mut self.calls,
+            defined_identifiers: HashMap::default(),
+            in_handler: false,
+        }
+    }
+}
+
+#[test_case(indoc! {"
+        a = 1234
+        b = int64(2147483648)
+        c = 1.234
+        d = True
+    "},
+    &[("a", "int32"), ("b", "int64"), ("c", "float"), ("d", "bool")].into(),
+    &[]
+    ; "primitives test")]
+#[test_case(indoc! {"
+        a = lambda x, y: x
+        b = lambda x: a(x, x)
+        c = 1.234
+        d = b(c)
+    "},
+    &[("a", "fn[[x:float, y:float], float]"), ("b", "fn[[x:float], float]"), ("c", "float"), ("d", "float")].into(),
+    &[]
+    ; "lambda test")]
+#[test_case(indoc! {"
+        a = lambda x: x + x
+        b = lambda x: a(x) + x
+        a = b
+        c = b(1)
+    "},
+    &[("a", "fn[[x:int32], int32]"), ("b", "fn[[x:int32], int32]"), ("c", "int32")].into(),
+    &[]
+    ; "lambda test 2")]
+#[test_case(indoc! {"
+        a = lambda x: x
+        b = lambda x: x
+
+        foo1 = Foo()
+        foo2 = Foo()
+        c = a(foo1.a)
+        d = b(foo2.a)
+
+        a(True)
+        b(123)
+
+    "},
+    &[("a", "fn[[x:bool], bool]"), ("b", "fn[[x:int32], int32]"), ("c", "bool"),
+     ("d", "int32"), ("foo1", "Foo[bool]"), ("foo2", "Foo[int32]")].into(),
+    &[]
+    ; "obj test")]
+#[test_case(indoc! {"
+        a = [1, 2, 3]
+        b = [x + x for x in a]
+    "},
+    &[("a", "list[int32]"), ("b", "list[int32]")].into(),
+    &[]
+    ; "listcomp test")]
+#[test_case(indoc! {"
+        a = virtual(Bar(), Bar)
+        b = a.b()
+        a = virtual(Bar2())
+    "},
+    &[("a", "virtual[Bar]"), ("b", "int32")].into(),
+    &[("Bar", "Bar"), ("Bar2", "Bar")]
+    ; "virtual test")]
+#[test_case(indoc! {"
+        a = [virtual(Bar(), Bar), virtual(Bar2())]
+        b = [x.b() for x in a]
+    "},
+    &[("a", "list[virtual[Bar]]"), ("b", "list[int32]")].into(),
+    &[("Bar", "Bar"), ("Bar2", "Bar")]
+    ; "virtual list test")]
+fn test_basic(source: &str, mapping: &HashMap<&str, &str>, virtuals: &[(&str, &str)]) {
+    println!("source:\n{source}");
+    let mut env = TestEnvironment::new();
+    let id_to_name = std::mem::take(&mut env.id_to_name);
+    let mut defined_identifiers: HashMap<_, _> =
+        env.identifier_mapping.keys().copied().map(|id| (id, IdentifierInfo::default())).collect();
+    defined_identifiers.insert("virtual".into(), IdentifierInfo::default());
+    let mut inferencer = env.get_inferencer();
+    inferencer.defined_identifiers.clone_from(&defined_identifiers);
+    let statements = parse_program(source, FileName::default()).unwrap();
+    let statements = statements
+        .into_iter()
+        .map(|v| inferencer.fold_stmt(v))
+        .collect::<Result<Vec<_>, _>>()
+        .unwrap();
+
+    inferencer.check_block(&statements, &mut defined_identifiers).unwrap();
+
+    for (k, v) in &inferencer.variable_mapping {
+        let name = inferencer.unifier.internal_stringify(
+            *v,
+            &mut |v| (*id_to_name.get(&v).unwrap()).into(),
+            &mut |v| format!("v{v}"),
+            &mut None,
+        );
+        println!("{k}: {name}");
+    }
+    for (k, v) in mapping {
+        let ty = inferencer.variable_mapping.get(&(*k).into()).unwrap();
+        let name = inferencer.unifier.internal_stringify(
+            *ty,
+            &mut |v| (*id_to_name.get(&v).unwrap()).into(),
+            &mut |v| format!("v{v}"),
+            &mut None,
+        );
+        assert_eq!(format!("{k}: {v}"), format!("{k}: {name}"));
+    }
+    assert_eq!(inferencer.virtual_checks.len(), virtuals.len());
+    for ((a, b, _), (x, y)) in zip(inferencer.virtual_checks.iter(), virtuals) {
+        let a = inferencer.unifier.internal_stringify(
+            *a,
+            &mut |v| (*id_to_name.get(&v).unwrap()).into(),
+            &mut |v| format!("v{v}"),
+            &mut None,
+        );
+        let b = inferencer.unifier.internal_stringify(
+            *b,
+            &mut |v| (*id_to_name.get(&v).unwrap()).into(),
+            &mut |v| format!("v{v}"),
+            &mut None,
+        );
+
+        assert_eq!(&a, x);
+        assert_eq!(&b, y);
+    }
+}
+
+#[test_case(indoc! {"
+        a = 2
+        b = 2
+        c = a + b
+        d = a - b
+        e = a * b
+        f = a / b
+        g = a // b
+        h = a % b
+    "},
+    &[("a", "int32"),
+    ("b", "int32"),
+    ("c", "int32"),
+    ("d", "int32"),
+    ("e", "int32"),
+    ("f", "float"),
+    ("g", "int32"),
+    ("h", "int32")].into()
+    ; "int32")]
+#[test_case(
+    indoc! {"
+        a = 2.4
+        b = 3.6
+        c = a + b
+        d = a - b
+        e = a * b
+        f = a / b
+        g = a // b
+        h = a % b
+        i = a ** b
+        ii = 3
+        j = a ** b
+    "},
+    &[("a", "float"),
+    ("b", "float"),
+    ("c", "float"),
+    ("d", "float"),
+    ("e", "float"),
+    ("f", "float"),
+    ("g", "float"),
+    ("h", "float"),
+    ("i", "float"),
+    ("ii", "int32"),
+    ("j", "float")].into()
+    ; "float"
+)]
+#[test_case(
+    indoc! {"
+        a = int64(12312312312)
+        b = int64(24242424424)
+        c = a + b
+        d = a - b
+        e = a * b
+        f = a / b
+        g = a // b
+        h = a % b
+        i = a == b
+        j = a > b
+        k = a < b
+        l = a != b
+    "},
+    &[("a", "int64"),
+    ("b", "int64"),
+    ("c", "int64"),
+    ("d", "int64"),
+    ("e", "int64"),
+    ("f", "float"),
+    ("g", "int64"),
+    ("h", "int64"),
+    ("i", "bool"),
+    ("j", "bool"),
+    ("k", "bool"),
+    ("l", "bool")].into()
+    ; "int64"
+)]
+#[test_case(
+    indoc! {"
+        a = True
+        b = False
+        c = a == b
+        d = not a
+        e = a != b
+    "},
+    &[("a", "bool"),
+    ("b", "bool"),
+    ("c", "bool"),
+    ("d", "bool"),
+    ("e", "bool")].into()
+    ; "boolean"
+)]
+fn test_primitive_magic_methods(source: &str, mapping: &HashMap<&str, &str>) {
+    println!("source:\n{source}");
+    let mut env = TestEnvironment::basic_test_env();
+    let id_to_name = std::mem::take(&mut env.id_to_name);
+    let mut defined_identifiers: HashMap<_, _> =
+        env.identifier_mapping.keys().copied().map(|id| (id, IdentifierInfo::default())).collect();
+    defined_identifiers.insert("virtual".into(), IdentifierInfo::default());
+    let mut inferencer = env.get_inferencer();
+    inferencer.defined_identifiers.clone_from(&defined_identifiers);
+    let statements = parse_program(source, FileName::default()).unwrap();
+    let statements = statements
+        .into_iter()
+        .map(|v| inferencer.fold_stmt(v))
+        .collect::<Result<Vec<_>, _>>()
+        .unwrap();
+
+    inferencer.check_block(&statements, &mut defined_identifiers).unwrap();
+
+    for (k, v) in &inferencer.variable_mapping {
+        let name = inferencer.unifier.internal_stringify(
+            *v,
+            &mut |v| (*id_to_name.get(&v).unwrap()).into(),
+            &mut |v| format!("v{v}"),
+            &mut None,
+        );
+        println!("{k}: {name}");
+    }
+    for (k, v) in mapping {
+        let ty = inferencer.variable_mapping.get(&(*k).into()).unwrap();
+        let name = inferencer.unifier.internal_stringify(
+            *ty,
+            &mut |v| (*id_to_name.get(&v).unwrap()).into(),
+            &mut |v| format!("v{v}"),
+            &mut None,
+        );
+        assert_eq!(format!("{k}: {v}"), format!("{k}: {name}"));
+    }
+}

nac3core/src/typecheck/typedef.rs

@@ -1,60 +0,0 @@
-use std::collections::HashMap;
-use std::rc::Rc;
-
-#[derive(PartialEq, Eq, Copy, Clone, Hash, Debug)]
-pub struct PrimitiveId(pub(crate) usize);
-
-#[derive(PartialEq, Eq, Copy, Clone, Hash, Debug)]
-pub struct ClassId(pub(crate) usize);
-
-#[derive(PartialEq, Eq, Copy, Clone, Hash, Debug)]
-pub struct ParamId(pub(crate) usize);
-
-#[derive(PartialEq, Eq, Copy, Clone, Hash, Debug)]
-pub struct VariableId(pub(crate) usize);
-
-#[derive(PartialEq, Eq, Clone, Hash, Debug)]
-pub enum TypeEnum {
-    BotType,
-    SelfType,
-    PrimitiveType(PrimitiveId),
-    ClassType(ClassId),
-    VirtualClassType(ClassId),
-    ParametricType(ParamId, Vec<Rc<TypeEnum>>),
-    TypeVariable(VariableId),
-}
-
-pub type Type = Rc<TypeEnum>;
-
-#[derive(Clone)]
-pub struct FnDef {
-    // we assume methods first argument to be SelfType,
-    // so the first argument is not contained here
-    pub args: Vec<Type>,
-    pub result: Option<Type>,
-}
-
-#[derive(Clone)]
-pub struct TypeDef<'a> {
-    pub name: &'a str,
-    pub fields: HashMap<&'a str, Type>,
-    pub methods: HashMap<&'a str, FnDef>,
-}
-
-#[derive(Clone)]
-pub struct ClassDef<'a> {
-    pub base: TypeDef<'a>,
-    pub parents: Vec<ClassId>,
-}
-
-#[derive(Clone)]
-pub struct ParametricDef<'a> {
-    pub base: TypeDef<'a>,
-    pub params: Vec<VariableId>,
-}
-
-#[derive(Clone)]
-pub struct VarDef<'a> {
-    pub name: &'a str,
-    pub bound: Vec<Type>,
-}

nac3core/src/typecheck/typedef/test.rs

@@ -0,0 +1,656 @@
+use std::collections::HashMap;
+
+use indoc::indoc;
+use itertools::Itertools;
+use test_case::test_case;
+
+use super::*;
+use crate::typecheck::magic_methods::with_fields;
+
+impl Unifier {
+    /// Check whether two types are equal.
+    fn eq(&mut self, a: Type, b: Type) -> bool {
+        if a == b {
+            return true;
+        }
+        let (ty_a, ty_b) = {
+            let table = &mut self.unification_table;
+            if table.unioned(a, b) {
+                return true;
+            }
+            (table.probe_value(a).clone(), table.probe_value(b).clone())
+        };
+
+        match (&*ty_a, &*ty_b) {
+            (
+                TypeEnum::TVar { fields: None, id: id1, .. },
+                TypeEnum::TVar { fields: None, id: id2, .. },
+            ) => id1 == id2,
+            (
+                TypeEnum::TVar { fields: Some(map1), .. },
+                TypeEnum::TVar { fields: Some(map2), .. },
+            ) => self.map_eq2(map1, map2),
+            (
+                TypeEnum::TTuple { ty: ty1, is_vararg_ctx: false },
+                TypeEnum::TTuple { ty: ty2, is_vararg_ctx: false },
+            ) => {
+                ty1.len() == ty2.len()
+                    && ty1.iter().zip(ty2.iter()).all(|(t1, t2)| self.eq(*t1, *t2))
+            }
+            (TypeEnum::TVirtual { ty: ty1 }, TypeEnum::TVirtual { ty: ty2 }) => self.eq(*ty1, *ty2),
+            (
+                TypeEnum::TObj { obj_id: id1, params: params1, .. },
+                TypeEnum::TObj { obj_id: id2, params: params2, .. },
+            ) => id1 == id2 && self.map_eq(params1, params2),
+            // TLiteral, TCall and TFunc are not yet implemented
+            _ => false,
+        }
+    }
+
+    fn map_eq<K>(&mut self, map1: &IndexMapping<K>, map2: &IndexMapping<K>) -> bool
+    where
+        K: std::hash::Hash + Eq + Clone,
+    {
+        if map1.len() != map2.len() {
+            return false;
+        }
+        for (k, v) in map1 {
+            if !map2.get(k).is_some_and(|v1| self.eq(*v, *v1)) {
+                return false;
+            }
+        }
+        true
+    }
+
+    fn map_eq2<K>(&mut self, map1: &Mapping<K, RecordField>, map2: &Mapping<K, RecordField>) -> bool
+    where
+        K: std::hash::Hash + Eq + Clone,
+    {
+        if map1.len() != map2.len() {
+            return false;
+        }
+        for (k, v) in map1 {
+            if !map2.get(k).is_some_and(|v1| self.eq(v.ty, v1.ty)) {
+                return false;
+            }
+        }
+        true
+    }
+}
+
+struct TestEnvironment {
+    pub unifier: Unifier,
+    pub type_mapping: HashMap<String, Type>,
+}
+
+impl TestEnvironment {
+    fn new() -> TestEnvironment {
+        let mut unifier = Unifier::new();
+        let mut type_mapping = HashMap::new();
+
+        type_mapping.insert(
+            "int".into(),
+            unifier.add_ty(TypeEnum::TObj {
+                obj_id: DefinitionId(0),
+                fields: HashMap::new(),
+                params: VarMap::new(),
+            }),
+        );
+        type_mapping.insert(
+            "float".into(),
+            unifier.add_ty(TypeEnum::TObj {
+                obj_id: DefinitionId(1),
+                fields: HashMap::new(),
+                params: VarMap::new(),
+            }),
+        );
+        type_mapping.insert(
+            "bool".into(),
+            unifier.add_ty(TypeEnum::TObj {
+                obj_id: DefinitionId(2),
+                fields: HashMap::new(),
+                params: VarMap::new(),
+            }),
+        );
+        let tvar = unifier.get_dummy_var();
+        type_mapping.insert(
+            "Foo".into(),
+            unifier.add_ty(TypeEnum::TObj {
+                obj_id: DefinitionId(3),
+                fields: [("a".into(), (tvar.ty, true))].into(),
+                params: into_var_map([tvar]),
+            }),
+        );
+        let tvar = unifier.get_dummy_var();
+        type_mapping.insert(
+            "list".into(),
+            unifier.add_ty(TypeEnum::TObj {
+                obj_id: PrimDef::List.id(),
+                fields: HashMap::new(),
+                params: into_var_map([tvar]),
+            }),
+        );
+
+        TestEnvironment { unifier, type_mapping }
+    }
+
+    fn parse(&mut self, typ: &str, mapping: &Mapping<String>) -> Type {
+        let result = self.internal_parse(typ, mapping);
+        assert!(result.1.is_empty());
+        result.0
+    }
+
+    fn internal_parse<'b>(&mut self, typ: &'b str, mapping: &Mapping<String>) -> (Type, &'b str) {
+        // for testing only, so we can just panic when the input is malformed
+        let end = typ.find(|c| ['[', ',', ']', '='].contains(&c)).unwrap_or(typ.len());
+        match &typ[..end] {
+            "list" => {
+                let mut s = &typ[end..];
+                assert_eq!(&s[0..1], "[");
+                let mut ty = Vec::new();
+                while &s[0..1] != "]" {
+                    let result = self.internal_parse(&s[1..], mapping);
+                    ty.push(result.0);
+                    s = result.1;
+                }
+
+                assert_eq!(ty.len(), 1);
+
+                let list_elem_tvar = if let TypeEnum::TObj { params, .. } =
+                    &*self.unifier.get_ty_immutable(self.type_mapping["list"])
+                {
+                    iter_type_vars(params).next().unwrap()
+                } else {
+                    unreachable!()
+                };
+
+                (
+                    self.unifier
+                        .subst(
+                            self.type_mapping["list"],
+                            &into_var_map([TypeVar { id: list_elem_tvar.id, ty: ty[0] }]),
+                        )
+                        .unwrap(),
+                    &s[1..],
+                )
+            }
+            "tuple" => {
+                let mut s = &typ[end..];
+                assert_eq!(&s[0..1], "[");
+                let mut ty = Vec::new();
+                while &s[0..1] != "]" {
+                    let result = self.internal_parse(&s[1..], mapping);
+                    ty.push(result.0);
+                    s = result.1;
+                }
+                (self.unifier.add_ty(TypeEnum::TTuple { ty, is_vararg_ctx: false }), &s[1..])
+            }
+            "Record" => {
+                let mut s = &typ[end..];
+                assert_eq!(&s[0..1], "[");
+                let mut fields = HashMap::new();
+                while &s[0..1] != "]" {
+                    let eq = s.find('=').unwrap();
+                    let key = s[1..eq].into();
+                    let result = self.internal_parse(&s[eq + 1..], mapping);
+                    fields.insert(key, RecordField::new(result.0, true, None));
+                    s = result.1;
+                }
+                (self.unifier.add_record(fields), &s[1..])
+            }
+            x => {
+                let mut s = &typ[end..];
+                let ty = mapping.get(x).copied().unwrap_or_else(|| {
+                    // mapping should be type variables, type_mapping should be concrete types
+                    // we should not resolve the type of type variables.
+                    let mut ty = *self.type_mapping.get(x).unwrap();
+                    let te = self.unifier.get_ty(ty);
+                    if let TypeEnum::TObj { params, .. } = &*te {
+                        if !params.is_empty() {
+                            assert_eq!(&s[0..1], "[");
+                            let mut p = Vec::new();
+                            while &s[0..1] != "]" {
+                                let result = self.internal_parse(&s[1..], mapping);
+                                p.push(result.0);
+                                s = result.1;
+                            }
+                            s = &s[1..];
+                            ty = self
+                                .unifier
+                                .subst(ty, &params.keys().copied().zip(p).collect())
+                                .unwrap_or(ty);
+                        }
+                    }
+                    ty
+                });
+                (ty, s)
+            }
+        }
+    }
+
+    fn unify(&mut self, typ1: Type, typ2: Type) -> Result<(), String> {
+        self.unifier.unify(typ1, typ2).map_err(|e| e.to_display(&self.unifier).to_string())
+    }
+}
+
+#[test_case(2,
+    &[("v1", "v2"), ("v2", "float")],
+    &[("v1", "float"), ("v2", "float")]
+    ; "simple variable"
+)]
+#[test_case(2,
+    &[("v1", "list[v2]"), ("v1", "list[float]")],
+    &[("v1", "list[float]"), ("v2", "float")]
+    ; "list element"
+)]
+#[test_case(3,
+    &[
+        ("v1", "Record[a=v3,b=v3]"),
+        ("v2", "Record[b=float,c=v3]"),
+        ("v1", "v2")
+    ],
+    &[
+        ("v1", "Record[a=float,b=float,c=float]"),
+        ("v2", "Record[a=float,b=float,c=float]"),
+        ("v3", "float")
+    ]
+    ; "record merge"
+)]
+#[test_case(3,
+    &[
+        ("v1", "Record[a=float]"),
+        ("v2", "Foo[v3]"),
+        ("v1", "v2")
+    ],
+    &[
+        ("v1", "Foo[float]"),
+        ("v3", "float")
+    ]
+    ; "record obj merge"
+)]
+/// Test cases for valid unifications.
+fn test_unify(
+    variable_count: u32,
+    unify_pairs: &[(&'static str, &'static str)],
+    verify_pairs: &[(&'static str, &'static str)],
+) {
+    let unify_count = unify_pairs.len();
+    // test all permutations...
+    for perm in unify_pairs.iter().permutations(unify_count) {
+        let mut env = TestEnvironment::new();
+        let mut mapping = HashMap::new();
+        for i in 1..=variable_count {
+            let v = env.unifier.get_dummy_var();
+            mapping.insert(format!("v{i}"), v.ty);
+        }
+        // unification may have side effect when we do type resolution, so freeze the types
+        // before doing unification.
+        let mut pairs = Vec::new();
+        for (a, b) in &perm {
+            let t1 = env.parse(a, &mapping);
+            let t2 = env.parse(b, &mapping);
+            pairs.push((t1, t2));
+        }
+        for (t1, t2) in pairs {
+            env.unifier.unify(t1, t2).unwrap();
+        }
+        for (a, b) in verify_pairs {
+            println!("{a} = {b}");
+            let t1 = env.parse(a, &mapping);
+            let t2 = env.parse(b, &mapping);
+            println!("a = {}, b = {}", env.unifier.stringify(t1), env.unifier.stringify(t2));
+            assert!(env.unifier.eq(t1, t2));
+        }
+    }
+}
+
+#[test_case(2,
+    &[
+        ("v1", "tuple[int]"),
+        ("v2", "list[int]"),
+    ],
+    (("v1", "v2"), "Incompatible types: 11[0] and tuple[0]")
+    ; "type mismatch"
+)]
+#[test_case(2,
+    &[
+        ("v1", "tuple[int]"),
+        ("v2", "tuple[float]"),
+    ],
+    (("v1", "v2"), "Incompatible types: tuple[0] and tuple[1]")
+    ; "tuple parameter mismatch"
+)]
+#[test_case(2,
+    &[
+        ("v1", "tuple[int,int]"),
+        ("v2", "tuple[int]"),
+    ],
+    (("v1", "v2"), "Tuple length mismatch: got tuple[0, 0] and tuple[0]")
+    ; "tuple length mismatch"
+)]
+#[test_case(3,
+    &[
+        ("v1", "Record[a=float,b=int]"),
+        ("v2", "Foo[v3]"),
+    ],
+    (("v1", "v2"), "`3[typevar5]::b` field/method does not exist")
+    ; "record obj merge"
+)]
+/// Test cases for invalid unifications.
+fn test_invalid_unification(
+    variable_count: u32,
+    unify_pairs: &[(&'static str, &'static str)],
+    erroneous_pair: ((&'static str, &'static str), &'static str),
+) {
+    let mut env = TestEnvironment::new();
+    let mut mapping = HashMap::new();
+    for i in 1..=variable_count {
+        let v = env.unifier.get_dummy_var();
+        mapping.insert(format!("v{i}"), v.ty);
+    }
+    // unification may have side effect when we do type resolution, so freeze the types
+    // before doing unification.
+    let mut pairs = Vec::new();
+    for (a, b) in unify_pairs {
+        let t1 = env.parse(a, &mapping);
+        let t2 = env.parse(b, &mapping);
+        pairs.push((t1, t2));
+    }
+    let (t1, t2) =
+        (env.parse(erroneous_pair.0 .0, &mapping), env.parse(erroneous_pair.0 .1, &mapping));
+    for (a, b) in pairs {
+        env.unifier.unify(a, b).unwrap();
+    }
+    assert_eq!(env.unify(t1, t2), Err(erroneous_pair.1.to_string()));
+}
+
+#[test]
+fn test_recursive_subst() {
+    let mut env = TestEnvironment::new();
+    let int = *env.type_mapping.get("int").unwrap();
+    let foo_id = *env.type_mapping.get("Foo").unwrap();
+    let foo_ty = env.unifier.get_ty(foo_id);
+    with_fields(&mut env.unifier, foo_id, |_unifier, fields| {
+        fields.insert("rec".into(), (foo_id, true));
+    });
+    let TypeEnum::TObj { params, .. } = &*foo_ty else { unreachable!() };
+    let mapping = params.iter().map(|(id, _)| (*id, int)).collect();
+    let instantiated = env.unifier.subst(foo_id, &mapping).unwrap();
+    let instantiated_ty = env.unifier.get_ty(instantiated);
+
+    let TypeEnum::TObj { fields, .. } = &*instantiated_ty else { unreachable!() };
+    assert!(env.unifier.unioned(fields.get(&"a".into()).unwrap().0, int));
+    assert!(env.unifier.unioned(fields.get(&"rec".into()).unwrap().0, instantiated));
+}
+
+#[test]
+fn test_virtual() {
+    let mut env = TestEnvironment::new();
+    let int = env.parse("int", &HashMap::new());
+    let fun = env.unifier.add_ty(TypeEnum::TFunc(FunSignature {
+        args: vec![],
+        ret: int,
+        vars: VarMap::new(),
+    }));
+    let bar = env.unifier.add_ty(TypeEnum::TObj {
+        obj_id: DefinitionId(5),
+        fields: [("f".into(), (fun, false)), ("a".into(), (int, false))].into(),
+        params: VarMap::new(),
+    });
+    let v0 = env.unifier.get_dummy_var().ty;
+    let v1 = env.unifier.get_dummy_var().ty;
+
+    let a = env.unifier.add_ty(TypeEnum::TVirtual { ty: bar });
+    let b = env.unifier.add_ty(TypeEnum::TVirtual { ty: v0 });
+    let c = env.unifier.add_record([("f".into(), RecordField::new(v1, false, None))].into());
+    env.unifier.unify(a, b).unwrap();
+    env.unifier.unify(b, c).unwrap();
+    assert!(env.unifier.eq(v1, fun));
+
+    let d = env.unifier.add_record([("a".into(), RecordField::new(v1, true, None))].into());
+    assert_eq!(env.unify(b, d), Err("`virtual[5]::a` field/method does not exist".to_string()));
+
+    let d = env.unifier.add_record([("b".into(), RecordField::new(v1, true, None))].into());
+    assert_eq!(env.unify(b, d), Err("`virtual[5]::b` field/method does not exist".to_string()));
+}
+
+#[test]
+fn test_typevar_range() {
+    let mut env = TestEnvironment::new();
+    let int = env.parse("int", &HashMap::new());
+    let boolean = env.parse("bool", &HashMap::new());
+    let float = env.parse("float", &HashMap::new());
+    let int_list = env.parse("list[int]", &HashMap::new());
+    let float_list = env.parse("list[float]", &HashMap::new());
+
+    let list_elem_tvar = if let TypeEnum::TObj { params, .. } =
+        &*env.unifier.get_ty_immutable(env.type_mapping["list"])
+    {
+        iter_type_vars(params).next().unwrap()
+    } else {
+        unreachable!()
+    };
+
+    // unification between v and int
+    // where v in (int, bool)
+    let v = env.unifier.get_fresh_var_with_range(&[int, boolean], None, None).ty;
+    env.unifier.unify(int, v).unwrap();
+
+    // unification between v and list[int]
+    // where v in (int, bool)
+    let v = env.unifier.get_fresh_var_with_range(&[int, boolean], None, None).ty;
+    assert_eq!(
+        env.unify(int_list, v),
+        Err("Expected any one of these types: 0, 2, but got 11[0]".to_string())
+    );
+
+    // unification between v and float
+    // where v in (int, bool)
+    let v = env.unifier.get_fresh_var_with_range(&[int, boolean], None, None).ty;
+    assert_eq!(
+        env.unify(float, v),
+        Err("Expected any one of these types: 0, 2, but got 1".to_string())
+    );
+
+    let v1 = env.unifier.get_fresh_var_with_range(&[int, boolean], None, None).ty;
+    let v1_list = env.unifier.add_ty(TypeEnum::TObj {
+        obj_id: env.type_mapping["list"].obj_id(&env.unifier).unwrap(),
+        fields: Mapping::default(),
+        params: into_var_map([TypeVar { id: list_elem_tvar.id, ty: v1 }]),
+    });
+    let v = env.unifier.get_fresh_var_with_range(&[int, v1_list], None, None).ty;
+    // unification between v and int
+    // where v in (int, list[v1]), v1 in (int, bool)
+    env.unifier.unify(int, v).unwrap();
+
+    let v = env.unifier.get_fresh_var_with_range(&[int, v1_list], None, None).ty;
+    // unification between v and list[int]
+    // where v in (int, list[v1]), v1 in (int, bool)
+    env.unifier.unify(int_list, v).unwrap();
+
+    let v = env.unifier.get_fresh_var_with_range(&[int, v1_list], None, None).ty;
+    // unification between v and list[float]
+    // where v in (int, list[v1]), v1 in (int, bool)
+    println!("float_list: {}, v: {}", env.unifier.stringify(float_list), env.unifier.stringify(v));
+    assert_eq!(
+        env.unify(float_list, v),
+        Err("Expected any one of these types: 0, 11[typevar6], but got 11[1]\n\nNotes:\n    typevar6 ∈ {0, 2}".to_string())
+    );
+
+    let a = env.unifier.get_fresh_var_with_range(&[int, float], None, None).ty;
+    let b = env.unifier.get_fresh_var_with_range(&[boolean, float], None, None).ty;
+    env.unifier.unify(a, b).unwrap();
+    env.unifier.unify(a, float).unwrap();
+
+    let a = env.unifier.get_fresh_var_with_range(&[int, float], None, None).ty;
+    let b = env.unifier.get_fresh_var_with_range(&[boolean, float], None, None).ty;
+    env.unifier.unify(a, b).unwrap();
+    assert_eq!(env.unify(a, int), Err("Expected any one of these types: 1, but got 0".into()));
+
+    let a = env.unifier.get_fresh_var_with_range(&[int, float], None, None).ty;
+    let b = env.unifier.get_fresh_var_with_range(&[boolean, float], None, None).ty;
+    let a_list = env.unifier.add_ty(TypeEnum::TObj {
+        obj_id: env.type_mapping["list"].obj_id(&env.unifier).unwrap(),
+        fields: Mapping::default(),
+        params: into_var_map([TypeVar { id: list_elem_tvar.id, ty: a }]),
+    });
+    let a_list = env.unifier.get_fresh_var_with_range(&[a_list], None, None).ty;
+    let b_list = env.unifier.add_ty(TypeEnum::TObj {
+        obj_id: env.type_mapping["list"].obj_id(&env.unifier).unwrap(),
+        fields: Mapping::default(),
+        params: into_var_map([TypeVar { id: list_elem_tvar.id, ty: b }]),
+    });
+    let b_list = env.unifier.get_fresh_var_with_range(&[b_list], None, None).ty;
+    env.unifier.unify(a_list, b_list).unwrap();
+    let float_list = env.unifier.add_ty(TypeEnum::TObj {
+        obj_id: env.type_mapping["list"].obj_id(&env.unifier).unwrap(),
+        fields: Mapping::default(),
+        params: into_var_map([TypeVar { id: list_elem_tvar.id, ty: float }]),
+    });
+    env.unifier.unify(a_list, float_list).unwrap();
+    // previous unifications should not affect a and b
+    env.unifier.unify(a, int).unwrap();
+
+    let a = env.unifier.get_fresh_var_with_range(&[int, float], None, None).ty;
+    let b = env.unifier.get_fresh_var_with_range(&[boolean, float], None, None).ty;
+    let a_list = env.unifier.add_ty(TypeEnum::TObj {
+        obj_id: env.type_mapping["list"].obj_id(&env.unifier).unwrap(),
+        fields: Mapping::default(),
+        params: into_var_map([TypeVar { id: list_elem_tvar.id, ty: a }]),
+    });
+    let b_list = env.unifier.add_ty(TypeEnum::TObj {
+        obj_id: env.type_mapping["list"].obj_id(&env.unifier).unwrap(),
+        fields: Mapping::default(),
+        params: into_var_map([TypeVar { id: list_elem_tvar.id, ty: b }]),
+    });
+    env.unifier.unify(a_list, b_list).unwrap();
+    let int_list = env.unifier.add_ty(TypeEnum::TObj {
+        obj_id: env.type_mapping["list"].obj_id(&env.unifier).unwrap(),
+        fields: Mapping::default(),
+        params: into_var_map([TypeVar { id: list_elem_tvar.id, ty: int }]),
+    });
+    assert_eq!(
+        env.unify(a_list, int_list),
+        Err("Incompatible types: 11[typevar23] and 11[0]\
+            \n\nNotes:\n    typevar23 ∈ {1}"
+            .into())
+    );
+
+    let a = env.unifier.get_fresh_var_with_range(&[int, float], None, None).ty;
+    let b = env.unifier.get_dummy_var().ty;
+    let a_list = env.unifier.add_ty(TypeEnum::TObj {
+        obj_id: env.type_mapping["list"].obj_id(&env.unifier).unwrap(),
+        fields: Mapping::default(),
+        params: into_var_map([TypeVar { id: list_elem_tvar.id, ty: a }]),
+    });
+    let a_list = env.unifier.get_fresh_var_with_range(&[a_list], None, None).ty;
+    let b_list = env.unifier.add_ty(TypeEnum::TObj {
+        obj_id: env.type_mapping["list"].obj_id(&env.unifier).unwrap(),
+        fields: Mapping::default(),
+        params: into_var_map([TypeVar { id: list_elem_tvar.id, ty: b }]),
+    });
+    env.unifier.unify(a_list, b_list).unwrap();
+    assert_eq!(
+        env.unify(b, boolean),
+        Err("Expected any one of these types: 0, 1, but got 2".into())
+    );
+}
+
+#[test]
+fn test_rigid_var() {
+    let mut env = TestEnvironment::new();
+    let a = env.unifier.get_fresh_rigid_var(None, None).ty;
+    let b = env.unifier.get_fresh_rigid_var(None, None).ty;
+    let x = env.unifier.get_dummy_var().ty;
+    let list_elem_tvar = env.unifier.get_fresh_var(Some("list_elem".into()), None);
+    let list_a = env.unifier.add_ty(TypeEnum::TObj {
+        obj_id: env.type_mapping["list"].obj_id(&env.unifier).unwrap(),
+        fields: Mapping::default(),
+        params: into_var_map([TypeVar { id: list_elem_tvar.id, ty: a }]),
+    });
+    let list_x = env.unifier.add_ty(TypeEnum::TObj {
+        obj_id: env.type_mapping["list"].obj_id(&env.unifier).unwrap(),
+        fields: Mapping::default(),
+        params: into_var_map([TypeVar { id: list_elem_tvar.id, ty: x }]),
+    });
+    let int = env.parse("int", &HashMap::new());
+    let list_int = env.parse("list[int]", &HashMap::new());
+
+    assert_eq!(env.unify(a, b), Err("Incompatible types: typevar4 and typevar3".to_string()));
+    env.unifier.unify(list_a, list_x).unwrap();
+    assert_eq!(
+        env.unify(list_x, list_int),
+        Err("Incompatible types: 11[typevar3] and 11[0]".to_string())
+    );
+
+    env.unifier.replace_rigid_var(a, int);
+    env.unifier.unify(list_x, list_int).unwrap();
+}
+
+#[test]
+fn test_instantiation() {
+    let mut env = TestEnvironment::new();
+    let int = env.parse("int", &HashMap::new());
+    let boolean = env.parse("bool", &HashMap::new());
+    let float = env.parse("float", &HashMap::new());
+    let list_int = env.parse("list[int]", &HashMap::new());
+
+    let list_elem_tvar = if let TypeEnum::TObj { params, .. } =
+        &*env.unifier.get_ty_immutable(env.type_mapping["list"])
+    {
+        iter_type_vars(params).next().unwrap()
+    } else {
+        unreachable!()
+    };
+
+    let obj_map: HashMap<_, _> = [(0usize, "int"), (1, "float"), (2, "bool"), (11, "list")].into();
+
+    let v = env.unifier.get_fresh_var_with_range(&[int, boolean], None, None).ty;
+    let list_v = env
+        .unifier
+        .subst(env.type_mapping["list"], &into_var_map([TypeVar { id: list_elem_tvar.id, ty: v }]))
+        .unwrap();
+    let v1 = env.unifier.get_fresh_var_with_range(&[list_v, int], None, None).ty;
+    let v2 = env.unifier.get_fresh_var_with_range(&[list_int, float], None, None).ty;
+    let t = env.unifier.get_dummy_var().ty;
+    let tuple = env.unifier.add_ty(TypeEnum::TTuple { ty: vec![v, v1, v2], is_vararg_ctx: false });
+    let v3 = env.unifier.get_fresh_var_with_range(&[tuple, t], None, None).ty;
+    // t = TypeVar('t')
+    // v = TypeVar('v', int, bool)
+    // v1 = TypeVar('v1', 'list[v]', int)
+    // v2 = TypeVar('v2', 'list[int]', float)
+    // v3 = TypeVar('v3', tuple[v, v1, v2], t)
+    // what values can v3 take?
+
+    let types = env.unifier.get_instantiations(v3).unwrap();
+    let expected_types = indoc! {"
+        tuple[bool, int, float]
+        tuple[bool, int, list[int]]
+        tuple[bool, list[bool], float]
+        tuple[bool, list[bool], list[int]]
+        tuple[bool, list[int], float]
+        tuple[bool, list[int], list[int]]
+        tuple[int, int, float]
+        tuple[int, int, list[int]]
+        tuple[int, list[bool], float]
+        tuple[int, list[bool], list[int]]
+        tuple[int, list[int], float]
+        tuple[int, list[int], list[int]]
+        v6"
+    }
+    .split('\n')
+    .collect_vec();
+    let types = types
+        .iter()
+        .map(|ty| {
+            env.unifier.internal_stringify(
+                *ty,
+                &mut |i| (*obj_map.get(&i).unwrap()).to_string(),
+                &mut |i| format!("v{i}"),
+                &mut None,
+            )
+        })
+        .sorted()
+        .collect_vec();
+    assert_eq!(expected_types, types);
+}

nac3core/src/typecheck/unification_table.rs

@@ -0,0 +1,182 @@
+use std::rc::Rc;
+
+use itertools::izip;
+
+#[derive(Copy, Clone, PartialEq, Eq, Debug, Hash)]
+pub struct UnificationKey(usize);
+
+#[derive(Clone)]
+pub struct UnificationTable<V> {
+    parents: Vec<usize>,
+    ranks: Vec<u32>,
+    values: Vec<Option<V>>,
+    log: Vec<Action<V>>,
+    generation: u32,
+}
+
+#[derive(Clone, Debug)]
+enum Action<V> {
+    Parent { key: usize, original_parent: usize },
+    Value { key: usize, original_value: Option<V> },
+    Rank { key: usize, original_rank: u32 },
+    Marker { generation: u32 },
+}
+
+impl<V> Default for UnificationTable<V> {
+    fn default() -> Self {
+        Self::new()
+    }
+}
+
+impl<V> UnificationTable<V> {
+    pub fn new() -> UnificationTable<V> {
+        UnificationTable {
+            parents: Vec::new(),
+            ranks: Vec::new(),
+            values: Vec::new(),
+            log: Vec::new(),
+            generation: 0,
+        }
+    }
+
+    pub fn new_key(&mut self, v: V) -> UnificationKey {
+        let index = self.parents.len();
+        self.parents.push(index);
+        self.ranks.push(0);
+        self.values.push(Some(v));
+        UnificationKey(index)
+    }
+
+    pub fn unify(&mut self, a: UnificationKey, b: UnificationKey) {
+        let mut a = self.find(a);
+        let mut b = self.find(b);
+        if a == b {
+            return;
+        }
+        if self.ranks[a] < self.ranks[b] {
+            std::mem::swap(&mut a, &mut b);
+        }
+        self.log.push(Action::Parent { key: b, original_parent: self.parents[b] });
+        self.parents[b] = a;
+        if self.ranks[a] == self.ranks[b] {
+            self.log.push(Action::Rank { key: a, original_rank: self.ranks[a] });
+            self.ranks[a] += 1;
+        }
+    }
+
+    pub fn probe_value_immutable(&self, key: UnificationKey) -> &V {
+        let mut root = key.0;
+        let mut parent = self.parents[root];
+        while root != parent {
+            root = parent;
+            // parent = root.parent
+            parent = self.parents[parent];
+        }
+        self.values[parent].as_ref().unwrap()
+    }
+
+    pub fn probe_value(&mut self, a: UnificationKey) -> &V {
+        let index = self.find(a);
+        self.values[index].as_ref().unwrap()
+    }
+
+    pub fn set_value(&mut self, a: UnificationKey, v: V) {
+        let index = self.find(a);
+        let original_value = self.values[index].replace(v);
+        self.log.push(Action::Value { key: index, original_value });
+    }
+
+    pub fn unioned(&mut self, a: UnificationKey, b: UnificationKey) -> bool {
+        self.find(a) == self.find(b)
+    }
+
+    pub fn get_representative(&mut self, key: UnificationKey) -> UnificationKey {
+        UnificationKey(self.find(key))
+    }
+
+    fn find(&mut self, key: UnificationKey) -> usize {
+        let mut root = key.0;
+        let mut parent = self.parents[root];
+        while root != parent {
+            // a = parent.parent
+            let a = self.parents[parent];
+            // root.parent = parent.parent
+            self.log.push(Action::Parent { key: root, original_parent: self.parents[root] });
+            self.parents[root] = a;
+            root = parent;
+            // parent = root.parent
+            parent = a;
+        }
+        parent
+    }
+
+    pub fn get_snapshot(&mut self) -> (usize, u32) {
+        let generation = self.generation;
+        self.log.push(Action::Marker { generation });
+        self.generation += 1;
+        (self.log.len(), generation)
+    }
+
+    pub fn restore_snapshot(&mut self, snapshot: (usize, u32)) {
+        let (log_len, generation) = snapshot;
+        assert!(self.log.len() >= log_len, "snapshot restoration error");
+        assert!(
+            matches!(self.log[log_len - 1], Action::Marker { generation: gen } if gen == generation),
+            "snapshot restoration error"
+        );
+        for action in self.log.drain(log_len - 1..).rev() {
+            match action {
+                Action::Parent { key, original_parent } => {
+                    self.parents[key] = original_parent;
+                }
+                Action::Value { key, original_value } => {
+                    self.values[key] = original_value;
+                }
+                Action::Rank { key, original_rank } => {
+                    self.ranks[key] = original_rank;
+                }
+                Action::Marker { .. } => {}
+            }
+        }
+    }
+
+    pub fn discard_snapshot(&mut self, snapshot: (usize, u32)) {
+        let (log_len, generation) = snapshot;
+        assert!(self.log.len() >= log_len, "snapshot discard error");
+        assert!(
+            matches!(self.log[log_len - 1], Action::Marker { generation: gen } if gen == generation),
+            "snapshot discard error"
+        );
+        self.log.clear();
+    }
+}
+
+impl<V> UnificationTable<Rc<V>>
+where
+    V: Clone,
+{
+    pub fn get_send(&self) -> UnificationTable<V> {
+        let values = izip!(self.values.iter(), self.parents.iter())
+            .enumerate()
+            .map(|(i, (v, p))| if *p == i { v.as_ref().map(|v| v.as_ref().clone()) } else { None })
+            .collect();
+        UnificationTable {
+            parents: self.parents.clone(),
+            ranks: self.ranks.clone(),
+            values,
+            log: Vec::new(),
+            generation: 0,
+        }
+    }
+
+    pub fn from_send(table: &UnificationTable<V>) -> UnificationTable<Rc<V>> {
+        let values = table.values.iter().cloned().map(|v| v.map(Rc::new)).collect();
+        UnificationTable {
+            parents: table.parents.clone(),
+            ranks: table.ranks.clone(),
+            values,
+            log: Vec::new(),
+            generation: 0,
+        }
+    }
+}

nac3embedded/Cargo.toml

@@ -1,15 +0,0 @@
-[package]
-name = "nac3embedded"
-version = "0.1.0"
-authors = ["M-Labs"]
-edition = "2018"
-
-[lib]
-name = "nac3embedded"
-crate-type = ["cdylib"]
-
-[dependencies]
-pyo3 = { version = "0.12.4", features = ["extension-module"] }
-inkwell = { git = "https://github.com/TheDan64/inkwell", branch = "master", features = ["llvm10-0"] }
-rustpython-parser = { git = "https://github.com/RustPython/RustPython", branch = "master" }
-nac3core = { path = "../nac3core" }

nac3embedded/demo.py

@@ -1,11 +0,0 @@
-from language import *
-
-
-class Demo:
-    @kernel
-    def run(self: bool) -> bool:
-        return False
-
-
-if __name__ == "__main__":
-    Demo().run()

nac3embedded/language.py

@@ -1,19 +0,0 @@
-from functools import wraps
-
-import nac3embedded
-
-
-__all__ = ["kernel", "portable"]
-
-
-def kernel(function):
-    @wraps(function)
-    def run_on_core(self, *args, **kwargs):
-        nac3 = nac3embedded.NAC3()
-        nac3.register_host_object(self)
-        nac3.compile_method(self, function.__name__)
-    return run_on_core
-
-
-def portable(function):
-    return fn

nac3embedded/nac3embedded.so

@@ -1 +0,0 @@
-../target/release/libnac3embedded.so

nac3embedded/src/lib.rs

@@ -1,116 +0,0 @@
-use std::collections::HashMap;
-use std::collections::hash_map::Entry;
-
-use pyo3::prelude::*;
-use pyo3::exceptions;
-use rustpython_parser::{ast, parser};
-use inkwell::context::Context;
-use inkwell::targets::*;
-
-use nac3core::CodeGen;
-
-fn runs_on_core(decorator_list: &[ast::Expression]) -> bool {
-    for decorator in decorator_list.iter() {
-        if let ast::ExpressionType::Identifier { name } = &decorator.node {
-            if name == "kernel" || name == "portable" {
-                return true
-            }
-        }
-    }
-    false
-}
-
-#[pyclass(name=NAC3)]
-struct Nac3 {
-    type_definitions: HashMap<i64, ast::Program>,
-    host_objects: HashMap<i64, i64>,
-}
-
-#[pymethods]
-impl Nac3 {
-    #[new]
-    fn new() -> Self {
-        Nac3 {
-            type_definitions: HashMap::new(),
-            host_objects: HashMap::new(),
-        }
-    }
-
-    fn register_host_object(&mut self, obj: PyObject) -> PyResult<()> {
-        Python::with_gil(|py| -> PyResult<()> {
-            let obj: &PyAny = obj.extract(py)?;
-            let obj_type = obj.get_type();
-
-            let builtins = PyModule::import(py, "builtins")?;
-            let type_id = builtins.call1("id", (obj_type, ))?.extract()?;
-
-            let entry = self.type_definitions.entry(type_id);
-            if let Entry::Vacant(entry) = entry {
-                let source = PyModule::import(py, "inspect")?.call1("getsource", (obj_type, ))?;
-                let ast = parser::parse_program(source.extract()?).map_err(|e|
-                    exceptions::PySyntaxError::new_err(format!("failed to parse host object source: {}", e)))?;
-                entry.insert(ast);
-                // TODO: examine AST and recursively register dependencies
-            };
-
-            let obj_id = builtins.call1("id", (obj, ))?.extract()?;
-            match self.host_objects.entry(obj_id) {
-                Entry::Vacant(entry) => entry.insert(type_id),
-                Entry::Occupied(_) => return Err(
-                    exceptions::PyValueError::new_err("host object registered twice")),
-            };
-            // TODO: collect other information about host object, e.g. value of fields
-
-            Ok(())
-        })
-    }
-
-    fn compile_method(&self, obj: PyObject, name: String) -> PyResult<()> {
-        Python::with_gil(|py| -> PyResult<()> {
-            let obj: &PyAny = obj.extract(py)?;
-            let builtins = PyModule::import(py, "builtins")?;
-            let obj_id = builtins.call1("id", (obj, ))?.extract()?;
-
-            let type_id = self.host_objects.get(&obj_id).ok_or_else(||
-                exceptions::PyKeyError::new_err("type of host object not found"))?;
-            let ast = self.type_definitions.get(&type_id).ok_or_else(||
-                exceptions::PyKeyError::new_err("type definition not found"))?;
-
-            if let ast::StatementType::ClassDef {
-                    name: _,
-                    body,
-                    bases: _,
-                    keywords: _,
-                    decorator_list: _ } = &ast.statements[0].node {
-                for statement in body.iter() {
-                    if let ast::StatementType::FunctionDef {
-                            is_async: _,
-                            name: funcdef_name,
-                            args: _,
-                            body: _,
-                            decorator_list,
-                            returns: _ } = &statement.node {
-                        if runs_on_core(decorator_list) && funcdef_name == &name {
-                            let context = Context::create();
-                            let mut codegen = CodeGen::new(&context);
-                            codegen.compile_toplevel(&body[0]).map_err(|e|
-                                exceptions::PyRuntimeError::new_err(format!("compilation failed: {}", e)))?;
-                            codegen.print_ir();
-                        }
-                    }
-                }
-            } else {
-                return Err(exceptions::PyValueError::new_err("expected ClassDef for type definition"));
-            }
-
-            Ok(())
-        })
-    }
-}
-
-#[pymodule]
-fn nac3embedded(_py: Python, m: &PyModule) -> PyResult<()> {
-    Target::initialize_all(&InitializationConfig::default());
-    m.add_class::<Nac3>()?;
-    Ok(())
-}

nac3ld/Cargo.toml

@@ -0,0 +1,8 @@
+[package]
+name = "nac3ld"
+version = "0.1.0"
+authors = ["M-Labs"]
+edition = "2021"
+
+[dependencies]
+byteorder = { version = "1.5", default-features = false }

nac3ld/src/dwarf.rs

@@ -0,0 +1,509 @@
+#![allow(non_camel_case_types, non_upper_case_globals)]
+
+use std::mem;
+
+use byteorder::{ByteOrder, LittleEndian};
+
+pub const DW_EH_PE_omit: u8 = 0xFF;
+pub const DW_EH_PE_absptr: u8 = 0x00;
+
+pub const DW_EH_PE_uleb128: u8 = 0x01;
+pub const DW_EH_PE_udata2: u8 = 0x02;
+pub const DW_EH_PE_udata4: u8 = 0x03;
+pub const DW_EH_PE_udata8: u8 = 0x04;
+pub const DW_EH_PE_sleb128: u8 = 0x09;
+pub const DW_EH_PE_sdata2: u8 = 0x0A;
+pub const DW_EH_PE_sdata4: u8 = 0x0B;
+pub const DW_EH_PE_sdata8: u8 = 0x0C;
+
+pub const DW_EH_PE_pcrel: u8 = 0x10;
+pub const DW_EH_PE_textrel: u8 = 0x20;
+pub const DW_EH_PE_datarel: u8 = 0x30;
+pub const DW_EH_PE_funcrel: u8 = 0x40;
+pub const DW_EH_PE_aligned: u8 = 0x50;
+
+pub const DW_EH_PE_indirect: u8 = 0x80;
+
+pub struct DwarfReader<'a> {
+    pub slice: &'a [u8],
+    pub virt_addr: u32,
+    base_slice: &'a [u8],
+    base_virt_addr: u32,
+}
+
+impl<'a> DwarfReader<'a> {
+    pub fn new(slice: &[u8], virt_addr: u32) -> DwarfReader {
+        DwarfReader { slice, virt_addr, base_slice: slice, base_virt_addr: virt_addr }
+    }
+
+    /// Creates a new instance from another instance of [DwarfReader], optionally removing any
+    /// offsets previously applied to the other instance.
+    pub fn from_reader(other: &DwarfReader<'a>, reset_offset: bool) -> DwarfReader<'a> {
+        if reset_offset {
+            DwarfReader::new(other.base_slice, other.base_virt_addr)
+        } else {
+            DwarfReader::new(other.slice, other.virt_addr)
+        }
+    }
+
+    pub fn offset(&mut self, offset: u32) {
+        self.slice = &self.slice[offset as usize..];
+        self.virt_addr = self.virt_addr.wrapping_add(offset);
+    }
+
+    /// ULEB128 and SLEB128 encodings are defined in Section 7.6 - "Variable Length Data" of the
+    /// [DWARF-4 Manual](https://dwarfstd.org/doc/DWARF4.pdf).
+    pub fn read_uleb128(&mut self) -> u64 {
+        let mut shift: usize = 0;
+        let mut result: u64 = 0;
+        let mut byte: u8;
+        loop {
+            byte = self.read_u8();
+            result |= u64::from(byte & 0x7F) << shift;
+            shift += 7;
+            if byte & 0x80 == 0 {
+                break;
+            }
+        }
+        result
+    }
+
+    pub fn read_sleb128(&mut self) -> i64 {
+        let mut shift: u32 = 0;
+        let mut result: u64 = 0;
+        let mut byte: u8;
+        loop {
+            byte = self.read_u8();
+            result |= u64::from(byte & 0x7F) << shift;
+            shift += 7;
+            if byte & 0x80 == 0 {
+                break;
+            }
+        }
+        // sign-extend
+        if shift < u64::BITS && (byte & 0x40) != 0 {
+            result |= (!0u64) << shift;
+        }
+        result as i64
+    }
+
+    pub fn read_u8(&mut self) -> u8 {
+        let val = self.slice[0];
+        self.slice = &self.slice[1..];
+        val
+    }
+}
+
+macro_rules! impl_read_fn {
+    ( $($type: ty, $byteorder_fn: ident);* ) => {
+        impl<'a> DwarfReader<'a> {
+            $(
+                pub fn $byteorder_fn(&mut self) -> $type {
+                    let val = LittleEndian::$byteorder_fn(self.slice);
+                    self.slice = &self.slice[mem::size_of::<$type>()..];
+                    val
+                }
+            )*
+        }
+    }
+}
+
+impl_read_fn!(
+    u16,    read_u16;
+    u32,    read_u32;
+    u64,    read_u64;
+    i16,    read_i16;
+    i32,    read_i32;
+    i64,    read_i64
+);
+
+pub struct DwarfWriter<'a> {
+    pub slice: &'a mut [u8],
+    pub offset: usize,
+}
+
+impl<'a> DwarfWriter<'a> {
+    pub fn new(slice: &mut [u8]) -> DwarfWriter {
+        DwarfWriter { slice, offset: 0 }
+    }
+
+    pub fn write_u8(&mut self, data: u8) {
+        self.slice[self.offset] = data;
+        self.offset += 1;
+    }
+
+    pub fn write_u32(&mut self, data: u32) {
+        LittleEndian::write_u32(&mut self.slice[self.offset..], data);
+        self.offset += 4;
+    }
+}
+
+fn read_encoded_pointer(reader: &mut DwarfReader, encoding: u8) -> Result<usize, ()> {
+    if encoding == DW_EH_PE_omit {
+        return Err(());
+    }
+
+    // DW_EH_PE_aligned implies it's an absolute pointer value
+    // However, we are linking library for 32-bits architecture
+    // The size of variable should be 4 bytes instead
+    if encoding == DW_EH_PE_aligned {
+        let shifted_virt_addr = round_up(reader.virt_addr as usize, mem::size_of::<u32>())?;
+        let addr_inc = shifted_virt_addr - reader.virt_addr as usize;
+
+        reader.slice = &reader.slice[addr_inc..];
+        reader.virt_addr = shifted_virt_addr as u32;
+        return Ok(reader.read_u32() as usize);
+    }
+
+    match encoding & 0x0F {
+        DW_EH_PE_absptr | DW_EH_PE_udata4 => Ok(reader.read_u32() as usize),
+        DW_EH_PE_uleb128 => Ok(reader.read_uleb128() as usize),
+        DW_EH_PE_udata2 => Ok(reader.read_u16() as usize),
+        DW_EH_PE_udata8 => Ok(reader.read_u64() as usize),
+        DW_EH_PE_sleb128 => Ok(reader.read_sleb128() as usize),
+        DW_EH_PE_sdata2 => Ok(reader.read_i16() as usize),
+        DW_EH_PE_sdata4 => Ok(reader.read_i32() as usize),
+        DW_EH_PE_sdata8 => Ok(reader.read_i64() as usize),
+        _ => Err(()),
+    }
+}
+
+fn read_encoded_pointer_with_pc(reader: &mut DwarfReader, encoding: u8) -> Result<usize, ()> {
+    let entry_virt_addr = reader.virt_addr;
+    let mut result = read_encoded_pointer(reader, encoding)?;
+
+    // DW_EH_PE_aligned implies it's an absolute pointer value
+    if encoding == DW_EH_PE_aligned {
+        return Ok(result);
+    }
+
+    result = match encoding & 0x70 {
+        DW_EH_PE_pcrel => result.wrapping_add(entry_virt_addr as usize),
+
+        // .eh_frame normally would not have these kinds of relocations
+        // These would not be supported by a dedicated linker relocation schemes for RISC-V
+        DW_EH_PE_textrel | DW_EH_PE_datarel | DW_EH_PE_funcrel | DW_EH_PE_aligned => {
+            unimplemented!()
+        }
+
+        // Other values should be impossible
+        _ => unreachable!(),
+    };
+
+    if encoding & DW_EH_PE_indirect != 0 {
+        // There should not be a need for indirect addressing, as assembly code from
+        // the dynamic library should not be freely moved relative to the EH frame.
+        unreachable!()
+    }
+
+    Ok(result)
+}
+
+#[inline]
+fn round_up(unrounded: usize, align: usize) -> Result<usize, ()> {
+    if align.is_power_of_two() {
+        Ok((unrounded + align - 1) & !(align - 1))
+    } else {
+        Err(())
+    }
+}
+
+/// Minimalistic structure to store everything needed for parsing FDEs to synthesize `.eh_frame_hdr`
+/// section.
+///
+/// Refer to [The Linux Standard Base Core Specification, Generic Part](https://refspecs.linuxfoundation.org/LSB_5.0.0/LSB-Core-generic/LSB-Core-generic/ehframechpt.html)
+/// for more information.
+pub struct EH_Frame<'a> {
+    reader: DwarfReader<'a>,
+}
+
+impl<'a> EH_Frame<'a> {
+    /// Creates an [EH_Frame] using the bytes in the `.eh_frame` section and its address in the ELF
+    /// file.
+    pub fn new(eh_frame_slice: &[u8], eh_frame_addr: u32) -> EH_Frame {
+        EH_Frame { reader: DwarfReader::new(eh_frame_slice, eh_frame_addr) }
+    }
+
+    /// Returns an [Iterator] over all Call Frame Information (CFI) records.
+    pub fn cfi_records(&self) -> CFI_Records<'a> {
+        let reader = DwarfReader::from_reader(&self.reader, true);
+        let len = reader.slice.len();
+
+        CFI_Records { reader, available: len }
+    }
+}
+
+/// A single Call Frame Information (CFI) record.
+///
+/// From the [specification](https://refspecs.linuxfoundation.org/LSB_5.0.0/LSB-Core-generic/LSB-Core-generic/ehframechpt.html):
+///
+/// > Each CFI record contains a Common Information Entry (CIE) record followed by 1 or more Frame
+/// > Description Entry (FDE) records.
+pub struct CFI_Record<'a> {
+    // It refers to the augmentation data that corresponds to 'R' in the augmentation string
+    fde_pointer_encoding: u8,
+    fde_reader: DwarfReader<'a>,
+}
+
+impl<'a> CFI_Record<'a> {
+    pub fn from_reader(cie_reader: &mut DwarfReader<'a>) -> Result<CFI_Record<'a>, ()> {
+        let length = cie_reader.read_u32();
+        let fde_reader = match length {
+            // eh_frame with 0 lengths means the CIE is terminated
+            0 => panic!("Cannot create an EH_Frame from a termination CIE"),
+
+            // length == u32::MAX means that the length is only representable with 64 bits,
+            // which does not make sense in a system with 32-bit address.
+            0xFFFF_FFFF => unimplemented!(),
+
+            _ => {
+                let mut fde_reader = DwarfReader::from_reader(cie_reader, false);
+                fde_reader.offset(length);
+                fde_reader
+            }
+        };
+
+        // Routine check on the .eh_frame well-formness, in terms of CIE ID & Version args.
+        let cie_ptr = cie_reader.read_u32();
+        assert_eq!(cie_ptr, 0);
+        assert_eq!(cie_reader.read_u8(), 1);
+
+        // Parse augmentation string
+        // The first character must be 'z', there is no way to proceed otherwise
+        assert_eq!(cie_reader.read_u8(), b'z');
+
+        // Establish a pointer that skips ahead of the string
+        // Skip code/data alignment factors & return address register along the way as well
+        // We only tackle the case where 'z' and 'R' are part of the augmentation string, otherwise
+        // we cannot get the addresses to make .eh_frame_hdr
+        let mut aug_data_reader = DwarfReader::from_reader(cie_reader, false);
+        let mut aug_str_len = 0;
+        loop {
+            if aug_data_reader.read_u8() == b'\0' {
+                break;
+            }
+            aug_str_len += 1;
+        }
+        if aug_str_len == 0 {
+            unimplemented!();
+        }
+        aug_data_reader.read_uleb128(); // Code alignment factor
+        aug_data_reader.read_sleb128(); // Data alignment factor
+        aug_data_reader.read_uleb128(); // Return address register
+        aug_data_reader.read_uleb128(); // Augmentation data length
+        let mut fde_pointer_encoding = DW_EH_PE_omit;
+        for _ in 0..aug_str_len {
+            match cie_reader.read_u8() {
+                b'L' => {
+                    aug_data_reader.read_u8();
+                }
+
+                b'P' => {
+                    let encoding = aug_data_reader.read_u8();
+                    read_encoded_pointer(&mut aug_data_reader, encoding)?;
+                }
+
+                b'R' => {
+                    fde_pointer_encoding = aug_data_reader.read_u8();
+                }
+
+                // Other characters are not supported
+                _ => unimplemented!(),
+            }
+        }
+        assert_ne!(fde_pointer_encoding, DW_EH_PE_omit);
+
+        Ok(CFI_Record { fde_pointer_encoding, fde_reader })
+    }
+
+    /// Returns a [DwarfReader] initialized to the first Frame Description Entry (FDE) of this CFI
+    /// record.
+    pub fn get_fde_reader(&self) -> DwarfReader<'a> {
+        DwarfReader::from_reader(&self.fde_reader, true)
+    }
+
+    /// Returns an [Iterator] over all Frame Description Entries (FDEs).
+    pub fn fde_records(&self) -> FDE_Records<'a> {
+        let reader = self.get_fde_reader();
+        let len = reader.slice.len();
+
+        FDE_Records { pointer_encoding: self.fde_pointer_encoding, reader, available: len }
+    }
+}
+
+/// [Iterator] over Call Frame Information (CFI) records in an
+/// [Exception Handling (EH) frame][EH_Frame].
+pub struct CFI_Records<'a> {
+    reader: DwarfReader<'a>,
+    available: usize,
+}
+
+impl<'a> Iterator for CFI_Records<'a> {
+    type Item = CFI_Record<'a>;
+
+    fn next(&mut self) -> Option<Self::Item> {
+        loop {
+            if self.available == 0 {
+                return None;
+            }
+
+            let mut this_reader = DwarfReader::from_reader(&self.reader, false);
+
+            // Remove the length of the header and the content from the counter
+            let length = self.reader.read_u32();
+            let length = match length {
+                // eh_frame with 0-length means the CIE is terminated
+                0 => return None,
+                0xFFFF_FFFF => unimplemented!("CIE entries larger than 4 bytes not supported"),
+                other => other,
+            } as usize;
+
+            // Remove the length of the header and the content from the counter
+            self.available -= length + mem::size_of::<u32>();
+            let mut next_reader = DwarfReader::from_reader(&self.reader, false);
+            next_reader.offset(length as u32);
+
+            let cie_ptr = self.reader.read_u32();
+
+            self.reader = next_reader;
+
+            // Skip this record if it is a FDE
+            if cie_ptr == 0 {
+                // Rewind back to the start of the CFI Record
+                return Some(CFI_Record::from_reader(&mut this_reader).ok().unwrap());
+            }
+        }
+    }
+}
+
+/// [Iterator] over Frame Description Entries (FDEs) in an
+/// [Exception Handling (EH) frame][EH_Frame].
+pub struct FDE_Records<'a> {
+    pointer_encoding: u8,
+    reader: DwarfReader<'a>,
+    available: usize,
+}
+
+impl<'a> Iterator for FDE_Records<'a> {
+    type Item = (u32, u32);
+
+    fn next(&mut self) -> Option<Self::Item> {
+        // Parse each FDE to obtain the starting address that the FDE applies to
+        // Send the FDE offset and the mentioned address to a callback that write up the
+        // .eh_frame_hdr section
+
+        if self.available == 0 {
+            return None;
+        }
+
+        // Remove the length of the header and the content from the counter
+        let length = match self.reader.read_u32() {
+            // eh_frame with 0-length means the CIE is terminated
+            0 => return None,
+            0xFFFF_FFFF => unimplemented!("CIE entries larger than 4 bytes not supported"),
+            other => other,
+        } as usize;
+
+        // Remove the length of the header and the content from the counter
+        self.available -= length + mem::size_of::<u32>();
+        let mut next_fde_reader = DwarfReader::from_reader(&self.reader, false);
+        next_fde_reader.offset(length as u32);
+
+        let cie_ptr = self.reader.read_u32();
+        let next_val = if cie_ptr != 0 {
+            let pc_begin = read_encoded_pointer_with_pc(&mut self.reader, self.pointer_encoding)
+                .expect("Failed to read PC Begin");
+            Some((pc_begin as u32, self.reader.virt_addr))
+        } else {
+            None
+        };
+
+        self.reader = next_fde_reader;
+
+        next_val
+    }
+}
+
+pub struct EH_Frame_Hdr<'a> {
+    fde_writer: DwarfWriter<'a>,
+    eh_frame_hdr_addr: u32,
+    fdes: Vec<(u32, u32)>,
+}
+
+impl<'a> EH_Frame_Hdr<'a> {
+    /// Create a [EH_Frame_Hdr] object, and write out the fixed fields of `.eh_frame_hdr` to memory.
+    ///
+    /// Load address is not known at this point.
+    pub fn new(
+        eh_frame_hdr_slice: &mut [u8],
+        eh_frame_hdr_addr: u32,
+        eh_frame_addr: u32,
+    ) -> EH_Frame_Hdr {
+        let mut writer = DwarfWriter::new(eh_frame_hdr_slice);
+
+        writer.write_u8(1); // version
+        writer.write_u8(0x1B); // eh_frame_ptr_enc - PC-relative 4-byte signed value
+        writer.write_u8(0x03); // fde_count_enc - 4-byte unsigned value
+        writer.write_u8(0x3B); // table_enc - .eh_frame_hdr section-relative 4-byte signed value
+
+        let eh_frame_offset = eh_frame_addr.wrapping_sub(
+            eh_frame_hdr_addr + writer.offset as u32 + ((mem::size_of::<u8>() as u32) * 4),
+        );
+        writer.write_u32(eh_frame_offset); // eh_frame_ptr
+        writer.write_u32(0); // `fde_count`, will be written in finalize_fde
+
+        EH_Frame_Hdr { fde_writer: writer, eh_frame_hdr_addr, fdes: Vec::new() }
+    }
+
+    /// The offset of the `fde_count` value relative to the start of the `.eh_frame_hdr` section in
+    /// bytes.
+    fn fde_count_offset() -> usize {
+        8
+    }
+
+    pub fn add_fde(&mut self, init_loc: u32, addr: u32) {
+        self.fdes.push((
+            init_loc.wrapping_sub(self.eh_frame_hdr_addr),
+            addr.wrapping_sub(self.eh_frame_hdr_addr),
+        ));
+    }
+
+    pub fn finalize_fde(mut self) {
+        self.fdes
+            .sort_by(|(left_init_loc, _), (right_init_loc, _)| left_init_loc.cmp(right_init_loc));
+        for (init_loc, addr) in &self.fdes {
+            self.fde_writer.write_u32(*init_loc);
+            self.fde_writer.write_u32(*addr);
+        }
+        LittleEndian::write_u32(
+            &mut self.fde_writer.slice[Self::fde_count_offset()..],
+            self.fdes.len() as u32,
+        );
+    }
+
+    pub fn size_from_eh_frame(eh_frame: &[u8]) -> usize {
+        // The virtual address of the EH frame does not matter in this case
+        // Calculation of size does not involve modifying any headers
+        let mut reader = DwarfReader::new(eh_frame, 0);
+        let mut fde_count = 0;
+        while !reader.slice.is_empty() {
+            // The original length field should be able to hold the entire value.
+            // The device memory space is limited to 32-bits addresses anyway.
+            let entry_length = reader.read_u32();
+            if entry_length == 0 || entry_length == 0xFFFF_FFFF {
+                unimplemented!()
+            }
+
+            // This slot stores the CIE ID (for CIE)/CIE Pointer (for FDE).
+            // This value must be non-zero for FDEs.
+            let cie_ptr = reader.read_u32();
+            if cie_ptr != 0 {
+                fde_count += 1;
+            }
+
+            reader.offset(entry_length - mem::size_of::<u32>() as u32);
+        }
+
+        12 + fde_count * 8
+    }
+}