lyken
/
nac3
forked from M-Labs/nac3
1
0
Fork 0
Code Pull Requests Activity

Compare commits

base: lyken:ndarray-strides
Branches Tags
lyken:ndarray-strides
lyken:to-models-tmptmp
lyken:to-models-save
lyken:ndstrides-wip
lyken:nds3
lyken:ndstrides-wip-9
lyken:ndstrides-wip-transpose
lyken:ndarray-strides-6
lyken:ndarray-strides-5
lyken:ndarray-strides-wip-2
lyken:ndarray-strides-wip
lyken:master
lyken:misc/impl-tracert
lyken:misc/mold
lyken:misc/update-pyo3
lyken:issue-313
lyken:refactor-primstore
lyken:numpy-anyall
lyken:enhance/cargo-test-standalone-demos
lyken:issue-136
lyken:rpc-obj-as-param
lyken:iterators
lyken:escape-analysis
lyken:refactor_anto
M-Labs:master
M-Labs:ndstrides-neo-next
M-Labs:misc/impl-tracert
M-Labs:ndstrides-14-end
M-Labs:ndstrides-13-tests
M-Labs:ndstrides-12-builtins
M-Labs:ndstrides-11-matmul
M-Labs:ndstrides-10-ops
M-Labs:ndstrides-9-subassign
M-Labs:ndstrides-8-transpose
M-Labs:ndstrides-7-broadcasting
M-Labs:ndstrides-6-reshape
M-Labs:ndstrides-5-miscfuncs
M-Labs:ndstrides-4-nparray
M-Labs:ndstrides-3-indexing
M-Labs:ndstrides-2-basic
M-Labs:ndstrides-1-model
M-Labs:ndstrides-base
M-Labs:misc/update-pyo3
M-Labs:misc/mold
M-Labs:ndstrides-1-model-lifetime
M-Labs:ndstrides-enhance-irrt
M-Labs:ndstrides-intro
M-Labs:issue-313
M-Labs:refactor-primstore
M-Labs:numpy-anyall
M-Labs:enhance/cargo-test-standalone-demos
M-Labs:issue-136
M-Labs:rpc-obj-as-param
M-Labs:iterators
M-Labs:escape-analysis
M-Labs:refactor_anto
..
compare: lyken:iterators
Branches Tags
lyken:to-models-tmptmp
lyken:to-models-save
lyken:ndstrides-wip
lyken:nds3
lyken:ndstrides-wip-9
lyken:ndstrides-wip-transpose
lyken:ndarray-strides-6
lyken:ndarray-strides-5
lyken:ndarray-strides-wip-2
lyken:ndarray-strides
lyken:ndarray-strides-wip
lyken:master
lyken:misc/impl-tracert
lyken:misc/mold
lyken:misc/update-pyo3
lyken:issue-313
lyken:refactor-primstore
lyken:numpy-anyall
lyken:enhance/cargo-test-standalone-demos
lyken:issue-136
lyken:rpc-obj-as-param
lyken:iterators
lyken:escape-analysis
lyken:refactor_anto
M-Labs:master
M-Labs:ndstrides-neo-next
M-Labs:misc/impl-tracert
M-Labs:ndstrides-14-end
M-Labs:ndstrides-13-tests
M-Labs:ndstrides-12-builtins
M-Labs:ndstrides-11-matmul
M-Labs:ndstrides-10-ops
M-Labs:ndstrides-9-subassign
M-Labs:ndstrides-8-transpose
M-Labs:ndstrides-7-broadcasting
M-Labs:ndstrides-6-reshape
M-Labs:ndstrides-5-miscfuncs
M-Labs:ndstrides-4-nparray
M-Labs:ndstrides-3-indexing
M-Labs:ndstrides-2-basic
M-Labs:ndstrides-1-model
M-Labs:ndstrides-base
M-Labs:misc/update-pyo3
M-Labs:misc/mold
M-Labs:ndstrides-1-model-lifetime
M-Labs:ndstrides-enhance-irrt
M-Labs:ndstrides-intro
M-Labs:issue-313
M-Labs:refactor-primstore
M-Labs:numpy-anyall
M-Labs:enhance/cargo-test-standalone-demos
M-Labs:issue-136
M-Labs:rpc-obj-as-param
M-Labs:iterators
M-Labs:escape-analysis
M-Labs:refactor_anto

2 Commits
ndarray-st ... iterators

Author SHA1 Message Date
wylited a5efba75ba nac3core: toplevel | add enumerate as a keyword for functions 2022-04-11 14:41:26 +08:00
wylited eeeccfd656 add extra builtin types for tuples and lists 2022-04-10 11:44:29 +08:00
127 changed files with 9397 additions and 33621 deletions
Show Stats Expand all files Collapse all files

1
.clippy.toml
View File

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

2
.gitignore vendored
View File

@ -1,3 +1,3 @@
__pycache__
/target
nix/windows/msys2
windows/msys2

24
.pre-commit-config.yaml
View File

@ -1,24 +0,0 @@
# 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: cargo
language: system
types: [file, rust]
pass_filenames: false
description: Runs cargo fmt on the codebase.
args: [fmt]
- id: nac3-cargo-clippy
name: nac3 cargo clippy
entry: cargo
language: system
types: [file, rust]
pass_filenames: false
description: Runs cargo clippy on the codebase.
args: [clippy, --tests]

1028
Cargo.lock generated
View File

File diff suppressed because it is too large Load Diff

2
Cargo.toml
View File

@ -1,6 +1,5 @@
[workspace]
members = [
"nac3ld",
"nac3ast",
"nac3parser",
"nac3core",
@ -8,7 +7,6 @@ members = [
"nac3artiq",
"runkernel",
]
resolver = "2"
[profile.release]
debug = true

33
README.md
View File

@ -1,10 +1,5 @@
<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.
@ -13,7 +8,7 @@ NAC3 has a modular design and its applicability reaches beyond ARTIQ. The ``nac3
## 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``).
NAC3 is packaged using the [Nix](https://nixos.org) Flakes system. Install Nix 2.4+ and enable flakes by adding ``experimental-features = nix-command flakes`` to ``nix.conf`` (e.g. ``~/.config/nix/nix.conf``).
## Try NAC3
@ -23,19 +18,27 @@ After setting up Nix as above, use ``nix shell git+https://github.com/m-labs/art
### Windows
Install [MSYS2](https://www.msys2.org/), and open "MSYS2 CLANG64". Edit ``/etc/pacman.conf`` to add:
Install [MSYS2](https://www.msys2.org/), and open "MSYS2 MinGW x64". Edit ``/etc/pacman.conf`` to add:
```
[artiq]
SigLevel = Optional TrustAll
Server = https://msys2.m-labs.hk/artiq-nac3
Server = https://lab.m-labs.hk/msys2
```
Then run the following commands:
```
pacman -Syu
pacman -S mingw-w64-clang-x86_64-artiq
pacman -S mingw-w64-x86_64-artiq
```
Install ``lld-msys2`` manually:
```
wget https://nixbld.m-labs.hk/build/115527/download/1/ld.lld.exe
mv ld.lld.exe C:/msys64/mingw64/bin
```
Note: This build of NAC3 cannot be used with Anaconda Python nor the python.org binaries for Windows. Those Python versions are compiled with Visual Studio (MSVC) and their ABI is incompatible with the GNU ABI used in this build. We have no plans to support Visual Studio nor the MSVC ABI. If you need a MSVC build, please install the requisite bloated spyware from Microsoft and compile NAC3 yourself.
## For developers
This repository contains:
@ -43,7 +46,6 @@ This repository contains:
- ``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.
@ -51,12 +53,3 @@ 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``.
Build NAC3 with ``cargo build --release``. See the demonstrations in ``nac3artiq`` and ``nac3standalone``.
### 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.

8
flake.lock
View File

@ -2,16 +2,16 @@
"nodes": {
"nixpkgs": {
"locked": {
"lastModified": 1720418205,
"narHash": "sha256-cPJoFPXU44GlhWg4pUk9oUPqurPlCFZ11ZQPk21GTPU=",
"lastModified": 1649024309,
"narHash": "sha256-AWbvj/NHZXVwAnHaVOFlxg7tcNerEKrKBmgGfztSHWM=",
"owner": "NixOS",
"repo": "nixpkgs",
"rev": "655a58a72a6601292512670343087c2d75d859c1",
"rev": "af0a9bc0e5341855518e9c1734d7ef913e5138b9",
"type": "github"
},
"original": {
"owner": "NixOS",
"ref": "nixos-unstable",
"ref": "nixos-21.11",
"repo": "nixpkgs",
"type": "github"
}

87
flake.nix
View File

@ -1,7 +1,7 @@
{
description = "The third-generation ARTIQ compiler";
inputs.nixpkgs.url = github:NixOS/nixpkgs/nixos-unstable;
inputs.nixpkgs.url = github:NixOS/nixpkgs/nixos-21.11;
outputs = { self, nixpkgs }:
let
@ -9,26 +9,15 @@
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.clang}/bin/clang $out/bin/clang-irrt-test
ln -s ${pkgs.llvmPackages_14.llvm.out}/bin/llvm-as $out/bin/llvm-as-irrt
'';
nac3artiq = pkgs.python3Packages.toPythonModule (
pkgs.rustPlatform.buildRustPackage rec {
pkgs.rustPlatform.buildRustPackage {
name = "nac3artiq";
outputs = [ "out" "runkernel" "standalone" ];
src = self;
cargoLock = {
lockFile = ./Cargo.lock;
};
cargoTestFlags = [ "--features" "test" ];
passthru.cargoLock = cargoLock;
nativeBuildInputs = [ pkgs.python3 pkgs.llvmPackages_14.clang llvm-tools-irrt pkgs.llvmPackages_14.llvm.out llvm-nac3 ];
cargoLock = { lockFile = ./Cargo.lock; };
nativeBuildInputs = [ pkgs.python3 pkgs.llvmPackages_13.clang-unwrapped llvm-nac3 ];
buildInputs = [ pkgs.python3 llvm-nac3 ];
checkInputs = [ (pkgs.python3.withPackages(ps: [ ps.numpy ps.scipy ])) ];
checkInputs = [ (pkgs.python3.withPackages(ps: [ ps.numpy ])) ];
checkPhase =
''
echo "Checking nac3standalone demos..."
@ -60,21 +49,21 @@
# LLVM PGO support
llvm-nac3-instrumented = pkgs.callPackage ./nix/llvm {
stdenv = pkgs.llvmPackages_14.stdenv;
stdenv = pkgs.llvmPackages_13.stdenv;
extraCmakeFlags = [ "-DLLVM_BUILD_INSTRUMENTED=IR" ];
};
nac3artiq-instrumented = pkgs.python3Packages.toPythonModule (
pkgs.rustPlatform.buildRustPackage {
name = "nac3artiq-instrumented";
src = self;
inherit (nac3artiq) cargoLock;
nativeBuildInputs = [ pkgs.python3 packages.x86_64-linux.llvm-tools-irrt llvm-nac3-instrumented ];
cargoLock = { lockFile = ./Cargo.lock; };
nativeBuildInputs = [ pkgs.python3 pkgs.llvmPackages_13.clang-unwrapped llvm-nac3-instrumented ];
buildInputs = [ pkgs.python3 llvm-nac3-instrumented ];
cargoBuildFlags = [ "--package" "nac3artiq" "--features" "init-llvm-profile" ];
doCheck = false;
configurePhase =
''
export CARGO_TARGET_X86_64_UNKNOWN_LINUX_GNU_RUSTFLAGS="-C link-arg=-L${pkgs.llvmPackages_14.compiler-rt}/lib/linux -C link-arg=-lclang_rt.profile-x86_64"
export CARGO_TARGET_X86_64_UNKNOWN_LINUX_GNU_RUSTFLAGS="-C link-arg=-L${pkgs.llvmPackages_13.compiler-rt}/lib/linux -C link-arg=-lclang_rt.profile-x86_64"
'';
installPhase =
''
@ -86,35 +75,11 @@
);
nac3artiq-profile = pkgs.stdenvNoCC.mkDerivation {
name = "nac3artiq-profile";
srcs = [
(pkgs.fetchFromGitHub {
owner = "m-labs";
repo = "sipyco";
rev = "939f84f9b5eef7efbf7423c735d1834783b6140e";
sha256 = "sha256-15Nun4EY35j+6SPZkjzZtyH/ncxLS60KuGJjFh5kSTc=";
})
(pkgs.fetchFromGitHub {
owner = "m-labs";
repo = "artiq";
rev = "923ca3377d42c815f979983134ec549dc39d3ca0";
sha256 = "sha256-oJoEeNEeNFSUyh6jXG8Tzp6qHVikeHS0CzfE+mODPgw=";
})
];
buildInputs = [
(python3-mimalloc.withPackages(ps: [ ps.numpy ps.scipy ps.jsonschema ps.lmdb nac3artiq-instrumented ]))
pkgs.llvmPackages_14.llvm.out
];
src = self;
buildInputs = [ (python3-mimalloc.withPackages(ps: [ ps.numpy nac3artiq-instrumented ])) pkgs.lld_13 pkgs.llvmPackages_13.libllvm ];
phases = [ "buildPhase" "installPhase" ];
buildPhase =
''
srcs=($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
'';
# TODO: get more representative code.
buildPhase = "python $src/nac3artiq/demo/demo.py";
installPhase =
''
mkdir $out
@ -122,15 +87,15 @@
'';
};
llvm-nac3-pgo = pkgs.callPackage ./nix/llvm {
stdenv = pkgs.llvmPackages_14.stdenv;
stdenv = pkgs.llvmPackages_13.stdenv;
extraCmakeFlags = [ "-DLLVM_PROFDATA_FILE=${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 ];
cargoLock = { lockFile = ./Cargo.lock; };
nativeBuildInputs = [ pkgs.python3 pkgs.llvmPackages_13.clang-unwrapped llvm-nac3-pgo ];
buildInputs = [ pkgs.python3 llvm-nac3-pgo ];
cargoBuildFlags = [ "--package" "nac3artiq" ];
cargoTestFlags = [ "--package" "nac3ast" "--package" "nac3parser" "--package" "nac3core" "--package" "nac3artiq" ];
@ -146,27 +111,22 @@
packages.x86_64-w64-mingw32 = import ./nix/windows { inherit pkgs; };
devShells.x86_64-linux.default = pkgs.mkShell {
devShell.x86_64-linux = pkgs.mkShell {
name = "nac3-dev-shell";
buildInputs = with pkgs; [
# build dependencies
packages.x86_64-linux.llvm-nac3
llvmPackages_14.clang llvmPackages_14.llvm.out # for running nac3standalone demos
packages.x86_64-linux.llvm-tools-irrt
llvmPackages_13.clang-unwrapped # 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 ]))
lld_13
(packages.x86_64-linux.python3-mimalloc.withPackages(ps: [ ps.numpy ]))
# development tools
cargo-insta
clippy
pre-commit
rustfmt
rust-analyzer
];
# https://nixos.wiki/wiki/Rust#Shell.nix_example
RUST_SRC_PATH = "${pkgs.rust.packages.stable.rustPlatform.rustLibSrc}";
};
devShells.x86_64-linux.msys2 = pkgs.mkShell {
name = "nac3-dev-shell-msys2";
@ -179,15 +139,16 @@
};
hydraJobs = {
inherit (packages.x86_64-linux) llvm-nac3 nac3artiq nac3artiq-pgo;
inherit (packages.x86_64-linux) llvm-nac3 nac3artiq;
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;
lld-msys2 = packages.x86_64-w64-mingw32.lld;
};
};
nixConfig = {
extra-trusted-public-keys = "nixbld.m-labs.hk-1:5aSRVA5b320xbNvu30tqxVPXpld73bhtOeH6uAjRyHc=";
extra-substituters = "https://nixbld.m-labs.hk";
binaryCachePublicKeys = ["nixbld.m-labs.hk-1:5aSRVA5b320xbNvu30tqxVPXpld73bhtOeH6uAjRyHc="];
binaryCaches = ["https://nixbld.m-labs.hk" "https://cache.nixos.org"];
};
}

56
nac3.svg
View File

@ -1,56 +0,0 @@
<?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>
Side by Side

Before

Width:  |  Height:  |  Size: 3.3 KiB

14
nac3artiq/Cargo.toml
View File

@ -2,25 +2,23 @@
name = "nac3artiq"
version = "0.1.0"
authors = ["M-Labs"]
edition = "2021"
edition = "2018"
[lib]
name = "nac3artiq"
crate-type = ["cdylib"]
[dependencies]
itertools = "0.13"
pyo3 = { version = "0.21", features = ["extension-module", "gil-refs"] }
parking_lot = "0.12"
tempfile = "3.10"
pyo3 = { version = "0.14", features = ["extension-module"] }
parking_lot = "0.11"
tempfile = "3"
nac3parser = { path = "../nac3parser" }
nac3core = { path = "../nac3core" }
nac3ld = { path = "../nac3ld" }
[dependencies.inkwell]
version = "0.4"
version = "0.1.0-beta.4"
default-features = false
features = ["llvm14-0", "target-x86", "target-arm", "target-riscv", "no-libffi-linking"]
features = ["llvm13-0", "target-x86", "target-arm", "target-riscv", "no-libffi-linking"]
[features]
init-llvm-profile = []

6
nac3artiq/demo/demo.py
View File

@ -1,4 +1,10 @@
from min_artiq import *
from numpy import int32, int64
@extern
def output_int(x: int32):
...
@nac3

7
nac3artiq/demo/embedding_map.py
View File

@ -18,13 +18,6 @@ class EmbeddingMap:
"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):

16
nac3artiq/demo/min_artiq.py
View File

@ -10,7 +10,7 @@ from embedding_map import EmbeddingMap
__all__ = [
"Kernel", "KernelInvariant", "virtual", "ConstGeneric",
"Kernel", "KernelInvariant", "virtual",
"Option", "Some", "none", "UnwrapNoneError",
"round64", "floor64", "ceil64",
"extern", "kernel", "portable", "nac3",
@ -67,12 +67,6 @@ def Some(v: T) -> Option[T]:
none = Option(None)
class _ConstGenericMarker:
pass
def ConstGeneric(name, constraint):
return TypeVar(name, _ConstGenericMarker, constraint)
def round64(x):
return round(x)
@ -86,13 +80,7 @@ def ceil64(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)
compiler = nac3artiq.NAC3(core_arguments["target"])
allow_registration = True
# Delay NAC3 analysis until all referenced variables are supposed to exist on the CPython side.
registered_functions = set()

24
nac3artiq/demo/string_attribute_issue337.py
View File

@ -1,24 +0,0 @@
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()

40
nac3artiq/demo/support_class_attr_issue102.py
View File

@ -1,40 +0,0 @@
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()

496
nac3artiq/src/codegen.rs
View File

@ -1,13 +1,12 @@
use nac3core::{
codegen::{
expr::gen_call,
llvm_intrinsics::{call_int_smax, call_stackrestore, call_stacksave},
stmt::{gen_block, gen_with},
CodeGenContext, CodeGenerator,
},
symbol_resolver::ValueEnum,
toplevel::{helper::PrimDef, numpy::unpack_ndarray_var_tys, DefinitionId, GenCall},
typecheck::typedef::{iter_type_vars, FunSignature, FuncArg, Type, TypeEnum, VarMap},
toplevel::{DefinitionId, GenCall},
typecheck::typedef::{FunSignature, FuncArg, Type, TypeEnum}
};
use nac3parser::ast::{Expr, ExprKind, Located, Stmt, StmtKind, StrRef};
@ -16,10 +15,7 @@ use inkwell::{
context::Context, module::Linkage, types::IntType, values::BasicValueEnum, AddressSpace,
};
use pyo3::{
types::{PyDict, PyList},
PyObject, PyResult, Python,
};
use pyo3::{PyObject, PyResult, Python, types::{PyDict, PyList}};
use crate::{symbol_resolver::InnerResolver, timeline::TimeFns};
@ -30,45 +26,13 @@ use std::{
sync::Arc,
};
/// The parallelism mode within a block.
#[derive(Copy, Clone, Eq, PartialEq)]
enum ParallelMode {
/// No parallelism is currently registered for this context.
None,
/// Legacy (or shallow) parallelism. Default before NAC3.
///
/// Each statement within the `with` block is treated as statements to be executed in parallel.
Legacy,
/// Deep parallelism. Default since NAC3.
///
/// Each function call within the `with` block (except those within a nested `sequential` block)
/// are treated to be executed in parallel.
Deep,
}
pub struct ArtiqCodeGenerator<'a> {
name: String,
/// The size of a `size_t` variable in bits.
size_t: u32,
/// Monotonic counter for naming `start`/`stop` variables used by `with parallel` blocks.
name_counter: u32,
/// Variable for tracking the start of a `with parallel` block.
start: Option<Expr<Option<Type>>>,
/// Variable for tracking the end of a `with parallel` block.
end: Option<Expr<Option<Type>>>,
timeline: &'a (dyn TimeFns + Sync),
/// The [`ParallelMode`] of the current parallel context.
///
/// The current parallel context refers to the nearest `with parallel` or `with legacy_parallel`
/// statement, which is used to determine when and how the timeline should be updated.
parallel_mode: ParallelMode,
}
impl<'a> ArtiqCodeGenerator<'a> {
@ -78,74 +42,7 @@ impl<'a> ArtiqCodeGenerator<'a> {
timeline: &'a (dyn TimeFns + Sync),
) -> ArtiqCodeGenerator<'a> {
assert!(size_t == 32 || size_t == 64);
ArtiqCodeGenerator {
name,
size_t,
name_counter: 0,
start: None,
end: None,
timeline,
parallel_mode: ParallelMode::None,
}
}
/// If the generator is currently in a direct-`parallel` block context, emits IR that resets the
/// position of the timeline to the initial timeline position before entering the `parallel`
/// block.
///
/// Direct-`parallel` block context refers to when the generator is generating statements whose
/// closest parent `with` statement is a `with parallel` block.
fn timeline_reset_start(&mut self, ctx: &mut CodeGenContext<'_, '_>) -> Result<(), String> {
if let Some(start) = self.start.clone() {
let start_val = self.gen_expr(ctx, &start)?.unwrap().to_basic_value_enum(
ctx,
self,
start.custom.unwrap(),
)?;
self.timeline.emit_at_mu(ctx, start_val);
}
Ok(())
}
/// If the generator is currently in a `parallel` block context, emits IR that updates the
/// maximum end position of the `parallel` block as specified by the timeline `end` value.
///
/// In general the `end` parameter should be set to `self.end` for updating the maximum end
/// position for the current `parallel` block. Other values can be passed in to update the
/// maximum end position for other `parallel` blocks.
///
/// `parallel`-block context refers to when the generator is generating statements within a
/// (possibly indirect) `parallel` block.
///
/// * `store_name` - The LLVM value name for the pointer to `end`. `.addr` will be appended to
/// the end of the provided value name.
fn timeline_update_end_max(
&mut self,
ctx: &mut CodeGenContext<'_, '_>,
end: Option<Expr<Option<Type>>>,
store_name: Option<&str>,
) -> Result<(), String> {
if let Some(end) = end {
let old_end = self.gen_expr(ctx, &end)?.unwrap().to_basic_value_enum(
ctx,
self,
end.custom.unwrap(),
)?;
let now = self.timeline.emit_now_mu(ctx);
let max =
call_int_smax(ctx, old_end.into_int_value(), now.into_int_value(), Some("smax"));
let end_store = self
.gen_store_target(
ctx,
&end,
store_name.map(|name| format!("{name}.addr")).as_deref(),
)?
.unwrap();
ctx.builder.build_store(end_store, max).unwrap();
}
Ok(())
ArtiqCodeGenerator { name, size_t, name_counter: 0, start: None, end: None, timeline }
}
}
@ -162,62 +59,49 @@ impl<'b> CodeGenerator for ArtiqCodeGenerator<'b> {
}
}
fn gen_block<'ctx, 'a, 'c, I: Iterator<Item = &'c Stmt<Option<Type>>>>(
fn gen_call<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
stmts: I,
) -> Result<(), String>
where
Self: Sized,
{
// Legacy parallel emits timeline end-update/timeline-reset after each top-level statement
// in the parallel block
if self.parallel_mode == ParallelMode::Legacy {
for stmt in stmts {
self.gen_stmt(ctx, stmt)?;
if ctx.is_terminated() {
break;
}
self.timeline_update_end_max(ctx, self.end.clone(), Some("end"))?;
self.timeline_reset_start(ctx)?;
}
Ok(())
} else {
gen_block(self, ctx, stmts)
}
}
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> {
let result = gen_call(self, ctx, obj, fun, params)?;
// Deep parallel emits timeline end-update/timeline-reset after each function call
if self.parallel_mode == ParallelMode::Deep {
self.timeline_update_end_max(ctx, self.end.clone(), Some("end"))?;
self.timeline_reset_start(ctx)?;
if let Some(end) = self.end.clone() {
let old_end = self.gen_expr(ctx, &end)?.unwrap().to_basic_value_enum(ctx, self)?;
let now = self.timeline.emit_now_mu(ctx);
let smax = ctx.module.get_function("llvm.smax.i64").unwrap_or_else(|| {
let i64 = ctx.ctx.i64_type();
ctx.module.add_function(
"llvm.smax.i64",
i64.fn_type(&[i64.into(), i64.into()], false),
None,
)
});
let max = ctx
.builder
.build_call(smax, &[old_end.into(), now.into()], "smax")
.try_as_basic_value()
.left()
.unwrap();
let end_store = self.gen_store_target(ctx, &end)?;
ctx.builder.build_store(end_store, max);
}
if let Some(start) = self.start.clone() {
let start_val = self.gen_expr(ctx, &start)?.unwrap().to_basic_value_enum(ctx, self)?;
self.timeline.emit_at_mu(ctx, start_val);
}
Ok(result)
}
fn gen_with(
fn gen_with<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'_, '_>,
ctx: &mut CodeGenContext<'ctx, 'a>,
stmt: &Stmt<Option<Type>>,
) -> Result<(), String> {
let StmtKind::With { items, body, .. } = &stmt.node else { unreachable!() };
if let StmtKind::With { items, body, .. } = &stmt.node {
if items.len() == 1 && items[0].optional_vars.is_none() {
let item = &items[0];
// Behavior of parallel and sequential:
// Each function call (indirectly, can be inside a sequential block) within a parallel
// block will update the end variable to the maximum now_mu in the block.
@ -232,21 +116,14 @@ impl<'b> CodeGenerator for ArtiqCodeGenerator<'b> {
// - If there is a end variable, it indicates that we are (indirectly) inside a
// parallel block, and we should update the max end value.
if let ExprKind::Name { id, ctx: name_ctx } = &item.context_expr.node {
if id == &"parallel".into() || id == &"legacy_parallel".into() {
if id == &"parallel".into() {
let old_start = self.start.take();
let old_end = self.end.take();
let old_parallel_mode = self.parallel_mode;
let now = if let Some(old_start) = &old_start {
self.gen_expr(ctx, old_start)?.unwrap().to_basic_value_enum(
ctx,
self,
old_start.custom.unwrap(),
)?
self.gen_expr(ctx, old_start)?.unwrap().to_basic_value_enum(ctx, self)?
} else {
self.timeline.emit_now_mu(ctx)
};
// Emulate variable allocation, as we need to use the CodeGenContext
// HashMap to store our variable due to lifetime limitation
// Note: we should be able to store variables directly if generic
@ -260,13 +137,11 @@ impl<'b> CodeGenerator for ArtiqCodeGenerator<'b> {
let start_expr = Located {
// location does not matter at this point
location: stmt.location,
node: ExprKind::Name { id: start, ctx: *name_ctx },
node: ExprKind::Name { id: start, ctx: name_ctx.clone() },
custom: Some(ctx.primitives.int64),
};
let start = self
.gen_store_target(ctx, &start_expr, Some("start.addr"))?
.unwrap();
ctx.builder.build_store(start, now).unwrap();
let start = self.gen_store_target(ctx, &start_expr)?;
ctx.builder.build_store(start, now);
Ok(Some(start_expr)) as Result<_, String>
},
|v| Ok(Some(v)),
@ -275,23 +150,15 @@ impl<'b> CodeGenerator for ArtiqCodeGenerator<'b> {
let end_expr = Located {
// location does not matter at this point
location: stmt.location,
node: ExprKind::Name { id: end, ctx: *name_ctx },
node: ExprKind::Name { id: end, ctx: name_ctx.clone() },
custom: Some(ctx.primitives.int64),
};
let end = self.gen_store_target(ctx, &end_expr, Some("end.addr"))?.unwrap();
ctx.builder.build_store(end, now).unwrap();
let end = self.gen_store_target(ctx, &end_expr)?;
ctx.builder.build_store(end, now);
self.end = Some(end_expr);
self.name_counter += 1;
self.parallel_mode = match id.to_string().as_str() {
"parallel" => ParallelMode::Deep,
"legacy_parallel" => ParallelMode::Legacy,
_ => unreachable!(),
};
self.gen_block(ctx, body.iter())?;
gen_block(self, ctx, body.iter())?;
let current = ctx.builder.get_insert_block().unwrap();
// if the current block is terminated, move before the terminator
// we want to set the timeline before reaching the terminator
// TODO: This may be unsound if there are multiple exit paths in the
@ -305,60 +172,68 @@ impl<'b> CodeGenerator for ArtiqCodeGenerator<'b> {
} else {
false
};
// set duration
let end_expr = self.end.take().unwrap();
let end_val = self.gen_expr(ctx, &end_expr)?.unwrap().to_basic_value_enum(
ctx,
self,
end_expr.custom.unwrap(),
)?;
let end_val =
self.gen_expr(ctx, &end_expr)?.unwrap().to_basic_value_enum(ctx, self)?;
// inside a sequential block
if old_start.is_none() {
self.timeline.emit_at_mu(ctx, end_val);
}
// inside a parallel block, should update the outer max now_mu
self.timeline_update_end_max(ctx, old_end.clone(), Some("outer.end"))?;
self.parallel_mode = old_parallel_mode;
self.end = old_end;
if let Some(old_end) = &old_end {
let outer_end_val = self
.gen_expr(ctx, old_end)?
.unwrap()
.to_basic_value_enum(ctx, self)?;
let smax =
ctx.module.get_function("llvm.smax.i64").unwrap_or_else(|| {
let i64 = ctx.ctx.i64_type();
ctx.module.add_function(
"llvm.smax.i64",
i64.fn_type(&[i64.into(), i64.into()], false),
None,
)
});
let max = ctx
.builder
.build_call(smax, &[end_val.into(), outer_end_val.into()], "smax")
.try_as_basic_value()
.left()
.unwrap();
let outer_end = self.gen_store_target(ctx, old_end)?;
ctx.builder.build_store(outer_end, max);
}
self.start = old_start;
self.end = old_end;
if reset_position {
ctx.builder.position_at_end(current);
}
return Ok(());
} else if id == &"sequential".into() {
// For deep parallel, temporarily take away start to avoid function calls in
// the block from resetting the timeline.
// This does not affect legacy parallel, as the timeline will be reset after
// this block finishes execution.
let start = self.start.take();
self.gen_block(ctx, body.iter())?;
self.start = start;
// Reset the timeline when we are exiting the sequential block
// Legacy parallel does not need this, since it will be reset after codegen
// for this statement is completed
if self.parallel_mode == ParallelMode::Deep {
self.timeline_reset_start(ctx)?;
for stmt in body.iter() {
self.gen_stmt(ctx, stmt)?;
if ctx.is_terminated() {
break;
}
}
self.start = start;
return Ok(());
}
}
}
// not parallel/sequential
gen_with(self, ctx, stmt)
} else {
unreachable!()
}
}
}
fn gen_rpc_tag(
ctx: &mut CodeGenContext<'_, '_>,
fn gen_rpc_tag<'ctx, 'a>(
ctx: &mut CodeGenContext<'ctx, 'a>,
ty: Type,
buffer: &mut Vec<u8>,
) -> Result<(), String> {
@ -393,32 +268,9 @@ fn gen_rpc_tag(
gen_rpc_tag(ctx, *ty, buffer)?;
}
}
TObj { obj_id, params, .. } if *obj_id == PrimDef::List.id() => {
let ty = iter_type_vars(params).next().unwrap().ty;
TList { ty } => {
buffer.push(b'l');
gen_rpc_tag(ctx, ty, buffer)?;
}
TObj { obj_id, .. } if *obj_id == PrimDef::NDArray.id() => {
let (ndarray_dtype, ndarray_ndims) = unpack_ndarray_var_tys(&mut ctx.unifier, ty);
let ndarray_ndims = if let TLiteral { values, .. } =
&*ctx.unifier.get_ty_immutable(ndarray_ndims)
{
if values.len() != 1 {
return Err(format!("NDArray types with multiple literal bounds for ndims is not supported: {}", ctx.unifier.stringify(ty)));
}
let value = values[0].clone();
u64::try_from(value.clone())
.map_err(|()| format!("Expected u64 for ndarray.ndims, got {value}"))?
} else {
unreachable!()
};
assert!((0u64..=u64::from(u8::MAX)).contains(&ndarray_ndims));
buffer.push(b'a');
buffer.push((ndarray_ndims & 0xFF) as u8);
gen_rpc_tag(ctx, ndarray_dtype, buffer)?;
gen_rpc_tag(ctx, *ty, buffer)?;
}
_ => return Err(format!("Unsupported type: {:?}", ctx.unifier.stringify(ty))),
}
@ -426,14 +278,14 @@ fn gen_rpc_tag(
Ok(())
}
fn rpc_codegen_callback_fn<'ctx>(
ctx: &mut CodeGenContext<'ctx, '_>,
fn rpc_codegen_callback_fn<'ctx, 'a>(
ctx: &mut CodeGenContext<'ctx, 'a>,
obj: Option<(Type, ValueEnum<'ctx>)>,
fun: (&FunSignature, DefinitionId),
args: Vec<(Option<StrRef>, ValueEnum<'ctx>)>,
generator: &mut dyn CodeGenerator,
) -> Result<Option<BasicValueEnum<'ctx>>, String> {
let ptr_type = ctx.ctx.i8_type().ptr_type(AddressSpace::default());
let ptr_type = ctx.ctx.i8_type().ptr_type(inkwell::AddressSpace::Generic);
let size_type = generator.get_size_type(ctx.ctx);
let int8 = ctx.ctx.i8_type();
let int32 = ctx.ctx.i32_type();
@ -445,7 +297,7 @@ fn rpc_codegen_callback_fn<'ctx>(
if obj.is_some() {
tag.push(b'O');
}
for arg in &fun.0.args {
for arg in fun.0.args.iter() {
gen_rpc_tag(ctx, arg.ty, &mut tag)?;
}
tag.push(b':');
@ -465,7 +317,7 @@ fn rpc_codegen_callback_fn<'ctx>(
format!("tagptr{}", fun.1 .0).as_str(),
);
tag_arr_ptr.set_initializer(&int8.const_array(
&tag.iter().map(|v| int8.const_int(u64::from(*v), false)).collect::<Vec<_>>(),
&tag.iter().map(|v| int8.const_int(*v as u64, false)).collect::<Vec<_>>(),
));
tag_arr_ptr.set_linkage(Linkage::Private);
let tag_ptr = ctx.module.add_global(tag_ptr_type, None, &hash);
@ -481,35 +333,45 @@ fn rpc_codegen_callback_fn<'ctx>(
})
.as_pointer_value();
let arg_length = args.len() + usize::from(obj.is_some());
let arg_length = args.len() + if obj.is_some() { 1 } else { 0 };
let stackptr = call_stacksave(ctx, Some("rpc.stack"));
let args_ptr = ctx
.builder
.build_array_alloca(
let stacksave = ctx.module.get_function("llvm.stacksave").unwrap_or_else(|| {
ctx.module.add_function("llvm.stacksave", ptr_type.fn_type(&[], false), None)
});
let stackrestore = ctx.module.get_function("llvm.stackrestore").unwrap_or_else(|| {
ctx.module.add_function(
"llvm.stackrestore",
ctx.ctx.void_type().fn_type(&[ptr_type.into()], false),
None,
)
});
let stackptr = ctx.builder.build_call(stacksave, &[], "rpc.stack");
let args_ptr = ctx.builder.build_array_alloca(
ptr_type,
ctx.ctx.i32_type().const_int(arg_length as u64, false),
"argptr",
)
.unwrap();
);
// -- rpc args handling
let mut keys = fun.0.args.clone();
let mut mapping = HashMap::new();
for (key, value) in args {
for (key, value) in args.into_iter() {
mapping.insert(key.unwrap_or_else(|| keys.remove(0).name), value);
}
// default value handling
for k in keys {
mapping
.insert(k.name, ctx.gen_symbol_val(generator, &k.default_value.unwrap(), k.ty).into());
for k in keys.into_iter() {
mapping.insert(
k.name,
ctx.gen_symbol_val(generator, &k.default_value.unwrap(), k.ty).into()
);
}
// reorder the parameters
let mut real_params = fun
.0
.args
.iter()
.map(|arg| mapping.remove(&arg.name).unwrap().to_basic_value_enum(ctx, generator, arg.ty))
.map(|arg| mapping.remove(&arg.name).unwrap().to_basic_value_enum(ctx, generator))
.collect::<Result<Vec<_>, _>>()?;
if let Some(obj) = obj {
if let ValueEnum::Static(obj) = obj.1 {
@ -521,19 +383,17 @@ fn rpc_codegen_callback_fn<'ctx>(
}
for (i, arg) in real_params.iter().enumerate() {
let arg_slot =
generator.gen_var_alloc(ctx, arg.get_type(), Some(&format!("rpc.arg{i}"))).unwrap();
ctx.builder.build_store(arg_slot, *arg).unwrap();
let arg_slot = ctx.builder.build_bitcast(arg_slot, ptr_type, "rpc.arg").unwrap();
let arg_slot = ctx.builder.build_alloca(arg.get_type(), &format!("rpc.arg{}", i));
ctx.builder.build_store(arg_slot, *arg);
let arg_slot = ctx.builder.build_bitcast(arg_slot, ptr_type, "rpc.arg");
let arg_ptr = unsafe {
ctx.builder.build_gep(
args_ptr,
&[int32.const_int(i as u64, false)],
&format!("rpc.arg{i}"),
&format!("rpc.arg{}", i),
)
}
.unwrap();
ctx.builder.build_store(arg_ptr, arg_slot).unwrap();
};
ctx.builder.build_store(arg_ptr, arg_slot);
}
// call
@ -543,20 +403,26 @@ fn rpc_codegen_callback_fn<'ctx>(
ctx.ctx.void_type().fn_type(
&[
int32.into(),
tag_ptr_type.ptr_type(AddressSpace::default()).into(),
ptr_type.ptr_type(AddressSpace::default()).into(),
tag_ptr_type.ptr_type(AddressSpace::Generic).into(),
ptr_type.ptr_type(AddressSpace::Generic).into(),
],
false,
),
None,
)
});
ctx.builder
.build_call(rpc_send, &[service_id.into(), tag_ptr.into(), args_ptr.into()], "rpc.send")
.unwrap();
ctx.builder.build_call(
rpc_send,
&[service_id.into(), tag_ptr.into(), args_ptr.into()],
"rpc.send",
);
// reclaim stack space used by arguments
call_stackrestore(ctx, stackptr);
ctx.builder.build_call(
stackrestore,
&[stackptr.try_as_basic_value().unwrap_left().into()],
"rpc.stackrestore",
);
// -- receive value:
// T result = {
@ -582,91 +448,84 @@ fn rpc_codegen_callback_fn<'ctx>(
let alloc_bb = ctx.ctx.append_basic_block(current_function, "rpc.continue");
let tail_bb = ctx.ctx.append_basic_block(current_function, "rpc.tail");
let ret_ty = ctx.get_llvm_abi_type(generator, fun.0.ret);
let ret_ty = ctx.get_llvm_type(generator, fun.0.ret);
let need_load = !ret_ty.is_pointer_type();
let slot = ctx.builder.build_alloca(ret_ty, "rpc.ret.slot").unwrap();
let slotgen = ctx.builder.build_bitcast(slot, ptr_type, "rpc.ret.ptr").unwrap();
ctx.builder.build_unconditional_branch(head_bb).unwrap();
let slot = ctx.builder.build_alloca(ret_ty, "rpc.ret.slot");
let slotgen = ctx.builder.build_bitcast(slot, ptr_type, "rpc.ret.ptr");
ctx.builder.build_unconditional_branch(head_bb);
ctx.builder.position_at_end(head_bb);
let phi = ctx.builder.build_phi(ptr_type, "rpc.ptr").unwrap();
let phi = ctx.builder.build_phi(ptr_type, "rpc.ptr");
phi.add_incoming(&[(&slotgen, prehead_bb)]);
let alloc_size = ctx
.build_call_or_invoke(rpc_recv, &[phi.as_basic_value()], "rpc.size.next")
.unwrap()
.into_int_value();
let is_done = ctx
.builder
.build_int_compare(inkwell::IntPredicate::EQ, int32.const_zero(), alloc_size, "rpc.done")
.unwrap();
let is_done = ctx.builder.build_int_compare(
inkwell::IntPredicate::EQ,
int32.const_zero(),
alloc_size,
"rpc.done",
);
ctx.builder.build_conditional_branch(is_done, tail_bb, alloc_bb).unwrap();
ctx.builder.build_conditional_branch(is_done, tail_bb, alloc_bb);
ctx.builder.position_at_end(alloc_bb);
let alloc_ptr = ctx.builder.build_array_alloca(ptr_type, alloc_size, "rpc.alloc").unwrap();
let alloc_ptr = ctx.builder.build_bitcast(alloc_ptr, ptr_type, "rpc.alloc.ptr").unwrap();
let alloc_ptr = ctx.builder.build_array_alloca(ptr_type, alloc_size, "rpc.alloc");
let alloc_ptr = ctx.builder.build_bitcast(alloc_ptr, ptr_type, "rpc.alloc.ptr");
phi.add_incoming(&[(&alloc_ptr, alloc_bb)]);
ctx.builder.build_unconditional_branch(head_bb).unwrap();
ctx.builder.build_unconditional_branch(head_bb);
ctx.builder.position_at_end(tail_bb);
let result = ctx.builder.build_load(slot, "rpc.result").unwrap();
let result = ctx.builder.build_load(slot, "rpc.result");
if need_load {
call_stackrestore(ctx, stackptr);
ctx.builder.build_call(
stackrestore,
&[stackptr.try_as_basic_value().unwrap_left().into()],
"rpc.stackrestore",
);
}
Ok(Some(result))
}
pub fn attributes_writeback(
ctx: &mut CodeGenContext<'_, '_>,
pub fn attributes_writeback<'ctx, 'a>(
ctx: &mut CodeGenContext<'ctx, 'a>,
generator: &mut dyn CodeGenerator,
inner_resolver: &InnerResolver,
host_attributes: &PyObject,
host_attributes: PyObject,
) -> Result<(), String> {
Python::with_gil(|py| -> PyResult<Result<(), String>> {
let host_attributes: &PyList = host_attributes.downcast(py)?;
let host_attributes = host_attributes.cast_as::<PyList>(py)?;
let top_levels = ctx.top_level.definitions.read();
let globals = inner_resolver.global_value_ids.read();
let int32 = ctx.ctx.i32_type();
let zero = int32.const_zero();
let mut values = Vec::new();
let mut scratch_buffer = Vec::new();
for val in (*globals).values() {
for (_, val) in globals.iter() {
let val = val.as_ref(py);
let ty = inner_resolver.get_obj_type(
py,
val,
&mut ctx.unifier,
&top_levels,
&ctx.primitives,
)?;
let ty = inner_resolver.get_obj_type(py, val, &mut ctx.unifier, &top_levels, &ctx.primitives)?;
if let Err(ty) = ty {
return Ok(Err(ty));
return Ok(Err(ty))
}
let ty = ty.unwrap();
match &*ctx.unifier.get_ty(ty) {
TypeEnum::TObj { fields, obj_id, .. }
if *obj_id != ctx.primitives.option.obj_id(&ctx.unifier).unwrap() =>
{
TypeEnum::TObj { fields, .. } => {
// we only care about primitive attributes
// for non-primitive attributes, they should be in another global
let mut attributes = Vec::new();
let obj = inner_resolver.get_obj_value(py, val, ctx, generator, ty)?.unwrap();
for (name, (field_ty, is_mutable)) in fields {
let obj = inner_resolver.get_obj_value(py, val, ctx, generator)?.unwrap();
for (name, (field_ty, is_mutable)) in fields.iter() {
if !is_mutable {
continue;
continue
}
if gen_rpc_tag(ctx, *field_ty, &mut scratch_buffer).is_ok() {
attributes.push(name.to_string());
let (index, _) = ctx.get_attr_index(ty, *name);
values.push((
*field_ty,
ctx.build_gep_and_load(
let index = ctx.get_attr_index(ty, *name);
values.push((*field_ty, ctx.build_gep_and_load(
obj.into_pointer_value(),
&[zero, int32.const_int(index as u64, false)],
None,
),
));
&[zero, int32.const_int(index as u64, false)])));
}
}
if !attributes.is_empty() {
@ -675,46 +534,33 @@ pub fn attributes_writeback(
pydict.set_item("fields", attributes)?;
host_attributes.append(pydict)?;
}
}
TypeEnum::TObj { obj_id, params, .. } if *obj_id == PrimDef::List.id() => {
let elem_ty = iter_type_vars(params).next().unwrap().ty;
if gen_rpc_tag(ctx, elem_ty, &mut scratch_buffer).is_ok() {
},
TypeEnum::TList { ty: elem_ty } => {
if gen_rpc_tag(ctx, *elem_ty, &mut scratch_buffer).is_ok() {
let pydict = PyDict::new(py);
pydict.set_item("obj", val)?;
host_attributes.append(pydict)?;
values.push((
ty,
inner_resolver.get_obj_value(py, val, ctx, generator, ty)?.unwrap(),
));
}
values.push((ty, inner_resolver.get_obj_value(py, val, ctx, generator)?.unwrap()));
}
},
_ => {}
}
}
let fun = FunSignature {
args: values
.iter()
.enumerate()
.map(|(i, (ty, _))| FuncArg {
args: values.iter().enumerate().map(|(i, (ty, _))| FuncArg {
name: i.to_string().into(),
ty: *ty,
default_value: None,
})
.collect(),
default_value: None
}).collect(),
ret: ctx.primitives.none,
vars: VarMap::default(),
vars: Default::default()
};
let args: Vec<_> =
values.into_iter().map(|(_, val)| (None, ValueEnum::Dynamic(val))).collect();
if let Err(e) =
rpc_codegen_callback_fn(ctx, None, (&fun, PrimDef::Int32.id()), args, generator)
{
let args: Vec<_> = values.into_iter().map(|(_, val)| (None, ValueEnum::Dynamic(val))).collect();
if let Err(e) = rpc_codegen_callback_fn(ctx, None, (&fun, DefinitionId(0)), args, generator) {
return Ok(Err(e));
}
Ok(Ok(()))
})
.unwrap()?;
}).unwrap()?;
Ok(())
}

872
nac3artiq/src/lib.rs
View File

File diff suppressed because it is too large Load Diff

973
nac3artiq/src/symbol_resolver.rs
View File

File diff suppressed because it is too large Load Diff

385
nac3artiq/src/timeline.rs
View File

@ -1,20 +1,10 @@
use inkwell::{
values::{BasicValueEnum, CallSiteValue},
AddressSpace, AtomicOrdering,
};
use itertools::Either;
use inkwell::{values::BasicValueEnum, AddressSpace, AtomicOrdering};
use nac3core::codegen::CodeGenContext;
/// 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>);
fn emit_now_mu<'ctx, 'a>(&self, ctx: &mut CodeGenContext<'ctx, 'a>) -> BasicValueEnum<'ctx>;
fn emit_at_mu<'ctx, 'a>(&self, ctx: &mut CodeGenContext<'ctx, 'a>, t: BasicValueEnum<'ctx>);
fn emit_delay_mu<'ctx, 'a>(&self, ctx: &mut CodeGenContext<'ctx, 'a>, dt: BasicValueEnum<'ctx>);
}
pub struct NowPinningTimeFns64 {}
@ -22,143 +12,141 @@ 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> {
fn emit_now_mu<'ctx, 'a>(&self, ctx: &mut CodeGenContext<'ctx, 'a>) -> 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_bitcast(now, i32_type.ptr_type(AddressSpace::default()), "now.hi.addr")
.map(BasicValueEnum::into_pointer_value)
.unwrap();
let now_hiptr =
ctx.builder.build_bitcast(now, i32_type.ptr_type(AddressSpace::Generic), "now_hiptr");
if let BasicValueEnum::PointerValue(now_hiptr) = now_hiptr {
let now_loptr = unsafe {
ctx.builder.build_gep(now_hiptr, &[i32_type.const_int(2, false)], "now.lo.addr")
ctx.builder.build_gep(now_hiptr, &[i32_type.const_int(2, false)], "now_gep")
};
if let (BasicValueEnum::IntValue(now_hi), BasicValueEnum::IntValue(now_lo)) = (
ctx.builder.build_load(now_hiptr, "now_hi"),
ctx.builder.build_load(now_loptr, "now_lo"),
) {
let zext_hi = ctx.builder.build_int_z_extend(now_hi, i64_type, "now_zext_hi");
let shifted_hi = ctx.builder.build_left_shift(
zext_hi,
i64_type.const_int(32, false),
"now_shifted_zext_hi",
);
let zext_lo = ctx.builder.build_int_z_extend(now_lo, i64_type, "now_zext_lo");
ctx.builder.build_or(shifted_hi, zext_lo, "now_or").into()
} else {
unreachable!();
}
} else {
unreachable!();
}
.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>) {
fn emit_at_mu<'ctx, 'a>(&self, ctx: &mut CodeGenContext<'ctx, 'a>, 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(),
if let BasicValueEnum::IntValue(time) = t {
let time_hi = ctx.builder.build_int_truncate(
ctx.builder.build_right_shift(time, i64_32, false, "now_lshr"),
i32_type,
"",
)
.unwrap();
let time_lo = ctx.builder.build_int_truncate(time, i32_type, "time.lo").unwrap();
"now_trunc",
);
let time_lo = ctx.builder.build_int_truncate(time, i32_type, "now_trunc");
let now = ctx
.module
.get_global("now")
.unwrap_or_else(|| ctx.module.add_global(i64_type, None, "now"));
let now_hiptr = ctx
.builder
.build_bitcast(now, i32_type.ptr_type(AddressSpace::default()), "now.hi.addr")
.map(BasicValueEnum::into_pointer_value)
.unwrap();
let now_hiptr = ctx.builder.build_bitcast(
now,
i32_type.ptr_type(AddressSpace::Generic),
"now_bitcast",
);
if let BasicValueEnum::PointerValue(now_hiptr) = now_hiptr {
let now_loptr = unsafe {
ctx.builder.build_gep(now_hiptr, &[i32_type.const_int(2, false)], "now.lo.addr")
}
.unwrap();
ctx.builder.build_gep(now_hiptr, &[i32_type.const_int(2, false)], "now_gep")
};
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();
} else {
unreachable!();
}
} else {
unreachable!();
}
}
fn emit_delay_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>, dt: BasicValueEnum<'ctx>) {
fn emit_delay_mu<'ctx, 'a>(
&self,
ctx: &mut CodeGenContext<'ctx, 'a>,
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_bitcast(now, i32_type.ptr_type(AddressSpace::default()), "now.hi.addr")
.map(BasicValueEnum::into_pointer_value)
.unwrap();
let now_hiptr =
ctx.builder.build_bitcast(now, i32_type.ptr_type(AddressSpace::Generic), "now_hiptr");
if let BasicValueEnum::PointerValue(now_hiptr) = now_hiptr {
let now_loptr = unsafe {
ctx.builder.build_gep(now_hiptr, &[i32_type.const_int(2, false)], "now.lo.addr")
}
.unwrap();
ctx.builder.build_gep(now_hiptr, &[i32_type.const_int(2, false)], "now_loptr")
};
if let (
BasicValueEnum::IntValue(now_hi),
BasicValueEnum::IntValue(now_lo),
BasicValueEnum::IntValue(dt),
) = (
ctx.builder.build_load(now_hiptr, "now_hi"),
ctx.builder.build_load(now_loptr, "now_lo"),
dt,
) {
let zext_hi = ctx.builder.build_int_z_extend(now_hi, i64_type, "now_zext_hi");
let shifted_hi = ctx.builder.build_left_shift(
zext_hi,
i64_type.const_int(32, false),
"now_shifted_zext_hi",
);
let zext_lo = ctx.builder.build_int_z_extend(now_lo, i64_type, "now_zext_lo");
let now_val = ctx.builder.build_or(shifted_hi, zext_lo, "now_or");
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(),
let time = ctx.builder.build_int_add(now_val, dt, "now_add");
let time_hi = ctx.builder.build_int_truncate(
ctx.builder.build_right_shift(
time,
i64_type.const_int(32, false),
false,
"now_lshr",
),
i32_type,
"time.hi",
)
.unwrap();
let time_lo = ctx.builder.build_int_truncate(time, i32_type, "time.lo").unwrap();
"now_trunc",
);
let time_lo = ctx.builder.build_int_truncate(time, i32_type, "now_trunc");
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();
} else {
unreachable!();
}
} else {
unreachable!();
};
}
}
@ -167,115 +155,110 @@ 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> {
fn emit_now_mu<'ctx, 'a>(&self, ctx: &mut CodeGenContext<'ctx, 'a>) -> 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 now_raw = ctx.builder.build_load(now.as_pointer_value(), "now");
if let BasicValueEnum::IntValue(now_raw) = now_raw {
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()
let now_lo = ctx.builder.build_left_shift(now_raw, i64_32, "now_shl");
let now_hi = ctx.builder.build_right_shift(now_raw, i64_32, false, "now_lshr");
ctx.builder.build_or(now_lo, now_hi, "now_or").into()
} else {
unreachable!();
}
}
fn emit_at_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>, t: BasicValueEnum<'ctx>) {
fn emit_at_mu<'ctx, 'a>(&self, ctx: &mut CodeGenContext<'ctx, 'a>, 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_bitcast(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(),
if let BasicValueEnum::IntValue(time) = t {
let time_hi = ctx.builder.build_int_truncate(
ctx.builder.build_right_shift(time, i64_32, false, "now_lshr"),
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_bitcast(now, i32_type.ptr_type(AddressSpace::default()), "now.hi.addr")
.map(BasicValueEnum::into_pointer_value)
.unwrap();
);
let time_lo = ctx.builder.build_int_truncate(time, i32_type, "now_trunc");
let now = ctx
.module
.get_global("now")
.unwrap_or_else(|| ctx.module.add_global(i64_type, None, "now"));
let now_hiptr = ctx.builder.build_bitcast(
now,
i32_type.ptr_type(AddressSpace::Generic),
"now_bitcast",
);
if let BasicValueEnum::PointerValue(now_hiptr) = now_hiptr {
let now_loptr = unsafe {
ctx.builder.build_gep(now_hiptr, &[i32_type.const_int(1, false)], "now.lo.addr")
}
.unwrap();
ctx.builder.build_gep(now_hiptr, &[i32_type.const_int(1, false)], "now_gep")
};
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();
} else {
unreachable!();
}
} else {
unreachable!();
}
}
fn emit_delay_mu<'ctx, 'a>(
&self,
ctx: &mut CodeGenContext<'ctx, 'a>,
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(), "now");
if let (BasicValueEnum::IntValue(now_raw), BasicValueEnum::IntValue(dt)) = (now_raw, dt) {
let now_lo = ctx.builder.build_left_shift(now_raw, i64_32, "now_shl");
let now_hi = ctx.builder.build_right_shift(now_raw, i64_32, false, "now_lshr");
let now_val = ctx.builder.build_or(now_lo, now_hi, "now_or");
let time = ctx.builder.build_int_add(now_val, dt, "now_add");
let time_hi = ctx.builder.build_int_truncate(
ctx.builder.build_right_shift(time, i64_32, false, "now_lshr"),
i32_type,
"now_trunc",
);
let time_lo = ctx.builder.build_int_truncate(time, i32_type, "now_trunc");
let now_hiptr = ctx.builder.build_bitcast(
now,
i32_type.ptr_type(AddressSpace::Generic),
"now_bitcast",
);
if let BasicValueEnum::PointerValue(now_hiptr) = now_hiptr {
let now_loptr = unsafe {
ctx.builder.build_gep(now_hiptr, &[i32_type.const_int(1, false)], "now_gep")
};
ctx.builder
.build_store(now_hiptr, time_hi)
.set_atomic_ordering(AtomicOrdering::SequentiallyConsistent)
.unwrap();
ctx.builder
.build_store(now_loptr, time_lo)
.set_atomic_ordering(AtomicOrdering::SequentiallyConsistent)
.unwrap();
} else {
unreachable!();
}
} else {
unreachable!();
}
}
}
@ -284,18 +267,14 @@ 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> {
fn emit_now_mu<'ctx, 'a>(&self, ctx: &mut CodeGenContext<'ctx, 'a>) -> 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()
ctx.builder.build_call(now_mu, &[], "now_mu").try_as_basic_value().left().unwrap()
}
fn emit_at_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>, t: BasicValueEnum<'ctx>) {
fn emit_at_mu<'ctx, 'a>(&self, ctx: &mut CodeGenContext<'ctx, 'a>, t: BasicValueEnum<'ctx>) {
let at_mu = ctx.module.get_function("at_mu").unwrap_or_else(|| {
ctx.module.add_function(
"at_mu",
@ -303,10 +282,14 @@ impl TimeFns for ExternTimeFns {
None,
)
});
ctx.builder.build_call(at_mu, &[t.into()], "at_mu").unwrap();
ctx.builder.build_call(at_mu, &[t.into()], "at_mu");
}
fn emit_delay_mu<'ctx>(&self, ctx: &mut CodeGenContext<'ctx, '_>, dt: BasicValueEnum<'ctx>) {
fn emit_delay_mu<'ctx, 'a>(
&self,
ctx: &mut CodeGenContext<'ctx, 'a>,
dt: BasicValueEnum<'ctx>,
) {
let delay_mu = ctx.module.get_function("delay_mu").unwrap_or_else(|| {
ctx.module.add_function(
"delay_mu",
@ -314,7 +297,7 @@ impl TimeFns for ExternTimeFns {
None,
)
});
ctx.builder.build_call(delay_mu, &[dt.into()], "delay_mu").unwrap();
ctx.builder.build_call(delay_mu, &[dt.into()], "delay_mu");
}
}

10
nac3ast/Cargo.toml
View File

@ -2,7 +2,7 @@
name = "nac3ast"
version = "0.1.0"
authors = ["RustPython Team", "M-Labs"]
edition = "2021"
edition = "2018"
[features]
default = ["constant-optimization", "fold"]
@ -10,7 +10,7 @@ constant-optimization = ["fold"]
fold = []
[dependencies]
lazy_static = "1.5"
parking_lot = "0.12"
string-interner = "0.17"
fxhash = "0.2"
lazy_static = "1.4.0"
parking_lot = "0.11.1"
string-interner = "0.13.0"
fxhash = "0.2.1"

730
nac3ast/src/ast_gen.rs
View File

File diff suppressed because it is too large Load Diff

66
nac3ast/src/constant.rs
View File

@ -28,12 +28,12 @@ impl From<bool> for Constant {
}
impl From<i32> for Constant {
fn from(i: i32) -> Constant {
Self::Int(i128::from(i))
Self::Int(i as i128)
}
}
impl From<i64> for Constant {
fn from(i: i64) -> Constant {
Self::Int(i128::from(i))
Self::Int(i as i128)
}
}
@ -50,7 +50,6 @@ pub enum ConversionFlag {
}
impl ConversionFlag {
#[must_use]
pub fn try_from_byte(b: u8) -> Option<Self> {
match b {
b's' => Some(Self::Str),
@ -70,7 +69,6 @@ pub struct ConstantOptimizer {
#[cfg(feature = "constant-optimization")]
impl ConstantOptimizer {
#[inline]
#[must_use]
pub fn new() -> Self {
Self { _priv: () }
}
@ -87,10 +85,14 @@ impl<U> crate::fold::Fold<U> for ConstantOptimizer {
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 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 {
@ -98,11 +100,18 @@ impl<U> crate::fold::Fold<U> for ConstantOptimizer {
_ => unreachable!(),
})
.collect();
crate::ExprKind::Constant { value: Constant::Tuple(tuple), kind: None }
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 })
Ok(crate::Expr {
node: expr,
custom: node.custom,
location: node.location,
})
}
_ => crate::fold::fold_expr(self, node),
}
@ -118,7 +127,7 @@ mod tests {
use crate::fold::Fold;
use crate::*;
let location = Location::new(0, 0, FileName::default());
let location = Location::new(0, 0, Default::default());
let custom = ();
let ast = Located {
location,
@ -129,12 +138,18 @@ mod tests {
Located {
location,
custom,
node: ExprKind::Constant { value: 1.into(), kind: None },
node: ExprKind::Constant {
value: 1.into(),
kind: None,
},
},
Located {
location,
custom,
node: ExprKind::Constant { value: 2.into(), kind: None },
node: ExprKind::Constant {
value: 2.into(),
kind: None,
},
},
Located {
location,
@ -145,17 +160,26 @@ mod tests {
Located {
location,
custom,
node: ExprKind::Constant { value: 3.into(), kind: None },
node: ExprKind::Constant {
value: 3.into(),
kind: None,
},
},
Located {
location,
custom,
node: ExprKind::Constant { value: 4.into(), kind: None },
node: ExprKind::Constant {
value: 4.into(),
kind: None,
},
},
Located {
location,
custom,
node: ExprKind::Constant { value: 5.into(), kind: None },
node: ExprKind::Constant {
value: 5.into(),
kind: None,
},
},
],
},
@ -163,7 +187,9 @@ mod tests {
],
},
};
let new_ast = ConstantOptimizer::new().fold_expr(ast).unwrap_or_else(|e| match e {});
let new_ast = ConstantOptimizer::new()
.fold_expr(ast)
.unwrap_or_else(|e| match e {});
assert_eq!(
new_ast,
Located {
@ -173,7 +199,11 @@ mod tests {
value: Constant::Tuple(vec![
1.into(),
2.into(),
Constant::Tuple(vec![3.into(), 4.into(), 5.into(),])
Constant::Tuple(vec![
3.into(),
4.into(),
5.into(),
])
]),
kind: None
},

9
nac3ast/src/fold_helpers.rs
View File

@ -64,4 +64,11 @@ macro_rules! simple_fold {
};
}
simple_fold!(usize, String, bool, StrRef, constant::Constant, constant::ConversionFlag);
simple_fold!(
usize,
String,
bool,
StrRef,
constant::Constant,
constant::ConversionFlag
);

6
nac3ast/src/impls.rs
View File

@ -2,7 +2,6 @@ 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 { .. } => {
@ -35,7 +34,10 @@ impl<U> ExprKind<U> {
ExprKind::Starred { .. } => "starred",
ExprKind::Slice { .. } => "slice",
ExprKind::JoinedStr { values } => {
if values.iter().any(|e| matches!(e.node, ExprKind::JoinedStr { .. })) {
if values
.iter()
.any(|e| matches!(e.node, ExprKind::JoinedStr { .. }))
{
"f-string expression"
} else {
"literal"

18
nac3ast/src/lib.rs
View File

@ -1,19 +1,3 @@
#![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;
@ -25,6 +9,6 @@ mod impls;
mod location;
pub use ast_gen::*;
pub use location::{FileName, Location};
pub use location::{Location, FileName};
pub type Suite<U = ()> = Vec<Stmt<U>>;

40
nac3ast/src/location.rs
View File

@ -1,9 +1,8 @@
//! Datatypes to support source location information.
use crate::ast_gen::StrRef;
use std::cmp::Ordering;
use std::fmt;
#[derive(Clone, Copy, Debug, Eq, PartialEq)]
#[derive(Clone, Copy, Debug, PartialEq)]
pub struct FileName(pub StrRef);
impl Default for FileName {
fn default() -> Self {
@ -18,38 +17,16 @@ impl From<String> for FileName {
}
/// A location somewhere in the sourcecode.
#[derive(Clone, Copy, Debug, Default, Eq, PartialEq)]
#[derive(Clone, Copy, Debug, Default, PartialEq)]
pub struct Location {
pub row: usize,
pub column: usize,
pub file: FileName,
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))
write!(f, "{}: line {} column {}", self.file.0, self.row, self.column)
}
}
@ -76,22 +53,23 @@ impl Location {
)
}
}
Visualize { loc: *self, line, desc }
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
}

24
nac3core/Cargo.toml
View File

@ -1,31 +1,25 @@
[features]
test = []
[package]
name = "nac3core"
version = "0.1.0"
authors = ["M-Labs"]
edition = "2021"
edition = "2018"
[dependencies]
itertools = "0.13"
crossbeam = "0.8"
indexmap = "2.2"
parking_lot = "0.12"
rayon = "1.8"
itertools = "0.10.1"
crossbeam = "0.8.1"
parking_lot = "0.11.1"
rayon = "1.5.1"
nac3parser = { path = "../nac3parser" }
strum = "0.26.2"
strum_macros = "0.26.4"
[dependencies.inkwell]
version = "0.4"
version = "0.1.0-beta.4"
default-features = false
features = ["llvm14-0", "target-x86", "target-arm", "target-riscv", "no-libffi-linking"]
features = ["llvm13-0", "target-x86", "target-arm", "target-riscv", "no-libffi-linking"]
[dev-dependencies]
test-case = "1.2.0"
indoc = "2.0"
indoc = "1.0"
insta = "=1.11.0"
[build-dependencies]
regex = "1.10"
regex = "1"

98
nac3core/build.rs
View File

@ -7,41 +7,30 @@ use std::{
process::{Command, Stdio},
};
fn compile_irrt(irrt_dir: &Path, out_dir: &Path) {
let irrt_cpp_path = irrt_dir.join("irrt.cpp");
fn main() {
const FILE: &str = "src/codegen/irrt/irrt.c";
println!("cargo:rerun-if-changed={}", FILE);
let out_dir = env::var("OUT_DIR").unwrap();
let out_path = Path::new(&out_dir);
/*
* 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 flags: &[&str] = &[
const FLAG: &[&str] = &[
"--target=wasm32",
irrt_cpp_path.to_str().unwrap(),
"-x",
"c++",
"-fno-discard-value-names",
"-fno-exceptions",
"-fno-rtti",
match env::var("PROFILE").as_deref() {
Ok("debug") => "-O0",
Ok("release") => "-O3",
flavor => panic!("Unknown or missing build flavor {flavor:?}"),
},
FILE,
"-O3",
"-emit-llvm",
"-S",
"-Wall",
"-Wextra",
"-Werror=return-type",
"-I",
irrt_dir.to_str().unwrap(),
"-o",
"-",
];
println!("cargo:rerun-if-changed={}", out_dir.to_str().unwrap());
let output = Command::new("clang-irrt")
.args(flags)
let output = Command::new("clang")
.args(FLAG)
.output()
.map(|o| {
assert!(o.status.success(), "{}", std::str::from_utf8(&o.stderr).unwrap());
@ -53,13 +42,9 @@ fn compile_irrt(irrt_dir: &Path, out_dir: &Path) {
let output = std::str::from_utf8(&output.stdout).unwrap().replace("\r\n", "\n");
let mut filtered_output = String::with_capacity(output.len());
// (?ms:^define.*?\}$) to capture `define` blocks
// (?m:^declare.*?$) to capture `declare` blocks
// (?m:^%.+?=\s*type\s*\{.+?\}$) to capture `type` declarations
let regex_filter =
Regex::new(r"(?ms:^define.*?\}$)|(?m:^declare.*?$)|(?m:^%.+?=\s*type\s*\{.+?\}$)").unwrap();
let regex_filter = regex::Regex::new(r"(?ms:^define.*?\}$)|(?m:^declare.*?$)").unwrap();
for f in regex_filter.captures_iter(&output) {
assert_eq!(f.len(), 1);
assert!(f.len() == 1);
filtered_output.push_str(&f[0]);
filtered_output.push('\n');
}
@ -70,65 +55,18 @@ fn compile_irrt(irrt_dir: &Path, out_dir: &Path) {
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();
let mut file = File::create(out_path.join("irrt.ll")).unwrap();
file.write_all(output.as_bytes()).unwrap();
let mut file = File::create(out_dir.join("irrt-filtered.ll")).unwrap();
let mut file = File::create(out_path.join("irrt-filtered.ll")).unwrap();
file.write_all(filtered_output.as_bytes()).unwrap();
}
let mut llvm_as = Command::new("llvm-as-irrt")
let mut llvm_as = Command::new("llvm-as")
.stdin(Stdio::piped())
.arg("-o")
.arg(out_dir.join("irrt.bc"))
.arg(out_path.join("irrt.bc"))
.spawn()
.unwrap();
llvm_as.stdin.as_mut().unwrap().write_all(filtered_output.as_bytes()).unwrap();
assert!(llvm_as.wait().unwrap().success());
}
fn compile_irrt_test(irrt_dir: &Path, out_dir: &Path) {
let irrt_test_cpp_path = irrt_dir.join("irrt_test.cpp");
let exe_path = out_dir.join("irrt_test.out");
let flags: &[&str] = &[
irrt_test_cpp_path.to_str().unwrap(),
"-x",
"c++",
"-I",
irrt_dir.to_str().unwrap(),
"-g",
"-fno-discard-value-names",
"-O0",
"-Wall",
"-Wextra",
"-Werror=return-type",
"-lm", // for `tgamma()`, `lgamma()`
"-o",
exe_path.to_str().unwrap(),
];
Command::new("clang-irrt-test")
.args(flags)
.output()
.map(|o| {
assert!(o.status.success(), "{}", std::str::from_utf8(&o.stderr).unwrap());
o
})
.unwrap();
println!("cargo:rerun-if-changed={}", out_dir.to_str().unwrap());
}
fn main() {
let out_dir = env::var("OUT_DIR").unwrap();
let out_dir = Path::new(&out_dir);
let irrt_dir = Path::new("./irrt");
compile_irrt(irrt_dir, out_dir);
// https://github.com/rust-lang/cargo/issues/2549
// `cargo test -F test` to also build `irrt_test.cpp
if cfg!(feature = "test") {
compile_irrt_test(irrt_dir, out_dir);
}
assert!(llvm_as.wait().unwrap().success())
}

5
nac3core/irrt/irrt.cpp
View File

@ -1,5 +0,0 @@
#include "irrt_everything.hpp"
/*
This file will be read by `clang-irrt` to conveniently produce LLVM IR for `nac3core/codegen`.
*/

437
nac3core/irrt/irrt.hpp
View File

@ -1,437 +0,0 @@
#ifndef IRRT_DONT_TYPEDEF_INTS
typedef _BitInt(8) int8_t;
typedef unsigned _BitInt(8) uint8_t;
typedef _BitInt(32) int32_t;
typedef unsigned _BitInt(32) uint32_t;
typedef _BitInt(64) int64_t;
typedef unsigned _BitInt(64) uint64_t;
#endif
// NDArray indices are always `uint32_t`.
typedef uint32_t NDIndex;
// The type of an index or a value describing the length of a range/slice is
// always `int32_t`.
typedef int32_t SliceIndex;
template <typename T>
static T max(T a, T b) {
return a > b ? a : b;
}
template <typename T>
static T min(T a, T b) {
return a > b ? b : a;
}
// 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>
static 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;
}
template <typename SizeT>
static 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>
static 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>
static 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>
static 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>
static 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];
}
}
template<typename SizeT>
static void __nac3_ndarray_strides_from_shape_impl(
SizeT ndims,
SizeT *shape,
SizeT *dst_strides
) {
SizeT stride_product = 1;
for (SizeT i = 0; i < ndims; i++) {
int dim_i = ndims - i - 1;
dst_strides[dim_i] = stride_product;
stride_product *= shape[dim_i];
}
}
extern "C" {
#define DEF_nac3_int_exp_(T) \
T __nac3_int_exp_##T(T base, T exp) {\
return __nac3_int_exp_impl(base, exp);\
}
DEF_nac3_int_exp_(int32_t)
DEF_nac3_int_exp_(int64_t)
DEF_nac3_int_exp_(uint32_t)
DEF_nac3_int_exp_(uint64_t)
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;
}
}
// 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;
}
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);
}
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);
}
void __nac3_ndarray_strides_from_shape(uint32_t ndims, uint32_t* shape, uint32_t* dst_strides) {
__nac3_ndarray_strides_from_shape_impl(ndims, shape, dst_strides);
}
void __nac3_ndarray_strides_from_shape64(uint64_t ndims, uint64_t* shape, uint64_t* dst_strides) {
__nac3_ndarray_strides_from_shape_impl(ndims, shape, dst_strides);
}
}

216
nac3core/irrt/irrt_basic.hpp
View File

@ -1,216 +0,0 @@
#pragma once
#include "irrt_utils.hpp"
#include "irrt_typedefs.hpp"
/*
This header contains IRRT implementations
that do not deserved to be categorized (e.g., into numpy, etc.)
Check out other *.hpp files before including them here!!
*/
// The type of an index or a value describing the length of a range/slice is
// always `int32_t`.
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;
}
}
extern "C" {
#define DEF_nac3_int_exp_(T) \
T __nac3_int_exp_##T(T base, T exp) {\
return __nac3_int_exp_impl(base, exp);\
}
DEF_nac3_int_exp_(int32_t)
DEF_nac3_int_exp_(int64_t)
DEF_nac3_int_exp_(uint32_t)
DEF_nac3_int_exp_(uint64_t)
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;
}
}
// 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;
}
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);
}
}

14
nac3core/irrt/irrt_everything.hpp
View File

@ -1,14 +0,0 @@
#pragma once
#include "irrt_utils.hpp"
#include "irrt_typedefs.hpp"
#include "irrt_basic.hpp"
#include "irrt_slice.hpp"
#include "irrt_numpy_ndarray.hpp"
/*
All IRRT implementations.
We don't have any pre-compiled objects, so we are writing all implementations in headers and
concatenate them with `#include` into one massive source file that contains all the IRRT stuff.
*/

466
nac3core/irrt/irrt_numpy_ndarray.hpp
View File

@ -1,466 +0,0 @@
#pragma once
#include "irrt_utils.hpp"
#include "irrt_typedefs.hpp"
#include "irrt_slice.hpp"
/*
NDArray-related implementations.
`*/
// NDArray indices are always `uint32_t`.
using NDIndex = uint32_t;
namespace {
namespace ndarray_util {
template <typename SizeT>
static void set_indices_by_nth(SizeT ndims, const SizeT* shape, SizeT* indices, SizeT nth) {
for (int32_t i = 0; i < ndims; i++) {
int32_t dim_i = ndims - i - 1;
int32_t dim = shape[dim_i];
indices[dim_i] = nth % dim;
nth /= dim;
}
}
// Compute the strides of an ndarray given an ndarray `shape`
// and assuming that the ndarray is *fully C-contagious*.
//
// You might want to read up on https://ajcr.net/stride-guide-part-1/.
template <typename SizeT>
static void set_strides_by_shape(SizeT itemsize, SizeT ndims, SizeT* dst_strides, const SizeT* shape) {
SizeT stride_product = 1;
for (SizeT i = 0; i < ndims; i++) {
int dim_i = ndims - i - 1;
dst_strides[dim_i] = stride_product * itemsize;
stride_product *= shape[dim_i];
}
}
// Compute the size/# of elements of an ndarray given its shape
template <typename SizeT>
static SizeT calc_size_from_shape(SizeT ndims, const SizeT* shape) {
SizeT size = 1;
for (SizeT dim_i = 0; dim_i < ndims; dim_i++) size *= shape[dim_i];
return size;
}
template <typename SizeT>
static bool can_broadcast_shape_to(
const SizeT target_ndims,
const SizeT *target_shape,
const SizeT src_ndims,
const SizeT *src_shape
) {
/*
// See https://numpy.org/doc/stable/user/basics.broadcasting.html
This function handles this example:
```
Image (3d array): 256 x 256 x 3
Scale (1d array): 3
Result (3d array): 256 x 256 x 3
```
Other interesting examples to consider:
- `can_broadcast_shape_to([3], [1, 1, 1, 1, 3]) == true`
- `can_broadcast_shape_to([3], [3, 1]) == false`
- `can_broadcast_shape_to([256, 256, 3], [256, 1, 3]) == true`
In cases when the shapes contain zero(es):
- `can_broadcast_shape_to([0], [1]) == true`
- `can_broadcast_shape_to([0], [2]) == false`
- `can_broadcast_shape_to([0, 4, 0, 0], [1]) == true`
- `can_broadcast_shape_to([0, 4, 0, 0], [1, 1, 1, 1]) == true`
- `can_broadcast_shape_to([0, 4, 0, 0], [1, 4, 1, 1]) == true`
- `can_broadcast_shape_to([4, 3], [0, 3]) == false`
- `can_broadcast_shape_to([4, 3], [0, 0]) == false`
*/
// This is essentially doing the following in Python:
// `for target_dim, src_dim in itertools.zip_longest(target_shape[::-1], src_shape[::-1], fillvalue=1)`
for (SizeT i = 0; i < max(target_ndims, src_ndims); i++) {
SizeT target_dim_i = target_ndims - i - 1;
SizeT src_dim_i = src_ndims - i - 1;
bool target_dim_exists = target_dim_i >= 0;
bool src_dim_exists = src_dim_i >= 0;
SizeT target_dim = target_dim_exists ? target_shape[target_dim_i] : 1;
SizeT src_dim = src_dim_exists ? src_shape[src_dim_i] : 1;
bool ok = src_dim == 1 || target_dim == src_dim;
if (!ok) return false;
}
return true;
}
}
typedef uint8_t NDSliceType;
extern "C" {
const NDSliceType INPUT_SLICE_TYPE_INDEX = 0;
const NDSliceType INPUT_SLICE_TYPE_SLICE = 1;
}
struct NDSlice {
// A poor-man's `std::variant<int, UserRange>`
NDSliceType type;
/*
if type == INPUT_SLICE_TYPE_INDEX => `slice` points to a single `SizeT`
if type == INPUT_SLICE_TYPE_SLICE => `slice` points to a single `UserRange`
*/
uint8_t *slice;
};
namespace ndarray_util {
template<typename SizeT>
SizeT deduce_ndims_after_slicing(SizeT ndims, SizeT num_slices, const NDSlice *slices) {
irrt_assert(num_slices <= ndims);
SizeT final_ndims = ndims;
for (SizeT i = 0; i < num_slices; i++) {
if (slices[i].type == INPUT_SLICE_TYPE_INDEX) {
final_ndims--; // An integer slice demotes the rank by 1
}
}
return final_ndims;
}
}
template <typename SizeT>
struct NDArrayIndicesIter {
SizeT ndims;
const SizeT *shape;
SizeT *indices;
void set_indices_zero() {
__builtin_memset(indices, 0, sizeof(SizeT) * ndims);
}
void next() {
for (SizeT i = 0; i < ndims; i++) {
SizeT dim_i = ndims - i - 1;
indices[dim_i]++;
if (indices[dim_i] < shape[dim_i]) {
break;
} else {
indices[dim_i] = 0;
}
}
}
};
// The NDArray object. `SizeT` is the *signed* size type of this ndarray.
//
// NOTE: The order of fields is IMPORTANT. DON'T TOUCH IT
//
// Some resources you might find helpful:
// - The official numpy implementations:
// - https://github.com/numpy/numpy/blob/735a477f0bc2b5b84d0e72d92f224bde78d4e069/doc/source/reference/c-api/types-and-structures.rst
// - On strides (about reshaping, slicing, C-contagiousness, etc)
// - https://ajcr.net/stride-guide-part-1/.
// - https://ajcr.net/stride-guide-part-2/.
// - https://ajcr.net/stride-guide-part-3/.
template <typename SizeT>
struct NDArray {
// The underlying data this `ndarray` is pointing to.
//
// NOTE: Formally this should be of type `void *`, but clang
// translates `void *` to `i8 *` when run with `-S -emit-llvm`,
// so we will put `uint8_t *` here for clarity.
uint8_t *data;
// The number of bytes of a single element in `data`.
//
// The `SizeT` is treated as `unsigned`.
SizeT itemsize;
// The number of dimensions of this shape.
//
// The `SizeT` is treated as `unsigned`.
SizeT ndims;
// Array shape, with length equal to `ndims`.
//
// The `SizeT` is treated as `unsigned`.
//
// NOTE: `shape` can contain 0.
// (those appear when the user makes an out of bounds slice into an ndarray, e.g., `np.zeros((3, 3))[400:].shape == (0, 3)`)
SizeT *shape;
// Array strides (stride value is in number of bytes, NOT number of elements), with length equal to `ndims`.
//
// The `SizeT` is treated as `signed`.
//
// NOTE: `strides` can have negative numbers.
// (those appear when there is a slice with a negative step, e.g., `my_array[::-1]`)
SizeT *strides;
// Calculate the size/# of elements of an `ndarray`.
// This function corresponds to `np.size(<ndarray>)` or `ndarray.size`
SizeT size() {
return ndarray_util::calc_size_from_shape(ndims, shape);
}
// Calculate the number of bytes of its content of an `ndarray` *in its view*.
// This function corresponds to `ndarray.nbytes`
SizeT nbytes() {
return this->size() * itemsize;
}
void set_value_at_pelement(uint8_t* pelement, const uint8_t* pvalue) {
__builtin_memcpy(pelement, pvalue, itemsize);
}
uint8_t* get_pelement(const SizeT *indices) {
uint8_t* element = data;
for (SizeT dim_i = 0; dim_i < ndims; dim_i++)
element += indices[dim_i] * strides[dim_i];
return element;
}
uint8_t* get_nth_pelement(SizeT nth) {
irrt_assert(0 <= nth);
irrt_assert(nth < this->size());
SizeT* indices = (SizeT*) __builtin_alloca(sizeof(SizeT) * this->ndims);
ndarray_util::set_indices_by_nth(this->ndims, this->shape, indices, nth);
return get_pelement(indices);
}
// Get pointer to the first element of this ndarray, assuming
// `this->size() > 0`, i.e., not "degenerate" due to zeroes in `this->shape`)
//
// This is particularly useful for when the ndarray is just containing a single scalar.
uint8_t* get_first_pelement() {
irrt_assert(this->size() > 0);
return this->data; // ...It is simply `this->data`
}
// Is the given `indices` valid/in-bounds?
bool in_bounds(const SizeT *indices) {
for (SizeT dim_i = 0; dim_i < ndims; dim_i++) {
bool dim_ok = indices[dim_i] < shape[dim_i];
if (!dim_ok) return false;
}
return true;
}
// Fill the ndarray with a value
void fill_generic(const uint8_t* pvalue) {
NDArrayIndicesIter<SizeT> iter;
iter.ndims = this->ndims;
iter.shape = this->shape;
iter.indices = (SizeT*) __builtin_alloca(sizeof(SizeT) * ndims);
iter.set_indices_zero();
for (SizeT i = 0; i < this->size(); i++, iter.next()) {
uint8_t* pelement = get_pelement(iter.indices);
set_value_at_pelement(pelement, pvalue);
}
}
// Set the strides of the ndarray with `ndarray_util::set_strides_by_shape`
void set_strides_by_shape() {
ndarray_util::set_strides_by_shape(itemsize, ndims, strides, shape);
}
// https://numpy.org/doc/stable/reference/generated/numpy.eye.html
void set_to_eye(SizeT k, const uint8_t* zero_pvalue, const uint8_t* one_pvalue) {
__builtin_assume(ndims == 2);
// TODO: Better implementation
fill_generic(zero_pvalue);
for (SizeT i = 0; i < min(shape[0], shape[1]); i++) {
SizeT row = i;
SizeT col = i + k;
SizeT indices[2] = { row, col };
if (!in_bounds(indices)) continue;
uint8_t* pelement = get_pelement(indices);
set_value_at_pelement(pelement, one_pvalue);
}
}
// To support numpy complex slices (e.g., `my_array[:50:2,4,:2:-1]`)
//
// Things assumed by this function:
// - `dst_ndarray` is allocated by the caller
// - `dst_ndarray.ndims` has the correct value (according to `ndarray_util::deduce_ndims_after_slicing`).
// - ... and `dst_ndarray.shape` and `dst_ndarray.strides` have been allocated by the caller as well
//
// Other notes:
// - `dst_ndarray->data` does not have to be set, it will be derived.
// - `dst_ndarray->itemsize` does not have to be set, it will be set to `this->itemsize`
// - `dst_ndarray->shape` and `dst_ndarray.strides` can contain empty values
void slice(SizeT num_ndslices, NDSlice* ndslices, NDArray<SizeT>* dst_ndarray) {
// REFERENCE CODE (check out `_index_helper` in `__getitem__`):
// https://github.com/wadetb/tinynumpy/blob/0d23d22e07062ffab2afa287374c7b366eebdda1/tinynumpy/tinynumpy.py#L652
irrt_assert(dst_ndarray->ndims == ndarray_util::deduce_ndims_after_slicing(this->ndims, num_ndslices, ndslices));
dst_ndarray->data = this->data;
SizeT this_axis = 0;
SizeT dst_axis = 0;
for (SizeT i = 0; i < num_ndslices; i++) {
NDSlice *ndslice = &ndslices[i];
if (ndslice->type == INPUT_SLICE_TYPE_INDEX) {
// Handle when the ndslice is just a single (possibly negative) integer
// e.g., `my_array[::2, -5, ::-1]`
// ^^------ like this
SizeT index_user = *((SizeT*) ndslice->slice);
SizeT index = resolve_index_in_length(this->shape[this_axis], index_user);
dst_ndarray->data += index * this->strides[this_axis]; // Add offset
// Next
this_axis++;
} else if (ndslice->type == INPUT_SLICE_TYPE_SLICE) {
// Handle when the ndslice is a slice (represented by UserSlice in IRRT)
// e.g., `my_array[::2, -5, ::-1]`
// ^^^------^^^^----- like these
UserSlice<SizeT>* user_slice = (UserSlice<SizeT>*) ndslice->slice;
Slice<SizeT> slice = user_slice->indices(this->shape[this_axis]); // To resolve negative indices and other funny stuff written by the user
// NOTE: There is no need to write special code to handle negative steps/strides.
// This simple implementation meticulously handles both positive and negative steps/strides.
// Check out the tinynumpy and IRRT's test cases if you are not convinced.
dst_ndarray->data += slice.start * this->strides[this_axis]; // Add offset (NOTE: no need to `* itemsize`, strides count in # of bytes)
dst_ndarray->strides[dst_axis] = slice.step * this->strides[this_axis]; // Determine stride
dst_ndarray->shape[dst_axis] = slice.len(); // Determine shape dimension
// Next
dst_axis++;
this_axis++;
} else {
__builtin_unreachable();
}
}
irrt_assert(dst_axis == dst_ndarray->ndims); // Sanity check on the implementation
}
// Similar to `np.broadcast_to(<ndarray>, <target_shape>)`
// Assumptions:
// - `this` has to be fully initialized.
// - `dst_ndarray->ndims` has to be set.
// - `dst_ndarray->shape` has to be set, this determines the shape `this` broadcasts to.
//
// Other notes:
// - `dst_ndarray->data` does not have to be set, it will be set to `this->data`.
// - `dst_ndarray->itemsize` does not have to be set, it will be set to `this->data`.
// - `dst_ndarray->strides` does not have to be set, it will be overwritten.
//
// Cautions:
// ```
// xs = np.zeros((4,))
// ys = np.zero((4, 1))
// ys[:] = xs # ok
//
// xs = np.zeros((1, 4))
// ys = np.zero((4,))
// ys[:] = xs # allowed
// # However `np.broadcast_to(xs, (4,))` would fails, as per numpy's broadcasting rule.
// # and apparently numpy will "deprecate" this? SEE https://github.com/numpy/numpy/issues/21744
// # This implementation will NOT support this assignment.
// ```
void broadcast_to(NDArray<SizeT>* dst_ndarray) {
dst_ndarray->data = this->data;
dst_ndarray->itemsize = this->itemsize;
irrt_assert(
ndarray_util::can_broadcast_shape_to(
dst_ndarray->ndims,
dst_ndarray->shape,
this->ndims,
this->shape
)
);
SizeT stride_product = 1;
for (SizeT i = 0; i < max(this->ndims, dst_ndarray->ndims); i++) {
SizeT this_dim_i = this->ndims - i - 1;
SizeT dst_dim_i = dst_ndarray->ndims - i - 1;
bool this_dim_exists = this_dim_i >= 0;
bool dst_dim_exists = dst_dim_i >= 0;
// TODO: Explain how this works
bool c1 = this_dim_exists && this->shape[this_dim_i] == 1;
bool c2 = dst_dim_exists && dst_ndarray->shape[dst_dim_i] != 1;
if (!this_dim_exists || (c1 && c2)) {
dst_ndarray->strides[dst_dim_i] = 0; // Freeze it in-place
} else {
dst_ndarray->strides[dst_dim_i] = stride_product * this->itemsize;
stride_product *= this->shape[this_dim_i]; // NOTE: this_dim_exist must be true here.
}
}
}
// Simulates `this_ndarray[:] = src_ndarray`, with automatic broadcasting.
// Caution on https://github.com/numpy/numpy/issues/21744
// Also see `NDArray::broadcast_to`
void assign_with(NDArray<SizeT>* src_ndarray) {
irrt_assert(
ndarray_util::can_broadcast_shape_to(
this->ndims,
this->shape,
src_ndarray->ndims,
src_ndarray->shape
)
);
// Broadcast the `src_ndarray` to make the reading process *much* easier
SizeT* broadcasted_src_ndarray_strides = __builtin_alloca(sizeof(SizeT) * this->ndims); // Remember to allocate strides beforehand
NDArray<SizeT> broadcasted_src_ndarray = {
.ndims = this->ndims,
.shape = this->shape,
.strides = broadcasted_src_ndarray_strides
};
src_ndarray->broadcast_to(&broadcasted_src_ndarray);
// Using iter instead of `get_nth_pelement` because it is slightly faster
SizeT* indices = __builtin_alloca(sizeof(SizeT) * this->ndims);
auto iter = NDArrayIndicesIter<SizeT> {
.ndims = this->ndims,
.shape = this->shape,
.indices = indices
};
const SizeT this_size = this->size();
for (SizeT i = 0; i < this_size; i++, iter.next()) {
uint8_t* src_pelement = broadcasted_src_ndarray_strides->get_pelement(indices);
uint8_t* this_pelement = this->get_pelement(indices);
this->set_value_at_pelement(src_pelement, src_pelement);
}
}
};
}
extern "C" {
uint32_t __nac3_ndarray_size(NDArray<int32_t>* ndarray) {
return ndarray->size();
}
uint64_t __nac3_ndarray_size64(NDArray<int64_t>* ndarray) {
return ndarray->size();
}
void __nac3_ndarray_fill_generic(NDArray<int32_t>* ndarray, uint8_t* pvalue) {
ndarray->fill_generic(pvalue);
}
void __nac3_ndarray_fill_generic64(NDArray<int64_t>* ndarray, uint8_t* pvalue) {
ndarray->fill_generic(pvalue);
}
// void __nac3_ndarray_slice(NDArray<int32_t>* ndarray, int32_t num_slices, NDSlice<int32_t> *slices, NDArray<int32_t> *dst_ndarray) {
// // ndarray->slice(num_slices, slices, dst_ndarray);
// }
}

80
nac3core/irrt/irrt_slice.hpp
View File

@ -1,80 +0,0 @@
#pragma once
#include "irrt_utils.hpp"
#include "irrt_typedefs.hpp"
namespace {
// A proper slice in IRRT, all negative indices have be resolved to absolute values.
// Even though nac3core's slices are always `int32_t`, we will template slice anyway
// since this struct is used as a general utility.
template <typename T>
struct Slice {
T start;
T stop;
T step;
// The length/The number of elements of the slice if it were a range,
// i.e., the value of `len(range(this->start, this->stop, this->end))`
T len() {
T diff = stop - 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;
}
}
};
template<typename T>
T resolve_index_in_length(T length, T index) {
irrt_assert(length >= 0);
if (index < 0) {
// Remember that index is negative, so do a plus here
return max(length + index, 0);
} else {
return min(length, index);
}
}
// NOTE: using a bitfield for the `*_defined` is better, at the
// cost of a more annoying implementation in nac3core inkwell
template <typename T>
struct UserSlice {
uint8_t start_defined;
T start;
uint8_t stop_defined;
T stop;
uint8_t step_defined;
T step;
// Like Python's `slice(start, stop, step).indices(length)`
Slice<T> indices(T length) {
// NOTE: This function implements Python's `slice.indices` *FAITHFULLY*.
// SEE: https://github.com/python/cpython/blob/f62161837e68c1c77961435f1b954412dd5c2b65/Objects/sliceobject.c#L546
irrt_assert(length >= 0);
irrt_assert(!step_defined || step != 0); // step_defined -> step != 0; step cannot be zero if specified by user
Slice<T> result;
result.step = step_defined ? step : 1;
bool step_is_negative = result.step < 0;
if (start_defined) {
result.start = resolve_index_in_length(length, start);
} else {
result.start = step_is_negative ? length - 1 : 0;
}
if (stop_defined) {
result.stop = resolve_index_in_length(length, stop);
} else {
result.stop = step_is_negative ? -1 : length;
}
return result;
}
};
}

658
nac3core/irrt/irrt_test.cpp
View File

@ -1,658 +0,0 @@
// This file will be compiled like a real C++ program,
// and we do have the luxury to use the standard libraries.
// That is if the nix flakes do not have issues... especially on msys2...
#include <cstdint>
#include <cstdio>
#include <cstdlib>
// Set `IRRT_DONT_TYPEDEF_INTS` because `cstdint` defines them
#define IRRT_DONT_TYPEDEF_INTS
#include "irrt_everything.hpp"
void test_fail() {
printf("[!] Test failed\n");
exit(1);
}
void __begin_test(const char* function_name, const char* file, int line) {
printf("######### Running %s @ %s:%d\n", function_name, file, line);
}
#define BEGIN_TEST() __begin_test(__FUNCTION__, __FILE__, __LINE__)
template <typename T>
void debug_print_array(const char* format, int len, T* as) {
printf("[");
for (int i = 0; i < len; i++) {
if (i != 0) printf(", ");
printf(format, as[i]);
}
printf("]");
}
template <typename T>
void assert_arrays_match(const char* label, const char* format, int len, T* expected, T* got) {
if (!arrays_match(len, expected, got)) {
printf(">>>>>>> %s\n", label);
printf(" Expecting = ");
debug_print_array(format, len, expected);
printf("\n");
printf(" Got = ");
debug_print_array(format, len, got);
printf("\n");
test_fail();
}
}
template <typename T>
void assert_values_match(const char* label, const char* format, T expected, T got) {
if (expected != got) {
printf(">>>>>>> %s\n", label);
printf(" Expecting = ");
printf(format, expected);
printf("\n");
printf(" Got = ");
printf(format, got);
printf("\n");
test_fail();
}
}
void print_repeated(const char *str, int count) {
for (int i = 0; i < count; i++) {
printf("%s", str);
}
}
template<typename SizeT, typename ElementT>
void __print_ndarray_aux(const char *format, bool first, bool last, SizeT* cursor, SizeT depth, NDArray<SizeT>* ndarray) {
// A really lazy recursive implementation
// Add left padding unless its the first entry (since there would be "[[[" before it)
if (!first) {
print_repeated(" ", depth);
}
const SizeT dim = ndarray->shape[depth];
if (depth + 1 == ndarray->ndims) {
// Recursed down to last dimension, print the values in a nice list
printf("[");
SizeT* indices = (SizeT*) __builtin_alloca(sizeof(SizeT) * ndarray->ndims);
for (SizeT i = 0; i < dim; i++) {
ndarray_util::set_indices_by_nth(ndarray->ndims, ndarray->shape, indices, *cursor);
ElementT* pelement = (ElementT*) ndarray->get_pelement(indices);
ElementT element = *pelement;
if (i != 0) printf(", "); // List delimiter
printf(format, element);
printf("(@");
debug_print_array("%d", ndarray->ndims, indices);
printf(")");
(*cursor)++;
}
printf("]");
} else {
printf("[");
for (SizeT i = 0; i < ndarray->shape[depth]; i++) {
__print_ndarray_aux<SizeT, ElementT>(
format,
i == 0, // first?
i + 1 == dim, // last?
cursor,
depth + 1,
ndarray
);
}
printf("]");
}
// Add newline unless its the last entry (since there will be "]]]" after it)
if (!last) {
print_repeated("\n", depth);
}
}
template<typename SizeT, typename ElementT>
void print_ndarray(const char *format, NDArray<SizeT>* ndarray) {
if (ndarray->ndims == 0) {
printf("<empty ndarray>");
} else {
SizeT cursor = 0;
__print_ndarray_aux<SizeT, ElementT>(format, true, true, &cursor, 0, ndarray);
}
printf("\n");
}
void test_calc_size_from_shape_normal() {
// Test shapes with normal values
BEGIN_TEST();
int32_t shape[4] = { 2, 3, 5, 7 };
assert_values_match("size", "%d", 210, ndarray_util::calc_size_from_shape<int32_t>(4, shape));
}
void test_calc_size_from_shape_has_zero() {
// Test shapes with 0 in them
BEGIN_TEST();
int32_t shape[4] = { 2, 0, 5, 7 };
assert_values_match("size", "%d", 0, ndarray_util::calc_size_from_shape<int32_t>(4, shape));
}
void test_set_strides_by_shape() {
// Test `set_strides_by_shape()`
BEGIN_TEST();
int32_t shape[4] = { 99, 3, 5, 7 };
int32_t strides[4] = { 0 };
ndarray_util::set_strides_by_shape((int32_t) sizeof(int32_t), 4, strides, shape);
int32_t expected_strides[4] = {
105 * sizeof(int32_t),
35 * sizeof(int32_t),
7 * sizeof(int32_t),
1 * sizeof(int32_t)
};
assert_arrays_match("strides", "%u", 4u, expected_strides, strides);
}
void test_ndarray_indices_iter_normal() {
// Test NDArrayIndicesIter normal behavior
BEGIN_TEST();
int32_t shape[3] = { 1, 2, 3 };
int32_t indices[3] = { 0, 0, 0 };
auto iter = NDArrayIndicesIter<int32_t> {
.ndims = 3,
.shape = shape,
.indices = indices
};
assert_arrays_match("indices #0", "%u", 3u, iter.indices, (int32_t[3]) { 0, 0, 0 });
iter.next();
assert_arrays_match("indices #1", "%u", 3u, iter.indices, (int32_t[3]) { 0, 0, 1 });
iter.next();
assert_arrays_match("indices #2", "%u", 3u, iter.indices, (int32_t[3]) { 0, 0, 2 });
iter.next();
assert_arrays_match("indices #3", "%u", 3u, iter.indices, (int32_t[3]) { 0, 1, 0 });
iter.next();
assert_arrays_match("indices #4", "%u", 3u, iter.indices, (int32_t[3]) { 0, 1, 1 });
iter.next();
assert_arrays_match("indices #5", "%u", 3u, iter.indices, (int32_t[3]) { 0, 1, 2 });
iter.next();
assert_arrays_match("indices #6", "%u", 3u, iter.indices, (int32_t[3]) { 0, 0, 0 }); // Loops back
iter.next();
assert_arrays_match("indices #7", "%u", 3u, iter.indices, (int32_t[3]) { 0, 0, 1 });
}
void test_ndarray_fill_generic() {
// Test ndarray fill_generic
BEGIN_TEST();
// Choose a type that's neither int32_t nor uint64_t (candidates of SizeT) to spice it up
// Also make all the octets non-zero, to see if `memcpy` in `fill_generic` is working perfectly.
uint16_t fill_value = 0xFACE;
uint16_t in_data[6] = { 100, 101, 102, 103, 104, 105 }; // Fill `data` with values that != `999`
int32_t in_itemsize = sizeof(uint16_t);
const int32_t in_ndims = 2;
int32_t in_shape[in_ndims] = { 2, 3 };
int32_t in_strides[in_ndims] = {};
NDArray<int32_t> ndarray = {
.data = (uint8_t*) in_data,
.itemsize = in_itemsize,
.ndims = in_ndims,
.shape = in_shape,
.strides = in_strides,
};
ndarray.set_strides_by_shape();
ndarray.fill_generic((uint8_t*) &fill_value); // `fill_generic` here
uint16_t expected_data[6] = { fill_value, fill_value, fill_value, fill_value, fill_value, fill_value };
assert_arrays_match("data", "0x%hX", 6, expected_data, in_data);
}
void test_ndarray_set_to_eye() {
// Test `set_to_eye` behavior (helper function to implement `np.eye()`)
BEGIN_TEST();
double in_data[9] = { 99.0, 99.0, 99.0, 99.0, 99.0, 99.0, 99.0, 99.0, 99.0 };
int32_t in_itemsize = sizeof(double);
const int32_t in_ndims = 2;
int32_t in_shape[in_ndims] = { 3, 3 };
int32_t in_strides[in_ndims] = {};
NDArray<int32_t> ndarray = {
.data = (uint8_t*) in_data,
.itemsize = in_itemsize,
.ndims = in_ndims,
.shape = in_shape,
.strides = in_strides,
};
ndarray.set_strides_by_shape();
double zero = 0.0;
double one = 1.0;
ndarray.set_to_eye(1, (uint8_t*) &zero, (uint8_t*) &one);
assert_values_match("in_data[0]", "%f", 0.0, in_data[0]);
assert_values_match("in_data[1]", "%f", 1.0, in_data[1]);
assert_values_match("in_data[2]", "%f", 0.0, in_data[2]);
assert_values_match("in_data[3]", "%f", 0.0, in_data[3]);
assert_values_match("in_data[4]", "%f", 0.0, in_data[4]);
assert_values_match("in_data[5]", "%f", 1.0, in_data[5]);
assert_values_match("in_data[6]", "%f", 0.0, in_data[6]);
assert_values_match("in_data[7]", "%f", 0.0, in_data[7]);
assert_values_match("in_data[8]", "%f", 0.0, in_data[8]);
}
void test_slice_1() {
// Test `slice(5, None, None).indices(100) == slice(5, 100, 1)`
BEGIN_TEST();
UserSlice<int> user_slice = {
.start_defined = 1,
.start = 5,
.stop_defined = 0,
.step_defined = 0,
};
auto slice = user_slice.indices(100);
assert_values_match("start", "%d", 5, slice.start);
assert_values_match("stop", "%d", 100, slice.stop);
assert_values_match("step", "%d", 1, slice.step);
}
void test_slice_2() {
// Test `slice(400, 999, None).indices(100) == slice(100, 100, 1)`
BEGIN_TEST();
UserSlice<int> user_slice = {
.start_defined = 1,
.start = 400,
.stop_defined = 0,
.step_defined = 0,
};
auto slice = user_slice.indices(100);
assert_values_match("start", "%d", 100, slice.start);
assert_values_match("stop", "%d", 100, slice.stop);
assert_values_match("step", "%d", 1, slice.step);
}
void test_slice_3() {
// Test `slice(-10, -5, None).indices(100) == slice(90, 95, 1)`
BEGIN_TEST();
UserSlice<int> user_slice = {
.start_defined = 1,
.start = -10,
.stop_defined = 1,
.stop = -5,
.step_defined = 0,
};
auto slice = user_slice.indices(100);
assert_values_match("start", "%d", 90, slice.start);
assert_values_match("stop", "%d", 95, slice.stop);
assert_values_match("step", "%d", 1, slice.step);
}
void test_slice_4() {
// Test `slice(None, None, -5).indices(100) == (99, -1, -5)`
BEGIN_TEST();
UserSlice<int> user_slice = {
.start_defined = 0,
.stop_defined = 0,
.step_defined = 1,
.step = -5
};
auto slice = user_slice.indices(100);
assert_values_match("start", "%d", 99, slice.start);
assert_values_match("stop", "%d", -1, slice.stop);
assert_values_match("step", "%d", -5, slice.step);
}
void test_ndslice_1() {
/*
Reference Python code:
```python
ndarray = np.arange(12, dtype=np.float64).reshape((3, 4));
# array([[ 0., 1., 2., 3.],
# [ 4., 5., 6., 7.],
# [ 8., 9., 10., 11.]])
dst_ndarray = ndarray[-2:, 1::2]
# array([[ 5., 7.],
# [ 9., 11.]])
assert dst_ndarray.shape == (2, 2)
assert dst_ndarray.strides == (32, 16)
assert dst_ndarray[0, 0] == 5.0
assert dst_ndarray[0, 1] == 7.0
assert dst_ndarray[1, 0] == 9.0
assert dst_ndarray[1, 1] == 11.0
```
*/
BEGIN_TEST();
double in_data[12] = { 0.0, 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0 };
int32_t in_itemsize = sizeof(double);
const int32_t in_ndims = 2;
int32_t in_shape[in_ndims] = { 3, 4 };
int32_t in_strides[in_ndims] = {};
NDArray<int32_t> ndarray = {
.data = (uint8_t*) in_data,
.itemsize = in_itemsize,
.ndims = in_ndims,
.shape = in_shape,
.strides = in_strides
};
ndarray.set_strides_by_shape();
// Destination ndarray
// As documented, ndims and shape & strides must be allocated and determined by the caller.
const int32_t dst_ndims = 2;
int32_t dst_shape[dst_ndims] = {999, 999}; // Empty values
int32_t dst_strides[dst_ndims] = {999, 999}; // Empty values
NDArray<int32_t> dst_ndarray = {
.data = nullptr,
.ndims = dst_ndims,
.shape = dst_shape,
.strides = dst_strides
};
// Create the slice in `ndarray[-2::, 1::2]`
UserSlice<int32_t> user_slice_1 = {
.start_defined = 1,
.start = -2,
.stop_defined = 0,
.step_defined = 0
};
UserSlice<int32_t> user_slice_2 = {
.start_defined = 1,
.start = 1,
.stop_defined = 0,
.step_defined = 1,
.step = 2
};
const int32_t num_ndslices = 2;
NDSlice ndslices[num_ndslices] = {
{ .type = INPUT_SLICE_TYPE_SLICE, .slice = (uint8_t*) &user_slice_1 },
{ .type = INPUT_SLICE_TYPE_SLICE, .slice = (uint8_t*) &user_slice_2 }
};
ndarray.slice(num_ndslices, ndslices, &dst_ndarray);
int32_t expected_shape[dst_ndims] = { 2, 2 };
int32_t expected_strides[dst_ndims] = { 32, 16 };
assert_arrays_match("shape", "%d", dst_ndims, expected_shape, dst_ndarray.shape);
assert_arrays_match("strides", "%d", dst_ndims, expected_strides, dst_ndarray.strides);
assert_values_match("dst_ndarray[0, 0]", "%f", 5.0, *((double *) dst_ndarray.get_pelement((int32_t[dst_ndims]) { 0, 0 })));
assert_values_match("dst_ndarray[0, 1]", "%f", 7.0, *((double *) dst_ndarray.get_pelement((int32_t[dst_ndims]) { 0, 1 })));
assert_values_match("dst_ndarray[1, 0]", "%f", 9.0, *((double *) dst_ndarray.get_pelement((int32_t[dst_ndims]) { 1, 0 })));
assert_values_match("dst_ndarray[1, 1]", "%f", 11.0, *((double *) dst_ndarray.get_pelement((int32_t[dst_ndims]) { 1, 1 })));
}
void test_ndslice_2() {
/*
```python
ndarray = np.arange(12, dtype=np.float64).reshape((3, 4))
# array([[ 0., 1., 2., 3.],
# [ 4., 5., 6., 7.],
# [ 8., 9., 10., 11.]])
dst_ndarray = ndarray[2, ::-2]
# array([11., 9.])
assert dst_ndarray.shape == (2,)
assert dst_ndarray.strides == (-16,)
assert dst_ndarray[0] == 11.0
assert dst_ndarray[1] == 9.0
dst_ndarray[1, 0] == 99 # If you write to `dst_ndarray`
assert ndarray[1, 3] == 99 # `ndarray` also updates!!
```
*/
BEGIN_TEST();
double in_data[12] = { 0.0, 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0 };
int32_t in_itemsize = sizeof(double);
const int32_t in_ndims = 2;
int32_t in_shape[in_ndims] = { 3, 4 };
int32_t in_strides[in_ndims] = {};
NDArray<int32_t> ndarray = {
.data = (uint8_t*) in_data,
.itemsize = in_itemsize,
.ndims = in_ndims,
.shape = in_shape,
.strides = in_strides
};
ndarray.set_strides_by_shape();
// Destination ndarray
// As documented, ndims and shape & strides must be allocated and determined by the caller.
const int32_t dst_ndims = 1;
int32_t dst_shape[dst_ndims] = {999}; // Empty values
int32_t dst_strides[dst_ndims] = {999}; // Empty values
NDArray<int32_t> dst_ndarray = {
.data = nullptr,
.ndims = dst_ndims,
.shape = dst_shape,
.strides = dst_strides
};
// Create the slice in `ndarray[2, ::-2]`
int32_t user_slice_1 = 2;
UserSlice<int32_t> user_slice_2 = {
.start_defined = 0,
.stop_defined = 0,
.step_defined = 1,
.step = -2
};
const int32_t num_ndslices = 2;
NDSlice ndslices[num_ndslices] = {
{ .type = INPUT_SLICE_TYPE_INDEX, .slice = (uint8_t*) &user_slice_1 },
{ .type = INPUT_SLICE_TYPE_SLICE, .slice = (uint8_t*) &user_slice_2 }
};
ndarray.slice(num_ndslices, ndslices, &dst_ndarray);
int32_t expected_shape[dst_ndims] = { 2 };
int32_t expected_strides[dst_ndims] = { -16 };
assert_arrays_match("shape", "%d", dst_ndims, expected_shape, dst_ndarray.shape);
assert_arrays_match("strides", "%d", dst_ndims, expected_strides, dst_ndarray.strides);
// [5.0, 3.0]
assert_values_match("dst_ndarray[0]", "%f", 11.0, *((double *) dst_ndarray.get_pelement((int32_t[dst_ndims]) { 0 })));
assert_values_match("dst_ndarray[1]", "%f", 9.0, *((double *) dst_ndarray.get_pelement((int32_t[dst_ndims]) { 1 })));
}
void test_can_broadcast_shape() {
BEGIN_TEST();
assert_values_match(
"can_broadcast_shape_to([3], [1, 1, 1, 1, 3]) == true",
"%d",
true,
ndarray_util::can_broadcast_shape_to(1, (int32_t[]) { 3 }, 5, (int32_t[]) { 1, 1, 1, 1, 3 })
);
assert_values_match(
"can_broadcast_shape_to([3], [3, 1]) == false",
"%d",
false,
ndarray_util::can_broadcast_shape_to(1, (int32_t[]) { 3 }, 2, (int32_t[]) { 3, 1 }));
assert_values_match(
"can_broadcast_shape_to([3], [3]) == true",
"%d",
true,
ndarray_util::can_broadcast_shape_to(1, (int32_t[]) { 3 }, 1, (int32_t[]) { 3 }));
assert_values_match(
"can_broadcast_shape_to([1], [3]) == false",
"%d",
false,
ndarray_util::can_broadcast_shape_to(1, (int32_t[]) { 1 }, 1, (int32_t[]) { 3 }));
assert_values_match(
"can_broadcast_shape_to([1], [1]) == true",
"%d",
true,
ndarray_util::can_broadcast_shape_to(1, (int32_t[]) { 1 }, 1, (int32_t[]) { 1 }));
assert_values_match(
"can_broadcast_shape_to([256, 256, 3], [256, 1, 3]) == true",
"%d",
true,
ndarray_util::can_broadcast_shape_to(3, (int32_t[]) { 256, 256, 3 }, 3, (int32_t[]) { 256, 1, 3 })
);
assert_values_match(
"can_broadcast_shape_to([256, 256, 3], [3]) == true",
"%d",
true,
ndarray_util::can_broadcast_shape_to(3, (int32_t[]) { 256, 256, 3 }, 1, (int32_t[]) { 3 })
);
assert_values_match(
"can_broadcast_shape_to([256, 256, 3], [2]) == false",
"%d",
false,
ndarray_util::can_broadcast_shape_to(3, (int32_t[]) { 256, 256, 3 }, 1, (int32_t[]) { 2 })
);
assert_values_match(
"can_broadcast_shape_to([256, 256, 3], [1]) == true",
"%d",
true,
ndarray_util::can_broadcast_shape_to(3, (int32_t[]) { 256, 256, 3 }, 1, (int32_t[]) { 1 })
);
// In cases when the shapes contain zero(es)
assert_values_match(
"can_broadcast_shape_to([0], [1]) == true",
"%d",
true,
ndarray_util::can_broadcast_shape_to(1, (int32_t[]) { 0 }, 1, (int32_t[]) { 1 })
);
assert_values_match(
"can_broadcast_shape_to([0], [2]) == false",
"%d",
false,
ndarray_util::can_broadcast_shape_to(1, (int32_t[]) { 0 }, 1, (int32_t[]) { 2 })
);
assert_values_match(
"can_broadcast_shape_to([0, 4, 0, 0], [1]) == true",
"%d",
true,
ndarray_util::can_broadcast_shape_to(4, (int32_t[]) { 0, 4, 0, 0 }, 1, (int32_t[]) { 1 })
);
assert_values_match(
"can_broadcast_shape_to([0, 4, 0, 0], [1, 1, 1, 1]) == true",
"%d",
true,
ndarray_util::can_broadcast_shape_to(4, (int32_t[]) { 0, 4, 0, 0 }, 4, (int32_t[]) { 1, 1, 1, 1 })
);
assert_values_match(
"can_broadcast_shape_to([0, 4, 0, 0], [1, 4, 1, 1]) == true",
"%d",
true,
ndarray_util::can_broadcast_shape_to(4, (int32_t[]) { 0, 4, 0, 0 }, 4, (int32_t[]) { 1, 4, 1, 1 })
);
assert_values_match(
"can_broadcast_shape_to([4, 3], [0, 3]) == false",
"%d",
false,
ndarray_util::can_broadcast_shape_to(2, (int32_t[]) { 4, 3 }, 2, (int32_t[]) { 0, 3 })
);
assert_values_match(
"can_broadcast_shape_to([4, 3], [0, 0]) == false",
"%d",
false,
ndarray_util::can_broadcast_shape_to(2, (int32_t[]) { 4, 3 }, 2, (int32_t[]) { 0, 0 })
);
}
void test_ndarray_broadcast_1() {
/*
# array = np.array([[19.9, 29.9, 39.9, 49.9]], dtype=np.float64)
# >>> [[19.9 29.9 39.9 49.9]]
#
# array = np.broadcast_to(array, (2, 3, 4))
# >>> [[[19.9 29.9 39.9 49.9]
# >>> [19.9 29.9 39.9 49.9]
# >>> [19.9 29.9 39.9 49.9]]
# >>> [[19.9 29.9 39.9 49.9]
# >>> [19.9 29.9 39.9 49.9]
# >>> [19.9 29.9 39.9 49.9]]]
#
# assery array.strides == (0, 0, 8)
*/
BEGIN_TEST();
double in_data[4] = { 19.9, 29.9, 39.9, 49.9 };
const int32_t in_ndims = 2;
int32_t in_shape[in_ndims] = {1, 4};
int32_t in_strides[in_ndims] = {};
NDArray<int32_t> ndarray = {
.data = (uint8_t*) in_data,
.itemsize = sizeof(double),
.ndims = in_ndims,
.shape = in_shape,
.strides = in_strides
};
ndarray.set_strides_by_shape();
const int32_t dst_ndims = 3;
int32_t dst_shape[dst_ndims] = {2, 3, 4};
int32_t dst_strides[dst_ndims] = {};
NDArray<int32_t> dst_ndarray = {
.ndims = dst_ndims,
.shape = dst_shape,
.strides = dst_strides
};
ndarray.broadcast_to(&dst_ndarray);
assert_arrays_match("dst_ndarray->strides", "%d", dst_ndims, (int32_t[]) { 0, 0, 8 }, dst_ndarray.strides);
assert_values_match("dst_ndarray[0, 0, 0]", "%f", 19.9, *((double*) dst_ndarray.get_pelement((int32_t[]) {0, 0, 0})));
assert_values_match("dst_ndarray[0, 0, 1]", "%f", 29.9, *((double*) dst_ndarray.get_pelement((int32_t[]) {0, 0, 1})));
assert_values_match("dst_ndarray[0, 0, 2]", "%f", 39.9, *((double*) dst_ndarray.get_pelement((int32_t[]) {0, 0, 2})));
assert_values_match("dst_ndarray[0, 0, 3]", "%f", 49.9, *((double*) dst_ndarray.get_pelement((int32_t[]) {0, 0, 3})));
assert_values_match("dst_ndarray[0, 1, 0]", "%f", 19.9, *((double*) dst_ndarray.get_pelement((int32_t[]) {0, 1, 0})));
assert_values_match("dst_ndarray[0, 1, 1]", "%f", 29.9, *((double*) dst_ndarray.get_pelement((int32_t[]) {0, 1, 1})));
assert_values_match("dst_ndarray[0, 1, 2]", "%f", 39.9, *((double*) dst_ndarray.get_pelement((int32_t[]) {0, 1, 2})));
assert_values_match("dst_ndarray[0, 1, 3]", "%f", 49.9, *((double*) dst_ndarray.get_pelement((int32_t[]) {0, 1, 3})));
assert_values_match("dst_ndarray[1, 2, 3]", "%f", 49.9, *((double*) dst_ndarray.get_pelement((int32_t[]) {1, 2, 3})));
}
void test_assign_with() {
/*
```
xs = np.array([[1.0, 2.0, 3.0], [4.0, 5.0, 6.0], [7.0, 8.0, 9.0]], dtype=np.float64)
ys = xs.shape
```
*/
}
int main() {
test_calc_size_from_shape_normal();
test_calc_size_from_shape_has_zero();
test_set_strides_by_shape();
test_ndarray_indices_iter_normal();
test_ndarray_fill_generic();
test_ndarray_set_to_eye();
test_slice_1();
test_slice_2();
test_slice_3();
test_slice_4();
test_ndslice_1();
test_ndslice_2();
test_can_broadcast_shape();
test_ndarray_broadcast_1();
test_assign_with();
return 0;
}

14
nac3core/irrt/irrt_typedefs.hpp
View File

@ -1,14 +0,0 @@
#pragma once
// This is made toggleable since `irrt_test.cpp` itself would include
// headers that define the `int_t` family.
#ifndef IRRT_DONT_TYPEDEF_INTS
typedef _BitInt(8) int8_t;
typedef unsigned _BitInt(8) uint8_t;
typedef _BitInt(32) int32_t;
typedef unsigned _BitInt(32) uint32_t;
typedef _BitInt(64) int64_t;
typedef unsigned _BitInt(64) uint64_t;
#endif
typedef int32_t SliceIndex;

37
nac3core/irrt/irrt_utils.hpp
View File

@ -1,37 +0,0 @@
#pragma once
#include "irrt_typedefs.hpp"
namespace {
template <typename T>
T max(T a, T b) {
return a > b ? a : b;
}
template <typename T>
T min(T a, T b) {
return a > b ? b : a;
}
template <typename T>
bool arrays_match(int len, T *as, T *bs) {
for (int i = 0; i < len; i++) {
if (as[i] != bs[i]) return false;
}
return true;
}
void irrt_panic() {
// Crash the program for now.
// TODO: Don't crash the program
// ... or at least produce a good message when doing testing IRRT
uint8_t* death = nullptr;
*death = 0; // TODO: address 0 on hardware might be writable?
}
// TODO: Make this a macro and allow it to be toggled on/off (e.g., debug vs release)
void irrt_assert(bool condition) {
if (!condition) irrt_panic();
}
}

1900
nac3core/src/codegen/builtin_fns.rs
View File

File diff suppressed because it is too large Load Diff

2069
nac3core/src/codegen/classes.rs
View File

File diff suppressed because it is too large Load Diff

51
nac3core/src/codegen/concrete_type.rs
View File

@ -3,13 +3,10 @@ use crate::{
toplevel::DefinitionId,
typecheck::{
type_inferencer::PrimitiveStore,
typedef::{
into_var_map, FunSignature, FuncArg, Type, TypeEnum, TypeVar, TypeVarId, Unifier,
},
typedef::{FunSignature, FuncArg, Type, TypeEnum, Unifier},
},
};
use indexmap::IndexMap;
use nac3parser::ast::StrRef;
use std::collections::HashMap;
@ -47,10 +44,13 @@ pub enum ConcreteTypeEnum {
TTuple {
ty: Vec<ConcreteType>,
},
TList {
ty: ConcreteType,
},
TObj {
obj_id: DefinitionId,
fields: HashMap<StrRef, (ConcreteType, bool)>,
params: IndexMap<TypeVarId, ConcreteType>,
params: HashMap<u32, ConcreteType>,
},
TVirtual {
ty: ConcreteType,
@ -58,15 +58,11 @@ pub enum ConcreteTypeEnum {
TFunc {
args: Vec<ConcreteFuncArg>,
ret: ConcreteType,
vars: HashMap<TypeVarId, ConcreteType>,
},
TLiteral {
values: Vec<SymbolValue>,
vars: HashMap<u32, ConcreteType>,
},
}
impl ConcreteTypeStore {
#[must_use]
pub fn new() -> ConcreteTypeStore {
ConcreteTypeStore {
store: vec![
@ -84,7 +80,6 @@ impl ConcreteTypeStore {
}
}
#[must_use]
pub fn get(&self, cty: ConcreteType) -> &ConcreteTypeEnum {
&self.store[cty.0]
}
@ -164,6 +159,9 @@ impl ConcreteTypeStore {
.map(|t| self.from_unifier_type(unifier, primitives, *t, cache))
.collect(),
},
TypeEnum::TList { ty } => ConcreteTypeEnum::TList {
ty: self.from_unifier_type(unifier, primitives, *ty, cache),
},
TypeEnum::TObj { obj_id, fields, params } => ConcreteTypeEnum::TObj {
obj_id: *obj_id,
fields: fields
@ -196,12 +194,9 @@ impl ConcreteTypeStore {
ty: self.from_unifier_type(unifier, primitives, *ty, cache),
},
TypeEnum::TFunc(signature) => {
self.from_signature(unifier, primitives, signature, cache)
self.from_signature(unifier, primitives, &*signature, cache)
}
TypeEnum::TLiteral { values, .. } => {
ConcreteTypeEnum::TLiteral { values: values.clone() }
}
_ => unreachable!("{:?}", ty_enum.get_type_name()),
_ => unreachable!(),
};
let index = if let Some(ConcreteType(index)) = cache.get(&ty).unwrap() {
self.store[*index] = result;
@ -226,7 +221,7 @@ impl ConcreteTypeStore {
return if let Some(ty) = ty {
*ty
} else {
*ty = Some(unifier.get_dummy_var().ty);
*ty = Some(unifier.get_dummy_var().0);
ty.unwrap()
};
}
@ -254,6 +249,9 @@ impl ConcreteTypeStore {
.map(|cty| self.to_unifier_type(unifier, primitives, *cty, cache))
.collect(),
},
ConcreteTypeEnum::TList { ty } => {
TypeEnum::TList { ty: self.to_unifier_type(unifier, primitives, *ty, cache) }
}
ConcreteTypeEnum::TVirtual { ty } => {
TypeEnum::TVirtual { ty: self.to_unifier_type(unifier, primitives, *ty, cache) }
}
@ -265,10 +263,10 @@ impl ConcreteTypeStore {
(*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 }
})),
params: params
.iter()
.map(|(id, cty)| (*id, self.to_unifier_type(unifier, primitives, *cty, cache)))
.collect::<HashMap<_, _>>(),
},
ConcreteTypeEnum::TFunc { args, ret, vars } => TypeEnum::TFunc(FunSignature {
args: args
@ -280,14 +278,11 @@ impl ConcreteTypeStore {
})
.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 }
})),
vars: vars
.iter()
.map(|(id, cty)| (*id, self.to_unifier_type(unifier, primitives, *cty, cache)))
.collect::<HashMap<_, _>>(),
}),
ConcreteTypeEnum::TLiteral { values, .. } => {
TypeEnum::TLiteral { values: values.clone(), loc: None }
}
};
let result = unifier.add_ty(result);
if let Some(ty) = cache.get(&cty).unwrap() {

3129
nac3core/src/codegen/expr.rs
View File

File diff suppressed because it is too large Load Diff

132
nac3core/src/codegen/extern_fns.rs
View File

@ -1,132 +0,0 @@
use inkwell::attributes::{Attribute, AttributeLoc};
use inkwell::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()
}

116
nac3core/src/codegen/generator.rs
View File

@ -1,5 +1,5 @@
use crate::{
codegen::{bool_to_i1, bool_to_i8, classes::ArraySliceValue, expr::*, stmt::*, CodeGenContext},
codegen::{expr::*, stmt::*, CodeGenContext},
symbol_resolver::ValueEnum,
toplevel::{DefinitionId, TopLevelDef},
typecheck::typedef::{FunSignature, Type},
@ -7,7 +7,7 @@ use crate::{
use inkwell::{
context::Context,
types::{BasicTypeEnum, IntType},
values::{BasicValueEnum, IntValue, PointerValue},
values::{BasicValueEnum, PointerValue},
};
use nac3parser::ast::{Expr, Stmt, StrRef};
@ -22,9 +22,9 @@ pub trait CodeGenerator {
/// - 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>(
fn gen_call<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'ctx, '_>,
ctx: &mut CodeGenContext<'ctx, 'a>,
obj: Option<(Type, ValueEnum<'ctx>)>,
fun: (&FunSignature, DefinitionId),
params: Vec<(Option<StrRef>, ValueEnum<'ctx>)>,
@ -36,12 +36,12 @@ pub trait CodeGenerator {
}
/// Generate object constructor and returns the constructed object.
/// - signature: Function signature of the constructor.
/// - signature: Function signature of the contructor.
/// - def: Class definition for the constructor class.
/// - params: Function parameters.
fn gen_constructor<'ctx>(
fn gen_constructor<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'ctx, '_>,
ctx: &mut CodeGenContext<'ctx, 'a>,
signature: &FunSignature,
def: &TopLevelDef,
params: Vec<(Option<StrRef>, ValueEnum<'ctx>)>,
@ -59,20 +59,20 @@ pub trait CodeGenerator {
/// 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>(
fn gen_func_instance<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'ctx, '_>,
ctx: &mut CodeGenContext<'ctx, 'a>,
obj: Option<(Type, ValueEnum<'ctx>)>,
fun: (&FunSignature, &mut TopLevelDef, String),
id: usize,
) -> Result<String, String> {
gen_func_instance(ctx, &obj, fun, id)
gen_func_instance(ctx, obj, fun, id)
}
/// Generate the code for an expression.
fn gen_expr<'ctx>(
fn gen_expr<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'ctx, '_>,
ctx: &mut CodeGenContext<'ctx, 'a>,
expr: &Expr<Option<Type>>,
) -> Result<Option<ValueEnum<'ctx>>, String>
where
@ -83,44 +83,30 @@ pub trait CodeGenerator {
/// 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>(
fn gen_var_alloc<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'ctx, '_>,
ctx: &mut CodeGenContext<'ctx, 'a>,
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)
gen_var(ctx, ty)
}
/// Return a pointer pointing to the target of the expression.
fn gen_store_target<'ctx>(
fn gen_store_target<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'ctx, '_>,
ctx: &mut CodeGenContext<'ctx, 'a>,
pattern: &Expr<Option<Type>>,
name: Option<&str>,
) -> Result<Option<PointerValue<'ctx>>, String>
) -> Result<PointerValue<'ctx>, String>
where
Self: Sized,
{
gen_store_target(self, ctx, pattern, name)
gen_store_target(self, ctx, pattern)
}
/// Generate code for an assignment expression.
fn gen_assign<'ctx>(
fn gen_assign<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'ctx, '_>,
ctx: &mut CodeGenContext<'ctx, 'a>,
target: &Expr<Option<Type>>,
value: ValueEnum<'ctx>,
) -> Result<(), String>
@ -132,9 +118,9 @@ pub trait CodeGenerator {
/// Generate code for a while expression.
/// Return true if the while loop must early return
fn gen_while(
fn gen_while<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'_, '_>,
ctx: &mut CodeGenContext<'ctx, 'a>,
stmt: &Stmt<Option<Type>>,
) -> Result<(), String>
where
@ -143,11 +129,11 @@ pub trait CodeGenerator {
gen_while(self, ctx, stmt)
}
/// Generate code for a for expression.
/// Return true if the for loop must early return
fn gen_for(
/// Generate code for a while expression.
/// Return true if the while loop must early return
fn gen_for<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'_, '_>,
ctx: &mut CodeGenContext<'ctx, 'a>,
stmt: &Stmt<Option<Type>>,
) -> Result<(), String>
where
@ -158,9 +144,9 @@ pub trait CodeGenerator {
/// Generate code for an if expression.
/// Return true if the statement must early return
fn gen_if(
fn gen_if<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'_, '_>,
ctx: &mut CodeGenContext<'ctx, 'a>,
stmt: &Stmt<Option<Type>>,
) -> Result<(), String>
where
@ -169,9 +155,9 @@ pub trait CodeGenerator {
gen_if(self, ctx, stmt)
}
fn gen_with(
fn gen_with<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'_, '_>,
ctx: &mut CodeGenContext<'ctx, 'a>,
stmt: &Stmt<Option<Type>>,
) -> Result<(), String>
where
@ -181,11 +167,10 @@ pub trait CodeGenerator {
}
/// Generate code for a statement
///
/// Return true if the statement must early return
fn gen_stmt(
fn gen_stmt<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'_, '_>,
ctx: &mut CodeGenContext<'ctx, 'a>,
stmt: &Stmt<Option<Type>>,
) -> Result<(), String>
where
@ -193,36 +178,6 @@ pub trait CodeGenerator {
{
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 {
@ -231,20 +186,17 @@ pub struct DefaultCodeGenerator {
}
impl DefaultCodeGenerator {
#[must_use]
pub fn new(name: String, size_t: u32) -> DefaultCodeGenerator {
assert!(matches!(size_t, 32 | 64));
assert!(size_t == 32 || size_t == 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...

140
nac3core/src/codegen/irrt/irrt.c Normal file
View File

@ -0,0 +1,140 @@
typedef _ExtInt(8) int8_t;
typedef unsigned _ExtInt(8) uint8_t;
typedef _ExtInt(32) int32_t;
typedef unsigned _ExtInt(32) uint32_t;
typedef _ExtInt(64) int64_t;
typedef unsigned _ExtInt(64) uint64_t;
# define MAX(a, b) (a > b ? a : b)
# define MIN(a, b) (a > b ? b : a)
// 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
#define DEF_INT_EXP(T) T __nac3_int_exp_##T( \
T base, \
T exp \
) { \
T res = (T)1; \
/* repeated squaring method */ \
do { \
if (exp & 1) res *= base; /* for n odd */ \
exp >>= 1; \
base *= base; \
} while (exp); \
return res; \
} \
DEF_INT_EXP(int32_t)
DEF_INT_EXP(int64_t)
DEF_INT_EXP(uint32_t)
DEF_INT_EXP(uint64_t)
int32_t __nac3_slice_index_bound(int32_t i, const int32_t len) {
if (i < 0) {
i = len + i;
}
if (i < 0) {
return 0;
} else if (i > len) {
return len;
}
return i;
}
int32_t __nac3_range_slice_len(const int32_t start, const int32_t end, const int32_t step) {
int32_t 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;
}
}
// 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)
int32_t __nac3_list_slice_assign_var_size(
int32_t dest_start,
int32_t dest_end,
int32_t dest_step,
uint8_t *dest_arr,
int32_t dest_arr_len,
int32_t src_start,
int32_t src_end,
int32_t src_step,
uint8_t *src_arr,
int32_t src_arr_len,
const int32_t 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 int32_t src_len = (src_end >= src_start) ? (src_end - src_start + 1) : 0;
const int32_t 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 = __builtin_alloca(src_arr_len * size);
__builtin_memcpy(tmp, src_arr, src_arr_len * size);
src_arr = tmp;
}
int32_t src_ind = src_start;
int32_t 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;
}

851
nac3core/src/codegen/irrt/mod.rs
View File

File diff suppressed because it is too large Load Diff

26
nac3core/src/codegen/irrt/test.rs
View File

@ -1,26 +0,0 @@
#[cfg(test)]
mod tests {
use std::{path::Path, process::Command};
#[test]
fn run_irrt_test() {
assert!(
cfg!(feature = "test"),
"Please do `cargo test -F test` to compile `irrt_test.out` and run test"
);
let irrt_test_out_path = Path::new(concat!(env!("OUT_DIR"), "/irrt_test.out"));
let output = Command::new(irrt_test_out_path.to_str().unwrap()).output().unwrap();
if !output.status.success() {
eprintln!("irrt_test failed with status {}:", output.status);
eprintln!("====== stdout ======");
eprintln!("{}", String::from_utf8(output.stdout).unwrap());
eprintln!("====== stderr ======");
eprintln!("{}", String::from_utf8(output.stderr).unwrap());
eprintln!("====================");
panic!("irrt_test failed");
}
}
}

308
nac3core/src/codegen/llvm_intrinsics.rs
View File

@ -1,308 +0,0 @@
use crate::codegen::CodeGenContext;
use inkwell::context::Context;
use inkwell::intrinsics::Intrinsic;
use inkwell::types::AnyTypeEnum::IntType;
use inkwell::types::FloatType;
use inkwell::values::{BasicValueEnum, CallSiteValue, FloatValue, IntValue, PointerValue};
use inkwell::AddressSpace;
use itertools::Either;
/// 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.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_bitcast(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_bitcast(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()
}

686
nac3core/src/codegen/mod.rs
View File

File diff suppressed because it is too large Load Diff

2017
nac3core/src/codegen/numpy.rs
View File

File diff suppressed because it is too large Load Diff

1359
nac3core/src/codegen/stmt.rs
View File

File diff suppressed because it is too large Load Diff

187
nac3core/src/codegen/test.rs
View File

@ -1,27 +1,18 @@
use crate::{
codegen::{
classes::{ListType, NDArrayType, ProxyType, RangeType},
concrete_type::ConcreteTypeStore,
CodeGenContext, CodeGenLLVMOptions, CodeGenTargetMachineOptions, CodeGenTask,
CodeGenerator, DefaultCodeGenerator, WithCall, WorkerRegistry,
concrete_type::ConcreteTypeStore, CodeGenContext, CodeGenTask, DefaultCodeGenerator,
WithCall, WorkerRegistry,
},
symbol_resolver::{SymbolResolver, ValueEnum},
toplevel::{
composer::{ComposerConfig, TopLevelComposer},
DefinitionId, FunInstance, TopLevelContext, TopLevelDef,
composer::TopLevelComposer, DefinitionId, FunInstance, TopLevelContext, TopLevelDef,
},
typecheck::{
type_inferencer::{FunctionData, Inferencer, PrimitiveStore},
typedef::{FunSignature, FuncArg, Type, TypeEnum, Unifier, VarMap},
typedef::{FunSignature, FuncArg, Type, TypeEnum, Unifier},
},
};
use indexmap::IndexMap;
use indoc::indoc;
use inkwell::{
targets::{InitializationConfig, Target},
OptimizationLevel,
};
use nac3parser::ast::FileName;
use nac3parser::{
ast::{fold::Fold, StrRef},
parser::parse_program,
@ -57,23 +48,23 @@ impl SymbolResolver for Resolver {
_: &PrimitiveStore,
str: StrRef,
) -> Result<Type, String> {
self.id_to_type.get(&str).copied().ok_or_else(|| format!("cannot find symbol `{str}`"))
self.id_to_type.get(&str).cloned().ok_or_else(|| format!("cannot find symbol `{}`", str))
}
fn get_symbol_value<'ctx>(
fn get_symbol_value<'ctx, 'a>(
&self,
_: StrRef,
_: &mut CodeGenContext<'ctx, '_>,
_: &mut CodeGenContext<'ctx, 'a>,
) -> Option<ValueEnum<'ctx>> {
unimplemented!()
}
fn get_identifier_def(&self, id: StrRef) -> Result<DefinitionId, HashSet<String>> {
fn get_identifier_def(&self, id: StrRef) -> Result<DefinitionId, String> {
self.id_to_def
.read()
.get(&id)
.copied()
.ok_or_else(|| HashSet::from([format!("cannot find symbol `{id}`")]))
.cloned()
.ok_or_else(|| format!("cannot find symbol `{}`", id))
}
fn get_string_id(&self, _: &str) -> i32 {
@ -92,9 +83,9 @@ fn test_primitives() {
d = a if c == 1 else 0
return d
"};
let statements = parse_program(source, FileName::default()).unwrap();
let statements = parse_program(source, Default::default()).unwrap();
let composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 32).0;
let composer: TopLevelComposer = Default::default();
let mut unifier = composer.unifier.clone();
let primitives = composer.primitives_ty;
let top_level = Arc::new(composer.make_top_level_context());
@ -103,7 +94,7 @@ fn test_primitives() {
let resolver = Arc::new(Resolver {
id_to_type: HashMap::new(),
id_to_def: RwLock::new(HashMap::new()),
class_names: HashMap::default(),
class_names: Default::default(),
}) as Arc<dyn SymbolResolver + Send + Sync>;
let threads = vec![DefaultCodeGenerator::new("test".into(), 32).into()];
@ -113,7 +104,7 @@ fn test_primitives() {
FuncArg { name: "b".into(), ty: primitives.int32, default_value: None },
],
ret: primitives.int32,
vars: VarMap::new(),
vars: HashMap::new(),
};
let mut store = ConcreteTypeStore::new();
@ -128,12 +119,12 @@ fn test_primitives() {
};
let mut virtual_checks = Vec::new();
let mut calls = HashMap::new();
let mut identifiers: HashSet<_> = ["a".into(), "b".into()].into();
let mut identifiers: HashSet<_> = ["a".into(), "b".into()].iter().cloned().collect();
let mut inferencer = Inferencer {
top_level: &top_level,
function_data: &mut function_data,
unifier: &mut unifier,
variable_mapping: HashMap::default(),
variable_mapping: Default::default(),
primitives: &primitives,
virtual_checks: &mut virtual_checks,
calls: &mut calls,
@ -157,7 +148,7 @@ fn test_primitives() {
});
let task = CodeGenTask {
subst: Vec::default(),
subst: Default::default(),
symbol_name: "testing".into(),
body: Arc::new(statements),
unifier_index: 0,
@ -190,45 +181,27 @@ fn test_primitives() {
; ModuleID = 'test'
source_filename = \"test\"
; Function Attrs: mustprogress nofree norecurse nosync nounwind readnone willreturn
define i32 @testing(i32 %0, i32 %1) local_unnamed_addr #0 !dbg !4 {
define i32 @testing(i32 %0, i32 %1) {
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
%add = add i32 %0, %1
%cmp = icmp eq i32 %add, 1
br i1 %cmp, label %then, label %else
then: ; preds = %init
br label %cont
else: ; preds = %init
br label %cont
cont: ; preds = %else, %then
%if_exp_result.0 = phi i32 [ %0, %then ], [ 0, %else ]
ret i32 %if_exp_result.0
}
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);
let (registry, handles) = WorkerRegistry::create_workers(threads, top_level, f);
registry.add_task(task);
registry.wait_tasks_complete(handles);
}
@ -239,14 +212,14 @@ fn test_simple_call() {
a = foo(a)
return a * 2
"};
let statements_1 = parse_program(source_1, FileName::default()).unwrap();
let statements_1 = parse_program(source_1, Default::default()).unwrap();
let source_2 = indoc! { "
return a + 1
"};
let statements_2 = parse_program(source_2, FileName::default()).unwrap();
let statements_2 = parse_program(source_2, Default::default()).unwrap();
let composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 32).0;
let composer: TopLevelComposer = Default::default();
let mut unifier = composer.unifier.clone();
let primitives = composer.primitives_ty;
let top_level = Arc::new(composer.make_top_level_context());
@ -255,7 +228,7 @@ fn test_simple_call() {
let signature = FunSignature {
args: vec![FuncArg { name: "a".into(), ty: primitives.int32, default_value: None }],
ret: primitives.int32,
vars: VarMap::new(),
vars: HashMap::new(),
};
let fun_ty = unifier.add_ty(TypeEnum::TFunc(signature.clone()));
let mut store = ConcreteTypeStore::new();
@ -279,7 +252,7 @@ fn test_simple_call() {
let resolver = Resolver {
id_to_type: HashMap::new(),
id_to_def: RwLock::new(HashMap::new()),
class_names: HashMap::default(),
class_names: Default::default(),
};
resolver.add_id_def("foo".into(), DefinitionId(foo_id));
let resolver = Arc::new(resolver) as Arc<dyn SymbolResolver + Send + Sync>;
@ -300,12 +273,12 @@ fn test_simple_call() {
};
let mut virtual_checks = Vec::new();
let mut calls = HashMap::new();
let mut identifiers: HashSet<_> = ["a".into(), "foo".into()].into();
let mut identifiers: HashSet<_> = ["a".into(), "foo".into()].iter().cloned().collect();
let mut inferencer = Inferencer {
top_level: &top_level,
function_data: &mut function_data,
unifier: &mut unifier,
variable_mapping: HashMap::default(),
variable_mapping: Default::default(),
primitives: &primitives,
virtual_checks: &mut virtual_checks,
calls: &mut calls,
@ -334,11 +307,11 @@ fn test_simple_call() {
&mut *top_level.definitions.read()[foo_id].write()
{
instance_to_stmt.insert(
String::new(),
"".to_string(),
FunInstance {
body: Arc::new(statements_2),
calls: Arc::new(inferencer.calls.clone()),
subst: IndexMap::default(),
subst: Default::default(),
unifier_id: 0,
},
);
@ -354,7 +327,7 @@ fn test_simple_call() {
});
let task = CodeGenTask {
subst: Vec::default(),
subst: Default::default(),
symbol_name: "testing".to_string(),
body: Arc::new(statements_1),
calls: Arc::new(calls1),
@ -369,83 +342,23 @@ fn test_simple_call() {
; ModuleID = 'test'
source_filename = \"test\"
; Function Attrs: mustprogress nofree norecurse nosync nounwind readnone willreturn
define i32 @testing(i32 %0) local_unnamed_addr #0 !dbg !5 {
define i32 @testing(i32 %0) {
init:
%add.i = shl i32 %0, 1, !dbg !10
%mul = add i32 %add.i, 2, !dbg !10
ret i32 %mul, !dbg !10
%call = call i32 @foo.0(i32 %0)
%mul = mul i32 %call, 2
ret i32 %mul
}
; Function Attrs: mustprogress nofree norecurse nosync nounwind readnone willreturn
define i32 @foo.0(i32 %0) local_unnamed_addr #0 !dbg !11 {
define i32 @foo.0(i32 %0) {
init:
%add = add i32 %0, 1, !dbg !12
ret i32 %add, !dbg !12
%add = add i32 %0, 1
ret i32 %add
}
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);
let (registry, handles) = WorkerRegistry::create_workers(threads, top_level, 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());
}

23
nac3core/src/lib.rs
View File

@ -1,26 +1,7 @@
#![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
)]
#![warn(clippy::all)]
#![allow(dead_code)]
pub mod codegen;
pub mod symbol_resolver;
pub mod toplevel;
pub mod typecheck;
pub mod util;

384
nac3core/src/symbol_resolver.rs
View File

@ -1,19 +1,22 @@
use std::fmt::Debug;
use std::rc::Rc;
use std::sync::Arc;
use std::{collections::HashMap, collections::HashSet, fmt::Display};
use std::{collections::HashMap, fmt::Display};
use crate::typecheck::typedef::TypeEnum;
use crate::{
codegen::{CodeGenContext, CodeGenerator},
toplevel::{type_annotation::TypeAnnotation, DefinitionId, TopLevelDef},
codegen::CodeGenContext,
toplevel::{DefinitionId, TopLevelDef},
};
use crate::{
codegen::CodeGenerator,
typecheck::{
type_inferencer::PrimitiveStore,
typedef::{Type, TypeEnum, Unifier, VarMap},
typedef::{Type, Unifier},
},
};
use inkwell::values::{BasicValueEnum, FloatValue, IntValue, PointerValue, StructValue};
use itertools::{chain, izip, Itertools};
use nac3parser::ast::{Constant, Expr, Location, StrRef};
use itertools::{chain, izip};
use nac3parser::ast::{Expr, Location, StrRef};
use parking_lot::RwLock;
#[derive(Clone, PartialEq, Debug)]
@ -30,192 +33,15 @@ pub enum 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 } = expected_ty.as_ref() else {
return Err(format!(
"Expected {:?}, but got Tuple",
expected_ty.get_type_name()
));
};
assert_eq!(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 })
}
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::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")
@ -224,82 +50,39 @@ impl Display for SymbolValue {
}
}
SymbolValue::Tuple(t) => {
write!(f, "({})", t.iter().map(|v| format!("{v}")).collect::<Vec<_>>().join(", "))
write!(f, "({})", t.iter().map(|v| format!("{}", v)).collect::<Vec<_>>().join(", "))
}
SymbolValue::OptionSome(v) => write!(f, "Some({v})"),
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>(
fn get_const_obj<'ctx, 'a>(
&self,
ctx: &mut CodeGenContext<'ctx, '_>,
ctx: &mut CodeGenContext<'ctx, 'a>,
generator: &mut dyn CodeGenerator,
) -> BasicValueEnum<'ctx>;
/// Converts this value to a LLVM [`BasicValueEnum`].
fn to_basic_value_enum<'ctx>(
fn to_basic_value_enum<'ctx, 'a>(
&self,
ctx: &mut CodeGenContext<'ctx, '_>,
ctx: &mut CodeGenContext<'ctx, 'a>,
generator: &mut dyn CodeGenerator,
expected_ty: Type,
) -> Result<BasicValueEnum<'ctx>, String>;
/// Returns a field within this value.
fn get_field<'ctx>(
fn get_field<'ctx, 'a>(
&self,
name: StrRef,
ctx: &mut CodeGenContext<'ctx, '_>,
ctx: &mut CodeGenContext<'ctx, 'a>,
) -> 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>),
}
@ -334,22 +117,20 @@ impl<'ctx> From<StructValue<'ctx>> for ValueEnum<'ctx> {
}
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::Static(v) => v.to_basic_value_enum(ctx, generator),
ValueEnum::Dynamic(v) => Ok(v),
}
}
}
pub trait SymbolResolver {
/// Get type of type variable identifier or top-level function type,
// get type of type variable identifier or top-level function type
fn get_symbol_type(
&self,
unifier: &mut Unifier,
@ -358,16 +139,16 @@ pub trait SymbolResolver {
str: StrRef,
) -> Result<Type, String>;
/// Get the top-level definition of identifiers.
fn get_identifier_def(&self, str: StrRef) -> Result<DefinitionId, HashSet<String>>;
// get the top-level definition of identifiers
fn get_identifier_def(&self, str: StrRef) -> Result<DefinitionId, String>;
fn get_symbol_value<'ctx>(
fn get_symbol_value<'ctx, 'a>(
&self,
str: StrRef,
ctx: &mut CodeGenContext<'ctx, '_>,
ctx: &mut CodeGenContext<'ctx, 'a>,
) -> Option<ValueEnum<'ctx>>;
fn get_default_param_value(&self, expr: &Expr) -> Option<SymbolValue>;
fn get_default_param_value(&self, expr: &nac3parser::ast::Expr) -> Option<SymbolValue>;
fn get_string_id(&self, s: &str) -> i32;
fn get_exception_id(&self, tyid: usize) -> usize;
@ -375,7 +156,7 @@ pub trait SymbolResolver {
&self,
_unifier: &mut Unifier,
_top_level_defs: &[Arc<RwLock<TopLevelDef>>],
_primitives: &PrimitiveStore,
_primitives: &PrimitiveStore
) -> Result<(), String> {
Ok(())
}
@ -388,23 +169,23 @@ thread_local! {
"float".into(),
"bool".into(),
"virtual".into(),
"list".into(),
"tuple".into(),
"str".into(),
"Exception".into(),
"uint32".into(),
"uint64".into(),
"Literal".into(),
];
}
/// Converts a type annotation into a [Type].
// convert type annotation into 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>> {
) -> Result<Type, String> {
use nac3parser::ast::ExprKind::*;
let ids = IDENTIFIER_ID.with(|ids| *ids);
let int32_id = ids[0];
@ -412,12 +193,12 @@ pub fn parse_type_annotation<T>(
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 list_id = ids[5];
let tuple_id = ids[6];
let str_id = ids[7];
let exn_id = ids[8];
let uint32_id = ids[9];
let uint64_id = ids[10];
let name_handling = |id: &StrRef, loc: Location, unifier: &mut Unifier| {
if *id == int32_id {
@ -438,33 +219,39 @@ pub fn parse_type_annotation<T>(
Ok(primitives.exception)
} else {
let obj_id = resolver.get_identifier_def(*id);
if let Ok(obj_id) = obj_id {
match obj_id {
Ok(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!(
return Err(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() }))
Ok(unifier.add_ty(TypeEnum::TObj {
obj_id,
fields,
params: Default::default(),
}))
} else {
Err(HashSet::from([format!("Cannot use function name as type at {loc}")]))
Err(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}")]),
)?;
}
Err(_) => {
let ty = resolver
.get_symbol_type(unifier, top_level_defs, primitives, *id)
.map_err(|e| 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}")]))
Err(format!("Unknown type annotation {} at {}", id, loc))
}
}
}
}
@ -474,6 +261,9 @@ pub fn parse_type_annotation<T>(
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 == list_id {
let ty = parse_type_annotation(resolver, top_level_defs, unifier, primitives, slice)?;
Ok(unifier.add_ty(TypeEnum::TList { ty }))
} else if *id == tuple_id {
if let Tuple { elts, .. } = &slice.node {
let ty = elts
@ -484,31 +274,8 @@ pub fn parse_type_annotation<T>(
.collect::<Result<Vec<_>, _>>()?;
Ok(unifier.add_ty(TypeEnum::TTuple { ty }))
} else {
Err(HashSet::from(["Expected multiple elements for tuple".into()]))
Err("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()
@ -524,13 +291,13 @@ pub fn parse_type_annotation<T>(
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!(
return Err(format!(
"Unexpected number of type parameters: expected {} but got {}",
type_vars.len(),
types.len()
)]));
));
}
let mut subst = VarMap::new();
let mut subst = HashMap::new();
for (var, ty) in izip!(type_vars.iter(), types.iter()) {
let id = if let TypeEnum::TVar { id, .. } = &*unifier.get_ty(*var) {
*id
@ -552,7 +319,7 @@ pub fn parse_type_annotation<T>(
}));
Ok(unifier.add_ty(TypeEnum::TObj { obj_id, fields, params: subst }))
} else {
Err(HashSet::from(["Cannot use function name as type".into()]))
Err("Cannot use function name as type".into())
}
}
};
@ -563,13 +330,10 @@ pub fn parse_type_annotation<T>(
if let Name { id, .. } = &value.node {
subscript_name_handle(id, slice, unifier)
} else {
Err(HashSet::from([format!("unsupported type expression at {}", expr.location)]))
Err(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)])),
_ => Err(format!("unsupported type expression at {}", expr.location)),
}
}
@ -580,7 +344,7 @@ impl dyn SymbolResolver + Send + Sync {
unifier: &mut Unifier,
primitives: &PrimitiveStore,
expr: &Expr<T>,
) -> Result<Type, HashSet<String>> {
) -> Result<Type, String> {
parse_type_annotation(self, top_level_defs, unifier, primitives, expr)
}
@ -593,13 +357,13 @@ impl dyn SymbolResolver + Send + Sync {
unifier.internal_stringify(
ty,
&mut |id| {
let TopLevelDef::Class { name, .. } = &*top_level_defs[id].read() else {
unreachable!("expected class definition")
};
if let TopLevelDef::Class { name, .. } = &*top_level_defs[id].read() {
name.to_string()
} else {
unreachable!("expected class definition")
}
},
&mut |id| format!("typevar{id}"),
&mut |id| format!("var{}", id),
&mut None,
)
}

2771
nac3core/src/toplevel/builtins.rs
View File

File diff suppressed because it is too large Load Diff

979
nac3core/src/toplevel/composer.rs
View File

File diff suppressed because it is too large Load Diff

683
nac3core/src/toplevel/helper.rs
View File

@ -1,298 +1,10 @@
use std::convert::TryInto;
use crate::symbol_resolver::SymbolValue;
use crate::toplevel::numpy::unpack_ndarray_var_tys;
use crate::typecheck::typedef::{into_var_map, iter_type_vars, Mapping, TypeVarId, VarMap};
use nac3parser::ast::{Constant, Location};
use strum::IntoEnumIterator;
use strum_macros::EnumIter;
use super::*;
/// All primitive types and functions in nac3core.
#[derive(Clone, Copy, Debug, EnumIter, PartialEq, Eq)]
pub enum PrimDef {
// Classes
Int32,
Int64,
Float,
Bool,
None,
Range,
Str,
Exception,
UInt32,
UInt64,
Option,
List,
NDArray,
// Member Functions
OptionIsSome,
OptionIsNone,
OptionUnwrap,
NDArrayCopy,
NDArrayFill,
FunInt32,
FunInt64,
FunUInt32,
FunUInt64,
FunFloat,
FunNpNDArray,
FunNpEmpty,
FunNpZeros,
FunNpOnes,
FunNpFull,
FunNpArray,
FunNpEye,
FunNpIdentity,
FunRound,
FunRound64,
FunNpRound,
FunRangeInit,
FunStr,
FunBool,
FunFloor,
FunFloor64,
FunNpFloor,
FunCeil,
FunCeil64,
FunNpCeil,
FunLen,
FunMin,
FunNpMin,
FunNpMinimum,
FunMax,
FunNpMax,
FunNpMaximum,
FunAbs,
FunNpIsNan,
FunNpIsInf,
FunNpSin,
FunNpCos,
FunNpExp,
FunNpExp2,
FunNpLog,
FunNpLog10,
FunNpLog2,
FunNpFabs,
FunNpSqrt,
FunNpRint,
FunNpTan,
FunNpArcsin,
FunNpArccos,
FunNpArctan,
FunNpSinh,
FunNpCosh,
FunNpTanh,
FunNpArcsinh,
FunNpArccosh,
FunNpArctanh,
FunNpExpm1,
FunNpCbrt,
FunSpSpecErf,
FunSpSpecErfc,
FunSpSpecGamma,
FunSpSpecGammaln,
FunSpSpecJ0,
FunSpSpecJ1,
FunNpArctan2,
FunNpCopysign,
FunNpFmax,
FunNpFmin,
FunNpLdExp,
FunNpHypot,
FunNpNextAfter,
// Top-Level Functions
FunSome,
}
/// Associated details of a [`PrimDef`]
pub enum PrimDefDetails {
PrimFunction { name: &'static str, simple_name: &'static str },
PrimClass { name: &'static str },
}
impl PrimDef {
/// Get the assigned [`DefinitionId`] of this [`PrimDef`].
///
/// The assigned definition ID is defined by the position this [`PrimDef`] enum unit variant is defined at,
/// with the first `PrimDef`'s definition id being `0`.
#[must_use]
pub fn id(&self) -> DefinitionId {
DefinitionId(*self as usize)
}
/// Check if a definition ID is that of a [`PrimDef`].
#[must_use]
pub fn contains_id(id: DefinitionId) -> bool {
Self::iter().any(|prim| prim.id() == id)
}
/// Get the definition "simple name" of this [`PrimDef`].
///
/// If the [`PrimDef`] is a function, this corresponds to [`TopLevelDef::Function::simple_name`].
///
/// If the [`PrimDef`] is a class, this returns [`None`].
#[must_use]
pub fn simple_name(&self) -> &'static str {
match self.details() {
PrimDefDetails::PrimFunction { simple_name, .. } => simple_name,
PrimDefDetails::PrimClass { .. } => {
panic!("PrimDef {self:?} has no simple_name as it is not a function.")
}
}
}
/// Get the definition "name" of this [`PrimDef`].
///
/// If the [`PrimDef`] is a function, this corresponds to [`TopLevelDef::Function::name`].
///
/// If the [`PrimDef`] is a class, this corresponds to [`TopLevelDef::Class::name`].
#[must_use]
pub fn name(&self) -> &'static str {
match self.details() {
PrimDefDetails::PrimFunction { name, .. } | PrimDefDetails::PrimClass { name } => name,
}
}
/// Get the associated details of this [`PrimDef`]
#[must_use]
pub fn details(self) -> PrimDefDetails {
fn class(name: &'static str) -> PrimDefDetails {
PrimDefDetails::PrimClass { name }
}
fn fun(name: &'static str, simple_name: Option<&'static str>) -> PrimDefDetails {
PrimDefDetails::PrimFunction { simple_name: simple_name.unwrap_or(name), name }
}
match self {
PrimDef::Int32 => class("int32"),
PrimDef::Int64 => class("int64"),
PrimDef::Float => class("float"),
PrimDef::Bool => class("bool"),
PrimDef::None => class("none"),
PrimDef::Range => class("range"),
PrimDef::Str => class("str"),
PrimDef::Exception => class("Exception"),
PrimDef::UInt32 => class("uint32"),
PrimDef::UInt64 => class("uint64"),
PrimDef::Option => class("Option"),
PrimDef::OptionIsSome => fun("Option.is_some", Some("is_some")),
PrimDef::OptionIsNone => fun("Option.is_none", Some("is_none")),
PrimDef::OptionUnwrap => fun("Option.unwrap", Some("unwrap")),
PrimDef::List => class("list"),
PrimDef::NDArray => class("ndarray"),
PrimDef::NDArrayCopy => fun("ndarray.copy", Some("copy")),
PrimDef::NDArrayFill => fun("ndarray.fill", Some("fill")),
PrimDef::FunInt32 => fun("int32", None),
PrimDef::FunInt64 => fun("int64", None),
PrimDef::FunUInt32 => fun("uint32", None),
PrimDef::FunUInt64 => fun("uint64", None),
PrimDef::FunFloat => fun("float", None),
PrimDef::FunNpNDArray => fun("np_ndarray", None),
PrimDef::FunNpEmpty => fun("np_empty", None),
PrimDef::FunNpZeros => fun("np_zeros", None),
PrimDef::FunNpOnes => fun("np_ones", None),
PrimDef::FunNpFull => fun("np_full", None),
PrimDef::FunNpArray => fun("np_array", None),
PrimDef::FunNpEye => fun("np_eye", None),
PrimDef::FunNpIdentity => fun("np_identity", None),
PrimDef::FunRound => fun("round", None),
PrimDef::FunRound64 => fun("round64", None),
PrimDef::FunNpRound => fun("np_round", None),
PrimDef::FunRangeInit => fun("range.__init__", Some("__init__")),
PrimDef::FunStr => fun("str", None),
PrimDef::FunBool => fun("bool", None),
PrimDef::FunFloor => fun("floor", None),
PrimDef::FunFloor64 => fun("floor64", None),
PrimDef::FunNpFloor => fun("np_floor", None),
PrimDef::FunCeil => fun("ceil", None),
PrimDef::FunCeil64 => fun("ceil64", None),
PrimDef::FunNpCeil => fun("np_ceil", None),
PrimDef::FunLen => fun("len", None),
PrimDef::FunMin => fun("min", None),
PrimDef::FunNpMin => fun("np_min", None),
PrimDef::FunNpMinimum => fun("np_minimum", None),
PrimDef::FunMax => fun("max", None),
PrimDef::FunNpMax => fun("np_max", None),
PrimDef::FunNpMaximum => fun("np_maximum", None),
PrimDef::FunAbs => fun("abs", None),
PrimDef::FunNpIsNan => fun("np_isnan", None),
PrimDef::FunNpIsInf => fun("np_isinf", None),
PrimDef::FunNpSin => fun("np_sin", None),
PrimDef::FunNpCos => fun("np_cos", None),
PrimDef::FunNpExp => fun("np_exp", None),
PrimDef::FunNpExp2 => fun("np_exp2", None),
PrimDef::FunNpLog => fun("np_log", None),
PrimDef::FunNpLog10 => fun("np_log10", None),
PrimDef::FunNpLog2 => fun("np_log2", None),
PrimDef::FunNpFabs => fun("np_fabs", None),
PrimDef::FunNpSqrt => fun("np_sqrt", None),
PrimDef::FunNpRint => fun("np_rint", None),
PrimDef::FunNpTan => fun("np_tan", None),
PrimDef::FunNpArcsin => fun("np_arcsin", None),
PrimDef::FunNpArccos => fun("np_arccos", None),
PrimDef::FunNpArctan => fun("np_arctan", None),
PrimDef::FunNpSinh => fun("np_sinh", None),
PrimDef::FunNpCosh => fun("np_cosh", None),
PrimDef::FunNpTanh => fun("np_tanh", None),
PrimDef::FunNpArcsinh => fun("np_arcsinh", None),
PrimDef::FunNpArccosh => fun("np_arccosh", None),
PrimDef::FunNpArctanh => fun("np_arctanh", None),
PrimDef::FunNpExpm1 => fun("np_expm1", None),
PrimDef::FunNpCbrt => fun("np_cbrt", None),
PrimDef::FunSpSpecErf => fun("sp_spec_erf", None),
PrimDef::FunSpSpecErfc => fun("sp_spec_erfc", None),
PrimDef::FunSpSpecGamma => fun("sp_spec_gamma", None),
PrimDef::FunSpSpecGammaln => fun("sp_spec_gammaln", None),
PrimDef::FunSpSpecJ0 => fun("sp_spec_j0", None),
PrimDef::FunSpSpecJ1 => fun("sp_spec_j1", None),
PrimDef::FunNpArctan2 => fun("np_arctan2", None),
PrimDef::FunNpCopysign => fun("np_copysign", None),
PrimDef::FunNpFmax => fun("np_fmax", None),
PrimDef::FunNpFmin => fun("np_fmin", None),
PrimDef::FunNpLdExp => fun("np_ldexp", None),
PrimDef::FunNpHypot => fun("np_hypot", None),
PrimDef::FunNpNextAfter => fun("np_nextafter", None),
PrimDef::FunSome => fun("Some", None),
}
}
}
/// Asserts that a [`PrimDef`] is in an allowlist.
///
/// Like `debug_assert!`, this statements of this function are only
/// enabled if `cfg!(debug_assertions)` is true.
pub fn debug_assert_prim_is_allowed(prim: PrimDef, allowlist: &[PrimDef]) {
if cfg!(debug_assertions) {
let allowed = allowlist.iter().any(|p| *p == prim);
assert!(
allowed,
"Disallowed primitive definition. Got {prim:?}, but expects it to be in {allowlist:?}"
);
}
}
/// Construct the fields of class `Exception`
/// See [`TypeEnum::TObj::fields`] and [`TopLevelDef::Class::fields`]
#[must_use]
pub fn make_exception_fields(int32: Type, int64: Type, str: Type) -> Vec<(StrRef, Type, bool)> {
vec![
("__name__".into(), int32, true),
("__file__".into(), str, true),
("__line__".into(), int32, true),
("__col__".into(), int32, true),
("__func__".into(), str, true),
("__message__".into(), str, true),
("__param0__".into(), int64, true),
("__param1__".into(), int64, true),
("__param2__".into(), int64, true),
]
}
impl TopLevelDef {
pub fn to_string(&self, unifier: &mut Unifier) -> String {
match self {
@ -331,160 +43,115 @@ impl TopLevelDef {
}
impl TopLevelComposer {
#[must_use]
pub fn make_primitives(size_t: u32) -> (PrimitiveStore, Unifier) {
pub fn make_primitives() -> (PrimitiveStore, Unifier) {
let mut unifier = Unifier::new();
let int32 = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Int32.id(),
obj_id: DefinitionId(0),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let int64 = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Int64.id(),
obj_id: DefinitionId(1),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let float = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Float.id(),
obj_id: DefinitionId(2),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let bool = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Bool.id(),
obj_id: DefinitionId(3),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let none = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::None.id(),
obj_id: DefinitionId(4),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let range = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Range.id(),
fields: [
("start".into(), (int32, true)),
("stop".into(), (int32, true)),
("step".into(), (int32, true)),
]
.into_iter()
.collect(),
params: VarMap::new(),
obj_id: DefinitionId(5),
fields: HashMap::new(),
params: HashMap::new(),
});
let str = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Str.id(),
obj_id: DefinitionId(6),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let exception = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Exception.id(),
fields: make_exception_fields(int32, int64, str)
obj_id: DefinitionId(7),
fields: vec![
("__name__".into(), (int32, true)),
("__file__".into(), (int32, true)),
("__line__".into(), (int32, true)),
("__col__".into(), (int32, true)),
("__func__".into(), (str, true)),
("__message__".into(), (str, true)),
("__param0__".into(), (int64, true)),
("__param1__".into(), (int64, true)),
("__param2__".into(), (int64, true)),
]
.into_iter()
.map(|(name, ty, mutable)| (name, (ty, mutable)))
.collect(),
params: VarMap::new(),
.collect::<HashMap<_, _>>(),
params: HashMap::new(),
});
let uint32 = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::UInt32.id(),
obj_id: DefinitionId(8),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let uint64 = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::UInt64.id(),
obj_id: DefinitionId(9),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let option_type_var = unifier.get_fresh_var(Some("option_type_var".into()), None);
let is_some_type_fun_ty = unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![],
ret: bool,
vars: into_var_map([option_type_var]),
vars: HashMap::from([(option_type_var.1, option_type_var.0)]),
}));
let unwrap_fun_ty = unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![],
ret: option_type_var.ty,
vars: into_var_map([option_type_var]),
ret: option_type_var.0,
vars: HashMap::from([(option_type_var.1, option_type_var.0)]),
}));
let option = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Option.id(),
obj_id: DefinitionId(10),
fields: vec![
(PrimDef::OptionIsSome.simple_name().into(), (is_some_type_fun_ty, true)),
(PrimDef::OptionIsNone.simple_name().into(), (is_some_type_fun_ty, true)),
(PrimDef::OptionUnwrap.simple_name().into(), (unwrap_fun_ty, true)),
("is_some".into(), (is_some_type_fun_ty, true)),
("is_none".into(), (is_some_type_fun_ty, true)),
("unwrap".into(), (unwrap_fun_ty, true)),
]
.into_iter()
.collect::<HashMap<_, _>>(),
params: into_var_map([option_type_var]),
params: HashMap::from([(option_type_var.1, option_type_var.0)]),
});
let size_t_ty = match size_t {
32 => uint32,
64 => uint64,
_ => unreachable!(),
};
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: Mapping::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(size_t_ty, Some("ndarray_ndims".into()), None);
let ndarray_copy_fun_ret_ty = unifier.get_fresh_var(None, None);
let ndarray_copy_fun_ty = unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![],
ret: ndarray_copy_fun_ret_ty.ty,
vars: into_var_map([ndarray_dtype_tvar, ndarray_ndims_tvar]),
}));
let ndarray_fill_fun_ty = unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![FuncArg {
name: "value".into(),
ty: ndarray_dtype_tvar.ty,
default_value: None,
}],
ret: none,
vars: into_var_map([ndarray_dtype_tvar, ndarray_ndims_tvar]),
}));
let ndarray = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::NDArray.id(),
fields: Mapping::from([
(PrimDef::NDArrayCopy.simple_name().into(), (ndarray_copy_fun_ty, true)),
(PrimDef::NDArrayFill.simple_name().into(), (ndarray_fill_fun_ty, true)),
]),
params: into_var_map([ndarray_dtype_tvar, ndarray_ndims_tvar]),
});
unifier.unify(ndarray_copy_fun_ret_ty.ty, ndarray).unwrap();
let primitives = PrimitiveStore {
int32,
int64,
uint32,
uint64,
float,
bool,
none,
range,
str,
exception,
uint32,
uint64,
option,
list,
ndarray,
size_t,
};
unifier.put_primitive_store(&primitives);
crate::typecheck::magic_methods::set_primitives_magic_methods(&primitives, &mut unifier);
(primitives, unifier)
}
/// already include the `definition_id` of itself inside the ancestors vector
/// when first registering, the `type_vars`, fields, methods, ancestors are invalid
#[must_use]
/// already include the definition_id of itself inside the ancestors vector
/// when first regitering, the type_vars, fields, methods, ancestors are invalid
pub fn make_top_level_class_def(
obj_id: DefinitionId,
index: usize,
resolver: Option<Arc<dyn SymbolResolver + Send + Sync>>,
name: StrRef,
constructor: Option<Type>,
@ -492,12 +159,11 @@ impl TopLevelComposer {
) -> TopLevelDef {
TopLevelDef::Class {
name,
object_id: obj_id,
type_vars: Vec::default(),
fields: Vec::default(),
attributes: Vec::default(),
methods: Vec::default(),
ancestors: Vec::default(),
object_id: DefinitionId(index),
type_vars: Default::default(),
fields: Default::default(),
methods: Default::default(),
ancestors: Default::default(),
constructor,
resolver,
loc,
@ -505,7 +171,6 @@ impl TopLevelComposer {
}
/// when first registering, the type is a invalid value
#[must_use]
pub fn make_top_level_function_def(
name: String,
simple_name: StrRef,
@ -517,16 +182,15 @@ impl TopLevelComposer {
name,
simple_name,
signature: ty,
var_id: Vec::default(),
instance_to_symbol: HashMap::default(),
instance_to_stmt: HashMap::default(),
var_id: Default::default(),
instance_to_symbol: Default::default(),
instance_to_stmt: Default::default(),
resolver,
codegen_callback: None,
loc,
}
}
#[must_use]
pub fn make_class_method_name(mut class_name: String, method_name: &str) -> String {
class_name.push('.');
class_name.push_str(method_name);
@ -536,13 +200,13 @@ impl TopLevelComposer {
pub fn get_class_method_def_info(
class_methods_def: &[(StrRef, Type, DefinitionId)],
method_name: StrRef,
) -> Result<(Type, DefinitionId), HashSet<String>> {
) -> Result<(Type, DefinitionId), String> {
for (name, ty, def_id) in class_methods_def {
if name == &method_name {
return Ok((*ty, *def_id));
}
}
Err(HashSet::from([format!("no method {method_name} in the current class")]))
Err(format!("no method {} in the current class", method_name))
}
/// get all base class def id of a class, excluding itself. \
@ -553,7 +217,7 @@ impl TopLevelComposer {
pub fn get_all_ancestors_helper(
child: &TypeAnnotation,
temp_def_list: &[Arc<RwLock<TopLevelDef>>],
) -> Result<Vec<TypeAnnotation>, HashSet<String>> {
) -> Result<Vec<TypeAnnotation>, String> {
let mut result: Vec<TypeAnnotation> = Vec::new();
let mut parent = Self::get_parent(child, temp_def_list);
while let Some(p) = parent {
@ -565,16 +229,16 @@ impl TopLevelComposer {
};
// check cycle
let no_cycle = result.iter().all(|x| {
let TypeAnnotation::CustomClass { id, .. } = x else {
unreachable!("must be class kind annotation")
};
if let TypeAnnotation::CustomClass { id, .. } = x {
id.0 != p_id.0
} else {
unreachable!("must be class kind annotation")
}
});
if no_cycle {
result.push(p);
} else {
return Err(HashSet::from(["cyclic inheritance detected".into()]));
return Err("cyclic inheritance detected".into());
}
}
Ok(result)
@ -592,23 +256,23 @@ impl TopLevelComposer {
};
let child_def = temp_def_list.get(child_id.0).unwrap();
let child_def = child_def.read();
let TopLevelDef::Class { ancestors, .. } = &*child_def else {
unreachable!("child must be top level class def")
};
if ancestors.is_empty() {
None
} else {
if let TopLevelDef::Class { ancestors, .. } = &*child_def {
if !ancestors.is_empty() {
Some(ancestors[0].clone())
} else {
None
}
} else {
unreachable!("child must be top level class def")
}
}
/// get the `var_id` of a given `TVar` type
pub fn get_var_id(var_ty: Type, unifier: &mut Unifier) -> Result<TypeVarId, HashSet<String>> {
/// get the var_id of a given TVar type
pub fn get_var_id(var_ty: Type, unifier: &mut Unifier) -> Result<u32, String> {
if let TypeEnum::TVar { id, .. } = unifier.get_ty(var_ty).as_ref() {
Ok(*id)
} else {
Err(HashSet::from(["not type var".to_string()]))
Err("not type var".to_string())
}
}
@ -622,17 +286,13 @@ impl TopLevelComposer {
let this = this.as_ref();
let other = unifier.get_ty(other);
let other = other.as_ref();
let (
if let (
TypeEnum::TFunc(FunSignature { args: this_args, ret: this_ret, .. }),
TypeEnum::TFunc(FunSignature { args: other_args, ret: other_ret, .. }),
) = (this, other)
else {
unreachable!("this function must be called with function type")
};
{
// check args
let args_ok =
this_args
let args_ok = this_args
.iter()
.map(|FuncArg { name, ty, .. }| (name, type_var_to_concrete_def.get(ty).unwrap()))
.zip(other_args.iter().map(|FuncArg { name, ty, .. }| {
@ -656,6 +316,9 @@ impl TopLevelComposer {
// return
args_ok && ret_ok
} else {
unreachable!("this function must be called with function type")
}
}
pub fn check_overload_field_type(
@ -671,7 +334,7 @@ impl TopLevelComposer {
)
}
pub fn get_all_assigned_field(stmts: &[Stmt<()>]) -> Result<HashSet<StrRef>, HashSet<String>> {
pub fn get_all_assigned_field(stmts: &[ast::Stmt<()>]) -> Result<HashSet<StrRef>, String> {
let mut result = HashSet::new();
for s in stmts {
match &s.node {
@ -688,10 +351,10 @@ impl TopLevelComposer {
}
} =>
{
return Err(HashSet::from([format!(
return Err(format!(
"redundant type annotation for class fields at {}",
s.location
)]))
))
}
ast::StmtKind::Assign { targets, .. } => {
for t in targets {
@ -713,14 +376,14 @@ impl TopLevelComposer {
ast::StmtKind::If { body, orelse, .. } => {
let inited_for_sure = Self::get_all_assigned_field(body.as_slice())?
.intersection(&Self::get_all_assigned_field(orelse.as_slice())?)
.copied()
.cloned()
.collect::<HashSet<_>>();
result.extend(inited_for_sure);
}
ast::StmtKind::Try { body, orelse, finalbody, .. } => {
let inited_for_sure = Self::get_all_assigned_field(body.as_slice())?
.intersection(&Self::get_all_assigned_field(orelse.as_slice())?)
.copied()
.cloned()
.collect::<HashSet<_>>();
result.extend(inited_for_sure);
result.extend(Self::get_all_assigned_field(finalbody.as_slice())?);
@ -728,9 +391,9 @@ impl TopLevelComposer {
ast::StmtKind::With { body, .. } => {
result.extend(Self::get_all_assigned_field(body.as_slice())?);
}
ast::StmtKind::Pass { .. }
| ast::StmtKind::Assert { .. }
| ast::StmtKind::Expr { .. } => {}
ast::StmtKind::Pass { .. } => {}
ast::StmtKind::Assert { .. } => {}
ast::StmtKind::Expr { .. } => {}
_ => {
unimplemented!()
@ -743,7 +406,7 @@ impl TopLevelComposer {
pub fn parse_parameter_default_value(
default: &ast::Expr,
resolver: &(dyn SymbolResolver + Send + Sync),
) -> Result<SymbolValue, HashSet<String>> {
) -> Result<SymbolValue, String> {
parse_parameter_default_value(default, resolver)
}
@ -753,6 +416,40 @@ impl TopLevelComposer {
primitive: &PrimitiveStore,
unifier: &mut Unifier,
) -> Result<(), String> {
fn type_default_param(
val: &SymbolValue,
primitive: &PrimitiveStore,
unifier: &mut Unifier,
) -> TypeAnnotation {
match val {
SymbolValue::Bool(..) => TypeAnnotation::Primitive(primitive.bool),
SymbolValue::Double(..) => TypeAnnotation::Primitive(primitive.float),
SymbolValue::I32(..) => TypeAnnotation::Primitive(primitive.int32),
SymbolValue::I64(..) => TypeAnnotation::Primitive(primitive.int64),
SymbolValue::U32(..) => TypeAnnotation::Primitive(primitive.uint32),
SymbolValue::U64(..) => TypeAnnotation::Primitive(primitive.uint64),
SymbolValue::Str(..) => TypeAnnotation::Primitive(primitive.str),
SymbolValue::Tuple(vs) => {
let vs_tys = vs
.iter()
.map(|v| type_default_param(v, primitive, unifier))
.collect::<Vec<_>>();
TypeAnnotation::Tuple(vs_tys)
}
SymbolValue::OptionNone => TypeAnnotation::CustomClass {
id: primitive.option.get_obj_id(unifier),
params: Default::default(),
},
SymbolValue::OptionSome(v) => {
let ty = type_default_param(v, primitive, unifier);
TypeAnnotation::CustomClass {
id: primitive.option.get_obj_id(unifier),
params: vec![ty],
}
}
}
}
fn is_compatible(
found: &TypeAnnotation,
expect: &TypeAnnotation,
@ -768,7 +465,7 @@ impl TopLevelComposer {
TypeAnnotation::CustomClass { id: e_id, params: e_param },
) => {
*f_id == *e_id
&& *f_id == primitive.option.obj_id(unifier).unwrap()
&& *f_id == primitive.option.get_obj_id(unifier)
&& (f_param.is_empty()
|| (f_param.len() == 1
&& e_param.len() == 1
@ -784,15 +481,15 @@ impl TopLevelComposer {
}
}
let found = val.get_type_annotation(primitive, unifier);
if is_compatible(&found, ty, unifier, primitive) {
Ok(())
} else {
let found = type_default_param(val, primitive, unifier);
if !is_compatible(&found, ty, unifier, primitive) {
Err(format!(
"incompatible default parameter type, expect {}, found {}",
ty.stringify(unifier),
found.stringify(unifier),
))
} else {
Ok(())
}
}
}
@ -800,14 +497,14 @@ impl TopLevelComposer {
pub fn parse_parameter_default_value(
default: &ast::Expr,
resolver: &(dyn SymbolResolver + Send + Sync),
) -> Result<SymbolValue, HashSet<String>> {
fn handle_constant(val: &Constant, loc: &Location) -> Result<SymbolValue, HashSet<String>> {
) -> Result<SymbolValue, String> {
fn handle_constant(val: &Constant, loc: &Location) -> Result<SymbolValue, String> {
match val {
Constant::Int(v) => {
if let Ok(v) = (*v).try_into() {
Ok(SymbolValue::I32(v))
} else {
Err(HashSet::from([format!("integer value out of range at {loc}")]))
Err(format!("integer value out of range at {}", loc))
}
}
Constant::Float(v) => Ok(SymbolValue::Double(*v)),
@ -815,126 +512,74 @@ pub fn parse_parameter_default_value(
Constant::Tuple(tuple) => Ok(SymbolValue::Tuple(
tuple.iter().map(|x| handle_constant(x, loc)).collect::<Result<Vec<_>, _>>()?,
)),
Constant::None => Err(HashSet::from([format!(
"`None` is not supported, use `none` for option type instead ({loc})"
)])),
Constant::None => Err(format!(
"`None` is not supported, use `none` for option type instead ({})",
loc
)),
_ => unimplemented!("this constant is not supported at {}", loc),
}
}
match &default.node {
ast::ExprKind::Constant { value, .. } => handle_constant(value, &default.location),
ast::ExprKind::Call { func, args, .. } if args.len() == 1 => match &func.node {
ast::ExprKind::Call { func, args, .. } if args.len() == 1 => {
match &func.node {
ast::ExprKind::Name { id, .. } if *id == "int64".into() => match &args[0].node {
ast::ExprKind::Constant { value: Constant::Int(v), .. } => {
let v: Result<i64, _> = (*v).try_into();
match v {
Ok(v) => Ok(SymbolValue::I64(v)),
_ => Err(HashSet::from([format!(
"default param value out of range at {}",
default.location
)])),
_ => Err(format!("default param value out of range at {}", default.location)),
}
}
_ => Err(HashSet::from([format!(
"only allow constant integer here at {}",
default.location
)])),
},
_ => Err(format!("only allow constant integer here at {}", default.location))
}
ast::ExprKind::Name { id, .. } if *id == "uint32".into() => match &args[0].node {
ast::ExprKind::Constant { value: Constant::Int(v), .. } => {
let v: Result<u32, _> = (*v).try_into();
match v {
Ok(v) => Ok(SymbolValue::U32(v)),
_ => Err(HashSet::from([format!(
"default param value out of range at {}",
default.location
)])),
_ => Err(format!("default param value out of range at {}", default.location)),
}
}
_ => Err(HashSet::from([format!(
"only allow constant integer here at {}",
default.location
)])),
},
_ => Err(format!("only allow constant integer here at {}", default.location))
}
ast::ExprKind::Name { id, .. } if *id == "uint64".into() => match &args[0].node {
ast::ExprKind::Constant { value: Constant::Int(v), .. } => {
let v: Result<u64, _> = (*v).try_into();
match v {
Ok(v) => Ok(SymbolValue::U64(v)),
_ => Err(HashSet::from([format!(
"default param value out of range at {}",
default.location
)])),
_ => Err(format!("default param value out of range at {}", default.location)),
}
}
_ => Err(HashSet::from([format!(
"only allow constant integer here at {}",
default.location
)])),
},
ast::ExprKind::Name { id, .. } if *id == "Some".into() => Ok(SymbolValue::OptionSome(
Box::new(parse_parameter_default_value(&args[0], resolver)?),
)),
_ => Err(HashSet::from([format!(
"unsupported default parameter at {}",
default.location
)])),
},
ast::ExprKind::Tuple { elts, .. } => Ok(SymbolValue::Tuple(
elts.iter()
_ => Err(format!("only allow constant integer here at {}", default.location))
}
ast::ExprKind::Name { id, .. } if *id == "Some".into() => Ok(
SymbolValue::OptionSome(
Box::new(parse_parameter_default_value(&args[0], resolver)?)
)
),
_ => Err(format!("unsupported default parameter at {}", default.location)),
}
}
ast::ExprKind::Tuple { elts, .. } => Ok(SymbolValue::Tuple(elts
.iter()
.map(|x| parse_parameter_default_value(x, resolver))
.collect::<Result<Vec<_>, _>>()?,
.collect::<Result<Vec<_>, _>>()?
)),
ast::ExprKind::Name { id, .. } if id == &"none".into() => Ok(SymbolValue::OptionNone),
ast::ExprKind::Name { id, .. } => {
resolver.get_default_param_value(default).ok_or_else(|| {
HashSet::from([format!(
resolver.get_default_param_value(default).ok_or_else(
|| format!(
"`{}` cannot be used as a default parameter at {} \
(not primitive type, option or tuple / not defined?)",
id, default.location
)])
})
id,
default.location
)
)
}
_ => Err(HashSet::from([format!(
_ => Err(format!(
"unsupported default parameter (not primitive type, option or tuple) at {}",
default.location
)])),
}
}
/// Obtains the element type of an array-like type.
pub fn arraylike_flatten_element_type(unifier: &mut Unifier, ty: Type) -> Type {
match &*unifier.get_ty(ty) {
TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::NDArray.id() => {
unpack_ndarray_var_tys(unifier, ty).0
}
TypeEnum::TObj { obj_id, params, .. } if *obj_id == PrimDef::List.id() => {
arraylike_flatten_element_type(unifier, iter_type_vars(params).next().unwrap().ty)
}
_ => ty,
}
}
/// Obtains the number of dimensions of an array-like type.
pub fn arraylike_get_ndims(unifier: &mut Unifier, ty: Type) -> u64 {
match &*unifier.get_ty(ty) {
TypeEnum::TObj { obj_id, .. } if *obj_id == PrimDef::NDArray.id() => {
let ndims = unpack_ndarray_var_tys(unifier, ty).1;
let TypeEnum::TLiteral { values, .. } = &*unifier.get_ty_immutable(ndims) else {
panic!("Expected TLiteral for ndarray.ndims, got {}", unifier.stringify(ndims))
};
if values.len() > 1 {
todo!("Getting num of dimensions for ndarray with more than one ndim bound is unimplemented")
}
u64::try_from(values[0].clone()).unwrap()
}
TypeEnum::TObj { obj_id, params, .. } if *obj_id == PrimDef::List.id() => {
arraylike_get_ndims(unifier, iter_type_vars(params).next().unwrap().ty) + 1
}
_ => 0,
))
}
}

87
nac3core/src/toplevel/mod.rs
View File

@ -3,24 +3,20 @@ use std::{
collections::{HashMap, HashSet},
fmt::Debug,
iter::FromIterator,
ops::{Deref, DerefMut},
sync::Arc,
};
use super::codegen::CodeGenContext;
use super::typecheck::type_inferencer::PrimitiveStore;
use super::typecheck::typedef::{
FunSignature, FuncArg, SharedUnifier, Type, TypeEnum, Unifier, VarMap,
};
use super::typecheck::typedef::{FunSignature, FuncArg, SharedUnifier, Type, TypeEnum, Unifier};
use crate::{
codegen::CodeGenerator,
symbol_resolver::{SymbolResolver, ValueEnum},
typecheck::{
type_inferencer::CodeLocation,
typedef::{CallId, TypeVarId},
},
typecheck::{type_inferencer::CodeLocation, typedef::CallId},
};
use inkwell::values::BasicValueEnum;
use itertools::Itertools;
use itertools::{izip, Itertools};
use nac3parser::ast::{self, Location, Stmt, StrRef};
use parking_lot::RwLock;
@ -30,14 +26,14 @@ 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(
type GenCallCallback = Box<
dyn for<'ctx, 'a> Fn(
&mut CodeGenContext<'ctx, 'a>,
Option<(Type, ValueEnum<'ctx>)>,
(&FunSignature, DefinitionId),
@ -45,28 +41,21 @@ type GenCallCallback = dyn for<'ctx, 'a> Fn(
&mut dyn CodeGenerator,
) -> Result<Option<BasicValueEnum<'ctx>>, String>
+ Send
+ Sync;
+ Sync,
>;
pub struct GenCall {
fp: Box<GenCallCallback>,
fp: GenCallCallback,
}
impl GenCall {
#[must_use]
pub fn new(fp: Box<GenCallCallback>) -> GenCall {
pub fn new(fp: 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>(
pub fn run<'ctx, 'a>(
&self,
ctx: &mut CodeGenContext<'ctx, '_>,
ctx: &mut CodeGenContext<'ctx, 'a>,
obj: Option<(Type, ValueEnum<'ctx>)>,
fun: (&FunSignature, DefinitionId),
args: Vec<(Option<StrRef>, ValueEnum<'ctx>)>,
@ -86,66 +75,58 @@ impl Debug for GenCall {
pub struct FunInstance {
pub body: Arc<Vec<Stmt<Option<Type>>>>,
pub calls: Arc<HashMap<CodeLocation, CallId>>,
pub subst: VarMap,
pub subst: HashMap<u32, Type>,
pub unifier_id: usize,
}
#[derive(Debug, Clone)]
pub enum TopLevelDef {
Class {
/// Name for error messages and symbols.
// name for error messages and symbols
name: StrRef,
/// Object ID used for [`TypeEnum`].
// 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.
// class fields
// name, 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.
// class methods, pointing to the corresponding function definition.
methods: Vec<(StrRef, Type, DefinitionId)>,
/// Ancestor classes, including itself.
// ancestor classes, including itself.
ancestors: Vec<TypeAnnotation>,
/// Symbol resolver of the module defined the class; [None] if it is built-in type.
// 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 type
constructor: Option<Type>,
/// Definition location.
// definition location
loc: Option<Location>,
},
Function {
/// Prefix for symbol, should be unique globally.
// prefix for symbol, should be unique globally
name: String,
/// Simple name, the same as in method/function definition.
// simple name, the same as in method/function definition
simple_name: StrRef,
/// Function signature.
// function signature.
signature: Type,
/// Instantiated type variable IDs.
var_id: Vec<TypeVarId>,
// instantiated type variable IDs
var_id: Vec<u32>,
/// Function instance to symbol mapping
///
/// * Key: String representation of type variable values, sorted by variable ID in ascending
/// 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.
/// 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
/// 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.
/// 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.
// symbol resolver of the module defined the class
resolver: Option<Arc<dyn SymbolResolver + Send + Sync>>,
/// Custom code generation callback.
// custom codegen callback
codegen_callback: Option<Arc<GenCall>>,
/// Definition location.
// definition location
loc: Option<Location>,
},
}

85
nac3core/src/toplevel/numpy.rs
View File

@ -1,85 +0,0 @@
use crate::{
toplevel::helper::PrimDef,
typecheck::{
type_inferencer::PrimitiveStore,
typedef::{Type, TypeEnum, TypeVarId, Unifier, VarMap},
},
};
use itertools::Itertools;
/// 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()
}

4
nac3core/src/toplevel/snapshots/nac3core__toplevel__test__test_analyze__generic_class.snap
View File

@ -1,11 +1,13 @@
---
source: nac3core/src/toplevel/test.rs
assertion_line: 549
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(245)]\n}\n",
"Function {\nname: \"Generic_A.fun\",\nsig: \"fn[[a:int32], V]\",\nvar_id: [18]\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",

4
nac3core/src/toplevel/snapshots/nac3core__toplevel__test__test_analyze__inheritance_override.snap
View File

@ -1,13 +1,15 @@
---
source: nac3core/src/toplevel/test.rs
assertion_line: 549
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[typevar234]\", \"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: [\"typevar234\"]\n}\n",
"Class {\nname: \"B\",\nancestors: [\"B[var7]\", \"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: [\"var7\"]\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",

8
nac3core/src/toplevel/snapshots/nac3core__toplevel__test__test_analyze__list_tuple_generic.snap
View File

@ -1,13 +1,15 @@
---
source: nac3core/src/toplevel/test.rs
assertion_line: 549
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(247)]\n}\n",
"Function {\nname: \"A.fun\",\nsig: \"fn[[a:T], V]\",\nvar_id: [TypeVarId(252)]\n}\n",
"Function {\nname: \"gfun\",\nsig: \"fn[[a:A[list[float], int32]], none]\",\nvar_id: []\n}\n",
"Function {\nname: \"A.__init__\",\nsig: \"fn[[v:V], none]\",\nvar_id: [20]\n}\n",
"Function {\nname: \"A.fun\",\nsig: \"fn[[a:T], V]\",\nvar_id: [25]\n}\n",
"Function {\nname: \"gfun\",\nsig: \"fn[[a:A[int32, list[float]]], 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",
]

12
nac3core/src/toplevel/snapshots/nac3core__toplevel__test__test_analyze__self1.snap
View File

@ -1,13 +1,15 @@
---
source: nac3core/src/toplevel/test.rs
assertion_line: 549
expression: res_vec
---
[
"Class {\nname: \"A\",\nancestors: [\"A[typevar233, typevar234]\"],\nfields: [\"a\", \"b\"],\nmethods: [(\"__init__\", \"fn[[a:A[float, bool], b:B], none]\"), (\"fun\", \"fn[[a:A[float, bool]], A[bool, int32]]\")],\ntype_vars: [\"typevar233\", \"typevar234\"]\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",
"Class {\nname: \"A\",\nancestors: [\"A[var6, var7]\"],\nfields: [\"a\", \"b\"],\nmethods: [(\"__init__\", \"fn[[a:A[bool, float], b:B], none]\"), (\"fun\", \"fn[[a:A[bool, float]], A[bool, int32]]\")],\ntype_vars: [\"var6\", \"var7\"]\n}\n",
"Function {\nname: \"A.__init__\",\nsig: \"fn[[a:A[bool, float], b:B], none]\",\nvar_id: []\n}\n",
"Function {\nname: \"A.fun\",\nsig: \"fn[[a:A[bool, float]], 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[bool, float]], A[bool, int32]]\"), (\"foo\", \"fn[[b:B], B]\"), (\"bar\", \"fn[[a:A[int32, list[B]]], tuple[A[bool, virtual[A[B, int32]]], 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",
"Function {\nname: \"B.bar\",\nsig: \"fn[[a:A[int32, list[B]]], tuple[A[bool, virtual[A[B, int32]]], B]]\",\nvar_id: []\n}\n",
]

6
nac3core/src/toplevel/snapshots/nac3core__toplevel__test__test_analyze__simple_class_compose.snap
View File

@ -1,17 +1,19 @@
---
source: nac3core/src/toplevel/test.rs
assertion_line: 549
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(253)]\n}\n",
"Function {\nname: \"A.foo\",\nsig: \"fn[[a:T, b:V], none]\",\nvar_id: [26]\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(261)]\n}\n",
"Function {\nname: \"ff\",\nsig: \"fn[[a:T], V]\",\nvar_id: [34]\n}\n",
]

190
nac3core/src/toplevel/test.rs
View File

@ -1,6 +1,3 @@
use super::*;
use crate::toplevel::helper::PrimDef;
use crate::typecheck::typedef::into_var_map;
use crate::{
codegen::CodeGenContext,
symbol_resolver::{SymbolResolver, ValueEnum},
@ -11,12 +8,13 @@ use crate::{
},
};
use indoc::indoc;
use nac3parser::ast::FileName;
use nac3parser::{ast::fold::Fold, parser::parse_program};
use parking_lot::Mutex;
use std::{collections::HashMap, sync::Arc};
use test_case::test_case;
use super::*;
struct ResolverInternal {
id_to_type: Mutex<HashMap<StrRef, Type>>,
id_to_def: Mutex<HashMap<StrRef, DefinitionId>>,
@ -38,7 +36,7 @@ struct Resolver(Arc<ResolverInternal>);
impl SymbolResolver for Resolver {
fn get_default_param_value(
&self,
_: &ast::Expr,
_: &nac3parser::ast::Expr,
) -> Option<crate::symbol_resolver::SymbolValue> {
unimplemented!()
}
@ -54,25 +52,20 @@ impl SymbolResolver for Resolver {
.id_to_type
.lock()
.get(&str)
.copied()
.ok_or_else(|| format!("cannot find symbol `{str}`"))
.cloned()
.ok_or_else(|| format!("cannot find symbol `{}`", str))
}
fn get_symbol_value<'ctx>(
fn get_symbol_value<'ctx, 'a>(
&self,
_: StrRef,
_: &mut CodeGenContext<'ctx, '_>,
_: &mut CodeGenContext<'ctx, 'a>,
) -> 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_identifier_def(&self, id: StrRef) -> Result<DefinitionId, String> {
self.0.id_to_def.lock().get(&id).cloned().ok_or_else(|| "Unknown identifier".to_string())
}
fn get_string_id(&self, _: &str) -> i32 {
@ -112,39 +105,23 @@ impl SymbolResolver for Resolver {
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(), ComposerConfig::default(), 64).0;
let mut composer: TopLevelComposer = Default::default();
for s in source {
let ast = parse_program(s, FileName::default()).unwrap();
let ast = parse_program(s, Default::default()).unwrap();
let ast = ast[0].clone();
composer.register_top_level(ast, None, "", false).unwrap();
composer.register_top_level(ast, None, "".into()).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(), 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(
&[
vec![
indoc! {"
def fun(a: int32) -> int32:
return a
@ -158,35 +135,35 @@ fn test_simple_register_without_constructor(source: &str) {
return 3
"},
],
&[
vec![
"fn[[a:0], 0]",
"fn[[a:2], 4]",
"fn[[b:1], 0]",
],
&[
vec![
"fun",
"foo",
"f"
];
"function compose"
)]
fn test_simple_function_analyze(source: &[&str], tys: &[&str], names: &[&str]) {
let mut composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 64).0;
fn test_simple_function_analyze(source: Vec<&str>, tys: Vec<&str>, names: Vec<&str>) {
let mut composer: TopLevelComposer = Default::default();
let internal_resolver = Arc::new(ResolverInternal {
id_to_def: Mutex::default(),
id_to_type: Mutex::default(),
class_names: Mutex::default(),
id_to_def: Default::default(),
id_to_type: Default::default(),
class_names: Default::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 = parse_program(s, Default::default()).unwrap();
let ast = ast[0].clone();
let (id, def_id, ty) =
composer.register_top_level(ast, Some(resolver.clone()), "", false).unwrap();
composer.register_top_level(ast, Some(resolver.clone()), "".into()).unwrap();
internal_resolver.add_id_def(id, def_id);
if let Some(ty) = ty {
internal_resolver.add_id_type(id, ty);
@ -212,7 +189,7 @@ fn test_simple_function_analyze(source: &[&str], tys: &[&str], names: &[&str]) {
}
#[test_case(
&[
vec![
indoc! {"
class A():
a: int32
@ -245,11 +222,11 @@ fn test_simple_function_analyze(source: &[&str], tys: &[&str], names: &[&str]) {
pass
"}
],
&[];
vec![];
"simple class compose"
)]
#[test_case(
&[
vec![
indoc! {"
class Generic_A(Generic[V], B):
a: int64
@ -267,11 +244,11 @@ fn test_simple_function_analyze(source: &[&str], tys: &[&str], names: &[&str]) {
pass
"}
],
&[];
vec![];
"generic class"
)]
#[test_case(
&[
vec![
indoc! {"
def foo(a: list[int32], b: tuple[T, float]) -> A[B, bool]:
pass
@ -296,11 +273,11 @@ fn test_simple_function_analyze(source: &[&str], tys: &[&str], names: &[&str]) {
pass
"}
],
&[];
vec![];
"list tuple generic"
)]
#[test_case(
&[
vec![
indoc! {"
class A(Generic[T, V]):
a: A[float, bool]
@ -321,11 +298,11 @@ fn test_simple_function_analyze(source: &[&str], tys: &[&str], names: &[&str]) {
pass
"}
],
&[];
vec![];
"self1"
)]
#[test_case(
&[
vec![
indoc! {"
class A(Generic[T]):
a: int32
@ -355,11 +332,11 @@ fn test_simple_function_analyze(source: &[&str], tys: &[&str], names: &[&str]) {
pass
"}
],
&[];
vec![];
"inheritance_override"
)]
#[test_case(
&[
vec![
indoc! {"
class A(Generic[T]):
def __init__(self):
@ -368,11 +345,11 @@ fn test_simple_function_analyze(source: &[&str], tys: &[&str], names: &[&str]) {
pass
"}
],
&["application of type vars to generic class is not currently supported (at unknown:4:24)"];
vec!["application of type vars to generic class is not currently supported (at unknown: line 4 column 24)"];
"err no type var in generic app"
)]
#[test_case(
&[
vec![
indoc! {"
class A(B):
def __init__(self):
@ -384,11 +361,11 @@ fn test_simple_function_analyze(source: &[&str], tys: &[&str], names: &[&str]) {
pass
"}
],
&["cyclic inheritance detected"];
vec!["cyclic inheritance detected"];
"cyclic1"
)]
#[test_case(
&[
vec![
indoc! {"
class A(B[bool, int64]):
def __init__(self):
@ -405,30 +382,30 @@ fn test_simple_function_analyze(source: &[&str], tys: &[&str], names: &[&str]) {
pass
"},
],
&["cyclic inheritance detected"];
vec!["cyclic inheritance detected"];
"cyclic2"
)]
#[test_case(
&[
vec![
indoc! {"
class A:
pass
"}
],
&["5: Class {\nname: \"A\",\ndef_id: DefinitionId(5),\nancestors: [CustomClassKind { id: DefinitionId(5), params: [] }],\nfields: [],\nmethods: [],\ntype_vars: []\n}"];
vec!["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! {"
vec![indoc! {"
class A:
def __init__():
pass
"}],
&["__init__ method must have a `self` parameter (at unknown:2:5)"];
vec!["__init__ method must have a `self` parameter (at unknown: line 2 column 5)"];
"err no self_1"
)]
#[test_case(
&[
vec![
indoc! {"
class A(B, Generic[T], C):
def __init__(self):
@ -446,11 +423,11 @@ fn test_simple_function_analyze(source: &[&str], tys: &[&str], names: &[&str]) {
"}
],
&["a class definition can only have at most one base class declaration and one generic declaration (at unknown:1:24)"];
vec!["a class definition can only have at most one base class declaration and one generic declaration (at unknown: line 1 column 24)"];
"err multiple inheritance"
)]
#[test_case(
&[
vec![
indoc! {"
class A(Generic[T]):
a: int32
@ -471,11 +448,11 @@ fn test_simple_function_analyze(source: &[&str], tys: &[&str], names: &[&str]) {
pass
"}
],
&["method fun has same name as ancestors' method, but incompatible type"];
vec!["method fun has same name as ancestors' method, but incompatible type"];
"err_incompatible_inheritance_method"
)]
#[test_case(
&[
vec![
indoc! {"
class A(Generic[T]):
a: int32
@ -497,11 +474,11 @@ fn test_simple_function_analyze(source: &[&str], tys: &[&str], names: &[&str]) {
pass
"}
],
&["field `a` has already declared in the ancestor classes"];
vec!["field `a` has already declared in the ancestor classes"];
"err_incompatible_inheritance_field"
)]
#[test_case(
&[
vec![
indoc! {"
class A:
def __init__(self):
@ -514,12 +491,12 @@ fn test_simple_function_analyze(source: &[&str], tys: &[&str], names: &[&str]) {
pass
"}
],
&["duplicate definition of class `A` (at unknown:1:1)"];
vec!["duplicate definition of class `A` (at unknown: line 1 column 1)"];
"class same name"
)]
fn test_analyze(source: &[&str], res: &[&str]) {
fn test_analyze(source: Vec<&str>, res: Vec<&str>) {
let print = false;
let mut composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 64).0;
let mut composer: TopLevelComposer = Default::default();
let internal_resolver = make_internal_resolver_with_tvar(
vec![
@ -534,15 +511,15 @@ fn test_analyze(source: &[&str], res: &[&str]) {
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 = parse_program(s, Default::default()).unwrap();
let ast = ast[0].clone();
let (id, def_id, ty) = {
match composer.register_top_level(ast, Some(resolver.clone()), "", false) {
match composer.register_top_level(ast, Some(resolver.clone()), "".into()) {
Ok(x) => x,
Err(msg) => {
if print {
println!("{msg}");
println!("{}", msg);
} else {
assert_eq!(res[0], msg);
}
@ -558,9 +535,9 @@ fn test_analyze(source: &[&str], res: &[&str]) {
if let Err(msg) = composer.start_analysis(false) {
if print {
println!("{}", msg.iter().sorted().join("\n----------\n"));
println!("{}", msg);
} else {
assert_eq!(res[0], msg.iter().next().unwrap());
assert_eq!(res[0], msg);
}
} else {
// skip 5 to skip primitives
@ -588,7 +565,7 @@ fn test_analyze(source: &[&str], res: &[&str]) {
return fib(n - 1)
"}
],
&[];
vec![];
"simple function"
)]
#[test_case(
@ -621,7 +598,7 @@ fn test_analyze(source: &[&str], res: &[&str]) {
return a.fun() + 2
"}
],
&[];
vec![];
"simple class body"
)]
#[test_case(
@ -646,7 +623,7 @@ fn test_analyze(source: &[&str], res: &[&str]) {
return [a, b]
"}
],
&[];
vec![];
"type var fun"
)]
#[test_case(
@ -667,7 +644,7 @@ fn test_analyze(source: &[&str], res: &[&str]) {
return ret if self.b else self.fun(self.a)
"}
],
&[];
vec![];
"type var class"
)]
#[test_case(
@ -691,12 +668,12 @@ fn test_analyze(source: &[&str], res: &[&str]) {
self.b = True
"}
],
&[];
vec![];
"no_init_inst_check"
)]
fn test_inference(source: Vec<&str>, res: &[&str]) {
fn test_inference(source: Vec<&str>, res: Vec<&str>) {
let print = true;
let mut composer = TopLevelComposer::new(Vec::new(), ComposerConfig::default(), 64).0;
let mut composer: TopLevelComposer = Default::default();
let internal_resolver = make_internal_resolver_with_tvar(
vec![
@ -718,15 +695,15 @@ fn test_inference(source: Vec<&str>, res: &[&str]) {
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 = parse_program(s, Default::default()).unwrap();
let ast = ast[0].clone();
let (id, def_id, ty) = {
match composer.register_top_level(ast, Some(resolver.clone()), "", false) {
match composer.register_top_level(ast, Some(resolver.clone()), "".into()) {
Ok(x) => x,
Err(msg) => {
if print {
println!("{msg}");
println!("{}", msg);
} else {
assert_eq!(res[0], msg);
}
@ -742,14 +719,16 @@ fn test_inference(source: Vec<&str>, res: &[&str]) {
if let Err(msg) = composer.start_analysis(true) {
if print {
println!("{}", msg.iter().sorted().join("\n----------\n"));
println!("{}", msg);
} else {
assert_eq!(res[0], msg.iter().next().unwrap());
assert_eq!(res[0], msg);
}
} 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) {
for (_i, (def, _)) in
composer.definition_ast_list.iter().skip(composer.builtin_num).enumerate()
{
let def = &*def.read();
if let TopLevelDef::Function { instance_to_stmt, name, .. } = def {
@ -758,7 +737,7 @@ fn test_inference(source: Vec<&str>, res: &[&str]) {
name,
instance_to_stmt.len()
);
for inst in instance_to_stmt {
for inst in instance_to_stmt.iter() {
let ast = &inst.1.body;
for b in ast.iter() {
println!("{:?}", stringify_folder.fold_stmt(b.clone()).unwrap());
@ -776,29 +755,22 @@ fn make_internal_resolver_with_tvar(
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_def: Default::default(),
id_to_type: tvars
.into_iter()
.map(|(name, range)| {
(name, {
let tvar = unifier.get_fresh_var_with_range(range.as_slice(), None, None);
let (ty, id) = unifier.get_fresh_var_with_range(range.as_slice(), None, None);
if print {
println!("{}: {:?}, typevar{}", name, tvar.ty, tvar.id);
println!("{}: {:?}, tvar{}", name, ty, id);
}
tvar.ty
ty
})
})
.collect::<HashMap<_, _>>()
.into(),
class_names: Mutex::new(HashMap::from([("list".into(), list)])),
class_names: Default::default(),
}
.into();
if print {
@ -818,8 +790,8 @@ impl<'a> Fold<Option<Type>> for TypeToStringFolder<'a> {
Ok(if let Some(ty) = user {
self.unifier.internal_stringify(
ty,
&mut |id| format!("class{id}"),
&mut |id| format!("typevar{id}"),
&mut |id| format!("class{}", id.to_string()),
&mut |id| format!("tvar{}", id.to_string()),
&mut None,
)
} else {

323
nac3core/src/toplevel/type_annotation.rs
View File

@ -1,8 +1,4 @@
use super::*;
use crate::symbol_resolver::SymbolValue;
use crate::toplevel::helper::PrimDef;
use crate::typecheck::typedef::VarMap;
use nac3parser::ast::Constant;
#[derive(Clone, Debug)]
pub enum TypeAnnotation {
@ -16,8 +12,7 @@ pub enum TypeAnnotation {
// can only be CustomClassKind
Virtual(Box<TypeAnnotation>),
TypeVar(Type),
/// A `Literal` allowing a subset of literals.
Literal(Vec<Constant>),
List(Box<TypeAnnotation>),
Tuple(Vec<TypeAnnotation>),
}
@ -27,57 +22,52 @@ impl 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() {
let class_name = match unifier.top_level {
Some(ref top) => {
if let TopLevelDef::Class { name, .. } =
&*top.definitions.read()[id.0].read()
{
(*name).into()
} else {
unreachable!()
}
} else {
format!("class_def_{}", id.0)
}
None => format!("class_def_{}", id.0),
};
format!("{}{}", class_name, {
let param_list =
params.iter().map(|p| p.stringify(unifier)).collect_vec().join(", ");
format!(
"{}{}",
class_name,
{
let param_list = params.iter().map(|p| p.stringify(unifier)).collect_vec().join(", ");
if param_list.is_empty() {
String::new()
"".into()
} else {
format!("[{param_list}]")
format!("[{}]", param_list)
}
})
}
Literal(values) => {
format!("Literal({})", values.iter().map(|v| format!("{v:?}")).join(", "))
)
}
Virtual(ty) => format!("virtual[{}]", ty.stringify(unifier)),
List(ty) => format!("list[{}]", ty.stringify(unifier)),
Tuple(types) => {
format!(
"tuple[{}]",
types.iter().map(|p| p.stringify(unifier)).collect_vec().join(", ")
)
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>(
pub fn parse_ast_to_type_annotation_kinds<T>(
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>> {
locked: HashMap<DefinitionId, Vec<Type>>,
) -> Result<TypeAnnotation, String> {
let name_handle = |id: &StrRef,
unifier: &mut Unifier,
locked: HashMap<DefinitionId, Vec<Type>, S>| {
locked: HashMap<DefinitionId, Vec<Type>>| {
if id == &"int32".into() {
Ok(TypeAnnotation::Primitive(primitives.int32))
} else if id == &"int64".into() {
@ -93,7 +83,7 @@ pub fn parse_ast_to_type_annotation_kinds<T, S: std::hash::BuildHasher + Clone>(
} 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() })
Ok(TypeAnnotation::CustomClass { id: DefinitionId(7), params: Default::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();
@ -101,10 +91,10 @@ pub fn parse_ast_to_type_annotation_kinds<T, S: std::hash::BuildHasher + Clone>(
if let TopLevelDef::Class { type_vars, .. } = &*def_read {
type_vars.clone()
} else {
return Err(HashSet::from([format!(
return Err(format!(
"function cannot be used as a type (at {})",
expr.location
)]));
));
}
} else {
locked.get(&obj_id).unwrap().clone()
@ -112,29 +102,23 @@ pub fn parse_ast_to_type_annotation_kinds<T, S: std::hash::BuildHasher + Clone>(
};
// check param number here
if !type_vars.is_empty() {
return Err(HashSet::from([format!(
return Err(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;
let var = unifier.get_fresh_var(Some(*id), Some(expr.location)).0;
unifier.unify(var, ty).unwrap();
Ok(TypeAnnotation::TypeVar(ty))
} else {
Err(HashSet::from([format!(
"`{}` is not a valid type annotation (at {})",
id, expr.location
)]))
Err(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
)]))
Err(format!("`{}` is not a valid type annotation (at {})", id, expr.location))
}
};
@ -142,22 +126,20 @@ pub fn parse_ast_to_type_annotation_kinds<T, S: std::hash::BuildHasher + Clone>(
|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
)]));
mut locked: HashMap<DefinitionId, Vec<Type>>| {
if vec!["virtual".into(), "Generic".into(), "list".into(), "tuple".into()].contains(id)
{
return Err(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")
};
if let TopLevelDef::Class { type_vars, .. } = &*def_read {
type_vars.clone()
} else {
unreachable!("must be class here")
}
} else {
locked.get(&obj_id).unwrap().clone()
}
@ -170,12 +152,12 @@ pub fn parse_ast_to_type_annotation_kinds<T, S: std::hash::BuildHasher + Clone>(
vec![slice]
};
if type_vars.len() != params_ast.len() {
return Err(HashSet::from([format!(
return Err(format!(
"expect {} type parameters but got {} (at {})",
type_vars.len(),
params_ast.len(),
params_ast[0].location,
)]));
));
}
let result = params_ast
.iter()
@ -199,17 +181,15 @@ pub fn parse_ast_to_type_annotation_kinds<T, S: std::hash::BuildHasher + Clone>(
if no_type_var {
result
} else {
return Err(HashSet::from([
format!(
"application of type vars to generic class is not currently supported (at {})",
return Err(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
@ -232,6 +212,23 @@ pub fn parse_ast_to_type_annotation_kinds<T, S: std::hash::BuildHasher + Clone>(
Ok(TypeAnnotation::Virtual(def.into()))
}
// list
ast::ExprKind::Subscript { value, slice, .. }
if {
matches!(&value.node, ast::ExprKind::Name { id, .. } if id == &"list".into())
} =>
{
let def_ann = parse_ast_to_type_annotation_kinds(
resolver,
top_level_defs,
unifier,
primitives,
slice.as_ref(),
locked,
)?;
Ok(TypeAnnotation::List(def_ann.into()))
}
// option
ast::ExprKind::Subscript { value, slice, .. }
if {
@ -284,70 +281,16 @@ pub fn parse_ast_to_type_annotation_kinds<T, S: std::hash::BuildHasher + Clone>(
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
)]))
Err(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
)])),
_ => Err(format!("unsupported expression for type annotation (at {})", expr.location)),
}
}
@ -357,46 +300,43 @@ pub fn parse_ast_to_type_annotation_kinds<T, S: std::hash::BuildHasher + Clone>(
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>> {
subst_list: &mut Option<Vec<Type>>
) -> Result<Type, 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")
};
let class_def: &TopLevelDef = def_read.deref();
if let TopLevelDef::Class { fields, methods, type_vars, .. } = class_def {
if type_vars.len() != params.len() {
return Err(HashSet::from([format!(
Err(format!(
"unexpected number of type parameters: expected {} but got {}",
type_vars.len(),
params.len()
)]));
}
))
} else {
let param_ty = params
.iter()
.map(|x| {
get_type_from_type_annotation_kinds(top_level_defs, unifier, x, subst_list)
get_type_from_type_annotation_kinds(
top_level_defs,
unifier,
primitives,
x,
subst_list
)
})
.collect::<Result<Vec<_>, _>>()?;
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();
let mut result: HashMap<u32, Type> = HashMap::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,
} => {
if let TypeEnum::TVar { id, range, fields: None, name, loc } =
unifier.get_ty(*tvar).as_ref()
{
let ok: bool = {
// create a temp type var and unify to check compatibility
p == *tvar || {
@ -405,51 +345,29 @@ pub fn get_type_from_type_annotation_kinds(
*name,
*loc,
);
unifier.unify(temp.ty, p).is_ok()
unifier.unify(temp.0, p).is_ok()
}
};
if ok {
result.insert(*id, p);
} else {
return Err(HashSet::from([format!(
return Err(format!(
"cannot apply type {} to type variable with id {:?}",
unifier.internal_stringify(
p,
&mut |id| format!("class{id}"),
&mut |id| format!("typevar{id}"),
&mut |id| format!("class{}", id),
&mut |id| format!("tvar{}", 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"),
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, _)| {
@ -469,37 +387,40 @@ pub fn get_type_from_type_annotation_kinds(
params: subst,
});
if need_subst {
if let Some(wl) = subst_list.as_mut() {
wl.push(ty);
}
subst_list.as_mut().map(|wl| wl.push(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)
} else {
unreachable!("should be class def here")
}
}
TypeAnnotation::Primitive(ty) | TypeAnnotation::TypeVar(ty) => Ok(*ty),
TypeAnnotation::Virtual(ty) => {
let ty = get_type_from_type_annotation_kinds(
top_level_defs,
unifier,
primitives,
ty.as_ref(),
subst_list,
subst_list
)?;
Ok(unifier.add_ty(TypeEnum::TVirtual { ty }))
}
TypeAnnotation::List(ty) => {
let ty = get_type_from_type_annotation_kinds(
top_level_defs,
unifier,
primitives,
ty.as_ref(),
subst_list
)?;
Ok(unifier.add_ty(TypeEnum::TList { ty }))
}
TypeAnnotation::Tuple(tys) => {
let tys = tys
.iter()
.map(|x| {
get_type_from_type_annotation_kinds(top_level_defs, unifier, x, subst_list)
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 }))
@ -515,11 +436,10 @@ pub fn get_type_from_type_annotation_kinds(
/// 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
/// But note that here we do not make a duplication of `T`, `V`, we direclty
/// 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,
@ -530,25 +450,27 @@ pub fn make_self_type_annotation(type_vars: &[Type], object_id: DefinitionId) ->
/// 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()));
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::List(ann) => {
result.extend(get_type_var_contained_in_type_annotation(ann.as_ref()))
}
TypeAnnotation::Tuple(anns) => {
for a in anns {
result.extend(get_type_var_contained_in_type_annotation(a));
}
}
TypeAnnotation::Primitive(..) | TypeAnnotation::Literal { .. } => {}
TypeAnnotation::Primitive(..) => {}
}
result
}
@ -563,20 +485,21 @@ pub fn check_overload_type_annotation_compatible(
(TypeAnnotation::Primitive(a), TypeAnnotation::Primitive(b)) => a == b,
(TypeAnnotation::TypeVar(a), TypeAnnotation::TypeVar(b)) => {
let a = unifier.get_ty(*a);
let a = &*a;
let a = a.deref();
let b = unifier.get_ty(*b);
let b = &*b;
let (
let b = b.deref();
if let (
TypeEnum::TVar { id: a, fields: None, .. },
TypeEnum::TVar { id: b, fields: None, .. },
) = (a, b)
else {
unreachable!("must be type var")
};
{
a == b
} else {
unreachable!("must be type var")
}
(TypeAnnotation::Virtual(a), TypeAnnotation::Virtual(b)) => {
}
(TypeAnnotation::Virtual(a), TypeAnnotation::Virtual(b))
| (TypeAnnotation::List(a), TypeAnnotation::List(b)) => {
check_overload_type_annotation_compatible(a.as_ref(), b.as_ref(), unifier)
}

130
nac3core/src/typecheck/function_check.rs
View File

@ -1,18 +1,14 @@
use crate::toplevel::helper::PrimDef;
use crate::typecheck::typedef::TypeEnum;
use super::type_inferencer::Inferencer;
use super::typedef::{Type, TypeEnum};
use nac3parser::ast::{
self, Constant, Expr, ExprKind,
Operator::{LShift, RShift},
Stmt, StmtKind, StrRef,
};
use super::typedef::Type;
use nac3parser::ast::{self, Expr, ExprKind, Stmt, StmtKind, StrRef};
use std::{collections::HashSet, iter::once};
impl<'a> Inferencer<'a> {
fn should_have_value(&mut self, expr: &Expr<Option<Type>>) -> Result<(), HashSet<String>> {
fn should_have_value(&mut self, expr: &Expr<Option<Type>>) -> Result<(), 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)]))
Err(format!("Error at {}: cannot have value none", expr.location))
} else {
Ok(())
}
@ -22,11 +18,10 @@ impl<'a> Inferencer<'a> {
&mut self,
pattern: &Expr<Option<Type>>,
defined_identifiers: &mut HashSet<StrRef>,
) -> Result<(), HashSet<String>> {
) -> Result<(), String> {
match &pattern.node {
ExprKind::Name { id, .. } if id == &"none".into() => {
Err(HashSet::from([format!("cannot assign to a `none` (at {})", pattern.location)]))
}
ast::ExprKind::Name { id, .. } if id == &"none".into() =>
Err(format!("cannot assign to a `none` (at {})", pattern.location)),
ExprKind::Name { id, .. } => {
if !defined_identifiers.contains(id) {
defined_identifiers.insert(*id);
@ -35,7 +30,7 @@ impl<'a> Inferencer<'a> {
Ok(())
}
ExprKind::Tuple { elts, .. } => {
for elt in elts {
for elt in elts.iter() {
self.check_pattern(elt, defined_identifiers)?;
self.should_have_value(elt)?;
}
@ -46,17 +41,16 @@ impl<'a> Inferencer<'a> {
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!(
return Err(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
)])),
ExprKind::Constant { .. } => {
Err(format!("cannot assign to a constant (at {})", pattern.location))
}
_ => self.check_expr(pattern, defined_identifiers),
}
}
@ -65,18 +59,15 @@ impl<'a> Inferencer<'a> {
&mut self,
expr: &Expr<Option<Type>>,
defined_identifiers: &mut HashSet<StrRef>,
) -> Result<(), HashSet<String>> {
) -> Result<(), 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!(
if !self.unifier.is_concrete(*ty, &self.function_data.bound_variables) {
return Err(format!(
"expected concrete type at {} but got {}",
expr.location,
self.unifier.get_ty(*ty).get_type_name()
)]));
));
}
}
match &expr.node {
@ -96,10 +87,10 @@ impl<'a> Inferencer<'a> {
self.defined_identifiers.insert(*id);
}
Err(e) => {
return Err(HashSet::from([format!(
return Err(format!(
"type error at identifier `{}` ({}) at {}",
id, e, expr.location
)]))
));
}
}
}
@ -107,7 +98,7 @@ impl<'a> Inferencer<'a> {
ExprKind::List { elts, .. }
| ExprKind::Tuple { elts, .. }
| ExprKind::BoolOp { values: elts, .. } => {
for elt in elts {
for elt in elts.iter() {
self.check_expr(elt, defined_identifiers)?;
self.should_have_value(elt)?;
}
@ -116,25 +107,11 @@ impl<'a> Inferencer<'a> {
self.check_expr(value, defined_identifiers)?;
self.should_have_value(value)?;
}
ExprKind::BinOp { left, op, right } => {
ExprKind::BinOp { left, 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)?;
@ -164,7 +141,7 @@ impl<'a> Inferencer<'a> {
}
ExprKind::Lambda { args, body } => {
let mut defined_identifiers = defined_identifiers.clone();
for arg in &args.args {
for arg in args.args.iter() {
// TODO: should we check the types here?
if !defined_identifiers.contains(&arg.node.arg) {
defined_identifiers.insert(arg.node.arg);
@ -202,45 +179,24 @@ impl<'a> Inferencer<'a> {
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 {
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 HashSet<StrRef>,
) -> Result<bool, HashSet<String>> {
) -> Result<bool, 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 {
for stmt in orelse.iter() {
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 {
for stmt in body.iter() {
self.check_stmt(stmt, &mut local_defined_identifiers)?;
}
Ok(false)
@ -253,7 +209,7 @@ impl<'a> Inferencer<'a> {
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 {
for ident in body_identifiers.iter() {
if !defined_identifiers.contains(ident) && orelse_identifiers.contains(ident) {
defined_identifiers.insert(*ident);
}
@ -270,7 +226,7 @@ impl<'a> Inferencer<'a> {
}
StmtKind::With { items, body, .. } => {
let mut new_defined_identifiers = defined_identifiers.clone();
for item in items {
for item in items.iter() {
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)?;
@ -282,7 +238,7 @@ impl<'a> Inferencer<'a> {
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 {
for handler in handlers.iter() {
let mut defined_identifiers = defined_identifiers.clone();
let ast::ExcepthandlerKind::ExceptHandler { name, body, .. } = &handler.node;
if let Some(name) = name {
@ -317,30 +273,6 @@ impl<'a> Inferencer<'a> {
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)
}
@ -359,11 +291,11 @@ impl<'a> Inferencer<'a> {
&mut self,
block: &[Stmt<Option<Type>>],
defined_identifiers: &mut HashSet<StrRef>,
) -> Result<bool, HashSet<String>> {
) -> Result<bool, String> {
let mut ret = false;
for stmt in block {
if ret {
eprintln!("warning: dead code at {}\n", stmt.location);
return Err(format!("dead code at {:?}", stmt.location));
}
if self.check_stmt(stmt, defined_identifiers)? {
ret = true;

720
nac3core/src/typecheck/magic_methods.rs
View File

@ -1,150 +1,69 @@
use crate::symbol_resolver::SymbolValue;
use crate::toplevel::helper::PrimDef;
use crate::toplevel::numpy::{make_ndarray_ty, unpack_ndarray_var_tys};
use crate::typecheck::{
type_inferencer::*,
typedef::{FunSignature, FuncArg, Type, TypeEnum, Unifier, VarMap},
typedef::{FunSignature, FuncArg, Type, TypeEnum, Unifier},
};
use itertools::{iproduct, Itertools};
use nac3parser::ast::StrRef;
use nac3parser::ast::{self, StrRef};
use nac3parser::ast::{Cmpop, Operator, Unaryop};
use std::cmp::max;
use std::collections::HashMap;
use std::rc::Rc;
use strum::IntoEnumIterator;
/// 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 }
}
}
/// 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,
}
/// 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 trait HasOpInfo {
fn op_info(&self) -> OpInfo;
}
fn try_get_cmpop_info(op: Cmpop) -> Option<OpInfo> {
pub fn binop_name(op: &Operator) -> &'static str {
match op {
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", "!=")),
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__",
}
}
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__",
}
}
pub fn unaryop_name(op: &Unaryop) -> &'static str {
match op {
Unaryop::UAdd => "__pos__",
Unaryop::USub => "__neg__",
Unaryop::Not => "__not__",
Unaryop::Invert => "__inv__",
}
}
pub fn comparison_name(op: &Cmpop) -> Option<&'static str> {
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__"),
_ => 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)>),
@ -164,31 +83,26 @@ where
pub fn impl_binop(
unifier: &mut Unifier,
_store: &PrimitiveStore,
store: &PrimitiveStore,
ty: Type,
other_ty: &[Type],
ret_ty: Option<Type>,
ops: &[Operator],
ret_ty: Type,
ops: &[ast::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 (ty, var_id) = unifier.get_fresh_var_with_range(other_ty, Some("N".into()), None);
(ty, Some(var_id))
};
let function_vars = if let Some(var_id) = other_var_id {
vec![(var_id, other_ty)].into_iter().collect::<VarMap>()
vec![(var_id, other_ty)].into_iter().collect::<HashMap<_, _>>()
} else {
VarMap::new()
HashMap::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(), {
for op in ops {
fields.insert(binop_name(op).into(), {
(
unifier.add_ty(TypeEnum::TFunc(FunSignature {
ret: ret_ty,
@ -202,21 +116,34 @@ pub fn impl_binop(
false,
)
});
fields.insert(binop_assign_name(op).into(), {
(
unifier.add_ty(TypeEnum::TFunc(FunSignature {
ret: store.none,
vars: function_vars.clone(),
args: vec![FuncArg {
ty: other_ty,
default_value: None,
name: "other".into(),
}],
})),
false,
)
});
}
});
}
pub fn impl_unaryop(unifier: &mut Unifier, ty: Type, ret_ty: Option<Type>, ops: &[Unaryop]) {
pub fn impl_unaryop(unifier: &mut Unifier, ty: Type, ret_ty: Type, ops: &[ast::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(),
unaryop_name(op).into(),
(
unifier.add_ty(TypeEnum::TFunc(FunSignature {
ret: ret_ty,
vars: VarMap::new(),
vars: HashMap::new(),
args: vec![],
})),
false,
@ -228,35 +155,19 @@ pub fn impl_unaryop(unifier: &mut Unifier, ty: Type, ret_ty: Option<Type>, ops:
pub fn impl_cmpop(
unifier: &mut Unifier,
_store: &PrimitiveStore,
store: &PrimitiveStore,
ty: Type,
other_ty: &[Type],
ops: &[Cmpop],
ret_ty: Option<Type>,
other_ty: Type,
ops: &[ast::Cmpop],
) {
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(),
comparison_name(op).unwrap().into(),
(
unifier.add_ty(TypeEnum::TFunc(FunSignature {
ret: ret_ty,
vars: function_vars.clone(),
ret: store.bool,
vars: HashMap::new(),
args: vec![FuncArg {
ty: other_ty,
default_value: None,
@ -270,13 +181,13 @@ pub fn impl_cmpop(
});
}
/// `Add`, `Sub`, `Mult`
/// Add, Sub, Mult
pub fn impl_basic_arithmetic(
unifier: &mut Unifier,
store: &PrimitiveStore,
ty: Type,
other_ty: &[Type],
ret_ty: Option<Type>,
ret_ty: Type,
) {
impl_binop(
unifier,
@ -284,390 +195,94 @@ pub fn impl_basic_arithmetic(
ty,
other_ty,
ret_ty,
&[Operator::Add, Operator::Sub, Operator::Mult],
);
&[ast::Operator::Add, ast::Operator::Sub, ast::Operator::Mult],
)
}
/// `Pow`
/// Pow
pub fn impl_pow(
unifier: &mut Unifier,
store: &PrimitiveStore,
ty: Type,
other_ty: &[Type],
ret_ty: Option<Type>,
ret_ty: Type,
) {
impl_binop(unifier, store, ty, other_ty, ret_ty, &[Operator::Pow]);
impl_binop(unifier, store, ty, other_ty, ret_ty, &[ast::Operator::Pow])
}
/// `BitOr`, `BitXor`, `BitAnd`
/// 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],
);
&[ast::Operator::BitAnd, ast::Operator::BitOr, ast::Operator::BitXor],
)
}
/// `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]);
/// LShift, RShift
pub fn impl_bitwise_shift(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type) {
impl_binop(unifier, store, ty, &[ty], ty, &[ast::Operator::LShift, ast::Operator::RShift])
}
/// `FloorDiv`
/// Div
pub fn impl_div(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type, other_ty: &[Type]) {
impl_binop(unifier, store, ty, other_ty, store.float, &[ast::Operator::Div])
}
/// FloorDiv
pub fn impl_floordiv(
unifier: &mut Unifier,
store: &PrimitiveStore,
ty: Type,
other_ty: &[Type],
ret_ty: Option<Type>,
ret_ty: Type,
) {
impl_binop(unifier, store, ty, other_ty, ret_ty, &[Operator::FloorDiv]);
impl_binop(unifier, store, ty, other_ty, ret_ty, &[ast::Operator::FloorDiv])
}
/// `Mod`
/// Mod
pub fn impl_mod(
unifier: &mut Unifier,
store: &PrimitiveStore,
ty: Type,
other_ty: &[Type],
ret_ty: Option<Type>,
ret_ty: Type,
) {
impl_binop(unifier, store, ty, other_ty, ret_ty, &[Operator::Mod]);
impl_binop(unifier, store, ty, other_ty, ret_ty, &[ast::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) {
impl_unaryop(unifier, ty, ty, &[ast::Unaryop::UAdd, ast::Unaryop::USub])
}
/// `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) {
impl_unaryop(unifier, ty, ty, &[ast::Unaryop::Invert])
}
/// `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) {
impl_unaryop(unifier, ty, store.bool, &[ast::Unaryop::Not])
}
/// `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>,
) {
/// Lt, LtE, Gt, GtE
pub fn impl_comparison(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type, other_ty: Type) {
impl_cmpop(
unifier,
store,
ty,
other_ty,
&[Cmpop::Lt, Cmpop::Gt, Cmpop::LtE, Cmpop::GtE],
ret_ty,
);
&[ast::Cmpop::Lt, ast::Cmpop::Gt, ast::Cmpop::LtE, ast::Cmpop::GtE],
)
}
/// `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 => {
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);
}))
/// Eq, NotEq
pub fn impl_eq(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type) {
impl_cmpop(unifier, store, ty, ty, &[ast::Cmpop::Eq, ast::Cmpop::NotEq])
}
pub fn set_primitives_magic_methods(store: &PrimitiveStore, unifier: &mut Unifier) {
@ -678,77 +293,38 @@ pub fn set_primitives_magic_methods(store: &PrimitiveStore, unifier: &mut Unifie
bool: bool_t,
uint32: uint32_t,
uint64: uint64_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_basic_arithmetic(unifier, store, t, &[t], t);
impl_pow(unifier, store, t, &[t], t);
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);
impl_div(unifier, store, t, &[t]);
impl_floordiv(unifier, store, t, &[t], t);
impl_mod(unifier, store, t, &[t], t);
impl_invert(unifier, store, t);
impl_not(unifier, store, t);
impl_comparison(unifier, store, t, t);
impl_eq(unifier, store, t);
}
for t in [int32_t, int64_t] {
impl_sign(unifier, store, t, Some(t));
impl_sign(unifier, store, 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);
impl_basic_arithmetic(unifier, store, float_t, &[float_t], float_t);
impl_pow(unifier, store, float_t, &[int32_t, float_t], float_t);
impl_div(unifier, store, float_t, &[float_t]);
impl_floordiv(unifier, store, float_t, &[float_t], float_t);
impl_mod(unifier, store, float_t, &[float_t], float_t);
impl_sign(unifier, store, float_t);
impl_not(unifier, store, float_t);
impl_comparison(unifier, store, float_t, float_t);
impl_eq(unifier, store, float_t);
/* 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);
/* 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);
impl_not(unifier, store, bool_t);
impl_eq(unifier, store, bool_t);
}

109
nac3core/src/typecheck/type_error.rs
View File

@ -1,46 +1,24 @@
use std::collections::HashMap;
use std::fmt::Display;
use crate::typecheck::{magic_methods::HasOpInfo, typedef::TypeEnum};
use crate::typecheck::typedef::TypeEnum;
use super::{
magic_methods::Binop,
typedef::{RecordKey, Type, Unifier},
};
use itertools::Itertools;
use nac3parser::ast::{Cmpop, Location, StrRef};
use super::typedef::{RecordKey, Type, Unifier};
use nac3parser::ast::{Location, StrRef};
#[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>,
expected: usize,
got: usize,
},
MissingArgs(String),
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),
@ -56,7 +34,6 @@ pub enum TypeErrorKind {
},
RequiresTypeAnn,
PolymorphicFunctionPointer,
NoSuchAttribute(RecordKey, Type),
}
#[derive(Debug, Clone)]
@ -66,18 +43,15 @@ pub struct TypeError {
}
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 }
}
@ -90,8 +64,8 @@ pub struct DisplayTypeError<'a> {
fn loc_to_str(loc: Option<Location>) -> String {
match loc {
Some(loc) => format!("(in {loc})"),
None => String::new(),
Some(loc) => format!("(in {})", loc),
None => "".to_string(),
}
}
@ -100,49 +74,23 @@ impl<'a> Display for DisplayTypeError<'a> {
use TypeErrorKind::*;
let mut notes = Some(HashMap::new());
match &self.err.kind {
GotMultipleValues { name } => {
write!(f, "For multiple values for parameter {name}")
TooManyArguments { expected, got } => {
write!(f, "Too many arguments. Expected {} but got {}", expected, got)
}
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})")
MissingArgs(args) => {
write!(f, "Missing arguments: {}", args)
}
UnknownArgName(name) => {
write!(f, "Unknown argument name: {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}")
write!(
f,
"Incorrect argument type for {}. Expected {}, but got {}",
name, expected, got
)
}
FieldUnificationError { field, types, loc } => {
let lhs = self.unifier.stringify_with_notes(types.0, &mut notes);
@ -178,7 +126,7 @@ impl<'a> Display for DisplayTypeError<'a> {
);
if let Some(loc) = loc {
result?;
write!(f, " (in {loc})")?;
write!(f, " (in {})", loc)?;
return Ok(());
}
result
@ -188,12 +136,12 @@ impl<'a> Display for DisplayTypeError<'a> {
{
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}")
write!(f, "Tuple length mismatch: got {} and {}", t1, 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}")
write!(f, "Incompatible types: {} and {}", t1, t2)
}
}
}
@ -202,21 +150,18 @@ impl<'a> Display for DisplayTypeError<'a> {
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")
write!(f, "Cannot assign to field {} of {}, which is immutable", name, t)
}
}
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")
write!(f, "`{}::{}` field/method does not exist", t, name)
}
TupleIndexOutOfBounds { index, len } => {
write!(
f,
"Tuple index out of bounds. Got {index} but tuple has only {len} elements"
"Tuple index out of bounds. Got {} but tuple has only {} elements",
index, len
)
}
RequiresTypeAnn => {
@ -227,13 +172,13 @@ impl<'a> Display for DisplayTypeError<'a> {
}
}?;
if let Some(loc) = self.err.loc {
write!(f, " at {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}")?;
write!(f, "\n {}", line)?;
}
}
Ok(())

1491
nac3core/src/typecheck/type_inferencer/mod.rs
View File

File diff suppressed because it is too large Load Diff

341
nac3core/src/typecheck/type_inferencer/test.rs
View File

@ -3,14 +3,12 @@ use super::*;
use crate::{
codegen::CodeGenContext,
symbol_resolver::ValueEnum,
toplevel::{helper::PrimDef, DefinitionId, TopLevelDef},
toplevel::{DefinitionId, TopLevelDef},
};
use indexmap::IndexMap;
use indoc::indoc;
use nac3parser::ast::FileName;
use itertools::zip;
use nac3parser::parser::parse_program;
use parking_lot::RwLock;
use std::iter::zip;
use test_case::test_case;
struct Resolver {
@ -22,7 +20,7 @@ struct Resolver {
impl SymbolResolver for Resolver {
fn get_default_param_value(
&self,
_: &ast::Expr,
_: &nac3parser::ast::Expr,
) -> Option<crate::symbol_resolver::SymbolValue> {
unimplemented!()
}
@ -34,22 +32,19 @@ impl SymbolResolver for Resolver {
_: &PrimitiveStore,
str: StrRef,
) -> Result<Type, String> {
self.id_to_type.get(&str).copied().ok_or_else(|| format!("cannot find symbol `{str}`"))
self.id_to_type.get(&str).cloned().ok_or_else(|| format!("cannot find symbol `{}`", str))
}
fn get_symbol_value<'ctx>(
fn get_symbol_value<'ctx, 'a>(
&self,
_: StrRef,
_: &mut CodeGenContext<'ctx, '_>,
_: &mut CodeGenContext<'ctx, 'a>,
) -> 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_identifier_def(&self, id: StrRef) -> Result<DefinitionId, String> {
self.id_to_def.get(&id).cloned().ok_or_else(|| "Unknown identifier".to_string())
}
fn get_string_id(&self, _: &str) -> i32 {
@ -67,7 +62,7 @@ struct TestEnvironment {
pub primitives: PrimitiveStore,
pub id_to_name: HashMap<usize, StrRef>,
pub identifier_mapping: HashMap<StrRef, Type>,
pub virtual_checks: Vec<(Type, Type, Location)>,
pub virtual_checks: Vec<(Type, Type, nac3parser::ast::Location)>,
pub calls: HashMap<CodeLocation, CallId>,
pub top_level: TopLevelContext,
}
@ -77,81 +72,67 @@ impl TestEnvironment {
let mut unifier = Unifier::new();
let int32 = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Int32.id(),
obj_id: DefinitionId(0),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::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 }],
ret: int32,
vars: VarMap::new(),
vars: HashMap::new(),
}));
fields.insert("__add__".into(), (add_ty, false));
});
let int64 = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Int64.id(),
obj_id: DefinitionId(1),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let float = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Float.id(),
obj_id: DefinitionId(2),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let bool = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Bool.id(),
obj_id: DefinitionId(3),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let none = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::None.id(),
obj_id: DefinitionId(4),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let range = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Range.id(),
obj_id: DefinitionId(5),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let str = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Str.id(),
obj_id: DefinitionId(6),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let exception = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Exception.id(),
obj_id: DefinitionId(7),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let uint32 = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::UInt32.id(),
obj_id: DefinitionId(8),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let uint64 = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::UInt64.id(),
obj_id: DefinitionId(9),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let option = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Option.id(),
obj_id: DefinitionId(10),
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]),
params: HashMap::new(),
});
let primitives = PrimitiveStore {
int32,
@ -165,14 +146,10 @@ impl TestEnvironment {
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<_, _> = [
let id_to_name = [
(0, "int32".into()),
(1, "int64".into()),
(2, "float".into()),
@ -182,21 +159,23 @@ impl TestEnvironment {
(6, "str".into()),
(7, "exception".into()),
]
.into();
.iter()
.cloned()
.collect();
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(),
id_to_def: Default::default(),
class_names: Default::default(),
}) as Arc<dyn SymbolResolver + Send + Sync>;
TestEnvironment {
top_level: TopLevelContext {
definitions: Arc::default(),
unifiers: Arc::default(),
definitions: Default::default(),
unifiers: Default::default(),
personality_symbol: None,
},
unifier,
@ -218,95 +197,70 @@ impl TestEnvironment {
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(),
obj_id: DefinitionId(0),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::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 }],
ret: int32,
vars: VarMap::new(),
vars: HashMap::new(),
}));
fields.insert("__add__".into(), (add_ty, false));
});
let int64 = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Int64.id(),
obj_id: DefinitionId(1),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let float = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Float.id(),
obj_id: DefinitionId(2),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let bool = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Bool.id(),
obj_id: DefinitionId(3),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let none = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::None.id(),
obj_id: DefinitionId(4),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let range = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Range.id(),
obj_id: DefinitionId(5),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let str = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Str.id(),
obj_id: DefinitionId(6),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let exception = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Exception.id(),
obj_id: DefinitionId(7),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let uint32 = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::UInt32.id(),
obj_id: DefinitionId(8),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let uint64 = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::UInt64.id(),
obj_id: DefinitionId(9),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
});
let option = unifier.add_ty(TypeEnum::TObj {
obj_id: PrimDef::Option.id(),
obj_id: DefinitionId(10),
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(),
params: HashMap::new(),
});
identifier_mapping.insert("None".into(), none);
for (i, name) in [
"int32",
"int64",
"float",
"bool",
"none",
"range",
"str",
"Exception",
"uint32",
"uint64",
"Option",
"list",
"ndarray",
]
for (i, name) in ["int32", "int64", "float", "bool", "none", "range", "str", "Exception"]
.iter()
.enumerate()
{
@ -314,11 +268,10 @@ impl TestEnvironment {
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(),
type_vars: Default::default(),
fields: Default::default(),
methods: Default::default(),
ancestors: Default::default(),
resolver: None,
constructor: None,
loc: None,
@ -326,7 +279,7 @@ impl TestEnvironment {
.into(),
);
}
let defs = 12;
let defs = 7;
let primitives = PrimitiveStore {
int32,
@ -340,29 +293,23 @@ impl TestEnvironment {
uint32,
uint64,
option,
list,
ndarray,
size_t: 64,
};
unifier.put_primitive_store(&primitives);
let tvar = unifier.get_dummy_var();
let (v0, id) = 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]),
fields: [("a".into(), (v0, true))].iter().cloned().collect::<HashMap<_, _>>(),
params: [(id, v0)].iter().cloned().collect::<HashMap<_, _>>(),
});
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(),
type_vars: vec![v0],
fields: [("a".into(), v0, true)].into(),
methods: Default::default(),
ancestors: Default::default(),
resolver: None,
constructor: None,
loc: None,
@ -375,29 +322,31 @@ impl TestEnvironment {
unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![],
ret: foo_ty,
vars: into_var_map([tvar]),
vars: [(id, v0)].iter().cloned().collect(),
})),
);
let fun = unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![],
ret: int32,
vars: IndexMap::default(),
vars: Default::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(),
fields: [("a".into(), (int32, true)), ("b".into(), (fun, true))]
.iter()
.cloned()
.collect::<HashMap<_, _>>(),
params: Default::default(),
});
top_level_defs.push(
RwLock::new(TopLevelDef::Class {
name: "Bar".into(),
object_id: DefinitionId(defs + 2),
type_vars: Vec::default(),
type_vars: Default::default(),
fields: [("a".into(), int32, true), ("b".into(), fun, true)].into(),
attributes: Vec::default(),
methods: Vec::default(),
ancestors: Vec::default(),
methods: Default::default(),
ancestors: Default::default(),
resolver: None,
constructor: None,
loc: None,
@ -409,24 +358,26 @@ impl TestEnvironment {
unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![],
ret: bar,
vars: IndexMap::default(),
vars: Default::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(),
fields: [("a".into(), (bool, true)), ("b".into(), (fun, false))]
.iter()
.cloned()
.collect::<HashMap<_, _>>(),
params: Default::default(),
});
top_level_defs.push(
RwLock::new(TopLevelDef::Class {
name: "Bar2".into(),
object_id: DefinitionId(defs + 3),
type_vars: Vec::default(),
type_vars: Default::default(),
fields: [("a".into(), bool, true), ("b".into(), fun, false)].into(),
attributes: Vec::default(),
methods: Vec::default(),
ancestors: Vec::default(),
methods: Default::default(),
ancestors: Default::default(),
resolver: None,
constructor: None,
loc: None,
@ -438,10 +389,10 @@ impl TestEnvironment {
unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![],
ret: bar2,
vars: IndexMap::default(),
vars: Default::default(),
})),
);
let class_names: HashMap<_, _> = [("Bar".into(), bar), ("Bar2".into(), bar2)].into();
let class_names = [("Bar".into(), bar), ("Bar2".into(), bar2)].iter().cloned().collect();
let id_to_name = [
"int32".into(),
@ -452,22 +403,18 @@ impl TestEnvironment {
"range".into(),
"str".into(),
"exception".into(),
"uint32".into(),
"uint64".into(),
"option".into(),
"list".into(),
"ndarray".into(),
"Foo".into(),
"Bar".into(),
"Bar2".into(),
]
.into_iter()
.iter()
.enumerate()
.map(|(a, b)| (a, *b))
.collect();
let top_level = TopLevelContext {
definitions: Arc::new(top_level_defs.into()),
unifiers: Arc::default(),
unifiers: Default::default(),
personality_symbol: None,
};
@ -478,7 +425,9 @@ impl TestEnvironment {
("Bar".into(), DefinitionId(defs + 2)),
("Bar2".into(), DefinitionId(defs + 3)),
]
.into(),
.iter()
.cloned()
.collect(),
class_names,
}) as Arc<dyn SymbolResolver + Send + Sync>;
@ -503,11 +452,11 @@ impl TestEnvironment {
top_level: &self.top_level,
function_data: &mut self.function_data,
unifier: &mut self.unifier,
variable_mapping: HashMap::default(),
variable_mapping: Default::default(),
primitives: &mut self.primitives,
virtual_checks: &mut self.virtual_checks,
calls: &mut self.calls,
defined_identifiers: HashSet::default(),
defined_identifiers: Default::default(),
in_handler: false,
}
}
@ -519,7 +468,7 @@ impl TestEnvironment {
c = 1.234
d = True
"},
&[("a", "int32"), ("b", "int64"), ("c", "float"), ("d", "bool")].into(),
[("a", "int32"), ("b", "int64"), ("c", "float"), ("d", "bool")].iter().cloned().collect(),
&[]
; "primitives test")]
#[test_case(indoc! {"
@ -528,7 +477,7 @@ impl TestEnvironment {
c = 1.234
d = b(c)
"},
&[("a", "fn[[x:float, y:float], float]"), ("b", "fn[[x:float], float]"), ("c", "float"), ("d", "float")].into(),
[("a", "fn[[x:float, y:float], float]"), ("b", "fn[[x:float], float]"), ("c", "float"), ("d", "float")].iter().cloned().collect(),
&[]
; "lambda test")]
#[test_case(indoc! {"
@ -537,7 +486,7 @@ impl TestEnvironment {
a = b
c = b(1)
"},
&[("a", "fn[[x:int32], int32]"), ("b", "fn[[x:int32], int32]"), ("c", "int32")].into(),
[("a", "fn[[x:int32], int32]"), ("b", "fn[[x:int32], int32]"), ("c", "int32")].iter().cloned().collect(),
&[]
; "lambda test 2")]
#[test_case(indoc! {"
@ -553,15 +502,15 @@ impl TestEnvironment {
b(123)
"},
&[("a", "fn[[x:bool], bool]"), ("b", "fn[[x:int32], int32]"), ("c", "bool"),
("d", "int32"), ("foo1", "Foo[bool]"), ("foo2", "Foo[int32]")].into(),
[("a", "fn[[x:bool], bool]"), ("b", "fn[[x:int32], int32]"), ("c", "bool"),
("d", "int32"), ("foo1", "Foo[bool]"), ("foo2", "Foo[int32]")].iter().cloned().collect(),
&[]
; "obj test")]
#[test_case(indoc! {"
a = [1, 2, 3]
b = [x + x for x in a]
"},
&[("a", "list[int32]"), ("b", "list[int32]")].into(),
[("a", "list[int32]"), ("b", "list[int32]")].iter().cloned().collect(),
&[]
; "listcomp test")]
#[test_case(indoc! {"
@ -569,25 +518,25 @@ impl TestEnvironment {
b = a.b()
a = virtual(Bar2())
"},
&[("a", "virtual[Bar]"), ("b", "int32")].into(),
[("a", "virtual[Bar]"), ("b", "int32")].iter().cloned().collect(),
&[("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(),
[("a", "list[virtual[Bar]]"), ("b", "list[int32]")].iter().cloned().collect(),
&[("Bar", "Bar"), ("Bar2", "Bar")]
; "virtual list test")]
fn test_basic(source: &str, mapping: &HashMap<&str, &str>, virtuals: &[(&str, &str)]) {
println!("source:\n{source}");
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: HashSet<_> = env.identifier_mapping.keys().copied().collect();
let mut defined_identifiers: HashSet<_> = env.identifier_mapping.keys().cloned().collect();
defined_identifiers.insert("virtual".into());
let mut inferencer = env.get_inferencer();
inferencer.defined_identifiers.clone_from(&defined_identifiers);
let statements = parse_program(source, FileName::default()).unwrap();
inferencer.defined_identifiers = defined_identifiers.clone();
let statements = parse_program(source, Default::default()).unwrap();
let statements = statements
.into_iter()
.map(|v| inferencer.fold_stmt(v))
@ -596,37 +545,37 @@ fn test_basic(source: &str, mapping: &HashMap<&str, &str>, virtuals: &[(&str, &s
inferencer.check_block(&statements, &mut defined_identifiers).unwrap();
for (k, v) in &inferencer.variable_mapping {
for (k, v) in inferencer.variable_mapping.iter() {
let name = inferencer.unifier.internal_stringify(
*v,
&mut |v| (*id_to_name.get(&v).unwrap()).into(),
&mut |v| format!("v{v}"),
&mut |v| format!("v{}", v),
&mut None,
);
println!("{k}: {name}");
println!("{}: {}", k, name);
}
for (k, v) in mapping {
for (k, v) in mapping.iter() {
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 |v| format!("v{}", v),
&mut None,
);
assert_eq!(format!("{k}: {v}"), format!("{k}: {name}"));
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 |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 |v| format!("v{}", v),
&mut None,
);
@ -645,14 +594,14 @@ fn test_basic(source: &str, mapping: &HashMap<&str, &str>, virtuals: &[(&str, &s
g = a // b
h = a % b
"},
&[("a", "int32"),
[("a", "int32"),
("b", "int32"),
("c", "int32"),
("d", "int32"),
("e", "int32"),
("f", "float"),
("g", "int32"),
("h", "int32")].into()
("h", "int32")].iter().cloned().collect()
; "int32")]
#[test_case(
indoc! {"
@ -668,7 +617,7 @@ fn test_basic(source: &str, mapping: &HashMap<&str, &str>, virtuals: &[(&str, &s
ii = 3
j = a ** b
"},
&[("a", "float"),
[("a", "float"),
("b", "float"),
("c", "float"),
("d", "float"),
@ -678,7 +627,7 @@ fn test_basic(source: &str, mapping: &HashMap<&str, &str>, virtuals: &[(&str, &s
("h", "float"),
("i", "float"),
("ii", "int32"),
("j", "float")].into()
("j", "float")].iter().cloned().collect()
; "float"
)]
#[test_case(
@ -696,7 +645,7 @@ fn test_basic(source: &str, mapping: &HashMap<&str, &str>, virtuals: &[(&str, &s
k = a < b
l = a != b
"},
&[("a", "int64"),
[("a", "int64"),
("b", "int64"),
("c", "int64"),
("d", "int64"),
@ -707,7 +656,7 @@ fn test_basic(source: &str, mapping: &HashMap<&str, &str>, virtuals: &[(&str, &s
("i", "bool"),
("j", "bool"),
("k", "bool"),
("l", "bool")].into()
("l", "bool")].iter().cloned().collect()
; "int64"
)]
#[test_case(
@ -718,22 +667,22 @@ fn test_basic(source: &str, mapping: &HashMap<&str, &str>, virtuals: &[(&str, &s
d = not a
e = a != b
"},
&[("a", "bool"),
[("a", "bool"),
("b", "bool"),
("c", "bool"),
("d", "bool"),
("e", "bool")].into()
("e", "bool")].iter().cloned().collect()
; "boolean"
)]
fn test_primitive_magic_methods(source: &str, mapping: &HashMap<&str, &str>) {
println!("source:\n{source}");
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: HashSet<_> = env.identifier_mapping.keys().copied().collect();
let mut defined_identifiers: HashSet<_> = env.identifier_mapping.keys().cloned().collect();
defined_identifiers.insert("virtual".into());
let mut inferencer = env.get_inferencer();
inferencer.defined_identifiers.clone_from(&defined_identifiers);
let statements = parse_program(source, FileName::default()).unwrap();
inferencer.defined_identifiers = defined_identifiers.clone();
let statements = parse_program(source, Default::default()).unwrap();
let statements = statements
.into_iter()
.map(|v| inferencer.fold_stmt(v))
@ -742,23 +691,23 @@ fn test_primitive_magic_methods(source: &str, mapping: &HashMap<&str, &str>) {
inferencer.check_block(&statements, &mut defined_identifiers).unwrap();
for (k, v) in &inferencer.variable_mapping {
for (k, v) in inferencer.variable_mapping.iter() {
let name = inferencer.unifier.internal_stringify(
*v,
&mut |v| (*id_to_name.get(&v).unwrap()).into(),
&mut |v| format!("v{v}"),
&mut |v| format!("v{}", v),
&mut None,
);
println!("{k}: {name}");
println!("{}: {}", k, name);
}
for (k, v) in mapping {
for (k, v) in mapping.iter() {
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 |v| format!("v{}", v),
&mut None,
);
assert_eq!(format!("{k}: {v}"), format!("{k}: {name}"));
assert_eq!(format!("{}: {}", k, v), format!("{}: {}", k, name));
}
}

827
nac3core/src/typecheck/typedef/mod.rs
View File

File diff suppressed because it is too large Load Diff

330
nac3core/src/typecheck/typedef/test.rs
View File

@ -32,25 +32,28 @@ impl Unifier {
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::TList { ty: ty1 }, TypeEnum::TList { ty: ty2 })
| (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
// TCall and TFunc are not yet implemented
_ => false,
}
}
fn map_eq<K>(&mut self, map1: &IndexMapping<K>, map2: &IndexMapping<K>) -> bool
fn map_eq<K>(&mut self, map1: &Mapping<K>, map2: &Mapping<K>) -> bool
where
K: std::hash::Hash + Eq + Clone,
K: std::hash::Hash + std::cmp::Eq + std::clone::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)) {
for (k, v) in map1.iter() {
if !map2.get(k).map(|v1| self.eq(*v, *v1)).unwrap_or(false) {
return false;
}
}
@ -59,13 +62,13 @@ impl Unifier {
fn map_eq2<K>(&mut self, map1: &Mapping<K, RecordField>, map2: &Mapping<K, RecordField>) -> bool
where
K: std::hash::Hash + Eq + Clone,
K: std::hash::Hash + std::cmp::Eq + std::clone::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)) {
for (k, v) in map1.iter() {
if !map2.get(k).map(|v1| self.eq(v.ty, v1.ty)).unwrap_or(false) {
return false;
}
}
@ -88,7 +91,7 @@ impl TestEnvironment {
unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(0),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
}),
);
type_mapping.insert(
@ -96,7 +99,7 @@ impl TestEnvironment {
unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(1),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
}),
);
type_mapping.insert(
@ -104,25 +107,16 @@ impl TestEnvironment {
unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(2),
fields: HashMap::new(),
params: VarMap::new(),
params: HashMap::new(),
}),
);
let tvar = unifier.get_dummy_var();
let (v0, id) = 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]),
fields: [("a".into(), (v0, true))].iter().cloned().collect::<HashMap<_, _>>(),
params: [(id, v0)].iter().cloned().collect::<HashMap<_, _>>(),
}),
);
@ -135,43 +129,17 @@ impl TestEnvironment {
result.0
}
fn internal_parse<'b>(&mut self, typ: &'b str, mapping: &Mapping<String>) -> (Type, &'b str) {
fn internal_parse<'a, 'b>(
&'a 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());
let end = typ.find(|c| ['[', ',', ']', '='].contains(&c)).unwrap_or_else(|| 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], "[");
assert!(&s[0..1] == "[");
let mut ty = Vec::new();
while &s[0..1] != "]" {
let result = self.internal_parse(&s[1..], mapping);
@ -180,9 +148,15 @@ impl TestEnvironment {
}
(self.unifier.add_ty(TypeEnum::TTuple { ty }), &s[1..])
}
"list" => {
assert!(&typ[end..end + 1] == "[");
let (ty, s) = self.internal_parse(&typ[end + 1..], mapping);
assert!(&s[0..1] == "]");
(self.unifier.add_ty(TypeEnum::TList { ty }), &s[1..])
}
"Record" => {
let mut s = &typ[end..];
assert_eq!(&s[0..1], "[");
assert!(&s[0..1] == "[");
let mut fields = HashMap::new();
while &s[0..1] != "]" {
let eq = s.find('=').unwrap();
@ -195,14 +169,14 @@ impl TestEnvironment {
}
x => {
let mut s = &typ[end..];
let ty = mapping.get(x).copied().unwrap_or_else(|| {
let ty = mapping.get(x).cloned().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 let TypeEnum::TObj { params, .. } = &*te.as_ref() {
if !params.is_empty() {
assert_eq!(&s[0..1], "[");
assert!(&s[0..1] == "[");
let mut p = Vec::new();
while &s[0..1] != "]" {
let result = self.internal_parse(&s[1..], mapping);
@ -212,7 +186,7 @@ impl TestEnvironment {
s = &s[1..];
ty = self
.unifier
.subst(ty, &params.keys().copied().zip(p).collect())
.subst(ty, &params.keys().cloned().zip(p.into_iter()).collect())
.unwrap_or(ty);
}
}
@ -276,12 +250,12 @@ fn test_unify(
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);
mapping.insert(format!("v{}", i), v.0);
}
// 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 {
for (a, b) in perm.iter() {
let t1 = env.parse(a, &mapping);
let t2 = env.parse(b, &mapping);
pairs.push((t1, t2));
@ -289,8 +263,8 @@ fn test_unify(
for (t1, t2) in pairs {
env.unifier.unify(t1, t2).unwrap();
}
for (a, b) in verify_pairs {
println!("{a} = {b}");
for (a, b) in verify_pairs.iter() {
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));
@ -304,7 +278,7 @@ fn test_unify(
("v1", "tuple[int]"),
("v2", "list[int]"),
],
(("v1", "v2"), "Incompatible types: 11[0] and tuple[0]")
(("v1", "v2"), "Incompatible types: list[0] and tuple[0]")
; "type mismatch"
)]
#[test_case(2,
@ -312,7 +286,7 @@ fn test_unify(
("v1", "tuple[int]"),
("v2", "tuple[float]"),
],
(("v1", "v2"), "Incompatible types: tuple[0] and tuple[1]")
(("v1", "v2"), "Incompatible types: 0 and 1")
; "tuple parameter mismatch"
)]
#[test_case(2,
@ -328,35 +302,35 @@ fn test_unify(
("v1", "Record[a=float,b=int]"),
("v2", "Foo[v3]"),
],
(("v1", "v2"), "`3[typevar5]::b` field/method does not exist")
(("v1", "v2"), "`3[var4]::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),
errornous_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);
mapping.insert(format!("v{}", i), v.0);
}
// 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 {
for (a, b) in unify_pairs.iter() {
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));
(env.parse(errornous_pair.0 .0, &mapping), env.parse(errornous_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()));
assert_eq!(env.unify(t1, t2), Err(errornous_pair.1.to_string()));
}
#[test]
@ -365,17 +339,23 @@ fn test_recursive_subst() {
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);
let mapping: HashMap<_, _>;
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();
if let TypeEnum::TObj { params, .. } = &*foo_ty {
mapping = params.iter().map(|(id, _)| (*id, int)).collect();
} else {
unreachable!()
}
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!() };
if let TypeEnum::TObj { fields, .. } = &*instantiated_ty {
assert!(env.unifier.unioned(fields.get(&"a".into()).unwrap().0, int));
assert!(env.unifier.unioned(fields.get(&"rec".into()).unwrap().0, instantiated));
} else {
unreachable!()
}
}
#[test]
@ -385,27 +365,36 @@ fn test_virtual() {
let fun = env.unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![],
ret: int,
vars: VarMap::new(),
vars: HashMap::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(),
fields: [("f".into(), (fun, false)), ("a".into(), (int, false))]
.iter()
.cloned()
.collect::<HashMap<StrRef, _>>(),
params: HashMap::new(),
});
let v0 = env.unifier.get_dummy_var().ty;
let v1 = env.unifier.get_dummy_var().ty;
let v0 = env.unifier.get_dummy_var().0;
let v1 = env.unifier.get_dummy_var().0;
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());
let c = env
.unifier
.add_record([("f".into(), RecordField::new(v1, false, None))].iter().cloned().collect());
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());
let d = env
.unifier
.add_record([("a".into(), RecordField::new(v1, true, None))].iter().cloned().collect());
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());
let d = env
.unifier
.add_record([("b".into(), RecordField::new(v1, true, None))].iter().cloned().collect());
assert_eq!(env.unify(b, d), Err("`virtual[5]::b` field/method does not exist".to_string()));
}
@ -418,132 +407,85 @@ fn test_typevar_range() {
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;
let v = env.unifier.get_fresh_var_with_range(&[int, boolean], None, None).0;
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;
let v = env.unifier.get_fresh_var_with_range(&[int, boolean], None, None).0;
assert_eq!(
env.unify(int_list, v),
Err("Expected any one of these types: 0, 2, but got 11[0]".to_string())
Err("Expected any one of these types: 0, 2, but got list[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;
let v = env.unifier.get_fresh_var_with_range(&[int, boolean], None, None).0;
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;
let v1 = env.unifier.get_fresh_var_with_range(&[int, boolean], None, None).0;
let v1_list = env.unifier.add_ty(TypeEnum::TList { ty: v1 });
let v = env.unifier.get_fresh_var_with_range(&[int, v1_list], None, None).0;
// 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;
let v = env.unifier.get_fresh_var_with_range(&[int, v1_list], None, None).0;
// 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;
let v = env.unifier.get_fresh_var_with_range(&[int, v1_list], None, None).0;
// 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())
Err("Expected any one of these types: 0, list[var5], but got list[1]\n\nNotes:\n var5 ∈ {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;
let a = env.unifier.get_fresh_var_with_range(&[int, float], None, None).0;
let b = env.unifier.get_fresh_var_with_range(&[boolean, float], None, None).0;
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;
let a = env.unifier.get_fresh_var_with_range(&[int, float], None, None).0;
let b = env.unifier.get_fresh_var_with_range(&[boolean, float], None, None).0;
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;
let a = env.unifier.get_fresh_var_with_range(&[int, float], None, None).0;
let b = env.unifier.get_fresh_var_with_range(&[boolean, float], None, None).0;
let a_list = env.unifier.add_ty(TypeEnum::TList { ty: a });
let a_list = env.unifier.get_fresh_var_with_range(&[a_list], None, None).0;
let b_list = env.unifier.add_ty(TypeEnum::TList { ty: b });
let b_list = env.unifier.get_fresh_var_with_range(&[b_list], None, None).0;
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 }]),
});
let float_list = env.unifier.add_ty(TypeEnum::TList { 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 }]),
});
let a = env.unifier.get_fresh_var_with_range(&[int, float], None, None).0;
let b = env.unifier.get_fresh_var_with_range(&[boolean, float], None, None).0;
let a_list = env.unifier.add_ty(TypeEnum::TList { ty: a });
let b_list = env.unifier.add_ty(TypeEnum::TList { 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 }]),
});
let int_list = env.unifier.add_ty(TypeEnum::TList { ty: int });
assert_eq!(
env.unify(a_list, int_list),
Err("Incompatible types: 11[typevar23] and 11[0]\
\n\nNotes:\n typevar23 ∈ {1}"
.into())
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_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 }]),
});
let a = env.unifier.get_fresh_var_with_range(&[int, float], None, None).0;
let b = env.unifier.get_dummy_var().0;
let a_list = env.unifier.add_ty(TypeEnum::TList { ty: a });
let a_list = env.unifier.get_fresh_var_with_range(&[a_list], None, None).0;
let b_list = env.unifier.add_ty(TypeEnum::TList { ty: b });
env.unifier.unify(a_list, b_list).unwrap();
assert_eq!(
env.unify(b, boolean),
@ -554,29 +496,17 @@ fn test_typevar_range() {
#[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 a = env.unifier.get_fresh_rigid_var(None, None).0;
let b = env.unifier.get_fresh_rigid_var(None, None).0;
let x = env.unifier.get_dummy_var().0;
let list_a = env.unifier.add_ty(TypeEnum::TList { ty: a });
let list_x = env.unifier.add_ty(TypeEnum::TList { 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()));
assert_eq!(env.unify(a, b), Err("Incompatible types: var3 and var2".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())
);
assert_eq!(env.unify(list_x, list_int), Err("Incompatible types: 0 and var2".to_string()));
env.unifier.replace_rigid_var(a, int);
env.unifier.unify(list_x, list_int).unwrap();
@ -590,26 +520,16 @@ fn test_instantiation() {
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")].iter().cloned().collect();
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 v = env.unifier.get_fresh_var_with_range(&[int, boolean], None, None).0;
let list_v = env.unifier.add_ty(TypeEnum::TList { ty: v });
let v1 = env.unifier.get_fresh_var_with_range(&[list_v, int], None, None).0;
let v2 = env.unifier.get_fresh_var_with_range(&[list_int, float], None, None).0;
let t = env.unifier.get_dummy_var().0;
let tuple = env.unifier.add_ty(TypeEnum::TTuple { ty: vec![v, v1, v2] });
let v3 = env.unifier.get_fresh_var_with_range(&[tuple, t], None, None).ty;
let v3 = env.unifier.get_fresh_var_with_range(&[tuple, t], None, None).0;
// t = TypeVar('t')
// v = TypeVar('v', int, bool)
// v1 = TypeVar('v1', 'list[v]', int)
@ -631,7 +551,7 @@ fn test_instantiation() {
tuple[int, list[bool], list[int]]
tuple[int, list[int], float]
tuple[int, list[int], list[int]]
v6"
v5"
}
.split('\n')
.collect_vec();
@ -640,8 +560,8 @@ fn test_instantiation() {
.map(|ty| {
env.unifier.internal_stringify(
*ty,
&mut |i| (*obj_map.get(&i).unwrap()).to_string(),
&mut |i| format!("v{i}"),
&mut |i| obj_map.get(&i).unwrap().to_string(),
&mut |i| format!("v{}", i),
&mut None,
)
})

53
nac3core/src/typecheck/unification_table.rs
View File

@ -16,10 +16,21 @@ pub struct UnificationTable<V> {
#[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 },
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> {
@ -30,13 +41,7 @@ impl<V> Default for UnificationTable<V> {
impl<V> UnificationTable<V> {
pub fn new() -> UnificationTable<V> {
UnificationTable {
parents: Vec::new(),
ranks: Vec::new(),
values: Vec::new(),
log: Vec::new(),
generation: 0,
}
UnificationTable { parents: Vec::new(), ranks: Vec::new(), values: Vec::new(), log: Vec::new(), generation: 0 }
}
pub fn new_key(&mut self, v: V) -> UnificationKey {
@ -120,10 +125,7 @@ impl<V> UnificationTable<V> {
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"
);
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 } => {
@ -143,10 +145,7 @@ impl<V> UnificationTable<V> {
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"
);
assert!(matches!(self.log[log_len - 1], Action::Marker { generation: gen } if gen == generation), "snapshot discard error");
self.log.clear();
}
}
@ -160,23 +159,11 @@ where
.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,
}
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,
}
UnificationTable { parents: table.parents.clone(), ranks: table.ranks.clone(), values, log: Vec::new(), generation: 0 }
}
}

8
nac3ld/Cargo.toml
View File

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

509
nac3ld/src/dwarf.rs
View File

@ -1,509 +0,0 @@
#![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
}
}

2893
nac3ld/src/elf.rs
View File

File diff suppressed because it is too large Load Diff

1503
nac3ld/src/lib.rs
View File

File diff suppressed because it is too large Load Diff

14
nac3parser/Cargo.toml
View File

@ -5,20 +5,20 @@ description = "Parser for python code."
authors = [ "RustPython Team", "M-Labs" ]
build = "build.rs"
license = "MIT"
edition = "2021"
edition = "2018"
[build-dependencies]
lalrpop = "0.20"
lalrpop = "0.19.6"
[dependencies]
nac3ast = { path = "../nac3ast" }
lalrpop-util = "0.20"
log = "0.4"
lalrpop-util = "0.19.6"
log = "0.4.1"
unic-emoji-char = "0.9"
unic-ucd-ident = "0.9"
unicode_names2 = "1.2"
phf = { version = "0.11", features = ["macros"] }
ahash = "0.8"
unicode_names2 = "0.4"
phf = { version = "0.9", features = ["macros"] }
ahash = "0.7.2"
[dev-dependencies]
insta = "=1.11.0"

44
nac3parser/src/config_comment_helper.rs
View File

@ -1,15 +1,15 @@
use crate::ast::Ident;
use crate::ast::Location;
use crate::error::*;
use crate::token::Tok;
use lalrpop_util::ParseError;
use nac3ast::*;
use crate::ast::Ident;
use crate::ast::Location;
use crate::token::Tok;
use crate::error::*;
pub fn make_config_comment(
com_loc: Location,
stmt_loc: Location,
nac3com_above: Vec<(Ident, Tok)>,
nac3com_end: Option<Ident>,
nac3com_end: Option<Ident>
) -> Result<Vec<Ident>, ParseError<Location, Tok, LexicalError>> {
if com_loc.column() != stmt_loc.column() && !nac3com_above.is_empty() {
return Err(ParseError::User {
@ -17,25 +17,24 @@ pub fn make_config_comment(
location: com_loc,
error: LexicalErrorType::OtherError(
format!(
"config comment at top must have the same indentation with what it applies (comment at {com_loc}, statement at {stmt_loc})",
"config comment at top must have the same indentation with what it applies (comment at {}, statement at {})",
com_loc,
stmt_loc,
)
)
}
});
})
};
Ok(nac3com_above
Ok(
nac3com_above
.into_iter()
.map(|(com, _)| com)
.chain(nac3com_end.map_or_else(|| vec![].into_iter(), |com| vec![com].into_iter()))
.collect())
.collect()
)
}
pub fn handle_small_stmt<U>(
stmts: &mut [Stmt<U>],
nac3com_above: Vec<(Ident, Tok)>,
nac3com_end: Option<Ident>,
com_above_loc: Location,
) -> Result<(), ParseError<Location, Tok, LexicalError>> {
pub fn handle_small_stmt<U>(stmts: &mut [Stmt<U>], nac3com_above: Vec<(Ident, Tok)>, nac3com_end: Option<Ident>, com_above_loc: Location) -> Result<(), ParseError<Location, Tok, LexicalError>> {
if com_above_loc.column() != stmts[0].location.column() && !nac3com_above.is_empty() {
return Err(ParseError::User {
error: LexicalError {
@ -48,12 +47,17 @@ pub fn handle_small_stmt<U>(
)
)
}
});
})
}
apply_config_comments(&mut stmts[0], nac3com_above.into_iter().map(|(com, _)| com).collect());
apply_config_comments(
&mut stmts[0],
nac3com_above
.into_iter()
.map(|(com, _)| com).collect()
);
apply_config_comments(
stmts.last_mut().unwrap(),
nac3com_end.map_or_else(Vec::new, |com| vec![com]),
nac3com_end.map_or_else(Vec::new, |com| vec![com])
);
Ok(())
}
@ -76,8 +80,6 @@ fn apply_config_comments<U>(stmt: &mut Stmt<U>, comments: Vec<Ident>) {
| StmtKind::Nonlocal { config_comment, .. }
| StmtKind::Assert { config_comment, .. } => config_comment.extend(comments),
_ => {
unreachable!("only small statements should call this function")
}
_ => { unreachable!("only small statements should call this function") }
}
}

54
nac3parser/src/error.rs
View File

@ -37,7 +37,7 @@ impl fmt::Display for LexicalErrorType {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
match self {
LexicalErrorType::StringError => write!(f, "Got unexpected string"),
LexicalErrorType::FStringError(error) => write!(f, "Got error in f-string: {error}"),
LexicalErrorType::FStringError(error) => write!(f, "Got error in f-string: {}", error),
LexicalErrorType::UnicodeError => write!(f, "Got unexpected unicode"),
LexicalErrorType::NestingError => write!(f, "Got unexpected nesting"),
LexicalErrorType::IndentationError => {
@ -59,13 +59,13 @@ impl fmt::Display for LexicalErrorType {
write!(f, "positional argument follows keyword argument")
}
LexicalErrorType::UnrecognizedToken { tok } => {
write!(f, "Got unexpected token {tok}")
write!(f, "Got unexpected token {}", tok)
}
LexicalErrorType::LineContinuationError => {
write!(f, "unexpected character after line continuation character")
}
LexicalErrorType::Eof => write!(f, "unexpected EOF while parsing"),
LexicalErrorType::OtherError(msg) => write!(f, "{msg}"),
LexicalErrorType::OtherError(msg) => write!(f, "{}", msg),
}
}
}
@ -96,7 +96,7 @@ impl fmt::Display for FStringErrorType {
FStringErrorType::UnopenedRbrace => write!(f, "Unopened '}}'"),
FStringErrorType::ExpectedRbrace => write!(f, "Expected '}}' after conversion flag."),
FStringErrorType::InvalidExpression(error) => {
write!(f, "Invalid expression: {error}")
write!(f, "Invalid expression: {}", error)
}
FStringErrorType::InvalidConversionFlag => write!(f, "Invalid conversion flag"),
FStringErrorType::EmptyExpression => write!(f, "Empty expression"),
@ -144,27 +144,36 @@ pub enum ParseErrorType {
impl From<LalrpopError<Location, Tok, LexicalError>> for ParseError {
fn from(err: LalrpopError<Location, Tok, LexicalError>) -> Self {
match err {
LalrpopError::ExtraToken { token } => {
ParseError { error: ParseErrorType::ExtraToken(token.1), location: token.0 }
}
LalrpopError::User { error } => {
ParseError { error: ParseErrorType::Lexical(error.error), location: error.location }
}
// TODO: Are there cases where this isn't an EOF?
LalrpopError::InvalidToken { location } => ParseError {
error: ParseErrorType::Eof,
location,
},
LalrpopError::ExtraToken { token } => ParseError {
error: ParseErrorType::ExtraToken(token.1),
location: token.0,
},
LalrpopError::User { error } => ParseError {
error: ParseErrorType::Lexical(error.error),
location: error.location,
},
LalrpopError::UnrecognizedToken { token, expected } => {
// Hacky, but it's how CPython does it. See PyParser_AddToken,
// in particular "Only one possible expected token" comment.
let expected = if expected.len() == 1 { Some(expected[0].clone()) } else { None };
let expected = if expected.len() == 1 {
Some(expected[0].clone())
} else {
None
};
ParseError {
error: ParseErrorType::UnrecognizedToken(token.1, expected),
location: token.0,
}
}
LalrpopError::UnrecognizedEof { location, .. }
// TODO: Are there cases where this isn't an EOF?
| LalrpopError::InvalidToken { location } => {
ParseError { error: ParseErrorType::Eof, location }
}
LalrpopError::UnrecognizedEOF { location, .. } => ParseError {
error: ParseErrorType::Eof,
location,
},
}
}
}
@ -179,7 +188,7 @@ impl fmt::Display for ParseErrorType {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
match *self {
ParseErrorType::Eof => write!(f, "Got unexpected EOF"),
ParseErrorType::ExtraToken(ref tok) => write!(f, "Got extraneous token: {tok:?}"),
ParseErrorType::ExtraToken(ref tok) => write!(f, "Got extraneous token: {:?}", tok),
ParseErrorType::InvalidToken => write!(f, "Got invalid token"),
ParseErrorType::UnrecognizedToken(ref tok, ref expected) => {
if *tok == Tok::Indent {
@ -187,10 +196,10 @@ impl fmt::Display for ParseErrorType {
} else if expected.as_deref() == Some("Indent") {
write!(f, "expected an indented block")
} else {
write!(f, "Got unexpected token {tok}")
write!(f, "Got unexpected token {}", tok)
}
}
ParseErrorType::Lexical(ref error) => write!(f, "{error}"),
ParseErrorType::Lexical(ref error) => write!(f, "{}", error),
}
}
}
@ -198,7 +207,6 @@ impl fmt::Display for ParseErrorType {
impl Error for ParseErrorType {}
impl ParseErrorType {
#[must_use]
pub fn is_indentation_error(&self) -> bool {
match self {
ParseErrorType::Lexical(LexicalErrorType::IndentationError) => true,
@ -208,11 +216,11 @@ impl ParseErrorType {
_ => false,
}
}
#[must_use]
pub fn is_tab_error(&self) -> bool {
matches!(
self,
ParseErrorType::Lexical(LexicalErrorType::TabError | LexicalErrorType::TabsAfterSpaces)
ParseErrorType::Lexical(LexicalErrorType::TabError)
| ParseErrorType::Lexical(LexicalErrorType::TabsAfterSpaces)
)
}
}

47
nac3parser/src/fstring.rs
View File

@ -15,7 +15,10 @@ struct FStringParser<'a> {
impl<'a> FStringParser<'a> {
fn new(source: &'a str, str_location: Location) -> Self {
Self { chars: source.chars().peekable(), str_location }
Self {
chars: source.chars().peekable(),
str_location,
}
}
#[inline]
@ -130,10 +133,10 @@ impl<'a> FStringParser<'a> {
)
} else {
Box::new(self.expr(ExprKind::Constant {
value: spec_expression.clone().into(),
value: spec_expression.to_owned().into(),
kind: None,
}))
});
})
}
'(' | '{' | '[' => {
expression.push(ch);
@ -248,11 +251,17 @@ impl<'a> FStringParser<'a> {
}
if !content.is_empty() {
values.push(self.expr(ExprKind::Constant { value: content.into(), kind: None }));
values.push(self.expr(ExprKind::Constant {
value: content.into(),
kind: None,
}))
}
let s = match values.len() {
0 => self.expr(ExprKind::Constant { value: String::new().into(), kind: None }),
0 => self.expr(ExprKind::Constant {
value: String::new().into(),
kind: None,
}),
1 => values.into_iter().next().unwrap(),
_ => self.expr(ExprKind::JoinedStr { values }),
};
@ -261,14 +270,16 @@ impl<'a> FStringParser<'a> {
}
fn parse_fstring_expr(source: &str) -> Result<Expr, ParseError> {
let fstring_body = format!("({source})");
let fstring_body = format!("({})", source);
parse_expression(&fstring_body)
}
/// Parse an fstring from a string, located at a certain position in the sourcecode.
/// In case of errors, we will get the location and the error returned.
pub fn parse_located_fstring(source: &str, location: Location) -> Result<Expr, FStringError> {
FStringParser::new(source, location).parse().map_err(|error| FStringError { error, location })
FStringParser::new(source, location)
.parse()
.map_err(|error| FStringError { error, location })
}
#[cfg(test)]
@ -282,7 +293,7 @@ mod tests {
#[test]
fn test_parse_fstring() {
let source = "{a}{ b }{{foo}}";
let parse_ast = parse_fstring(source).unwrap();
let parse_ast = parse_fstring(&source).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
@ -290,7 +301,7 @@ mod tests {
#[test]
fn test_parse_fstring_nested_spec() {
let source = "{foo:{spec}}";
let parse_ast = parse_fstring(source).unwrap();
let parse_ast = parse_fstring(&source).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
@ -298,7 +309,7 @@ mod tests {
#[test]
fn test_parse_fstring_not_nested_spec() {
let source = "{foo:spec}";
let parse_ast = parse_fstring(source).unwrap();
let parse_ast = parse_fstring(&source).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
@ -311,7 +322,7 @@ mod tests {
#[test]
fn test_fstring_parse_selfdocumenting_base() {
let src = "{user=}";
let parse_ast = parse_fstring(src).unwrap();
let parse_ast = parse_fstring(&src).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
@ -319,7 +330,7 @@ mod tests {
#[test]
fn test_fstring_parse_selfdocumenting_base_more() {
let src = "mix {user=} with text and {second=}";
let parse_ast = parse_fstring(src).unwrap();
let parse_ast = parse_fstring(&src).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
@ -327,7 +338,7 @@ mod tests {
#[test]
fn test_fstring_parse_selfdocumenting_format() {
let src = "{user=:>10}";
let parse_ast = parse_fstring(src).unwrap();
let parse_ast = parse_fstring(&src).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
@ -360,35 +371,35 @@ mod tests {
#[test]
fn test_parse_fstring_not_equals() {
let source = "{1 != 2}";
let parse_ast = parse_fstring(source).unwrap();
let parse_ast = parse_fstring(&source).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_fstring_equals() {
let source = "{42 == 42}";
let parse_ast = parse_fstring(source).unwrap();
let parse_ast = parse_fstring(&source).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_fstring_selfdoc_prec_space() {
let source = "{x =}";
let parse_ast = parse_fstring(source).unwrap();
let parse_ast = parse_fstring(&source).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_fstring_selfdoc_trailing_space() {
let source = "{x= }";
let parse_ast = parse_fstring(source).unwrap();
let parse_ast = parse_fstring(&source).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_fstring_yield_expr() {
let source = "{yield}";
let parse_ast = parse_fstring(source).unwrap();
let parse_ast = parse_fstring(&source).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
}

12
nac3parser/src/function.rs
View File

@ -54,7 +54,8 @@ pub fn parse_args(func_args: Vec<FunctionArgument>) -> Result<ArgumentList, Lexi
let mut keyword_names = HashSet::with_capacity_and_hasher(func_args.len(), RandomState::new());
for (name, value) in func_args {
if let Some((location, name)) = name {
match name {
Some((location, name)) => {
if let Some(keyword_name) = &name {
if keyword_names.contains(keyword_name) {
return Err(LexicalError {
@ -68,9 +69,13 @@ pub fn parse_args(func_args: Vec<FunctionArgument>) -> Result<ArgumentList, Lexi
keywords.push(ast::Keyword::new(
location,
ast::KeywordData { arg: name.map(String::into), value: Box::new(value) },
ast::KeywordData {
arg: name.map(|name| name.into()),
value: Box::new(value),
},
));
} else {
}
None => {
// Allow starred args after keyword arguments.
if !keywords.is_empty() && !is_starred(&value) {
return Err(LexicalError {
@ -82,6 +87,7 @@ pub fn parse_args(func_args: Vec<FunctionArgument>) -> Result<ArgumentList, Lexi
args.push(value);
}
}
}
Ok(ArgumentList { args, keywords })
}

301
nac3parser/src/lexer.rs
View File

@ -3,12 +3,12 @@
//! This means source code is translated into separate tokens.
pub use super::token::Tok;
use crate::ast::{FileName, Location};
use crate::ast::{Location, FileName};
use crate::error::{LexicalError, LexicalErrorType};
use std::char;
use std::cmp::Ordering;
use std::num::IntErrorKind;
use std::str::FromStr;
use std::num::IntErrorKind;
use unic_emoji_char::is_emoji_presentation;
use unic_ucd_ident::{is_xid_continue, is_xid_start};
@ -32,14 +32,20 @@ impl IndentationLevel {
if self.spaces <= other.spaces {
Ok(Ordering::Less)
} else {
Err(LexicalError { location, error: LexicalErrorType::TabError })
Err(LexicalError {
location,
error: LexicalErrorType::TabError,
})
}
}
Ordering::Greater => {
if self.spaces >= other.spaces {
Ok(Ordering::Greater)
} else {
Err(LexicalError { location, error: LexicalErrorType::TabError })
Err(LexicalError {
location,
error: LexicalErrorType::TabError,
})
}
}
Ordering::Equal => Ok(self.spaces.cmp(&other.spaces)),
@ -57,7 +63,7 @@ pub struct Lexer<T: Iterator<Item = char>> {
chr1: Option<char>,
chr2: Option<char>,
location: Location,
config_comment_prefix: Option<&'static str>,
config_comment_prefix: Option<&'static str>
}
pub static KEYWORDS: phf::Map<&'static str, Tok> = phf::phf_map! {
@ -130,7 +136,11 @@ where
T: Iterator<Item = char>,
{
pub fn new(source: T) -> Self {
let mut nlh = NewlineHandler { source, chr0: None, chr1: None };
let mut nlh = NewlineHandler {
source,
chr0: None,
chr1: None,
};
nlh.shift();
nlh.shift();
nlh
@ -159,7 +169,7 @@ where
self.shift();
} else {
// Transform MAC EOL into \n
self.chr0 = Some('\n');
self.chr0 = Some('\n')
}
} else {
break;
@ -179,13 +189,13 @@ where
chars: input,
at_begin_of_line: true,
nesting: 0,
indentation_stack: vec![IndentationLevel::default()],
indentation_stack: vec![Default::default()],
pending: Vec::new(),
chr0: None,
location: start,
chr1: None,
chr2: None,
config_comment_prefix: Some(" nac3:"),
config_comment_prefix: Some(" nac3:")
};
lxr.next_char();
lxr.next_char();
@ -207,9 +217,11 @@ where
let mut saw_f = false;
loop {
// Detect r"", f"", b"" and u""
if !(saw_b || saw_u || saw_f) && matches!(self.chr0, Some('b' | 'B')) {
if !(saw_b || saw_u || saw_f) && matches!(self.chr0, Some('b') | Some('B')) {
saw_b = true;
} else if !(saw_b || saw_r || saw_u || saw_f) && matches!(self.chr0, Some('u' | 'U')) {
} else if !(saw_b || saw_r || saw_u || saw_f)
&& matches!(self.chr0, Some('u') | Some('U'))
{
saw_u = true;
} else if !(saw_r || saw_u) && (self.chr0 == Some('r') || self.chr0 == Some('R')) {
saw_r = true;
@ -275,15 +287,15 @@ where
let end_pos = self.get_pos();
let value = match i128::from_str_radix(&value_text, radix) {
Ok(value) => value,
Err(e) => match e.kind() {
Err(e) => {
match e.kind() {
IntErrorKind::PosOverflow | IntErrorKind::NegOverflow => i128::MAX,
_ => {
return Err(LexicalError {
error: LexicalErrorType::OtherError(format!("{e:?}")),
_ => return Err(LexicalError {
error: LexicalErrorType::OtherError(format!("{:?}", e)),
location: start_pos,
})
}),
}
}
},
};
Ok((start_pos, Tok::Int { value }, end_pos))
}
@ -326,7 +338,14 @@ where
if self.chr0 == Some('j') || self.chr0 == Some('J') {
self.next_char();
let end_pos = self.get_pos();
Ok((start_pos, Tok::Complex { real: 0.0, imag: value }, end_pos))
Ok((
start_pos,
Tok::Complex {
real: 0.0,
imag: value,
},
end_pos,
))
} else {
let end_pos = self.get_pos();
Ok((start_pos, Tok::Float { value }, end_pos))
@ -345,7 +364,7 @@ where
let value = value_text.parse::<i128>().ok();
let nonzero = match value {
Some(value) => value != 0i128,
None => true,
None => true
};
if start_is_zero && nonzero {
return Err(LexicalError {
@ -360,7 +379,7 @@ where
/// Consume a sequence of numbers with the given radix,
/// the digits can be decorated with underscores
/// like this: `'1_2_3_4'` == `'1234'`
/// like this: '1_2_3_4' == '1234'
fn radix_run(&mut self, radix: u32) -> String {
let mut value_text = String::new();
@ -393,7 +412,7 @@ where
2 => matches!(c, Some('0'..='1')),
8 => matches!(c, Some('0'..='7')),
10 => matches!(c, Some('0'..='9')),
16 => matches!(c, Some('0'..='9' | 'a'..='f' | 'A'..='F')),
16 => matches!(c, Some('0'..='9') | Some('a'..='f') | Some('A'..='F')),
other => unimplemented!("Radix not implemented: {}", other),
}
}
@ -401,8 +420,8 @@ where
/// Test if we face '[eE][-+]?[0-9]+'
fn at_exponent(&self) -> bool {
match self.chr0 {
Some('e' | 'E') => match self.chr1 {
Some('+' | '-') => matches!(self.chr2, Some('0'..='9')),
Some('e') | Some('E') => match self.chr1 {
Some('+') | Some('-') => matches!(self.chr2, Some('0'..='9')),
Some('0'..='9') => true,
_ => false,
},
@ -414,17 +433,19 @@ where
fn lex_comment(&mut self) -> Option<Spanned> {
self.next_char();
// if possibly nac3 pseudocomment, special handling for `# nac3:`
let (mut prefix, mut is_comment) =
self.config_comment_prefix.map_or_else(|| ("".chars(), false), |v| (v.chars(), true));
let (mut prefix, mut is_comment) = self
.config_comment_prefix
.map_or_else(|| ("".chars(), false), |v| (v.chars(), true));
// for the correct location of config comment
let mut start_loc = self.location;
start_loc.go_left();
loop {
match self.chr0 {
Some('\n') | None => return None,
Some('\n') => return None,
None => return None,
Some(c) => {
if let (true, Some(p)) = (is_comment, prefix.next()) {
is_comment = is_comment && c == p;
is_comment = is_comment && c == p
} else {
// done checking prefix, if is comment then return the spanned
if is_comment {
@ -439,20 +460,22 @@ where
return Some((
start_loc,
Tok::ConfigComment { content: content.trim().into() },
self.location,
self.location
));
}
}
}
}
self.next_char();
}
};
}
fn unicode_literal(&mut self, literal_number: usize) -> Result<char, LexicalError> {
let mut p: u32 = 0u32;
let unicode_error =
LexicalError { error: LexicalErrorType::UnicodeError, location: self.get_pos() };
let unicode_error = LexicalError {
error: LexicalErrorType::UnicodeError,
location: self.get_pos(),
};
for i in 1..=literal_number {
match self.next_char() {
Some(c) => match c.to_digit(16) {
@ -463,8 +486,8 @@ where
}
}
match p {
0xD800..=0xDFFF => Ok(char::REPLACEMENT_CHARACTER),
_ => char::from_u32(p).ok_or(unicode_error),
0xD800..=0xDFFF => Ok(std::char::REPLACEMENT_CHARACTER),
_ => std::char::from_u32(p).ok_or(unicode_error),
}
}
@ -473,7 +496,7 @@ where
octet_content.push(first);
while octet_content.len() < 3 {
if let Some('0'..='7') = self.chr0 {
octet_content.push(self.next_char().unwrap());
octet_content.push(self.next_char().unwrap())
} else {
break;
}
@ -507,8 +530,10 @@ where
}
}
}
unicode_names2::character(&name)
.ok_or(LexicalError { error: LexicalErrorType::UnicodeError, location: start_pos })
unicode_names2::character(&name).ok_or(LexicalError {
error: LexicalErrorType::UnicodeError,
location: start_pos,
})
}
fn lex_string(
@ -541,7 +566,7 @@ where
} else if is_raw {
string_content.push('\\');
if let Some(c) = self.next_char() {
string_content.push(c);
string_content.push(c)
} else {
return Err(LexicalError {
error: LexicalErrorType::StringError,
@ -574,7 +599,7 @@ where
Some('u') if !is_bytes => string_content.push(self.unicode_literal(4)?),
Some('U') if !is_bytes => string_content.push(self.unicode_literal(8)?),
Some('N') if !is_bytes => {
string_content.push(self.parse_unicode_name()?);
string_content.push(self.parse_unicode_name()?)
}
Some(c) => {
string_content.push('\\');
@ -625,15 +650,20 @@ where
let end_pos = self.get_pos();
let tok = if is_bytes {
Tok::Bytes { value: string_content.chars().map(|c| c as u8).collect() }
Tok::Bytes {
value: string_content.chars().map(|c| c as u8).collect(),
}
} else {
Tok::String { value: string_content, is_fstring }
Tok::String {
value: string_content,
is_fstring,
}
};
Ok((start_pos, tok, end_pos))
}
fn is_identifier_start(c: char) -> bool {
fn is_identifier_start(&self, c: char) -> bool {
match c {
'_' | 'a'..='z' | 'A'..='Z' => true,
'+' | '-' | '*' | '/' | '=' | ' ' | '<' | '>' => false,
@ -805,14 +835,18 @@ where
// Check if we have some character:
if let Some(c) = self.chr0 {
// First check identifier:
if Self::is_identifier_start(c) {
if self.is_identifier_start(c) {
let identifier = self.lex_identifier()?;
self.emit(identifier);
} else if is_emoji_presentation(c) {
let tok_start = self.get_pos();
self.next_char();
let tok_end = self.get_pos();
self.emit((tok_start, Tok::Name { name: c.to_string().into() }, tok_end));
self.emit((
tok_start,
Tok::Name { name: c.to_string().into() },
tok_end,
));
} else {
self.consume_character(c)?;
}
@ -865,15 +899,18 @@ where
'=' => {
let tok_start = self.get_pos();
self.next_char();
if let Some('=') = self.chr0 {
match self.chr0 {
Some('=') => {
self.next_char();
let tok_end = self.get_pos();
self.emit((tok_start, Tok::EqEqual, tok_end));
} else {
}
_ => {
let tok_end = self.get_pos();
self.emit((tok_start, Tok::Equal, tok_end));
}
}
}
'+' => {
let tok_start = self.get_pos();
self.next_char();
@ -897,15 +934,18 @@ where
}
Some('*') => {
self.next_char();
if let Some('=') = self.chr0 {
match self.chr0 {
Some('=') => {
self.next_char();
let tok_end = self.get_pos();
self.emit((tok_start, Tok::DoubleStarEqual, tok_end));
} else {
}
_ => {
let tok_end = self.get_pos();
self.emit((tok_start, Tok::DoubleStar, tok_end));
}
}
}
_ => {
let tok_end = self.get_pos();
self.emit((tok_start, Tok::Star, tok_end));
@ -923,15 +963,18 @@ where
}
Some('/') => {
self.next_char();
if let Some('=') = self.chr0 {
match self.chr0 {
Some('=') => {
self.next_char();
let tok_end = self.get_pos();
self.emit((tok_start, Tok::DoubleSlashEqual, tok_end));
} else {
}
_ => {
let tok_end = self.get_pos();
self.emit((tok_start, Tok::DoubleSlash, tok_end));
}
}
}
_ => {
let tok_end = self.get_pos();
self.emit((tok_start, Tok::Slash, tok_end));
@ -1098,15 +1141,18 @@ where
match self.chr0 {
Some('<') => {
self.next_char();
if let Some('=') = self.chr0 {
match self.chr0 {
Some('=') => {
self.next_char();
let tok_end = self.get_pos();
self.emit((tok_start, Tok::LeftShiftEqual, tok_end));
} else {
}
_ => {
let tok_end = self.get_pos();
self.emit((tok_start, Tok::LeftShift, tok_end));
}
}
}
Some('=') => {
self.next_char();
let tok_end = self.get_pos();
@ -1124,15 +1170,18 @@ where
match self.chr0 {
Some('>') => {
self.next_char();
if let Some('=') = self.chr0 {
match self.chr0 {
Some('=') => {
self.next_char();
let tok_end = self.get_pos();
self.emit((tok_start, Tok::RightShiftEqual, tok_end));
} else {
}
_ => {
let tok_end = self.get_pos();
self.emit((tok_start, Tok::RightShift, tok_end));
}
}
}
Some('=') => {
self.next_char();
let tok_end = self.get_pos();
@ -1284,14 +1333,13 @@ where
#[cfg(test)]
mod tests {
use super::{make_tokenizer, NewlineHandler, Tok};
use nac3ast::FileName;
const WINDOWS_EOL: &str = "\r\n";
const MAC_EOL: &str = "\r";
const UNIX_EOL: &str = "\n";
pub fn lex_source(source: &str) -> Vec<Tok> {
let lexer = make_tokenizer(source, FileName::default());
let lexer = make_tokenizer(source, Default::default());
lexer.map(|x| x.unwrap().1).collect()
}
@ -1371,7 +1419,7 @@ class Foo(A, B):
Dedent,
Dedent
]
);
)
}
#[test]
@ -1391,8 +1439,14 @@ class Foo(A, B):
assert_eq!(
tokens,
vec![
Tok::String { value: "\\\\".to_owned(), is_fstring: false },
Tok::String { value: "\\".to_owned(), is_fstring: false },
Tok::String {
value: "\\\\".to_owned(),
is_fstring: false,
},
Tok::String {
value: "\\".to_owned(),
is_fstring: false,
},
Tok::Newline,
]
);
@ -1405,13 +1459,27 @@ class Foo(A, B):
assert_eq!(
tokens,
vec![
Tok::Int { value: 47i128 },
Tok::Int { value: 13i128 },
Tok::Int { value: 0i128 },
Tok::Int { value: 123i128 },
Tok::Int {
value: 47i128,
},
Tok::Int {
value: 13i128,
},
Tok::Int {
value: 0i128,
},
Tok::Int {
value: 123i128,
},
Tok::Float { value: 0.2 },
Tok::Complex { real: 0.0, imag: 2.0 },
Tok::Complex { real: 0.0, imag: 2.2 },
Tok::Complex {
real: 0.0,
imag: 2.0,
},
Tok::Complex {
real: 0.0,
imag: 2.2,
},
Tok::Newline,
]
);
@ -1471,13 +1539,21 @@ class Foo(A, B):
assert_eq!(
tokens,
vec![
Tok::Name { name: String::from("avariable").into() },
Tok::Name {
name: String::from("avariable").into(),
},
Tok::Equal,
Tok::Int { value: 99i128 },
Tok::Int {
value: 99i128
},
Tok::Plus,
Tok::Int { value: 2i128 },
Tok::Int {
value: 2i128
},
Tok::Minus,
Tok::Int { value: 0i128 },
Tok::Int {
value: 0i128
},
Tok::Newline,
]
);
@ -1664,15 +1740,42 @@ class Foo(A, B):
assert_eq!(
tokens,
vec![
Tok::String { value: String::from("double"), is_fstring: false },
Tok::String { value: String::from("single"), is_fstring: false },
Tok::String { value: String::from("can't"), is_fstring: false },
Tok::String { value: String::from("\\\""), is_fstring: false },
Tok::String { value: String::from("\t\r\n"), is_fstring: false },
Tok::String { value: String::from("\\g"), is_fstring: false },
Tok::String { value: String::from("raw\\'"), is_fstring: false },
Tok::String { value: String::from("Đ"), is_fstring: false },
Tok::String { value: String::from("\u{80}\u{0}a"), is_fstring: false },
Tok::String {
value: String::from("double"),
is_fstring: false,
},
Tok::String {
value: String::from("single"),
is_fstring: false,
},
Tok::String {
value: String::from("can't"),
is_fstring: false,
},
Tok::String {
value: String::from("\\\""),
is_fstring: false,
},
Tok::String {
value: String::from("\t\r\n"),
is_fstring: false,
},
Tok::String {
value: String::from("\\g"),
is_fstring: false,
},
Tok::String {
value: String::from("raw\\'"),
is_fstring: false,
},
Tok::String {
value: String::from("Đ"),
is_fstring: false,
},
Tok::String {
value: String::from("\u{80}\u{0}a"),
is_fstring: false,
},
Tok::Newline,
]
);
@ -1727,7 +1830,7 @@ class Foo(A, B):
#[test]
fn test_escape_char_in_byte_literal() {
// backslash does not escape
let source = r#"b"omkmok\Xaa""#;
let source = r##"b"omkmok\Xaa""##;
let tokens = lex_source(source);
let res = vec![111, 109, 107, 109, 111, 107, 92, 88, 97, 97];
assert_eq!(tokens, vec![Tok::Bytes { value: res }, Tok::Newline]);
@ -1737,17 +1840,41 @@ class Foo(A, B):
fn test_raw_byte_literal() {
let source = r"rb'\x1z'";
let tokens = lex_source(source);
assert_eq!(tokens, vec![Tok::Bytes { value: b"\\x1z".to_vec() }, Tok::Newline]);
assert_eq!(
tokens,
vec![
Tok::Bytes {
value: b"\\x1z".to_vec()
},
Tok::Newline
]
);
let source = r"rb'\\'";
let tokens = lex_source(source);
assert_eq!(tokens, vec![Tok::Bytes { value: b"\\\\".to_vec() }, Tok::Newline]);
assert_eq!(
tokens,
vec![
Tok::Bytes {
value: b"\\\\".to_vec()
},
Tok::Newline
]
)
}
#[test]
fn test_escape_octet() {
let source = r"b'\43a\4\1234'";
let source = r##"b'\43a\4\1234'"##;
let tokens = lex_source(source);
assert_eq!(tokens, vec![Tok::Bytes { value: b"#a\x04S4".to_vec() }, Tok::Newline]);
assert_eq!(
tokens,
vec![
Tok::Bytes {
value: b"#a\x04S4".to_vec()
},
Tok::Newline
]
)
}
#[test]
@ -1756,7 +1883,13 @@ class Foo(A, B):
let tokens = lex_source(source);
assert_eq!(
tokens,
vec![Tok::String { value: "\u{2002}".to_owned(), is_fstring: false }, Tok::Newline]
);
vec![
Tok::String {
value: "\u{2002}".to_owned(),
is_fstring: false,
},
Tok::Newline
]
)
}
}

31
nac3parser/src/lib.rs
View File

@ -15,24 +15,6 @@
//!
//! ```
#![deny(
future_incompatible,
let_underscore,
nonstandard_style,
rust_2024_compatibility,
clippy::all
)]
#![warn(clippy::pedantic)]
#![allow(
clippy::enum_glob_use,
clippy::fn_params_excessive_bools,
clippy::missing_errors_doc,
clippy::missing_panics_doc,
clippy::module_name_repetitions,
clippy::too_many_lines,
clippy::wildcard_imports
)]
#[macro_use]
extern crate log;
use lalrpop_util::lalrpop_mod;
@ -45,16 +27,9 @@ pub mod lexer;
pub mod mode;
pub mod parser;
lalrpop_mod!(
#[allow(
future_incompatible,
let_underscore,
nonstandard_style,
rust_2024_compatibility,
unused,
clippy::all,
clippy::pedantic
)]
#[allow(clippy::all)]
#[allow(unused)]
python
);
pub mod config_comment_helper;
pub mod token;
pub mod config_comment_helper;

33
nac3parser/src/parser.rs
View File

@ -5,7 +5,6 @@
//! parse a whole program, a single statement, or a single
//! expression.
use nac3ast::Location;
use std::iter;
use crate::ast::{self, FileName};
@ -64,7 +63,7 @@ pub fn parse_program(source: &str, file: FileName) -> Result<ast::Suite, ParseEr
///
/// ```
pub fn parse_expression(source: &str) -> Result<ast::Expr, ParseError> {
parse(source, Mode::Expression, FileName::default()).map(|top| match top {
parse(source, Mode::Expression, Default::default()).map(|top| match top {
ast::Mod::Expression { body } => *body,
_ => unreachable!(),
})
@ -73,10 +72,12 @@ pub fn parse_expression(source: &str) -> Result<ast::Expr, ParseError> {
// Parse a given source code
pub fn parse(source: &str, mode: Mode, file: FileName) -> Result<ast::Mod, ParseError> {
let lxr = lexer::make_tokenizer(source, file);
let marker_token = (Location::default(), mode.to_marker(), Location::default());
let marker_token = (Default::default(), mode.to_marker(), Default::default());
let tokenizer = iter::once(Ok(marker_token)).chain(lxr);
python::TopParser::new().parse(tokenizer).map_err(ParseError::from)
python::TopParser::new()
.parse(tokenizer)
.map_err(ParseError::from)
}
#[cfg(test)]
@ -85,42 +86,42 @@ mod tests {
#[test]
fn test_parse_empty() {
let parse_ast = parse_program("", FileName::default()).unwrap();
let parse_ast = parse_program("", Default::default()).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_print_hello() {
let source = String::from("print('Hello world')");
let parse_ast = parse_program(&source, FileName::default()).unwrap();
let parse_ast = parse_program(&source, Default::default()).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_print_2() {
let source = String::from("print('Hello world', 2)");
let parse_ast = parse_program(&source, FileName::default()).unwrap();
let parse_ast = parse_program(&source, Default::default()).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_kwargs() {
let source = String::from("my_func('positional', keyword=2)");
let parse_ast = parse_program(&source, FileName::default()).unwrap();
let parse_ast = parse_program(&source, Default::default()).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_if_elif_else() {
let source = String::from("if 1: 10\nelif 2: 20\nelse: 30");
let parse_ast = parse_program(&source, FileName::default()).unwrap();
let parse_ast = parse_program(&source, Default::default()).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_lambda() {
let source = "lambda x, y: x * y"; // lambda(x, y): x * y";
let parse_ast = parse_program(source, FileName::default()).unwrap();
let parse_ast = parse_program(source, Default::default()).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
@ -128,7 +129,7 @@ mod tests {
fn test_parse_tuples() {
let source = "a, b = 4, 5";
insta::assert_debug_snapshot!(parse_program(source, FileName::default()).unwrap());
insta::assert_debug_snapshot!(parse_program(source, Default::default()).unwrap());
}
#[test]
@ -139,7 +140,7 @@ class Foo(A, B):
pass
def method_with_default(self, arg='default'):
pass";
insta::assert_debug_snapshot!(parse_program(source, FileName::default()).unwrap());
insta::assert_debug_snapshot!(parse_program(source, Default::default()).unwrap());
}
#[test]
@ -182,7 +183,7 @@ while i < 2: # nac3: 4
# nac3: if1
if 1: # nac3: if2
3";
insta::assert_debug_snapshot!(parse_program(source, FileName::default()).unwrap());
insta::assert_debug_snapshot!(parse_program(source, Default::default()).unwrap());
}
#[test]
@ -195,7 +196,7 @@ while test: # nac3: while3
# nac3: simple assign0
a = 3 # nac3: simple assign1
";
insta::assert_debug_snapshot!(parse_program(source, FileName::default()).unwrap());
insta::assert_debug_snapshot!(parse_program(source, Default::default()).unwrap());
}
#[test]
@ -214,7 +215,7 @@ if a: # nac3: small2
for i in a: # nac3: for1
pass
";
insta::assert_debug_snapshot!(parse_program(source, FileName::default()).unwrap());
insta::assert_debug_snapshot!(parse_program(source, Default::default()).unwrap());
}
#[test]
@ -223,6 +224,6 @@ for i in a: # nac3: for1
if a: # nac3: something
a = 3
";
assert!(parse_program(source, FileName::default()).is_err());
assert!(parse_program(source, Default::default()).is_err());
}
}

30
nac3parser/src/token.rs
View File

@ -1,7 +1,7 @@
//! Different token definitions.
//! Loosely based on token.h from CPython source:
use crate::ast;
use std::fmt::{self, Write};
use crate::ast;
/// Python source code can be tokenized in a sequence of these tokens.
#[derive(Clone, Debug, PartialEq)]
@ -111,23 +111,15 @@ impl fmt::Display for Tok {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
use Tok::*;
match self {
Name { name } => {
write!(f, "'{}'", ast::get_str_from_ref(&ast::get_str_ref_lock(), *name))
}
Int { value } => {
if *value == i128::MAX {
write!(f, "'#OFL#'")
} else {
write!(f, "'{value}'")
}
}
Float { value } => write!(f, "'{value}'"),
Complex { real, imag } => write!(f, "{real}j{imag}"),
Name { name } => write!(f, "'{}'", ast::get_str_from_ref(&ast::get_str_ref_lock(), *name)),
Int { value } => if *value != i128::MAX { write!(f, "'{}'", value) } else { write!(f, "'#OFL#'") },
Float { value } => write!(f, "'{}'", value),
Complex { real, imag } => write!(f, "{}j{}", real, imag),
String { value, is_fstring } => {
if *is_fstring {
write!(f, "f")?;
write!(f, "f")?
}
write!(f, "{value:?}")
write!(f, "{:?}", value)
}
Bytes { value } => {
write!(f, "b\"")?;
@ -137,16 +129,12 @@ impl fmt::Display for Tok {
10 => f.write_str("\\n")?,
13 => f.write_str("\\r")?,
32..=126 => f.write_char(*i as char)?,
_ => write!(f, "\\x{i:02x}")?,
_ => write!(f, "\\x{:02x}", i)?,
}
}
f.write_str("\"")
}
ConfigComment { content } => write!(
f,
"ConfigComment: '{}'",
ast::get_str_from_ref(&ast::get_str_ref_lock(), *content)
),
ConfigComment { content } => write!(f, "ConfigComment: '{}'", ast::get_str_from_ref(&ast::get_str_ref_lock(), *content)),
Newline => f.write_str("Newline"),
Indent => f.write_str("Indent"),
Dedent => f.write_str("Dedent"),

12
nac3standalone/Cargo.toml
View File

@ -2,18 +2,14 @@
name = "nac3standalone"
version = "0.1.0"
authors = ["M-Labs"]
edition = "2021"
edition = "2018"
[dependencies]
parking_lot = "0.12"
parking_lot = "0.11.1"
nac3parser = { path = "../nac3parser" }
nac3core = { path = "../nac3core" }
[dependencies.clap]
version = "4.5"
features = ["derive"]
[dependencies.inkwell]
version = "0.4"
version = "0.1.0-beta.4"
default-features = false
features = ["llvm14-0", "target-x86", "target-arm", "target-riscv", "no-libffi-linking"]
features = ["llvm13-0", "target-x86", "target-arm", "target-riscv", "no-libffi-linking"]

4
nac3standalone/demo/.gitignore vendored
View File

@ -1,4 +0,0 @@
*.bc
*.ll
*.o
/demo

25
nac3standalone/demo/check_demo.sh
View File

@ -1,25 +0,0 @@
#!/usr/bin/env bash
set -e
if [ -z "$1" ]; then
echo "Requires at least one argument"
exit 1
fi
declare -a nac3args
while [ $# -gt 1 ]; do
nac3args+=("$1")
shift
done
demo="$1"
echo -n "Checking $demo... "
./interpret_demo.py "$demo" > interpreted.log
./run_demo.sh --out run.log "${nac3args[@]}" "$demo"
./run_demo.sh --lli --out run_lli.log "${nac3args[@]}" "$demo"
diff -Nau interpreted.log run.log
diff -Nau interpreted.log run_lli.log
echo "ok"
rm -f interpreted.log run.log run_lli.log

8
nac3standalone/demo/check_demos.sh
View File

@ -4,8 +4,12 @@ set -e
count=0
for demo in src/*.py; do
./check_demo.sh "$@" "$demo"
((count += 1))
echo -n "checking $demo... "
./interpret_demo.py $demo > interpreted.log
./run_demo.sh $demo > run.log
diff -Nau interpreted.log run.log
echo "ok"
let "count+=1"
done
echo "Ran $count demo checks - PASSED"

122
nac3standalone/demo/demo.c
View File

@ -1,122 +0,0 @@
#include <inttypes.h>
#include <math.h>
#include <stdbool.h>
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#define usize size_t
double dbl_nan(void) {
return NAN;
}
double dbl_inf(void) {
return INFINITY;
}
void output_bool(bool x) {
puts(x ? "True" : "False");
}
void output_int32(int32_t x) {
printf("%"PRId32"\n", x);
}
void output_int64(int64_t x) {
printf("%"PRId64"\n", x);
}
void output_uint32(uint32_t x) {
printf("%"PRIu32"\n", x);
}
void output_uint64(uint64_t x) {
printf("%"PRIu64"\n", x);
}
void output_float64(double x) {
if (isnan(x)) {
puts("nan");
} else {
printf("%f\n", x);
}
}
void output_range(int32_t range[3]) {
printf("range(");
printf("%d, %d", range[0], range[1]);
if (range[2] != 1) {
printf(", %d", range[2]);
}
puts(")");
}
void output_asciiart(int32_t x) {
static const char *chars = " .,-:;i+hHM$*#@ ";
if (x < 0) {
putchar('\n');
} else {
putchar(chars[x]);
}
}
struct cslice {
void *data;
usize len;
};
void output_int32_list(struct cslice *slice) {
const int32_t *data = (int32_t *) slice->data;
putchar('[');
for (usize i = 0; i < slice->len; ++i) {
if (i == slice->len - 1) {
printf("%d", data[i]);
} else {
printf("%d, ", data[i]);
}
}
putchar(']');
putchar('\n');
}
void output_str(struct cslice *slice) {
const char *data = (const char *) slice->data;
for (usize i = 0; i < slice->len; ++i) {
putchar(data[i]);
}
}
void output_strln(struct cslice *slice) {
output_str(slice);
putchar('\n');
}
uint64_t dbg_stack_address(__attribute__((unused)) struct cslice *slice) {
int i;
void *ptr = (void *) &i;
return (uintptr_t) ptr;
}
uint32_t __nac3_personality(uint32_t state, uint32_t exception_object, uint32_t context) {
printf("__nac3_personality(state: %u, exception_object: %u, context: %u)\n", state, exception_object, context);
exit(101);
__builtin_unreachable();
}
uint32_t __nac3_raise(uint32_t state, uint32_t exception_object, uint32_t context) {
printf("__nac3_raise(state: %u, exception_object: %u, context: %u)\n", state, exception_object, context);
exit(101);
__builtin_unreachable();
}
void __nac3_end_catch(void) {}
extern int32_t run(void);
int main(void) {
run();
}

90
nac3standalone/demo/demo.rs Normal file
View File

@ -0,0 +1,90 @@
mod cslice {
// copied from https://github.com/dherman/cslice
use std::marker::PhantomData;
use std::slice;
#[repr(C)]
#[derive(Clone, Copy)]
pub struct CSlice<'a, T> {
base: *const T,
len: usize,
marker: PhantomData<&'a ()>,
}
impl<'a, T> AsRef<[T]> for CSlice<'a, T> {
fn as_ref(&self) -> &[T] {
unsafe { slice::from_raw_parts(self.base, self.len) }
}
}
}
#[no_mangle]
pub extern "C" fn output_int32(x: i32) {
println!("{}", x);
}
#[no_mangle]
pub extern "C" fn output_int64(x: i64) {
println!("{}", x);
}
#[no_mangle]
pub extern "C" fn output_uint32(x: u32) {
println!("{}", x);
}
#[no_mangle]
pub extern "C" fn output_uint64(x: u64) {
println!("{}", x);
}
#[no_mangle]
pub extern "C" fn output_float64(x: f64) {
// debug output to preserve the digits after the decimal points
// to match python `print` function
println!("{:?}", x);
}
#[no_mangle]
pub extern "C" fn output_asciiart(x: i32) {
let chars = " .,-:;i+hHM$*#@ ";
if x < 0 {
println!("");
} else {
print!("{}", chars.chars().nth(x as usize).unwrap());
}
}
#[no_mangle]
pub extern "C" fn output_int32_list(x: &cslice::CSlice<i32>) {
print!("[");
let mut it = x.as_ref().iter().peekable();
while let Some(e) = it.next() {
if it.peek().is_none() {
print!("{}", e);
} else {
print!("{}, ", e);
}
}
println!("]");
}
#[no_mangle]
pub extern "C" fn __nac3_personality(_state: u32, _exception_object: u32, _context: u32) -> u32 {
unimplemented!();
}
#[no_mangle]
pub extern "C" fn __nac3_raise(_state: u32, _exception_object: u32, _context: u32) -> u32 {
unimplemented!();
}
extern "C" {
fn run() -> i32;
}
fn main() {
unsafe {
run();
}
}

163
nac3standalone/demo/interpret_demo.py
View File

@ -3,14 +3,10 @@
import sys
import importlib.util
import importlib.machinery
import math
import numpy as np
import numpy.typing as npt
import pathlib
from numpy import int32, int64, uint32, uint64
from scipy import special
from typing import TypeVar, Generic, Literal, Union
from typing import TypeVar, Generic
T = TypeVar('T')
class Option(Generic[T]):
@ -45,99 +41,26 @@ def Some(v: T) -> Option[T]:
none = Option(None)
class _ConstGenericMarker:
pass
def ConstGeneric(name, constraint):
return TypeVar(name, _ConstGenericMarker, constraint)
N = TypeVar("N", bound=np.uint64)
class _NDArrayDummy(Generic[T, N]):
pass
# https://stackoverflow.com/questions/67803260/how-to-create-a-type-alias-with-a-throw-away-generic
NDArray = Union[npt.NDArray[T], _NDArrayDummy[T, N]]
def _bool(x):
if isinstance(x, np.ndarray):
return np.bool_(x)
else:
return bool(x)
def _float(x):
if isinstance(x, np.ndarray):
return np.float_(x)
else:
return float(x)
def round_away_zero(x):
if isinstance(x, np.ndarray):
return np.vectorize(round_away_zero)(x)
else:
if x >= 0.0:
return math.floor(x + 0.5)
else:
return math.ceil(x - 0.5)
def _floor(x):
if isinstance(x, np.ndarray):
return np.vectorize(_floor)(x)
else:
return math.floor(x)
def _ceil(x):
if isinstance(x, np.ndarray):
return np.vectorize(_ceil)(x)
else:
return math.ceil(x)
def patch(module):
def dbl_nan():
return np.nan
def dbl_inf():
return np.inf
def output_asciiart(x):
if x < 0:
sys.stdout.write("\n")
else:
sys.stdout.write(" .,-:;i+hHM$*#@ "[x])
def output_float(x):
print("%f" % x)
def output_strln(x):
print(x, end='')
def dbg_stack_address(_):
return 0
def extern(fun):
name = fun.__name__
if name == "dbl_nan":
return dbl_nan
elif name == "dbl_inf":
return dbl_inf
elif name == "output_asciiart":
if name == "output_asciiart":
return output_asciiart
elif name == "output_float64":
return output_float
elif name == "output_str":
return output_strln
elif name in {
"output_bool",
"output_int32",
"output_int64",
"output_int32_list",
"output_uint32",
"output_uint64",
"output_strln",
"output_range",
"output_float64"
}:
return print
elif name == "dbg_stack_address":
return dbg_stack_address
else:
raise NotImplementedError
@ -145,93 +68,13 @@ def patch(module):
module.int64 = int64
module.uint32 = uint32
module.uint64 = uint64
module.bool = _bool
module.float = _float
module.TypeVar = TypeVar
module.ConstGeneric = ConstGeneric
module.Generic = Generic
module.Literal = Literal
module.extern = extern
module.Option = Option
module.Some = Some
module.none = none
# Builtin Math functions
module.round = round_away_zero
module.round64 = round_away_zero
module.np_round = np.round
module.floor = _floor
module.floor64 = _floor
module.np_floor = np.floor
module.ceil = _ceil
module.ceil64 = _ceil
module.np_ceil = np.ceil
# NumPy ndarray functions
module.ndarray = NDArray
module.np_ndarray = np.ndarray
module.np_empty = np.empty
module.np_zeros = np.zeros
module.np_ones = np.ones
module.np_full = np.full
module.np_eye = np.eye
module.np_identity = np.identity
module.np_array = np.array
# NumPy Math functions
module.np_isnan = np.isnan
module.np_isinf = np.isinf
module.np_min = np.min
module.np_minimum = np.minimum
module.np_max = np.max
module.np_maximum = np.maximum
module.np_sin = np.sin
module.np_cos = np.cos
module.np_exp = np.exp
module.np_exp2 = np.exp2
module.np_log = np.log
module.np_log10 = np.log10
module.np_log2 = np.log2
module.np_fabs = np.fabs
module.np_trunc = np.trunc
module.np_sqrt = np.sqrt
module.np_rint = np.rint
module.np_tan = np.tan
module.np_arcsin = np.arcsin
module.np_arccos = np.arccos
module.np_arctan = np.arctan
module.np_sinh = np.sinh
module.np_cosh = np.cosh
module.np_tanh = np.tanh
module.np_arcsinh = np.arcsinh
module.np_arccosh = np.arccosh
module.np_arctanh = np.arctanh
module.np_expm1 = np.expm1
module.np_cbrt = np.cbrt
module.np_arctan2 = np.arctan2
module.np_copysign = np.copysign
module.np_fmax = np.fmax
module.np_fmin = np.fmin
module.np_ldexp = np.ldexp
module.np_hypot = np.hypot
module.np_nextafter = np.nextafter
# SciPy Math Functions
module.sp_spec_erf = special.erf
module.sp_spec_erfc = special.erfc
module.sp_spec_gamma = special.gamma
module.sp_spec_gammaln = special.gammaln
module.sp_spec_j0 = special.j0
module.sp_spec_j1 = special.j1
# NumPy NDArray Functions
module.np_ndarray = np.ndarray
module.np_empty = np.empty
module.np_zeros = np.zeros
module.np_ones = np.ones
module.np_full = np.full
module.np_eye = np.eye
module.np_identity = np.identity
def file_import(filename, prefix="file_import_"):
filename = pathlib.Path(filename)

66
nac3standalone/demo/run_demo.sh
View File

@ -7,72 +7,14 @@ if [ -z "$1" ]; then
exit 1
fi
declare -a nac3args
while [ $# -ge 1 ]; do
case "$1" in
--help)
echo "Usage: run_demo.sh [--help] [--out OUTFILE] [--lli] [--debug] -- [NAC3ARGS...]"
exit
;;
--out)
shift
outfile="$1"
;;
--lli)
use_lli=1
;;
--debug)
debug=1
;;
--)
shift
break
;;
*)
break
;;
esac
shift
done
while [ $# -ge 1 ]; do
nac3args+=("$1")
shift
done
if [ -n "$debug" ] && [ -e ../../target/debug/nac3standalone ]; then
nac3standalone=../../target/debug/nac3standalone
elif [ -e ../../target/release/nac3standalone ]; then
if [ -e ../../target/release/nac3standalone ]; then
nac3standalone=../../target/release/nac3standalone
else
# used by Nix builds
nac3standalone=../../target/x86_64-unknown-linux-gnu/release/nac3standalone
fi
rm -f ./*.o ./*.bc demo
if [ -z "$use_lli" ]; then
$nac3standalone "${nac3args[@]}"
clang -c -std=gnu11 -Wall -Wextra -O3 -o demo.o demo.c
clang -lm -o demo module.o demo.o
if [ -z "$outfile" ]; then
rm -f *.o
$nac3standalone $1
rustc -o demo demo.rs -Crelocation-model=static -Clink-arg=./module.o
./demo
else
./demo > "$outfile"
fi
else
$nac3standalone --emit-llvm "${nac3args[@]}"
clang -c -std=gnu11 -Wall -Wextra -O3 -emit-llvm -o demo.bc demo.c
shopt -s nullglob
llvm-link -o nac3out.bc module*.bc main.bc
shopt -u nullglob
if [ -z "$outfile" ]; then
lli --extra-module demo.bc --extra-module irrt.bc nac3out.bc
else
lli --extra-module demo.bc --extra-module irrt.bc nac3out.bc > "$outfile"
fi
fi

30
nac3standalone/demo/src/bool_literal_condition.py
View File

@ -1,30 +0,0 @@
# Different cases for using boolean variables in boolean contexts.
# Tests whether all boolean variables (expressed as i8s) are lowered into i1s before used in branching instruction (`br`)
def bfunc(b: bool) -> bool:
return not b
def run() -> int32:
b1 = True
b2 = False
if b1:
pass
if not b2:
pass
while b2:
pass
l = [i for i in range(10) if b2]
b_and = True and False
b_or = True or False
b_and = b1 and b2
b_or = b1 or b2
bfunc(b1)
return 0

16
nac3standalone/demo/src/classes.py
View File

@ -6,10 +6,6 @@ def output_int32(x: int32):
def output_int64(x: int64):
...
@extern
def output_strln(x: str):
...
class B:
b: int32
@ -27,14 +23,10 @@ class A:
def get_a(self) -> int32:
return self.a
# def get_b(self) -> B:
# return self.b
def get_b(self) -> B:
return self.b
class Initless:
def foo(self):
output_strln("hello")
def run() -> int32:
a = A(10)
output_int32(a.a)
@ -43,8 +35,4 @@ def run() -> int32:
output_int32(a.a)
output_int32(a.get_a())
# output_int32(a.get_b().b) FIXME: NAC3 prints garbage
initless = Initless()
initless.foo()
return 0

50
nac3standalone/demo/src/const_generic.py
View File

@ -1,50 +0,0 @@
A = ConstGeneric("A", int32)
B = ConstGeneric("B", uint32)
T = TypeVar("T")
class ConstGenericClass(Generic[A]):
def __init__(self):
pass
class ConstGeneric2Class(Generic[A, B]):
def __init__(self):
pass
class HybridGenericClass2(Generic[A, T]):
pass
class HybridGenericClass3(Generic[T, A, B]):
pass
def make_generic_2() -> ConstGenericClass[Literal[2]]:
return ...
def make_generic2_1_2() -> ConstGeneric2Class[Literal[1], Literal[2]]:
return ...
def make_hybrid_class_2_int32() -> HybridGenericClass2[Literal[2], int32]:
return ...
def make_hybrid_class_i32_0_1() -> HybridGenericClass3[int32, Literal[0], Literal[1]]:
return ...
def consume_generic_2(instance: ConstGenericClass[Literal[2]]):
pass
def consume_generic2_1_2(instance: ConstGeneric2Class[Literal[1], Literal[2]]):
pass
def consume_hybrid_class_2_i32(instance: HybridGenericClass2[Literal[2], int32]):
pass
def consume_hybrid_class_i32_0_1(instance: HybridGenericClass3[int32, Literal[0], Literal[1]]):
pass
def f():
consume_generic_2(make_generic_2())
consume_generic2_1_2(make_generic2_1_2())
consume_hybrid_class_2_i32(make_hybrid_class_2_int32())
consume_hybrid_class_i32_0_1(make_hybrid_class_i32_0_1())
def run() -> int32:
return 0

8
nac3standalone/demo/src/dead_code_issue118.py
View File

@ -1,8 +0,0 @@
def f():
return
return
def run() -> int32:
f()
return 0

102
nac3standalone/demo/src/demo_test.py
View File

@ -1,102 +0,0 @@
@extern
def output_bool(x: bool):
...
@extern
def output_int32(x: int32):
...
@extern
def output_int64(x: int64):
...
@extern
def output_uint32(x: uint32):
...
@extern
def output_uint64(x: uint64):
...
@extern
def output_float64(x: float):
...
@extern
def output_range(x: range):
...
@extern
def output_int32_list(x: list[int32]):
...
@extern
def output_asciiart(x: int32):
...
@extern
def output_str(x: str):
...
@extern
def output_strln(x: str):
...
def test_output_bool():
output_bool(True)
output_bool(False)
def test_output_int32():
output_int32(-128)
def test_output_int64():
output_int64(int64(-256))
def test_output_uint32():
output_uint32(uint32(128))
def test_output_uint64():
output_uint64(uint64(256))
def test_output_float64():
output_float64(0.0)
output_float64(1.0)
output_float64(-1.0)
output_float64(128.0)
output_float64(-128.0)
output_float64(16.25)
output_float64(-16.25)
def test_output_range():
r = range(1, 100, 5)
output_int32(r.start)
output_int32(r.stop)
output_int32(r.step)
output_range(range(10))
output_range(range(1, 10))
output_range(range(1, 10, 2))
def test_output_asciiart():
for i in range(17):
output_asciiart(i)
output_asciiart(0)
def test_output_int32_list():
output_int32_list([0, 1, 3, 5, 10])
def test_output_str_family():
output_str("hello")
output_strln(" world")
def run() -> int32:
test_output_bool()
test_output_int32()
test_output_int64()
test_output_uint32()
test_output_uint64()
test_output_float64()
test_output_range()
test_output_asciiart()
test_output_int32_list()
test_output_str_family()
return 0

32
nac3standalone/demo/src/inheritance.py
View File

@ -1,32 +0,0 @@
from __future__ import annotations
@extern
def output_int32(x: int32):
...
class A:
a: int32
def __init__(self, a: int32):
self.a = a
def f1(self):
self.f2()
def f2(self):
output_int32(self.a)
class B(A):
b: int32
def __init__(self, b: int32):
self.a = b + 1
self.b = b
def run() -> int32:
aaa = A(5)
bbb = B(2)
aaa.f1()
bbb.f1()
return 0

Some files were not shown because too many files have changed in this diff Show More