Compare commits

..

685 Commits

Author SHA1 Message Date
wylited 35b6459c58 nac3core: replace paramter with parameter 2022-04-13 15:42:26 +08:00
wylited e94b25f544 spelling (#264)
Co-authored-by: wylited <ds@m-labs.hk>
Co-committed-by: wylited <ds@m-labs.hk>
2022-04-13 11:32:31 +08:00
Sebastien Bourdeauducq 6972689469 nac3artiq: cleanup demo 2022-04-12 10:34:14 +08:00
Sebastien Bourdeauducq 3fb22c9182 nac3artiq: treat host numpy.float64 as float. Closes #90 2022-04-12 10:33:28 +08:00
Sebastien Bourdeauducq 1e7abf0268 fix tests 2022-04-12 10:06:41 +08:00
Sebastien Bourdeauducq f5a6d29106 update insta snapshots 2022-04-12 09:56:49 +08:00
Sebastien Bourdeauducq ca07cb66cd format typevars consistently 2022-04-12 09:28:17 +08:00
Sebastien Bourdeauducq 93e9a6a38a update dependencies 2022-04-12 09:13:04 +08:00
ychenfo 722e3df086 nac3core, artiq: optimize kernel invariant for tuple index 2022-04-11 14:58:40 +08:00
ychenfo ad9ad22cb8 nac3core: optimize unwrap KernelInvariant 2022-04-11 14:58:35 +08:00
ychenfo f66f66b3a4 nac3core, artiq: remove unnecessary ptr casts 2022-04-10 01:28:46 +08:00
ychenfo 6c485bc9dc nac3artiq: skip attribute writeback for option
option types do not have any fields to be written back to the host so it is ok to skip. If we do not skip, there will be error when getting the value of it since it can be `none`, whose type is not concrete
2022-04-10 01:28:30 +08:00
ychenfo 089bba96a3 nac3artiq: get_obj_value take an additional argument for expected type 2022-04-10 01:28:30 +08:00
ychenfo 0e0871bc38 nac3core, artiq: to_basic_value_enum takes an argument indicating the expected type 2022-04-10 01:28:22 +08:00
ychenfo 26187bff0b nac3core: add missing bound check and negative index handling for list subscription assignment 2022-04-09 04:56:31 +08:00
ychenfo 86ce513cb5 nac3standalone: fix broken test
previously this test unexpectedly passed because it is a slice assignment to extend the list, which is valid in CPython and hence in interpret_demo, and which also happened to give the same output in nac3 by memmove the elements in the list of bool
2022-04-05 18:21:46 +08:00
ychenfo c29cbf6ddd nac3core: add bound check for list slice 2022-04-05 18:21:46 +08:00
ychenfo 7443c5ea0f nac3core: add location information to codegen context 2022-04-05 18:21:46 +08:00
Sebastien Bourdeauducq f55b077e60 README: update Windows instructions 2022-04-05 18:07:38 +08:00
Sebastien Bourdeauducq e05b0bf5dc flake: update nixpkgs 2022-04-05 10:10:08 +08:00
Sebastien Bourdeauducq 8eda0affc9 windows: add wine-msys2-build 2022-04-05 10:06:36 +08:00
Sebastien Bourdeauducq 75c53b40a3 windows: update msys2 packages, add setuptools to environment 2022-04-05 10:06:14 +08:00
pca006132 0d10044d66 Merge pull request 'Fix float**int with negative power' (#254) from neg_powi_fix into master
Reviewed-on: M-Labs/nac3#254
2022-04-04 22:43:20 +08:00
ychenfo 23b7f4ef18 nac3standalone: add tests for power 2022-04-04 22:10:56 +08:00
ychenfo 710904f975 nac3core: fix powi with negative integer power 2022-04-04 22:10:56 +08:00
Sebastien Bourdeauducq 4bf452ec5a windows: do not check dependencies when making package 2022-04-04 16:03:59 +08:00
Sebastien Bourdeauducq 9fdce11efe windows: depend on python 2022-04-04 15:21:34 +08:00
Sebastien Bourdeauducq f24ef85aed hydra: use msys2 type 2022-04-04 15:03:53 +08:00
Sebastien Bourdeauducq 4a19787f10 README: update 2022-04-04 15:03:44 +08:00
Sebastien Bourdeauducq 8209c0a475 windows: create MSYS2 package 2022-04-04 14:24:47 +08:00
pca006132 4f66bdeda9 Merge pull request 'nac3core: do not get llvm value too eagerly for kernel invariant' (#253) from kernel_invariant_fix into master
Reviewed-on: M-Labs/nac3#253
2022-03-31 12:48:49 +08:00
Sebastien Bourdeauducq 57369896d7 update dependencies 2022-03-31 10:40:18 +08:00
ychenfo 2edeb31d21 nac3core: do not get llvm value too eagerly for kernel invariant 2022-03-31 10:28:16 +08:00
ychenfo b8ef44d64e nac3standalone: add test for default param 2022-03-30 04:05:47 +08:00
ychenfo c3156afebd nac3core: fix broken tests 2022-03-30 04:05:47 +08:00
ychenfo 388c9b7241 nac3core: better check and err msg for default param 2022-03-30 04:05:47 +08:00
ychenfo e52d7fc97a nac3artiq: resolve unsigned int host variable as defautl param 2022-03-30 04:05:47 +08:00
ychenfo 6ab73a223c nac3core/artiq: support default param of option type 2022-03-30 04:05:47 +08:00
ychenfo a38cc04444 nac3core: assert statement 2022-03-29 06:56:40 +08:00
ychenfo 1f5826d352 fix ternary if (#250)
Use store and load to handle if expression as the blocks might be changed when generating sub-expressions.

Reviewed-on: M-Labs/nac3#250
Co-authored-by: ychenfo <yc@m-labs.hk>
Co-committed-by: ychenfo <yc@m-labs.hk>
2022-03-29 06:54:00 +08:00
Sebastien Bourdeauducq 94eebde4ea README: add note about MSVC Python 2022-03-28 10:45:01 +08:00
Sebastien Bourdeauducq 63ec382673 README: update Windows instructions 2022-03-27 19:36:02 +08:00
Sebastien Bourdeauducq 0ca1a7bedb windows: work around broken LLD install script 2022-03-27 19:14:02 +08:00
Sebastien Bourdeauducq 201ca3f63d Revert "nac3artiq: use lld.exe on Windows"
This reverts commit 19182759cd.
2022-03-27 19:09:11 +08:00
Sebastien Bourdeauducq 19182759cd nac3artiq: use lld.exe on Windows 2022-03-27 18:41:38 +08:00
Sebastien Bourdeauducq edd039abdc windows: build LLD 2022-03-27 18:41:23 +08:00
Sebastien Bourdeauducq 3852cc1058 windows: don't fixup LLVM 2022-03-27 18:38:23 +08:00
Sebastien Bourdeauducq 0600ee8efa nac3artiq: use correct lld invokation on Windows 2022-03-27 18:25:14 +08:00
ychenfo bed33a7421 nac3standalone: add tests for tuple 2022-03-27 10:31:20 +08:00
ychenfo 0d2b844a2e nac3artiq: avoid getting tuple as pointer value 2022-03-27 10:31:20 +08:00
ychenfo 8d7e300a4a nac3core: do not use const struct for tuple 2022-03-27 10:13:17 +08:00
ychenfo 10d623e36f nac3core/artiq: fix tuple representation 2022-03-27 07:47:14 +08:00
ychenfo 000b128551 nac3artiq: cast none to correct ptr type (#241) 2022-03-26 23:32:50 +08:00
Sebastien Bourdeauducq e4581a6d9b nac3standalone/demo: fix return type in loop.py 2022-03-26 21:10:12 +08:00
pca006132 1a82d296e7 nac3core/codegen: prevent users from modifying loop counter
Fixes #211
2022-03-26 20:58:37 +08:00
pca006132 bf067e2481 nac3artiq: implement attribute writeback
We will only writeback attributes that are supported by the current RPC
implementation: primitives, tuple and lists of lists... of primitives.
2022-03-26 20:13:58 +08:00
ychenfo ba8ed6c663 nac3artiq: handle recursive types properly 2022-03-26 18:54:21 +08:00
ychenfo 26a4834254 fix warnings 2022-03-26 18:52:08 +08:00
Sebastien Bourdeauducq 1ad4b0227c windows: fix src location 2022-03-26 15:46:21 +08:00
Sebastien Bourdeauducq 6288a66dc5 windows: fix cargo lockfile location 2022-03-26 15:23:31 +08:00
Sebastien Bourdeauducq de4320eefb improve package names 2022-03-26 15:15:59 +08:00
Sebastien Bourdeauducq a380cd5010 move all Nix files to one folder 2022-03-26 15:13:43 +08:00
ychenfo 80631fc92b Option type support (#224)
Co-authored-by: ychenfo <yc@m-labs.hk>
Co-committed-by: ychenfo <yc@m-labs.hk>
2022-03-26 15:09:15 +08:00
Sebastien Bourdeauducq 55db05fdbb update dependencies 2022-03-24 22:30:15 +08:00
pca006132 24a26b53ae nac3core/toplevel: fixed broken tests
We should not include class type variables in functions type variables.
2022-03-24 21:33:09 +08:00
pca006132 1084ba2158 nac3core: fixed typevar with finite range
1. Function type variables should not include class type variables,
   because they are not bound to the function.
2. Defer type variable constraint evaluation until we get all fields
   definition.
2022-03-24 21:31:51 +08:00
ychenfo be75fa7368 nac3core: fix assign to constant 2022-03-24 07:13:13 +08:00
Sebastien Bourdeauducq ec52128a4a indentation 2022-03-23 10:45:28 +08:00
Sebastien Bourdeauducq b10b49e39a windows: run cargo tests 2022-03-23 09:53:45 +08:00
Sebastien Bourdeauducq d92ce201d3 runkernel: fix windows build 2022-03-23 09:32:58 +08:00
Sebastien Bourdeauducq 8b485f552b windows: set PYO3_CONFIG_FILE, use exec in wine-msys2 2022-03-23 09:22:33 +08:00
pca006132 d9be8d3978 nac3core/typecheck/unification_table: fixed snapshot restore bug
Closes 229
2022-03-23 00:25:10 +08:00
pca006132 41d62f7325 nac3core/toplevel: fixed typevar substitution bug 2022-03-23 00:25:10 +08:00
Sebastien Bourdeauducq 4400d9b57d windows: attempt to disable libffi further 2022-03-22 22:52:53 +08:00
Sebastien Bourdeauducq 8ee5db7462 Revert "windows: numpy is not necessary for build"
...but it is nice to have in the development shell.

This reverts commit 1114d11b34.
2022-03-22 22:15:18 +08:00
Sebastien Bourdeauducq 6d9b3abcd7 nicer MSYS2 development shell 2022-03-22 22:14:46 +08:00
Sebastien Bourdeauducq f11a0776e7 README: fix build name 2022-03-22 22:12:23 +08:00
Sebastien Bourdeauducq f2dc03dfa1 windows: finalize nac3artiq packaging 2022-03-22 19:58:31 +08:00
Sebastien Bourdeauducq 1c807ebe08 windows: use Z: consistently for Nix store paths 2022-03-22 19:43:31 +08:00
Sebastien Bourdeauducq 9e0b5187dd windows: make LLVM accessible to nac3artiq 2022-03-22 19:40:40 +08:00
Sebastien Bourdeauducq 1887a337ff windows: attempt to fix 'encodings' python errors 2022-03-22 19:38:32 +08:00
Sebastien Bourdeauducq 03f5b80153 windows: add nac3artiq derivation (WIP) 2022-03-22 19:19:06 +08:00
Sebastien Bourdeauducq 1114d11b34 windows: numpy is not necessary for build 2022-03-22 19:05:58 +08:00
Sebastien Bourdeauducq a7a188da76 windows: work around -DLLVM_BUILD_TOOLS=OFF not disabling llvm-lto 2022-03-22 18:35:06 +08:00
Sebastien Bourdeauducq eb6ceefdcd build LLVM with Wine + MSYS2 2022-03-22 18:03:25 +08:00
Sebastien Bourdeauducq 9332d1643c irrt: normalize end-of-line characters. Closes #231 2022-03-22 16:04:48 +08:00
Sebastien Bourdeauducq 718b076e50 irrt: use __builtin_alloca 2022-03-22 15:41:25 +08:00
Sebastien Bourdeauducq 9d86b46e86 nac3core: add DEBUG_DUMP_IRRT 2022-03-22 15:39:15 +08:00
ychenfo 263bc82434 nac3artiq: remove debug print 2022-03-21 04:23:40 +08:00
Sebastien Bourdeauducq 3f890f183c nac3standalone/demo: handle imports consistently 2022-03-19 09:14:27 +08:00
pca006132 234823c51a nac3standalone: added typevar test 2022-03-18 16:52:52 +08:00
pca006132 b97c016629 nac3core: fixed test breakage 2022-03-18 16:52:28 +08:00
Sebastien Bourdeauducq 14a5c7981e Revert "Revert "update dependencies""
This reverts commit 93af337ed3.
2022-03-18 08:06:13 +08:00
pca006132 35ac5cb6f6 nac3core: fixed typevar bug 2022-03-18 01:07:44 +08:00
pca006132 93af337ed3 Revert "update dependencies"
This reverts commit 9ccdc0180d.
2022-03-17 21:53:58 +08:00
Sebastien Bourdeauducq 0ca2797428 fix compilation warning 2022-03-17 21:31:45 +08:00
Sebastien Bourdeauducq 9ccdc0180d update dependencies 2022-03-17 21:18:07 +08:00
Sebastien Bourdeauducq c5993c2a58 composer: improve class field typevar error message 2022-03-17 21:04:42 +08:00
pca006132 fb8553311c nac3artiq: remove accidentally added print 2022-03-17 15:13:00 +08:00
pca006132 04e7a7eb4b nac3artiq: support more exceptions 2022-03-17 15:03:22 +08:00
pca006132 642e3b2bad nac3core: moved all builtin errors to nac3artiq code
This remove the need for hard-coding those definition IDs.
2022-03-17 00:04:49 +08:00
pca006132 e126fef012 nac3artiq: support more builtin errors 2022-03-16 23:42:08 +08:00
Sebastien Bourdeauducq 8fd868a673 update dependencies 2022-03-10 17:28:56 +08:00
pca006132 94aac16cc5 nac3artiq: fixed RPC codegen for lists 2022-03-10 16:48:28 +08:00
pca006132 2f85bb3837 nac3core: impl call attributes
sret for returning large structs, and byval for struct args in extern
function calls.
2022-03-09 22:09:36 +08:00
ychenfo e266d3c2b0 nac3parser: modify to handle UAdd in front of int constant 2022-03-09 10:46:58 +08:00
ychenfo 60b3807ab3 nac3standalone: add test for abs function 2022-03-08 23:26:01 +08:00
ychenfo 5006028e2d nac3core: abs builtin function 2022-03-08 23:23:36 +08:00
ychenfo 1cc276cb43 nac3standalone: add test for max function 2022-03-08 22:23:13 +08:00
ychenfo 8241a29908 nac3core: max builtin function 2022-03-08 22:22:00 +08:00
ychenfo e9a17cf8f8 nac3standalone: add test for min function 2022-03-08 21:59:42 +08:00
ychenfo adb5c69e67 nac3core: min builtin function 2022-03-08 21:59:37 +08:00
ychenfo d848c2284e nac3parser: fix broken tests 2022-03-08 18:21:19 +08:00
ychenfo f7e62ab5b7 nac3ast/parser/core: use i128 for u64 constants 2022-03-08 18:21:14 +08:00
ychenfo 9f6c7b3359 nac3core: type conversion to/from uint 2022-03-08 13:42:45 +08:00
ychenfo 142e99a0f1 nac3core: fix broken tests 2022-03-08 13:34:08 +08:00
ychenfo 79c469301a basic unsigned integer support 2022-03-08 13:34:02 +08:00
ychenfo 8602852241 nac3core: use signed extension to convert i32 to i64 2022-03-06 04:49:02 +08:00
ychenfo 42fbe8e383 nac3core: fix err msg of too many args 2022-03-05 03:59:45 +08:00
pca006132 63b0f29728 Fix broken tests 2022-03-05 00:27:51 +08:00
pca006132 a5e1da0b92 nac3artiq/demo/embedding_map: avoid key 0
Object key 0 is reserved for builtin exceptions.
2022-03-05 00:27:23 +08:00
pca006132 294943e303 nac3core: get exception ID from symbol resolver
We need to store the exception class somewhere in order to create them
back in the host. Fixes #200
2022-03-05 00:26:35 +08:00
ychenfo 84b4bd920b nac3artiq: remove cached pyid_to_type if error 2022-03-04 16:23:25 +08:00
Sebastien Bourdeauducq 317eb80005 update dependencies 2022-03-03 17:10:22 +08:00
Sebastien Bourdeauducq 59ac5aae8a fix error message string (2) 2022-03-02 08:33:13 +08:00
Sebastien Bourdeauducq da039e3acf fix error message string 2022-03-02 08:04:15 +08:00
pca006132 d1e172501d nac3artiq: remove debug messages 2022-02-28 23:10:05 +08:00
pca006132 323d77a455 nac3artiq: improve error message for out of range error 2022-02-28 23:09:14 +08:00
pca006132 d41c923cfd nac3artiq: handle recursive types properly 2022-02-28 23:08:42 +08:00
Sebastien Bourdeauducq 5d8e87d923 more readable type annotation error string 2022-02-28 16:24:03 +08:00
Sebastien Bourdeauducq a9c73a4915 fix some error strings 2022-02-28 11:10:33 +08:00
Sebastien Bourdeauducq 804d5db27e nac3artiq: make CompileError importable from Python 2022-02-26 17:29:13 +08:00
Sebastien Bourdeauducq cbc77dddb0 nac3artiq: raise specific exception on error 2022-02-26 17:17:06 +08:00
pca006132 846d1726ef nac3core: fixed keyword arguments handling 2022-02-26 16:34:30 +08:00
pca006132 0686e83f4c nac3core/typecheck: fixed incorrect rollback 2022-02-25 20:01:11 +08:00
pca006132 e710b6c320 nac3core: fix exception final branch handling
According to https://github.com/m-labs/artiq/pull/1855
Passed the test cases from 1855.
Fixes #196.
2022-02-25 17:42:47 +08:00
pca006132 cc769a7006 nac3core: reset unification table state before printing errors
Fixes nondeterministic error messages due to nondeterministic
unification order. As all unification operations will be restored, the
error messages should not be affected by the unification order before
the failure operation.
2022-02-25 14:47:19 +08:00
Sebastien Bourdeauducq 5cd4fe6507 update tests 2022-02-23 11:50:03 +08:00
Sebastien Bourdeauducq aa79c8d8b7 rename exception symbols in host code 2022-02-23 11:43:41 +08:00
Sebastien Bourdeauducq 75fde1bbf7 update tests 2022-02-23 11:39:47 +08:00
Sebastien Bourdeauducq 17792b76b7 rename exception symbols 2022-02-23 11:04:35 +08:00
Sebastien Bourdeauducq 6ae770d5eb update dependencies 2022-02-23 10:59:13 +08:00
pca006132 d3cb5d6e52 Fixed type error messages 2022-02-22 17:22:15 +08:00
Sebastien Bourdeauducq bb7c0a2d79 nac3artiq: remove errors from demo 2022-02-22 16:00:37 +08:00
pca006132 3ad25c8f07 nac3core: sort error messages for determinism 2022-02-22 14:33:43 +08:00
pca006132 ede3706ca8 type_inferencer: special case tuple index error message 2022-02-21 18:41:42 +08:00
pca006132 f97f93d92c applied rustfmt and clippy auto fix 2022-02-21 18:27:46 +08:00
pca006132 d9cb506f6a nac3core: refactored for better error messages 2022-02-21 18:24:19 +08:00
pca006132 352831b2ca nac3core: removed legacy location definition 2022-02-13 22:39:24 +08:00
pca006132 21d9182ba2 nac3core: disallow methods/fields in Exception subclass
Fixes #192
2022-02-13 21:45:22 +08:00
Sebastien Bourdeauducq 91f41052fe test: remove outdated comment 2022-02-13 17:24:47 +08:00
pca006132 14d25b3b9d Fixed broken tests 2022-02-13 17:21:42 +08:00
Sebastien Bourdeauducq 265d234266 update LLVM 2022-02-13 13:20:08 +08:00
Sebastien Bourdeauducq 2e44745933 runkernel: add dummy artiq_personality function 2022-02-13 13:03:38 +08:00
Sebastien Bourdeauducq 4b8e70f746 nac3standalone: disable broken tests (#188) 2022-02-13 11:41:42 +08:00
Sebastien Bourdeauducq 31e76ca3b6 nac3standalone: add dummy support for artiq_personality
So existing tests can run again
2022-02-13 11:35:02 +08:00
Sebastien Bourdeauducq 343f6fd067 update dependencies 2022-02-13 10:51:03 +08:00
Sebastien Bourdeauducq f1ebf8f96e flake: update nixpkgs 2022-02-13 10:47:22 +08:00
pca006132 b18626b149 Fix compilation and test failures 2022-02-12 22:50:32 +08:00
pca006132 750d912eb4 nac3core: do list bound check and negative index handling
Raise error when index out of range. Note that we use llvm.expect to
tell the optimizer that we expect not to raise an exception, so the
normal path performance would be better. If this assumption is violated,
the exception overhead might be slightly larger, but the percentage
increase in overhead should not be high since exception unwinding is
already pretty slow.
2022-02-12 22:50:32 +08:00
pca006132 bf52e294ee nac3artiq: RPC support 2022-02-12 22:50:32 +08:00
pca006132 e303248261 nac3core: exception type check and codegen 2022-02-12 22:50:32 +08:00
pca006132 7ea5a5f84d nac3core: codegen refactoring
- No longer check if the statement will return. Instead, we check if
  the current basic block is terminated, which is simpler and handles
  exception/break/continue better.
- Use invoke statement when unwind is needed.
- Moved codegen for a block of statements into a separate function.
2022-02-12 22:13:59 +08:00
pca006132 b267a656a8 nac3core: added exception type and fixed primitive representation
- Added `Exception` primitive type and some builtin exception types.
  Note that all exception types share the same layout, and should
  inherit from the base `Exception` type. There are some hacks in the
  toplevel module for handling exception types, we should revisit and
  fix them later.
- Added new primitive types to concrete type module, otherwise there
  would be some weird type errors.
- Changed the representation of strings to CSlice<u8>, instead of
  CString.
2022-02-12 22:13:59 +08:00
pca006132 050c862c1a nac3core: function codegen callback changes
Added code generator argument to the callback, so it would be easier to
write complicated codegen with that callback. To prepare for RPC
codegen.
2022-02-12 21:24:41 +08:00
Sebastien Bourdeauducq ffe89eec86 llvm: disable threads 2022-02-08 14:52:09 +08:00
ychenfo d6ab73afb0 nac3core: style 2022-02-07 02:18:56 +08:00
ychenfo 6f9f455152 nac3core: list slice irrt use one function to handle var size 2022-02-07 02:09:50 +08:00
ychenfo e50f1017fa nac3core: irrt list of tuple use struct value representation 2022-02-07 02:09:50 +08:00
ychenfo 77608346b1 nac3core: handle tuple by value 2022-02-07 02:09:50 +08:00
Sebastien Bourdeauducq f5ce7376e3 flake: fix Windows build 2022-02-05 16:53:47 +08:00
Sebastien Bourdeauducq 1288624218 lock insta version (#179) 2022-01-31 15:18:49 +08:00
Sebastien Bourdeauducq 0124bcd26c update dependencies (missing part of previous commit) 2022-01-31 14:15:05 +08:00
Sebastien Bourdeauducq de065cfa14 update dependencies 2022-01-31 12:28:40 +08:00
pca006132 304181fd8c Merge pull request 'fix errors of non-primitive host object when running multiple kernels' (#171) from multiple_kernel_err into master
Reviewed-on: M-Labs/nac3#171
2022-01-27 14:46:22 +08:00
ychenfo 43048eb8d8 nac3standalone: add tests for list slice and len 2022-01-26 03:58:27 +08:00
ychenfo ace0e2a2c6 nac3core: fix use of size_t in list comprehension, cleanup 2022-01-25 03:35:59 +08:00
Sebastien Bourdeauducq e891683f2e flake: hack-link libstdc++ statically on Windows. Closes #175 2022-01-24 16:54:05 +08:00
Sebastien Bourdeauducq 8e01a20ac3 README: add Windows instructions 2022-01-24 15:54:01 +08:00
Sebastien Bourdeauducq 465514ca7a flake: fix mcfgthread filename 2022-01-24 15:52:04 +08:00
Sebastien Bourdeauducq 9c34dd9c80 flake: distribute mcfgthreads-12.dll on hydra 2022-01-24 15:49:32 +08:00
Sebastien Bourdeauducq ced7acd871 check_demos: improve output 2022-01-24 11:38:43 +08:00
Sebastien Bourdeauducq 6ea40809b3 README: fix nix shell URL 2022-01-24 11:35:39 +08:00
Sebastien Bourdeauducq f8e3f7a4ca add some basic list tests 2022-01-23 14:28:08 +08:00
Sebastien Bourdeauducq ba997ae094 flake: run nac3standalone demo checks
also keep auxiliary projects in separate Nix outputs
2022-01-23 11:32:34 +08:00
Sebastien Bourdeauducq 2a0caf931f nac3standalone: work around bash mess with exit codes of substituted processes
https://unix.stackexchange.com/questions/376114/how-to-detect-an-error-using-process-substitution
2022-01-23 11:15:11 +08:00
Sebastien Bourdeauducq 64b94955fe nac3standalone: reorganize demos, compare against cpython 2022-01-23 10:35:06 +08:00
Sebastien Bourdeauducq f478c6afcc update dependencies 2022-01-19 21:17:07 +08:00
ychenfo 0439bf6aef nac3artiq: fix errors of non-primitive object when running multiple kernels 2022-01-15 04:43:39 +08:00
Sebastien Bourdeauducq fd4bf12808 fix grammar of some type error messages 2022-01-14 16:56:23 +08:00
Sebastien Bourdeauducq d7b14dd705 update dependencies 2022-01-14 16:55:10 +08:00
ychenfo 9d342d9f0f nac3artiq: error msg improvement for synthesized __modinit__ 2022-01-14 16:28:37 +08:00
ychenfo ae8f82ccb0 nac3core: fix broken tests 2022-01-14 16:28:37 +08:00
ychenfo 4a1a4dc076 nac3core/artiq/standalone: symbol resolver return error msg for type error of host variables 2022-01-14 16:28:34 +08:00
ychenfo eba9fc8a69 nac3core: add missing location for type inference 2022-01-14 03:05:11 +08:00
ychenfo 4976e89ae2 nac3core: list slice support 2022-01-13 16:53:32 +08:00
Sebastien Bourdeauducq 82509d60ec remove obvious comment 2022-01-13 12:31:28 +08:00
ychenfo 2579ecbd19 nac3core: irrt module get attribute id using name instead of hard code 2022-01-11 17:25:07 +08:00
ychenfo 44f4c4f028 nac3core: build script use Path::join 2022-01-09 12:06:45 +08:00
Sebastien Bourdeauducq 8ef9e74aaf move rustfmt.toml upper 2022-01-09 11:31:06 +08:00
Sebastien Bourdeauducq 9c20e84c84 flake: fix/cleanup 2022-01-09 11:30:36 +08:00
Sebastien Bourdeauducq b88f17ed42 switch to clang-unwrapped, build IRRT with wasm32 2022-01-09 10:56:28 +08:00
Sebastien Bourdeauducq 096193f7ab demo: rewrite in Rust 2022-01-09 10:51:10 +08:00
ychenfo 4760851638 nac3standalone: link modules and load irrt like in nac3artiq 2022-01-09 02:17:58 +08:00
ychenfo 1ee857de6a nac3core: format, fix clippy warning 2022-01-09 01:12:18 +08:00
Sebastien Bourdeauducq 4a65d82db5 introduce IRRT, implement power
based on code by Yijia
M-Labs/nac3#160
2022-01-09 00:57:50 +08:00
Sebastien Bourdeauducq b638d1b4b0 nac3standalone: set up LLVM inliner like in nac3artiq 2022-01-08 21:03:58 +08:00
Sebastien Bourdeauducq 52ccf31bb1 update dependencies 2022-01-04 22:00:29 +08:00
Sebastien Bourdeauducq 4904610dc6 flake: provide mimalloc-enabled Python
The Linux linker and the libc are garbage, so there isn't much of an alternative to using the Nix wrapper and LD_PRELOAD.
2022-01-04 21:54:55 +08:00
ychenfo 7193e3f328 nac3core: codegen fix empty list llvm type 2021-12-30 05:09:21 +08:00
Sebastien Bourdeauducq 2822c613ef llvm: fix TLI-musl.patch 2021-12-29 20:52:59 +08:00
Sebastien Bourdeauducq a0bf6da6c2 update dependencies 2021-12-28 12:08:55 +08:00
Sebastien Bourdeauducq 9cc9a0284a nac3standalone: style 2021-12-28 10:59:17 +08:00
ychenfo 85e06d431a nac3core: improve some type annotation error messages (#87) 2021-12-28 10:49:14 +08:00
ychenfo 9b3b47ce50 fix broken tests 2021-12-28 01:38:16 +08:00
ychenfo 88f0da7bdd add file name to AST node location 2021-12-28 01:28:55 +08:00
pca006132 1bd966965e fixed M-Labs/nac3#146 2021-12-27 22:56:50 +08:00
pca006132 521f136f2e redo "nac3artiq: fixed compilation error"
This reverts commit 3b5328d3cd.
2021-12-27 22:56:30 +08:00
pca006132 fa04768a77 redo "nac3core: fix #84"
This reverts commit 86005da8e1.
2021-12-27 22:56:26 +08:00
Sebastien Bourdeauducq 6162d21a5b LLVM PGO support 2021-12-26 21:11:14 +08:00
Sebastien Bourdeauducq 8101483ebd flake: style 2021-12-26 18:57:02 +08:00
Sebastien Bourdeauducq dc5e42c5eb flake: use LLVM 13 throughout 2021-12-26 18:56:23 +08:00
Sebastien Bourdeauducq 86005da8e1 Revert "nac3core: fix #84"
This reverts commit 0902d8adf4.
2021-12-26 08:35:27 +08:00
Sebastien Bourdeauducq 3b5328d3cd Revert "nac3artiq: fixed compilation error"
This reverts commit 34cabe0e55.
2021-12-26 08:31:37 +08:00
Sebastien Bourdeauducq 5aa6749241 remove num-traits 2021-12-26 00:32:08 +08:00
Sebastien Bourdeauducq 80d3ab1b0f remove bigints 2021-12-26 00:23:54 +08:00
Sebastien Bourdeauducq ec986dfdf3 update dependencies 2021-12-25 23:03:53 +08:00
Sebastien Bourdeauducq d2a5cd6d57 update to LLVM 13 2021-12-25 22:49:47 +08:00
Sebastien Bourdeauducq 9e3f75255e update inkwell. Closes #67 2021-12-25 22:17:06 +08:00
Sebastien Bourdeauducq 53f13b44cf flake: update nixpkgs 2021-12-25 21:10:19 +08:00
pca006132 34cabe0e55 nac3artiq: fixed compilation error 2021-12-23 15:47:54 +08:00
pca006132 6e85f549f6 Merge pull request 'nac3core: fix #84' (#146) from fix_84 into master
Reviewed-on: M-Labs/nac3#146
2021-12-23 15:28:29 +08:00
pca006132 0902d8adf4 nac3core: fix #84 2021-12-23 15:26:48 +08:00
ychenfo 66320679be improve error messages
#112, #110, #108, #87

Reviewed-on: M-Labs/nac3#145
Co-authored-by: ychenfo <yc@m-labs.hk>
Co-committed-by: ychenfo <yc@m-labs.hk>
2021-12-22 08:52:19 +08:00
Sebastien Bourdeauducq 0ff995722c Revert "nac3core: add missing expr concrete type check"
This reverts commit cb450372d6.
2021-12-20 18:13:45 +08:00
Sebastien Bourdeauducq e2b44a066b return int32 in len(). Closes #141 2021-12-20 17:44:42 +08:00
Sebastien Bourdeauducq 2008db8097 nac3standalone: remove unused import 2021-12-20 17:39:16 +08:00
ychenfo cb450372d6 nac3core: add missing expr concrete type check 2021-12-19 18:01:49 +08:00
ychenfo ff27a1697e nac3core: fix for loop type inference 2021-12-19 18:01:49 +08:00
ychenfo 91625dd327 update kernel-only attribute annotation
Reviewed-on: M-Labs/nac3#127
Co-authored-by: ychenfo <yc@m-labs.hk>
Co-committed-by: ychenfo <yc@m-labs.hk>
2021-12-19 11:04:53 +08:00
Sebastien Bourdeauducq 7420ce185b README: update 2021-12-13 19:02:46 +08:00
Sebastien Bourdeauducq 69b9ac5152 nac3standalone: consistent naming 2021-12-13 11:19:11 +08:00
ychenfo ccfcba4066 nac3standalone: add output_long 2021-12-13 10:44:33 +08:00
ychenfo b5637a04e9 nac3core: use official implementation for len 2021-12-13 10:44:33 +08:00
ychenfo 2c6601d97c nac3core: fix len on range with step of different sign 2021-12-13 10:44:33 +08:00
ychenfo 82359b81a2 nac3core: fix bool to int conversion 2021-12-13 04:13:43 +08:00
ychenfo 4d2fd9582a nac3core: fix broken tests 2021-12-09 01:37:05 +08:00
ychenfo b7892ce952 nac3core: add len support for list and range 2021-12-09 01:37:00 +08:00
ychenfo 01d3249646 nac3core: add missing llvm range type 2021-12-09 01:16:05 +08:00
Sebastien Bourdeauducq d2ffdeeb47 flake: update nixpkgs and work around openssh cross compilation breakage. Closes #123 2021-12-08 21:21:37 +08:00
Sebastien Bourdeauducq ae902aac2f remove devshell inputs from hydraJobs
We are not recompiling packages that depend on LLVM anymore, llvm-nac3 is only used for static linking within NAC3.
2021-12-08 17:43:05 +08:00
Sebastien Bourdeauducq 3f73896477 remove a small amount of LLVM bloat
Also avoids libffi.dll dependency on Windows.
2021-12-08 17:41:34 +08:00
Sebastien Bourdeauducq ddb4c548ae add and use local copy of LLVM Nix files
Modifications accumulate and many are not suitable for nixpkgs upstream.

Based on nixpkgs 3f629e3dd5293bd3c211c4950c418f7cfb4501af
2021-12-08 16:55:25 +08:00
pca006132 6d00d4dabb nac3artiq: cache python data if possible 2021-12-05 20:30:03 +08:00
Sebastien Bourdeauducq baa713a3ca flake: don't attempt to fixup Windows build 2021-12-05 14:40:10 +08:00
Sebastien Bourdeauducq d2919b9620 Revert "flake: better shells"
llvm-config/llvm-sys hates pkgs.buildEnv.

This reverts commit e4f35372d3.
2021-12-05 14:35:58 +08:00
Sebastien Bourdeauducq 9ee2168932 Revert "flake: fix hydraJobs"
This reverts commit e8e1499478.
2021-12-05 14:35:58 +08:00
pca006132 65bc1e5fa4 nac3artiq: handle name_to_pyid in compilation
python variables can change between kernel invocations
2021-12-05 13:10:54 +08:00
pca006132 2938eacd16 nac3artiq: supports running multiple kernels 2021-12-05 13:10:54 +08:00
Sebastien Bourdeauducq e8e1499478 flake: fix hydraJobs 2021-12-05 13:03:44 +08:00
Sebastien Bourdeauducq e4f35372d3 flake: better shells 2021-12-05 12:56:47 +08:00
Sebastien Bourdeauducq 41f88095a5 min_artiq: add round64, floor64, ceil64 2021-12-04 20:35:52 +08:00
pca006132 c98f367f90 nac3artiq: enables inlining 2021-12-04 17:52:03 +08:00
ychenfo 1f3aa48361 nac3parser: modify parser to handle negative integer edge cases 2021-12-03 16:35:58 +08:00
Sebastien Bourdeauducq 8c05d8431d flake: use upstream nixpkgs patch
https://github.com/NixOS/nixpkgs/pull/148367
2021-12-03 11:57:01 +08:00
Sebastien Bourdeauducq 0ae2aae645 flake: publish zipfile with Windows Python module on Hydra 2021-12-02 22:47:35 +08:00
Sebastien Bourdeauducq b0eb7815da flake: consistent naming 2021-12-02 22:37:41 +08:00
Sebastien Bourdeauducq 26e60fca6e flake: cleanup tarball unpacking 2021-12-02 22:37:32 +08:00
Sebastien Bourdeauducq 22a509e7ce flake: add Hydra job for Windows build
This is a proof-of-concept; it works but requires manual fiddling with DLLs
(e.g. copy them from the Nix store into the Windows environment), and LLD
is not available on Windows.
2021-12-02 22:29:44 +08:00
Sebastien Bourdeauducq 4526c28edb Merge branch 'windows' 2021-12-02 22:26:55 +08:00
Sebastien Bourdeauducq 25fc9db66d cargo: specify inkwell LLVM target explicitly
Windows LLVM linking otherwise breaks on the non-existing targets.
2021-12-02 22:24:33 +08:00
Sebastien Bourdeauducq 6315027a8b flake: use *.pyd for Windows Python module 2021-12-02 22:24:23 +08:00
Sebastien Bourdeauducq c0f8d5c602 flake: Windows libs working 2021-12-02 22:01:19 +08:00
Sebastien Bourdeauducq 998f49261d flake: fix Windows libs further 2021-12-02 21:02:48 +08:00
Sebastien Bourdeauducq aab43b1c07 flake: unbreak Windows library link (WIP) 2021-12-02 20:00:50 +08:00
Sebastien Bourdeauducq a6275fbb57 flake: add libffi on Windows 2021-12-02 19:08:20 +08:00
Sebastien Bourdeauducq 8a46032f4c flake: unbreak llvm-config for cross-compilation of static libs 2021-12-02 18:46:04 +08:00
Sebastien Bourdeauducq 1c31aa6e8e consistent naming 2021-12-02 10:45:46 +08:00
sb10q b030aec191 Merge pull request 'Add floor and ceil, move built-in functions to a separate file' (#120) from built_in_floor_ceil into master
Reviewed-on: M-Labs/nac3#120
2021-12-02 10:40:50 +08:00
ychenfo aa2d79fea6 Merge branch 'master' into built_in_floor_ceil 2021-12-02 01:08:55 +08:00
ychenfo 1e6848ab92 nac3core: distinguish i64 and i32 in bool conversion 2021-12-02 01:02:42 +08:00
Sebastien Bourdeauducq a91b2d602c flake: switch to nixos- branch 2021-12-01 22:49:43 +08:00
Sebastien Bourdeauducq c683958e4a nac3artiq: clarify comment about virtual class 2021-12-01 22:49:20 +08:00
Sebastien Bourdeauducq 142f82f987 remove debug prints 2021-12-01 22:48:06 +08:00
ychenfo dfd3548ed2 TypeVar and virtual support in Symbol Resolver (#99)
Add `TypeVar` and `virtual` support for Symbol Resolver in nac3artiq and nac3standalone

Reviewed-on: M-Labs/nac3#99
Co-authored-by: ychenfo <yc@m-labs.hk>
Co-committed-by: ychenfo <yc@m-labs.hk>
2021-12-01 22:44:53 +08:00
Sebastien Bourdeauducq 31fba04cee flake: fix Windows build, now finding LLVM and Python 2021-12-01 18:30:26 +08:00
ychenfo fa2fe8ed5d nac3core: add ceil and floor 2021-12-01 03:23:58 +08:00
ychenfo 7ede4f15b6 nac3core: move builtin definitions to another file 2021-12-01 02:52:00 +08:00
ychenfo 0fe346106d nac3core: fix converting int to bool (#119) 2021-11-30 03:02:26 +08:00
Sebastien Bourdeauducq 681d85d3be remove debug print 2021-11-28 12:57:28 +08:00
sb10q 14119a2c80 Merge pull request 'KernelInvariant' (#114) from KernelInvariant into master
Reviewed-on: M-Labs/nac3#114
2021-11-28 12:49:31 +08:00
pca006132 b35075245b nac3artiq: remove debug print 2021-11-27 21:29:57 +08:00
pca006132 4b17511b4a Merge branch 'master' into KernelInvariant 2021-11-27 21:29:27 +08:00
pca006132 7ee82de312 nac3core: fixed weird type inference error 2021-11-27 20:27:46 +08:00
Sebastien Bourdeauducq 701ca36e99 flake: windows build WIP 2021-11-26 17:26:18 +08:00
Sebastien Bourdeauducq 5e1b0a10a0 flake: patch nixpkgs to fix mingw llvm_12 build 2021-11-26 17:01:44 +08:00
Sebastien Bourdeauducq 9f316a3294 flake: revert nixpkgs to unbreak rust cross-compilation 2021-11-26 17:00:20 +08:00
Sebastien Bourdeauducq 0ae1fe82b4 remove unnecessary cargo config
extra-link-arg has been stabilized
2021-11-23 15:35:33 +08:00
Sebastien Bourdeauducq de8fc264d7 fix unsupported default parameter error message 2021-11-23 15:34:44 +08:00
Sebastien Bourdeauducq 970f075490 flake: switch to nixpkgs 21.11 release 2021-11-23 11:22:08 +08:00
ychenfo 4587088835 Constant Default Parameter Support (#98)
Add support for constant default parameter

Reviewed-on: M-Labs/nac3#98
Co-authored-by: ychenfo <yc@m-labs.hk>
Co-committed-by: ychenfo <yc@m-labs.hk>
2021-11-23 07:32:09 +08:00
ychenfo 49476d06e1 nac3core: clearer comments 2021-11-22 15:06:16 +08:00
ychenfo 664e02cec4 nac3core: fix clippy warning 2021-11-22 14:55:39 +08:00
ychenfo c6f75c8bde nac3standalone: fix error message when no entry point is found 2021-11-22 14:52:52 +08:00
ychenfo 01b51b62ee nac3core: composer better error msg in for uninit field 2021-11-21 06:11:55 +08:00
ychenfo aae9925014 nac3standalone: report when entry point run function cannot be found 2021-11-21 06:11:55 +08:00
ychenfo d336200bf4 nac3core: fix broken tests due to the fix of rigid typevar handling 2021-11-21 06:11:55 +08:00
ychenfo a50df6560e nac3core: fix handling on rigid typevar 2021-11-21 06:11:55 +08:00
ychenfo a9635f0979 nac3core: top level use codegen official get_subst_key 2021-11-21 06:11:55 +08:00
ychenfo c2706fa720 nac3core: fix polymorphic class method partial instantiation 2021-11-21 06:11:55 +08:00
pca006132 f5ec103c82 nac3artiq: kernel invariant support 2021-11-20 21:15:15 +08:00
pca006132 ba08deada6 nac3core: refactor codegen 2021-11-20 19:50:25 +08:00
Sebastien Bourdeauducq 439cef636f runkernel: improve print_int debug functions 2021-11-19 12:39:57 +08:00
ychenfo 1e47b364c5 nac3artiq: support now-pinning on RISC-V with wide data bus (#97) 2021-11-16 17:37:40 +08:00
ychenfo 8ab3ee9cce nac3core: AugAssign support (#82) 2021-11-13 12:24:22 +08:00
Sebastien Bourdeauducq 9ae08d6e3d nac3artiq: add stubs for now-pinning on rv32g (#97) 2021-11-13 12:10:55 +08:00
Sebastien Bourdeauducq d6b92adf70 nac3artiq: add stack guard 2021-11-12 20:03:52 +08:00
Sebastien Bourdeauducq aa84fefa56 fix previous commit (again) 2021-11-12 15:47:21 +08:00
Sebastien Bourdeauducq 5ad7aa5a93 flake: fix previous commit 2021-11-12 15:37:39 +08:00
Sebastien Bourdeauducq b64d2399f2 flake: build devShell dependencies on Hydra 2021-11-12 15:08:24 +08:00
Sebastien Bourdeauducq ebca596be6 flake: update nixpkgs 2021-11-12 15:04:21 +08:00
Sebastien Bourdeauducq 4aeea87702 flake: export nixpkgs-patched 2021-11-12 14:57:46 +08:00
Sebastien Bourdeauducq e25a9bbcda typo 2021-11-11 23:44:14 +08:00
Sebastien Bourdeauducq 978eaf16a4 nac3artiq: support RISC-V with and without FPU. Closes #83 2021-11-11 23:43:50 +08:00
Sebastien Bourdeauducq 4547eee82a llvm: switch RISC-V ABI when FPU is present
Patch is a bit of a hack and ignores 64-bit CPUs.

Also only build the LLVM targets we need.
2021-11-11 23:42:32 +08:00
Sebastien Bourdeauducq 96607432c1 nac3core: use Python 3.9 list/tuple annotations in test
Closes #85
2021-11-11 20:05:08 +08:00
Sebastien Bourdeauducq dba1a8b3d4 nac3standalone: link libm in demo runner 2021-11-11 19:44:18 +08:00
Sebastien Bourdeauducq 612b6768c0 nac3artiq: bail early on non-NAC3 classes 2021-11-11 16:35:40 +08:00
Sebastien Bourdeauducq c004da85f7 nac3artiq: drop host-only base classes. Closes #80 2021-11-11 16:08:29 +08:00
Sebastien Bourdeauducq 7fc04936cb runkernel: add print_int debug function 2021-11-10 17:34:13 +08:00
Sebastien Bourdeauducq b57b869c49 min_artiq: remove unused imports 2021-11-10 14:01:39 +08:00
Sebastien Bourdeauducq 50f1aca1aa nac3artiq: move module registration list to CPython side
In ARTIQ, we cannot create a global NAC3 object because we do not
know the ISA in advance.
2021-11-07 10:29:14 +08:00
pca006132 ffa89e9308 fix clippy warnings 2021-11-06 23:00:18 +08:00
pca006132 34cf303e6c nac3artiq: modified demo to use KernelInvariants 2021-11-06 22:50:28 +08:00
pca006132 b1e83a1fd4 nac3core: type check invariants
This rejects code that tries to assign to KernelInvariant fields and
methods.
2021-11-06 22:48:08 +08:00
Sebastien Bourdeauducq 7385b91113 nac3artiq: support kernel entry short form from original ARTIQ 2021-11-06 18:41:59 +08:00
Sebastien Bourdeauducq 016cbf2b90 nac3artiq: return bytes in compile_method_to_mem 2021-11-06 14:29:23 +08:00
Sebastien Bourdeauducq 37eae090e5 nac3artiq: fix linker inputs 2021-11-06 14:23:54 +08:00
Sebastien Bourdeauducq 204baabfd2 nac3artiq: add compile_method_to_mem 2021-11-06 14:14:53 +08:00
Sebastien Bourdeauducq 597857ccd0 typo 2021-11-06 14:14:40 +08:00
ychenfo efc9edbc14 nac3parser: fix decorator and above comments 2021-11-06 14:13:17 +08:00
Sebastien Bourdeauducq 7d66195eae nac3artiq: embed linker script, put intermediate objects in temp dir 2021-11-06 13:04:00 +08:00
pca006132 1fea51d9b3 Merge pull request 'nac3parser: add comment support' (#68) from with_nac3comment into master
Reviewed-on: M-Labs/nac3#68
2021-11-05 20:46:42 +08:00
pca006132 99b29d8ded Merge branch 'master' into with_nac3comment 2021-11-05 20:46:29 +08:00
pca006132 3db95b120b nac3core: implements bool conversion function 2021-11-05 20:34:34 +08:00
pca006132 8dbb4ad58a nac3core/toplevel: make test less fragile
test results should not depend on internal states if possible
2021-11-05 20:28:21 +08:00
ychenfo ee67b22ebc Merge branch 'master' into with_nac3comment 2021-11-05 20:01:36 +08:00
Sebastien Bourdeauducq afb94dd299 nac3artiq: move demo to dedicated folder 2021-11-05 18:28:31 +08:00
Sebastien Bourdeauducq d6f0607ff0 nac3artiq: rename class decorator to nac3 2021-11-05 18:08:36 +08:00
Sebastien Bourdeauducq 610448fa73 nac3artiq: include parallel in demo 2021-11-05 18:07:18 +08:00
Sebastien Bourdeauducq e8228710e7 min_artiq: remove unnecessary definitions 2021-11-05 17:50:26 +08:00
ychenfo 032e1d84cf nac3parser: add and fix tests due to comment support 2021-11-04 15:03:34 +08:00
ychenfo b239806558 nac3core: adapt to ast change due to comment support 2021-11-04 15:02:51 +08:00
ychenfo 694c7e945c nac3ast: generated ast with comment fields 2021-11-04 15:01:50 +08:00
ychenfo 3b1cc02d06 nac3parser, ast: add comment support core changes 2021-11-04 15:00:27 +08:00
Sebastien Bourdeauducq 32d1fe811b flake: update nixpkgs 2021-11-03 21:54:27 +08:00
Sebastien Bourdeauducq 36e4028f5b fix and run parser tests 2021-11-03 17:39:48 +08:00
Sebastien Bourdeauducq b6ff46c39e README: update 2021-11-03 17:22:14 +08:00
Sebastien Bourdeauducq bf7e2c295a integrate nac3parser 2021-11-03 17:11:00 +08:00
pca006132 48ce6bb6c5 rustpython-parser: string interner, optimizations, thread local cache
corresponds to M-Labs RustPython fork at efdf7829ba1a5f87d30df8eaff12a330544f3cbd
branch parser-mod
2021-11-03 16:36:28 +08:00
Sebastien Bourdeauducq 80c7bc1cbd add RustPython parser
based on RustPython 67b338863ee6f16b0df0a7d1aa9debba55284651
2021-11-03 16:34:10 +08:00
Sebastien Bourdeauducq e89bc93b5f ignore expressions in class definition body (#26) 2021-11-02 23:30:12 +08:00
Sebastien Bourdeauducq 47f563908a basic string support (#30) 2021-11-02 23:22:49 +08:00
Sebastien Bourdeauducq 0e914ab7e9 composer: add range keyword 2021-11-02 18:56:14 +08:00
Sebastien Bourdeauducq 613020a717 test: add missing id_to_name entry 2021-11-02 18:34:48 +08:00
Sebastien Bourdeauducq ee2c0d8bab flake: add cargo-insta to dev shell 2021-11-02 18:13:59 +08:00
Sebastien Bourdeauducq 0d1e9262af flake: update cargoSha256 2021-11-02 15:17:10 +08:00
Sebastien Bourdeauducq bc0f82cad8 Revert "nac3artiq/codegen: fixed smax problem"
We have LLVM 12 now and can use the intrinsic.

This reverts commit 98d9f73afb.
2021-11-02 14:00:28 +08:00
Sebastien Bourdeauducq 624dfe8cd1 upgrade to LLVM 12 2021-11-02 14:00:20 +08:00
pca006132 e47597bb8a Merge branch 'context-manager' 2021-11-02 11:17:00 +08:00
pca006132 98d9f73afb nac3artiq/codegen: fixed smax problem
It turns out the smax intrinsic I use is a new one that is not supported
in LLVM11. Now implemented with signed integer compare and select.
2021-11-02 11:10:21 +08:00
Sebastien Bourdeauducq da2886565b runkernel: sort out cargo rigmarole with rustc link arg 2021-11-01 00:28:27 +08:00
Sebastien Bourdeauducq b37cf6de08 nac3artiq: share isa->time_fns map 2021-11-01 00:03:15 +08:00
pca006132 083eacc268 with parallel/sequential support
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.
Each function call directly inside a parallel block will reset the timeline after
execution. A parallel block within a sequential block (or not within any block) will
set the timeline to the max now_mu within the block (and the outer max now_mu will also
be updated).

Implementation: We track the start and end separately.
- If there is a start variable, it indicates that we are directly inside a
parallel block and we have to reset the timeline after every function call.
- If there is a end variable, it indicates that we are (indirectly) inside a
parallel block, and we should update the max end value.

Note: requires testing, it is difficult to inspect the output IR
2021-10-31 23:54:37 +08:00
Sebastien Bourdeauducq 137efebb33 runkernel: add simple host kernel runner 2021-10-31 23:52:43 +08:00
Sebastien Bourdeauducq 443b95d909 nac3artiq: do not use custom linker script when targeting host 2021-10-31 23:51:50 +08:00
Sebastien Bourdeauducq 8b73a123cc nac3artiq: support compiling for the host 2021-10-31 23:02:21 +08:00
pca006132 84c5201243 with parallel/sequential support
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.
Each function call directly inside a parallel block will reset the timeline after
execution. A parallel block within a sequential block (or not within any block) will
set the timeline to the max now_mu within the block (and the outer max now_mu will also
be updated).

Implementation: We track the start and end separately.
- If there is a start variable, it indicates that we are directly inside a
parallel block and we have to reset the timeline after every function call.
- If there is a end variable, it indicates that we are (indirectly) inside a
parallel block, and we should update the max end value.

Note: requires testing, it is difficult to inspect the output IR
2021-10-31 17:16:21 +08:00
pca006132 558c3f03ef nac3core/codegen: list comprehension support 2021-10-24 16:53:43 +08:00
pca006132 45673b0ecc nac3core/codegen: cleanup 2021-10-24 16:53:43 +08:00
pca006132 181607008d nac3core/codegen: supports list iter 2021-10-24 14:39:50 +08:00
pca006132 fb92b6d364 nac3core: supports range iterator 2021-10-23 23:53:36 +08:00
pca006132 2f6ba69770 nac3core/typecheck: check if value is none 2021-10-23 21:31:14 +08:00
pca006132 cc83bbc63a nac3core/codegen: fix broken test 2021-10-17 13:07:45 +08:00
pca006132 279f47f633 nac3core/codegen: avoid sending unifiers
Previously, we have to copy types from one unification table to another,
and make the table sendable. This requires cloning (processing) the
whole table 3 times per function call which is not efficient and uses
more memory than required when the unification table is large.

We now use a concrete type table to only copy the type we need. This
reduces the overhead as we only need to process the unification table
for once (when we do the function codegen), and reduces memory usage by
a bit (but not noticeable when the unification table is small, i.e. the
types are simple).
2021-10-17 13:02:18 +08:00
pca006132 9850cbe313 nac3core/codegen: optimize for every function
This speeds up compilation and reduces memory usage.
2021-10-17 12:56:11 +08:00
pca006132 1f5bea2448 nac3core/codegen: refactor according to #23 2021-10-16 22:17:36 +08:00
pca006132 c4259d14d1 fixed some clippy warnings 2021-10-16 18:08:13 +08:00
pca006132 26076c37ba nac3core/typecheck: supports recursive type inference 2021-10-16 15:56:49 +08:00
Sebastien Bourdeauducq fd0b11087e nac3core: use round instead of rint. Closes #61 2021-10-11 08:18:52 +08:00
Sebastien Bourdeauducq 3a1dd893a1 nac3artiq/demo: get closer to regular ARTIQ 2021-10-10 17:45:38 +08:00
pca006132 a4ccac2329 nac3artiq: implements #55, #56 2021-10-10 16:26:01 +08:00
pca006132 77542170fd nac3core: fixes #60 2021-10-10 15:01:06 +08:00
pca006132 a3ce5be10b nac3core: fixes #32 and #57 2021-10-09 16:20:49 +08:00
Sebastien Bourdeauducq a22552a012 nac3artiq: work around #56 2021-10-09 15:52:45 +08:00
Sebastien Bourdeauducq 6ba74ed9f6 nac3artiq: allow creating drivers on device 2021-10-09 15:51:47 +08:00
Sebastien Bourdeauducq 8b32c8270d nac3artiq: explain delayed registration 2021-10-09 15:21:41 +08:00
Sebastien Bourdeauducq 5749141efb nac3artiq: add simple KernelInvariant CPython wrapper 2021-10-08 23:46:46 +08:00
Sebastien Bourdeauducq 3b10172810 nac3artiq: get closer to original ARTIQ semantics in demo
Currently crashes the compiler with:
thread '<unnamed>' panicked at 'called `Option::unwrap()` on a `None` value', nac3core/src/codegen/expr.rs:395:58
2021-10-08 23:41:41 +08:00
Sebastien Bourdeauducq 82efb0e720 missing file from previous commit 2021-10-08 23:21:36 +08:00
Sebastien Bourdeauducq d3a21d75fa handle time cursor functions on the Rust side entirely
This is preparation for with sequential/with parallel support.
2021-10-08 23:14:22 +08:00
pca006132 a07674a042 nac3artiq: host object supports typevar 2021-10-08 22:45:08 +08:00
Sebastien Bourdeauducq c5bcd352a5 flake: add hydraJobs 2021-10-08 08:18:42 +08:00
Sebastien Bourdeauducq 79d3c5caae switch to Nix flakes 2021-10-08 00:04:22 +08:00
pca006132 c697e522d3 nac3artiq: #33 demo
The python API is changed a bit to allow running constructor with
@kernel annotation.
2021-10-07 15:58:19 +08:00
pca006132 08947d20c2 nac3artiq: implements #33 2021-10-07 15:57:45 +08:00
pca006132 62673cf608 nac3artiq: add back builtins (fix #53) 2021-10-06 16:46:41 +08:00
pca006132 11144301ca nac3artiq: added simple host value support 2021-10-06 16:07:42 +08:00
ychenfo 4fcb54e463 nac3core: fix #46, better toplevel return type error msg 2021-10-03 18:07:45 +08:00
ychenfo 24b2111c64 nac3core: fix #45 toplevel better error msg for methods/functions 2021-10-03 17:25:28 +08:00
ychenfo f5ce1afe0b fix tests and switch to insta
Use a library called 'insta' to better organize those longer correct test outputs in toplevel tests. 'insta' creates `.snap` files as snapshots of the test output, and will automatically do the diff if the output is different. This makes maintaining test cases with larger outputs a lot easier.

Reviewed-on: M-Labs/nac3#42
Co-authored-by: ychenfo <yc@m-labs.hk>
Co-committed-by: ychenfo <yc@m-labs.hk>
2021-10-03 16:39:12 +08:00
Sebastien Bourdeauducq 915460ecb7 nac3artiq: fix typo 2021-10-03 16:18:15 +08:00
Sebastien Bourdeauducq b2c7f51d57 nac3artiq: guarantee ordering of pinned now stores 2021-10-03 10:00:42 +08:00
Sebastien Bourdeauducq 248d8cbece nac3artiq: handle now-pinning depending on target 2021-10-02 23:40:06 +08:00
Sebastien Bourdeauducq c429a86586 nac3artiq: refactor timeline functions 2021-10-02 23:35:28 +08:00
Sebastien Bourdeauducq c5e731f16d nac3artiq: implement timeline functions 2021-10-02 23:22:46 +08:00
Sebastien Bourdeauducq 0cbe4778d2 nac3standalone: demonstrate scalar conversion functions in mandelbrot 2021-10-02 19:21:59 +08:00
Sebastien Bourdeauducq c93305739d nac3artiq: fix misleading error message 2021-10-02 19:17:33 +08:00
Sebastien Bourdeauducq ba93931758 implement timeline functions for RISC-V (WIP) 2021-10-02 19:05:35 +08:00
Sebastien Bourdeauducq 3dd916b6ac nac3artiq: update unsendable comment 2021-10-02 18:28:44 +08:00
pca006132 8447aa3000 nac3artiq: allows kernel function 2021-10-01 00:02:15 +08:00
pca006132 1d2a32b140 nac3core/toplevel: impl scalar conversion
Implemented scalar conversion functions as builtin functions.
`round` for int64 is now implemented as `round64`.
2021-09-30 23:39:29 +08:00
pca006132 07a9229d52 nac3artiq: implements #36 2021-09-30 22:30:54 +08:00
pca006132 f0fdfe42cb nac3core: better impl of #24 2021-09-30 17:07:48 +08:00
Sebastien Bourdeauducq 928b5bafb5 nac3artiq: add missing symlink 2021-09-29 15:36:30 +08:00
Sebastien Bourdeauducq dceaf42500 nac3artiq: support ISA selection 2021-09-29 15:33:12 +08:00
Sebastien Bourdeauducq bfd041d361 nac3artiq: filter class definitions 2021-09-27 22:25:19 +08:00
Sebastien Bourdeauducq 6141f01180 nac3artiq: parse whole Python module, filter ast 2021-09-27 22:12:25 +08:00
Sebastien Bourdeauducq 8d839db553 typo 2021-09-27 19:12:18 +08:00
Sebastien Bourdeauducq 316db42940 nac3embedded -> nac3artiq, README cleanup 2021-09-27 10:30:54 +08:00
Sebastien Bourdeauducq 64404bba20 syscall -> extern (#21) 2021-09-27 10:13:03 +08:00
pca006132 d4ed76d76e nac3core: implementing #24 2021-09-26 22:17:09 +08:00
pca006132 3c121dfcda nac3core/toplevel/composer: fixes #29 2021-09-25 22:02:19 +08:00
pca006132 693ac7d336 nac3core/toplevel: added personality symbol config 2021-09-25 21:44:00 +08:00
Sebastien Bourdeauducq dd998c8afc nac3embedded: RTIO LED blinking demo 2021-09-25 14:17:11 +08:00
Sebastien Bourdeauducq 7ab762a174 demo: add more syscalls and features 2021-09-24 14:45:44 +08:00
Sebastien Bourdeauducq 7ab2114882 nac3embedded: switch to Zynq (#24) 2021-09-24 14:45:09 +08:00
Sebastien Bourdeauducq 4535b60fc0 nac3embedded: add device_db for artiq_run of the compiler output 2021-09-24 13:26:23 +08:00
Sebastien Bourdeauducq bf48151748 nac3embedded: do not use *.so for output to avoid confusing cpython 2021-09-24 13:25:48 +08:00
Sebastien Bourdeauducq bed8ce1f26 nac3embedded: set ELF entry point symbol 2021-09-24 13:25:18 +08:00
Sebastien Bourdeauducq c26689c7a7 shell.nix: fix syntax 2021-09-24 12:53:48 +08:00
Sebastien Bourdeauducq ac17bf50f8 nac3embedded: RTIO syscalls 2021-09-24 11:39:26 +08:00
Sebastien Bourdeauducq 13bd18bfcb nac3embedded: produce final shared library 2021-09-24 11:07:52 +08:00
Sebastien Bourdeauducq 5c236271c3 nac3embedded: string interning 2021-09-24 09:58:58 +08:00
Sebastien Bourdeauducq 14662a66dc nac3embedded: run linker (WIP) 2021-09-23 21:30:47 +08:00
pca006132 c4fbfeaca9 nac3standalone: added thread number arg 2021-09-23 20:02:56 +08:00
pca006132 20a752fd3a Merge pull request 'optimization (#13)' (#15) from optimization into master
Reviewed-on: M-Labs/nac3#15
2021-09-23 19:58:43 +08:00
pca006132 6a69211c55 Merge remote-tracking branch 'origin/master' into optimization 2021-09-23 19:58:24 +08:00
Sebastien Bourdeauducq 59dac8bdf5 nac3embedded: compile for RISC-V ARTIQ coredevice 2021-09-23 19:38:48 +08:00
Sebastien Bourdeauducq edd60e3f9a nac3embedded: compile again 2021-09-23 19:30:03 +08:00
pca006132 799ed58d21 nac3core/type_inferencer: avoid type var for assign 2021-09-22 19:25:47 +08:00
pca006132 105d605e6d nac3core: fix clippy warnings 2021-09-22 18:04:06 +08:00
pca006132 97f5b7c324 fixed performance regression 2021-09-22 18:04:06 +08:00
pca006132 7d48883583 fixed tests 2021-09-22 18:04:06 +08:00
pca006132 084efe92af nac3core: use string interning 2021-09-22 18:04:06 +08:00
pca006132 891056631f nac3core: use Arc to reduce copy 2021-09-22 18:04:06 +08:00
pca006132 1b5ac3cd25 nac3core: do not alloc call if it is monomorphic 2021-09-22 18:04:06 +08:00
pca006132 5ed2b450d3 nac3core/typecheck: no type variable for monomorphic functions 2021-09-22 18:04:06 +08:00
pca006132 a508baae20 added syscall annotation
and temporarily disabled the keyword check for top-level functions
2021-09-22 17:58:46 +08:00
Sebastien Bourdeauducq 013e7cfc2a codegen: support pass statement. Closes #10 2021-09-22 15:17:13 +08:00
Sebastien Bourdeauducq db14b4b635 demo: remove old obj files 2021-09-22 15:02:05 +08:00
Sebastien Bourdeauducq 8acb39f31f fix parallel compilation 2021-09-22 15:00:32 +08:00
Sebastien Bourdeauducq d561450bf5 demo: fix classes example 2021-09-22 14:57:24 +08:00
Sebastien Bourdeauducq 956cca6ac8 cleanup demos 2021-09-22 14:57:13 +08:00
pca006132 4a5f2d495e added time measurement to track performance 2021-09-22 14:45:56 +08:00
pca006132 4fe643f45b allows function ending with a number 2021-09-22 14:45:42 +08:00
ychenfo 1c170f5c41 nac3core: type inferencer fix the pass statement 2021-09-21 13:19:13 +08:00
pca006132 df6c9c8a35 fix #11 2021-09-21 11:29:51 +08:00
ychenfo 20905a9b67 nac3core: better field initialization check 2021-09-21 03:02:12 +08:00
ychenfo e66693282c nac3core: change the place to unify constructor type for function body type check
add really basic field initialize check
2021-09-20 23:44:39 +08:00
ychenfo dd1be541b8 nac3core: allow class to have no __init__, function/method name with module path added to ensure uniqueness 2021-09-20 23:36:19 +08:00
pca006132 3c930ae9ab fixed #12 2021-09-20 15:51:42 +08:00
ychenfo 35a94a8fc0 nac3core: fix broken test 2021-09-20 01:58:07 +08:00
pca006132 4939ff4dbd simple implementation of classes 2021-09-19 22:54:06 +08:00
ychenfo bf1769cef6 nac3standalone: more tests 2021-09-19 17:50:01 +08:00
ychenfo 2b74895b71 nac3standalone, nac3core: can use top level composer to compile and run mandelbrot 2021-09-19 16:19:49 +08:00
ychenfo 1b0f3d07cc nac3core: top level fix field of funinstance 2021-09-17 22:32:13 +08:00
ychenfo ed5dfd4100 nac3core: top level inferencer call with type var more test 2021-09-17 16:31:33 +08:00
ychenfo 41e63f24d0 nac3core: top level add test utility to print stringfied type 2021-09-17 16:31:21 +08:00
ychenfo d0df705c5a nac3core: toplevel type var test 2021-09-17 00:39:42 +08:00
ychenfo a0662c58e6 nac3core: fix recursive top level function call 2021-09-17 00:39:42 +08:00
ychenfo 526c18bda0 nac3core: top level inferencer without type var should be ok 2021-09-17 00:39:42 +08:00
ychenfo a10ab81ee7 toplevel composer: add ast to class methods, suppress warning 2021-09-17 00:39:42 +08:00
pca006132 f5353419ac codegen: minimized lock holding time
The previous way of holding the lock would prohibit multithread code
generation for llvm
2021-09-16 21:36:42 +08:00
pca006132 180392e2ab typecheck: fixed recursive substitution 2021-09-12 21:33:21 +08:00
ychenfo 471547855e nac3core: toplevel change class method name handling, cleanup comments 2021-09-12 13:14:46 +08:00
ychenfo 2ac3f9a176 nac3core: separate top level compoer to a new file 2021-09-12 05:10:10 +08:00
ychenfo cb310965b8 nac3core: toplevel register consider module path 2021-09-12 05:00:26 +08:00
ychenfo 118f19762a nac3core: toplevel format 2021-09-12 04:40:40 +08:00
ychenfo b419634f8a nac3core: top level fields inheritance check, more tests 2021-09-12 04:34:30 +08:00
ychenfo 147298ff40 nac3core: top level fix class fields as nac3 spec 2021-09-12 03:49:21 +08:00
ychenfo c7cb02b0f3 nac3core: toplevel fix parse type annotation dead lock 2021-09-12 03:01:56 +08:00
ychenfo 03b5e51822 nac3standalone: cleanup 2021-09-10 21:27:08 +08:00
ychenfo 4eacd1aa9e nac3core: top level err test 2021-09-10 21:26:39 +08:00
ychenfo 9eef51f29f nac3core: top level class method self parameter fixed 2021-09-10 16:14:08 +08:00
ychenfo 917d447605 nac3core: clean up, fix broken test 2021-09-09 02:09:35 +08:00
ychenfo f1013d9a17 nac3core: top level fix type var within list tuple, test of type var application compatibility 2021-09-09 02:03:44 +08:00
ychenfo 2ce507964c nac3core: fix broken top level test due to hashmap order 2021-09-09 00:44:56 +08:00
ychenfo 5a1a8ecee3 nac3core: self is not not allowed to explicitly appear in method type annotations 2021-09-08 21:53:54 +08:00
ychenfo 1300b5ebdd nac3core: clean up and format 2021-09-08 19:45:36 +08:00
ychenfo 87f25e1c5d nac3core: remove mutex on dyn symbol resolve 2021-09-08 19:27:32 +08:00
ychenfo 55335fc05d nac3core: top level simple type var handled 2021-09-08 02:27:12 +08:00
ychenfo 247b364191 nac3core: top level fix cyclic ancestor analysis, add tests 2021-09-07 17:30:15 +08:00
ychenfo bbcec6ae6f nac3core: toplevel fix bug in make self annotation and return type check 2021-09-07 10:03:31 +08:00
ychenfo 235b6e34d1 nac3core: top level derive fmt::Debug, fix dead lock 2021-09-07 00:20:40 +08:00
ychenfo 54b4572c5f nac3core: allow interior mutability to dyn symbolresolver, add add_id_def to symbolresolver trait, remove primitive from top level def list 2021-09-06 19:23:04 +08:00
ychenfo dc7c014b10 nac3core: top level more test 2021-08-31 17:40:38 +08:00
ychenfo 1ae6acc061 nac3core: top level fix function/methods none return type 2021-08-31 15:41:48 +08:00
ychenfo 98d032b72a nac3core: top level fix duplicate def, start adding tests 2021-08-31 15:23:57 +08:00
ychenfo 7bbd608492 nac3core: top level cleanup, rewrite ancestors handling, __init__ occruence check 2021-08-31 13:54:16 +08:00
ychenfo 4a9593efa3 nac3core: top level clean up and fix ancestors analysis 2021-08-30 22:47:55 +08:00
ychenfo 098bd1e6e6 nac3core: top level check inheritance method overload 2021-08-30 17:39:29 +08:00
ychenfo 82c2edcf8d nac3core: toplevel cleanup and add list and tuple support 2021-08-30 14:16:11 +08:00
ychenfo 40e58d02ed nac3core: toplevel all ancestors are put into the def list, disallow generic base class for now 2021-08-30 14:15:57 +08:00
ychenfo e2a9bdd8bc nac3core: toplevel no duplicate type var too early 2021-08-30 14:15:36 +08:00
ychenfo 236989defc nac3core: remove unnecessary inline and function parameters 2021-08-27 16:52:31 +08:00
pca006132 22a728833d nac3core: fixed broken test 2021-08-27 16:50:53 +08:00
pca006132 2223c86d9b nac3standalone: compile multiple functions 2021-08-27 16:25:59 +08:00
pca006132 72aebed559 nac3core: unification table optimization
avoid cloning values that we no longer need.
2021-08-27 13:05:06 +08:00
pca006132 8c1c7fcfc3 nac3core: fixed broken tests 2021-08-27 13:04:51 +08:00
pca006132 6633eabb89 nac3core: optimized by using HashSet 2021-08-27 12:36:51 +08:00
pca006132 d81249cabe nac3standalone: enabled optimization and recorded time 2021-08-27 12:35:21 +08:00
pca006132 bf4e0009c0 codegen: do not generate cont_bb if unreachable 2021-08-27 11:46:12 +08:00
pca006132 52dd792b3e nac3standalone: added return check 2021-08-27 11:39:36 +08:00
pca006132 a24e204824 type_inferencer: check defined identifiers during inference 2021-08-27 11:13:43 +08:00
ychenfo 35ef0386db move helper function to another file 2021-08-27 10:21:51 +08:00
ychenfo b9a580d271 handle class method/fields type var 2021-08-27 09:53:09 +08:00
ychenfo 018d6643e1 top level: top level function type var handled
top level: class methods/fields type var handling
2021-08-27 01:41:34 +08:00
ychenfo 935e7410fd check type params in class generic base declaration 2021-08-26 11:54:37 +08:00
ychenfo 1a21fb1072 fix codegen test about top level composer return type change 2021-08-25 18:00:01 +08:00
ychenfo 35a331552b cyclic inheritance check added 2021-08-25 17:47:26 +08:00
ychenfo 0bab477ab0 get rid of nested tuple in type annotation helper function 2021-08-25 17:47:26 +08:00
ychenfo 862d205f67 remove self kind and extra primitive info in the return of top level composer constructor, adding some helper function for type annotation 2021-08-25 17:47:26 +08:00
pca006132 e2b11c3fee codegen: fixed deadlock and added call test 2021-08-25 17:44:01 +08:00
pca006132 0608fd9659 fixed test and nac3standalone 2021-08-25 15:30:36 +08:00
pca006132 173102fc56 codegen/expr: function codegen and refactoring 2021-08-25 15:29:58 +08:00
pca006132 93270d7227 use forked version of rustpython
to implement Clone for AST nodes
2021-08-25 15:28:32 +08:00
pca006132 1ffa1a8bb0 typecheck/typedef: added copy_from
This function would copy concrete type from one unifier to another
unifier, and can handle recursive types. This would be used in code
generation for moving types from one unification table to another one.
2021-08-25 12:02:10 +08:00
ychenfo 01f7a31aae put parse ast into type annotation into one function 2021-08-24 17:44:37 +08:00
ychenfo 32773c14e0 move top level related things to a separate module 2021-08-24 17:19:17 +08:00
pca006132 c356062239 symbol_resolver: handle list and tuples 2021-08-24 14:58:19 +08:00
ychenfo 56f082ca7c handle type var associated with class/function partially, change llvm version of nac3embedded to 11
format
2021-08-23 17:00:32 +08:00
ychenfo 39f300b62a clean up and add duplicate function/parameter/class name test
formatted
2021-08-23 14:22:46 +08:00
ychenfo 7b1fe36e90 formatted 2021-08-23 11:13:45 +08:00
ychenfo fb5b4697a9 fix rebase conflict and some test failure with unifier's error message 2021-08-23 10:34:11 +08:00
ychenfo 364054331c handle class fields and methods 2021-08-23 02:54:45 +08:00
ychenfo 40b062ce0f change the parse type annotation parameter type, refactoring top level 2021-08-23 02:54:45 +08:00
pca006132 f5b8b58826 added tuple assignment check 2021-08-21 15:11:01 +08:00
pca006132 c4d6b3691a codegen/expr: fixed warnings 2021-08-21 15:10:50 +08:00
pca006132 957ceb74e4 nac3core/typecheck: added basic location information 2021-08-21 14:51:46 +08:00
pca006132 e47d063efc codegen: store to list element 2021-08-19 17:14:38 +08:00
pca006132 0e2da0d180 codegen: gep related fixes
we can now compile simple programs that uses tuples and lists
2021-08-19 16:54:15 +08:00
pca006132 39545c0005 nac3standalone usable 2021-08-19 15:30:52 +08:00
pca006132 3279f7a776 codegen for simple function call, and various fixes 2021-08-19 15:30:15 +08:00
sb10q f205a8282a Merge pull request 'hm-inference' (#6) from hm-inference into master
Reviewed-on: M-Labs/nac3#6
2021-08-19 11:46:50 +08:00
pca006132 d1215bf5ac nac3core/codegen/expr: fixed typo 2021-08-19 11:45:33 +08:00
pca006132 6e424a6a3e fixed codegen test 2021-08-19 11:32:22 +08:00
pca006132 9a07ef3301 Merge remote-tracking branch 'origin/hm-inference_anto' into hm-inference 2021-08-19 11:32:04 +08:00
ychenfo c238c264e7 add type vars to the primitive binop function def 2021-08-19 11:18:58 +08:00
Sebastien Bourdeauducq f8a697e3d4 switch to LLVM 11 2021-08-19 11:14:35 +08:00
ychenfo 4b38fe66a2 format 2021-08-18 17:33:48 +08:00
ychenfo 9cb07e6f04 start to handle base inheritance methods, fields 2021-08-18 17:32:55 +08:00
ychenfo 6279dbb589 formating 2021-08-18 16:33:50 +08:00
ychenfo 529442590f some parsing of top level class fields and methods 2021-08-18 16:28:17 +08:00
ychenfo 4fcd48e4c8 try to use def list ast tuple and remove method_to_def_id map 2021-08-18 10:01:11 +08:00
ychenfo 619963dc8c removed locks in toplevelcomposer 2021-08-17 16:36:06 +08:00
ychenfo 276daa03f7 start refactorinng for less redundancy 2021-08-17 14:01:18 +08:00
ychenfo a94145348a fix on comments and redundant code, start handling 'self' things 2021-08-17 11:07:16 +08:00
ychenfo fa40fd73c6 formatted 2021-08-16 20:17:08 +08:00
ychenfo 79ce13722a partially parsed class methods nad fields 2021-08-16 17:40:12 +08:00
ychenfo eb814dd8c3 clean unused use 2021-08-16 13:57:21 +08:00
ychenfo 3734663188 add RefCell to FunSignature in TypeEnum 2021-08-16 13:53:45 +08:00
ychenfo d8c3c063ec split top level handling in several functions 2021-08-16 13:53:39 +08:00
pca006132 d3ad894521 removed code comment 2021-08-13 16:30:33 +08:00
pca006132 784111fdbe Merge remote-tracking branch 'origin/hm-inference_anto' into hm-inference 2021-08-13 16:28:04 +08:00
pca006132 d30918bea0 worker thread panic handling 2021-08-13 16:20:14 +08:00
pca006132 e2adf82229 threadpool for parallel code generation 2021-08-13 14:48:46 +08:00
ychenfo 33391c55c2 add Sync bound to Symbol resolver in top level 2021-08-13 14:22:49 +08:00
ychenfo 3f65e1b133 start refactor top_level 2021-08-13 13:57:24 +08:00
ychenfo ba5bb78f11 top level parse class base/generic 2021-08-13 13:57:24 +08:00
ychenfo e176aa660d commit for pull new symbol resolver 2021-08-13 13:57:24 +08:00
pca006132 cb01c79603 removed Arc from TypeEnum 2021-08-13 13:33:59 +08:00
pca006132 1db8378f60 formatting 2021-08-12 16:36:23 +08:00
pca006132 8c7ccb626b fixed symbol_resolver blanket implementation 2021-08-12 14:44:50 +08:00
pca006132 1f6c16e08b fixed compilation failure 2021-08-12 13:56:51 +08:00
pca006132 77943a8117 added primitive codegen test 2021-08-12 13:56:06 +08:00
ychenfo 3a93e2b048 TypeEnum::TObj.param is now RefCell for interior mutability 2021-08-12 13:17:51 +08:00
ychenfo 824a5cb01a register top level clean up 2021-08-12 10:51:41 +08:00
ychenfo 17ee8fe6d0 starting cleaning up and further add Arc<Mutex> 2021-08-12 10:51:41 +08:00
pca006132 d46a4b2d38 symbol_resolver: fixed type variable handling 2021-08-12 10:25:32 +08:00
pca006132 de8b67b605 refactored symbol resolver 2021-08-11 17:28:29 +08:00
pca006132 0af4e95914 Merge remote-tracking branch 'origin/hm-inference_anto' into hm-inference 2021-08-11 15:42:32 +08:00
ychenfo 99276c8f31 formatted 2021-08-11 15:18:21 +08:00
ychenfo 42a636b4ce add Arc<Mutex<dyn SymbolResolver>> and change from Box<SymbolResolve> to Arc<SymbolResolver>, need format and cleanup 2021-08-11 15:11:51 +08:00
pca006132 e112354d25 codegen refactored 2021-08-11 14:37:26 +08:00
ychenfo 43236db9bd update some previous work on top level with the clean up 2021-08-11 13:31:59 +08:00
ychenfo 1bec6cf2db continue working on the top level 2021-08-11 11:16:53 +08:00
pca006132 a73ab922e2 cleanup 2021-08-10 21:57:31 +08:00
ychenfo 82ce816177 refactored top level parsing, need review 2021-08-10 10:37:06 +08:00
ychenfo 6ad953f877 top level class roughly handled, push for review 2021-08-10 10:37:06 +08:00
pca006132 4db871c244 put alloca in init block 2021-08-09 16:37:28 +08:00
pca006132 cc0692a34c modified alloca 2021-08-09 16:19:20 +08:00
pca006132 7a90ff5791 while loop constructs 2021-08-09 16:10:17 +08:00
pca006132 d8c713ce3d assignment statement 2021-08-09 15:39:50 +08:00
pca006132 1ffb792000 make tuple a ptr to a struct instead of a struct 2021-08-07 17:41:48 +08:00
pca006132 057fcfe3df default parameter value generation 2021-08-07 17:31:01 +08:00
pca006132 86ca02796b function parameter handling 2021-08-07 17:25:14 +08:00
pca006132 711482d09c expr codegen cleanup 2021-08-07 15:30:03 +08:00
pca006132 7a38ab3119 codegen for function call 2021-08-07 15:06:39 +08:00
pca006132 34d3317ea0 store operation method signature 2021-08-07 10:41:53 +08:00
pca006132 c405e46b00 moving location and symbol_resolver out from typecheck 2021-08-07 10:28:41 +08:00
ychenfo 18db2ddd53 change the type TypeEnum::TObj {object_id} to DefinitionId as with top_level
change TopLevelDef::Class {object_id} to DefinitionId
2021-08-06 10:57:01 +08:00
ychenfo fe26070364 cleanup basic_test_env 2021-08-06 10:57:01 +08:00
pca006132 095f28468b added if expr 2021-08-05 16:52:41 +08:00
pca006132 29286210b5 implementing codegen 2021-08-05 14:56:09 +08:00
pca006132 b01d0f6fbb formatting 2021-08-05 14:56:09 +08:00
pca006132 3dcd846302 added rayon dependency 2021-08-05 14:56:09 +08:00
ychenfo c0227210df bit shift lhs rhs same type; float ** int and float ** float both supported 2021-08-05 11:55:46 +08:00
CrescentonC 99c71687a6 fixed: bitwise shift rhs can only be int32; better structured code 2021-08-04 16:46:16 +08:00
CrescentonC d052f007fb fix typo of primitives method 2021-08-04 12:03:56 +08:00
pca006132 8452579c67 use parking_lot RwLock
The std::sync::RwLock is platform dependent, and is unfair on Linux
(may starve writer)
2021-08-03 14:11:41 +08:00
pca006132 f00c1813e3 top-level related changes 2021-08-03 13:38:27 +08:00
pca006132 d4d12a9d1d added crossbeam dependency 2021-08-03 12:38:55 +08:00
pca006132 a3acf09bda typedef: make it send
Rc in calls is not send, so we use Arc instead.
2021-08-03 12:38:12 +08:00
pca006132 52dc112410 unification table: modified conversion impl
from UnificationTable<Rc<RefCell<T>> <==> UnificationTable<T>
to UnificationTable<Rc<T>> <==> UnificationTable<T>
2021-08-03 12:35:58 +08:00
CrescentonC d4807293b0 clean up unused variabls and comments 2021-08-03 10:41:52 +08:00
CrescentonC d4721db4a3 not creating temp for borrow, more concise code 2021-08-03 09:45:39 +08:00
CrescentonC a7e3eeea0d add primitive magic method support; change from TypeEnum::TObj { fields: Mapping<String>, ..} to TypeEnum::TObj {fields: RefCell<Mapping<String>>, .. } for interior mutability 2021-08-02 17:36:37 +08:00
CrescentonC f7bbc3e10d Merge branch 'hm-inference' into hm-inference_anto 2021-08-02 11:33:36 +08:00
CrescentonC 7e0d55443a better structured primitive magic methods impl 2021-08-02 11:28:05 +08:00
pca006132 197a72c658 added comment 2021-07-30 16:43:25 +08:00
pca006132 eba92ed8bd added method to get all instantiations 2021-07-30 16:32:50 +08:00
CrescentonC b87c627c41 updated with field in the test environment 2021-07-30 15:46:57 +08:00
CrescentonC ae79533cfd Merge remote-tracking branch 'origin/hm-inference' into hm-inference_anto 2021-07-30 15:41:53 +08:00
CrescentonC 9983aa62e6 add primitive magic methods 2021-07-30 15:40:14 +08:00
pca006132 7ad8e2d81d cleanup some error reporting code 2021-07-30 13:50:46 +08:00
pca006132 743a9384a3 added rigid type variable 2021-07-30 11:28:27 +08:00
pca006132 f2c5a9b352 added location -> call mapping
This allows code generation module to get function instantiation
parameter directly.
2021-07-30 11:01:11 +08:00
ychenfo 09e76efcf7 start adding primitive magic methods 2021-07-29 15:36:19 +08:00
pca006132 832513e210 new is_concrete type check 2021-07-28 17:25:19 +08:00
pca006132 f665ea358b fixed fold 2021-07-28 10:44:58 +08:00
pca006132 e15473d2c9 fixed pattern matching 2021-07-27 14:39:53 +08:00
pca006132 5f0490cd84 added virtual test 2021-07-27 11:58:35 +08:00
pca006132 1d13b16f94 updated function check 2021-07-26 16:00:29 +08:00
pca006132 8d0856a58d added documentation 2021-07-26 14:38:18 +08:00
pca006132 bf31c48bba fixed missing unification 2021-07-26 14:20:47 +08:00
Sebastien Bourdeauducq 0941de9ee1 Revert "shell.nix: set LLVM_SYS_100_PREFIX. Closes #4"
This reverts commit 53ebe8d8b2.
2021-07-26 12:29:16 +08:00
pca006132 8618837816 fixed range unification 2021-07-26 12:00:06 +08:00
Sebastien Bourdeauducq 53ebe8d8b2 shell.nix: set LLVM_SYS_100_PREFIX. Closes #4 2021-07-24 09:31:24 +08:00
pca006132 d7df93bef1 fixed range check 2021-07-23 17:22:05 +08:00
pca006132 d140164a38 fixed virtual unification 2021-07-23 16:19:00 +08:00
pca006132 ddcf4b7e39 refactored typedef 2021-07-23 15:57:37 +08:00
pca006132 88c45172b2 basic check for use-before-def 2021-07-22 17:07:49 +08:00
pca006132 c315227a28 init function check 2021-07-22 15:36:37 +08:00
pca006132 d484fa1e5c added return type check 2021-07-22 11:49:00 +08:00
pca006132 09c9218852 use custom unification table implementation
as the ena implementation did not expose the underlying vector store, we
cannot map over it to get a table without Rc<RefCell<T>> so that we can
send it around...
2021-07-22 11:37:29 +08:00
pca006132 4f81690128 modified occur check 2021-07-21 16:10:11 +08:00
pca006132 b3d849ea7a list test 2021-07-21 16:06:06 +08:00
pca006132 3e03398d9b obj test 2021-07-21 15:59:01 +08:00
pca006132 2f5c3b3cb7 more cleanup and started adding tests 2021-07-21 15:36:35 +08:00
pca006132 25ff24a320 modified interface 2021-07-21 14:24:46 +08:00
pca006132 0296844d5f cleanup 2021-07-21 13:28:05 +08:00
pca006132 e95bfe1d31 added statements 2021-07-20 16:56:04 +08:00
pca006132 bc9b453b3e function call implementation 2021-07-20 16:13:43 +08:00
pca006132 fa31e8f336 fold listcomp 2021-07-20 13:45:17 +08:00
pca006132 22455e43ac lambda fold 2021-07-20 11:47:19 +08:00
pca006132 016166de46 skeleton done 2021-07-19 17:26:51 +08:00
pca006132 eb4b2bb7f6 refactored using constrain
to allow easier modification later with subtyping
2021-07-19 17:05:48 +08:00
pca006132 e732f7e089 removed integer encoding 2021-07-19 16:51:58 +08:00
pca006132 d4b85d0bac expression type inference (WIP) 2021-07-19 13:35:01 +08:00
pca006132 c913fb28bd use signed integer for TSeq 2021-07-19 13:34:45 +08:00
pca006132 f51603f6da cleanup 2021-07-19 09:52:25 +08:00
pca006132 d67407716c function unification... 2021-07-16 15:55:52 +08:00
pca006132 f4121b570d added documentation 2021-07-16 14:34:52 +08:00
pca006132 8b078dfa1b naming 2021-07-16 13:59:08 +08:00
pca006132 62736bd4bf cleanup: we don't actually need arena 2021-07-16 13:58:02 +08:00
pca006132 c2d00aa762 occur check 2021-07-15 16:51:55 +08:00
pca006132 d94f25583b added tests 2021-07-15 16:00:23 +08:00
pca006132 1df3f4e757 most of unification... 2021-07-14 17:20:12 +08:00
pca006132 97fe450a0b occur check 2021-07-14 16:40:50 +08:00
pca006132 e8c5189fce simplified code with Rc<RefCell<T>> 2021-07-14 15:58:58 +08:00
pca006132 291e642699 partial implementation 2021-07-14 15:24:00 +08:00
pca006132 e554737b68 tmp 2021-07-14 08:12:47 +08:00
pca006132 84c980fed3 type scheme instantiation 2021-06-30 17:18:56 +08:00
pca006132 2985b88351 refactor for HM style inference... 2021-06-30 16:28:18 +08:00
148 changed files with 32728 additions and 2439 deletions

1
.gitignore vendored
View File

@ -1,2 +1,3 @@
__pycache__ __pycache__
/target /target
windows/msys2

823
Cargo.lock generated

File diff suppressed because it is too large Load Diff

View File

@ -1,6 +1,12 @@
[workspace] [workspace]
members = [ members = [
"nac3ast",
"nac3parser",
"nac3core", "nac3core",
"nac3standalone", "nac3standalone",
"nac3embedded", "nac3artiq",
"runkernel",
] ]
[profile.release]
debug = true

View File

@ -1,34 +1,55 @@
# nac3 compiler # NAC3
NAC3 is a major, backward-incompatible rewrite of the compiler for the [ARTIQ](https://m-labs.hk/artiq) physics experiment control and data acquisition system. It features greatly improved compilation speeds, a much better type system, and more predictable and transparent operation.
NAC3 has a modular design and its applicability reaches beyond ARTIQ. The ``nac3core`` module does not contain anything specific to ARTIQ, and can be used in any project that requires compiling Python to machine code.
**WARNING: NAC3 is currently experimental software and several important features are not implemented yet.**
## Packaging
NAC3 is packaged using the [Nix](https://nixos.org) Flakes system. Install Nix 2.4+ and enable flakes by adding ``experimental-features = nix-command flakes`` to ``nix.conf`` (e.g. ``~/.config/nix/nix.conf``).
## Try NAC3
### Linux
After setting up Nix as above, use ``nix shell git+https://github.com/m-labs/artiq.git?ref=nac3`` to get a shell with the NAC3 version of ARTIQ. See the ``examples`` directory in ARTIQ (``nac3`` Git branch) for some samples of NAC3 kernel code.
### Windows
Install [MSYS2](https://www.msys2.org/), and open "MSYS2 MinGW x64". Edit ``/etc/pacman.conf`` to add:
```
[artiq]
SigLevel = Optional TrustAll
Server = https://lab.m-labs.hk/msys2
```
Then run the following commands:
```
pacman -Syu
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: This repository contains:
- nac3core: Core compiler library, containing type-checking, static analysis (in - ``nac3ast``: Python abstract syntax tree definition (based on RustPython).
the future) and code generation. - ``nac3parser``: Python parser (based on RustPython).
- nac3embedded: Integration with CPython runtime. - ``nac3core``: Core compiler library, containing type-checking and code generation.
- nac3standalone: Standalone compiler tool. - ``nac3standalone``: Standalone compiler tool (core language only).
- ``nac3artiq``: Integration with ARTIQ and implementation of ARTIQ-specific extensions to the core language.
The core compiler would know nothing about symbol resolution, host variables - ``runkernel``: Simple program that runs compiled ARTIQ kernels on the host and displays RTIO operations. Useful for testing without hardware.
etc. The nac3embedded/nac3standalone library would provide (implement) the
symbol resolver to the core compiler for resolving the type and value for
unknown symbols. The core compiler would only type check classes and functions
requested by the nac3embedded/nac3standalone lib (the API should allow the
caller to specify which methods should be compiled). After type checking, the
compiler would analyse the set of functions/classes that are used and perform
code generation.
value could be integer values, boolean values, bytes (for memcpy), function ID
(full name + concrete type)
## Current Plan
Type checking:
- [x] Basic interface for symbol resolver.
- [x] Track location information in context object (for diagnostics).
- [ ] Refactor old expression and statement type inference code. (anto)
- [ ] Error diagnostics utilities. (pca)
- [ ] Move tests to external files, write scripts for testing. (pca)
- [ ] Implement function type checking (instantiate bounded type parameters),
loop unrolling, type inference for lists with virtual objects. (pca)
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``.

27
flake.lock Normal file
View File

@ -0,0 +1,27 @@
{
"nodes": {
"nixpkgs": {
"locked": {
"lastModified": 1649619156,
"narHash": "sha256-p0q4zpuKMwrzGF+5ZU7Thnpac5TinhDI9jr2mBxhV4w=",
"owner": "NixOS",
"repo": "nixpkgs",
"rev": "e7d63bd0d50df412f5a1d8acfa3caae75522e347",
"type": "github"
},
"original": {
"owner": "NixOS",
"ref": "nixos-21.11",
"repo": "nixpkgs",
"type": "github"
}
},
"root": {
"inputs": {
"nixpkgs": "nixpkgs"
}
}
},
"root": "root",
"version": 7
}

154
flake.nix Normal file
View File

@ -0,0 +1,154 @@
{
description = "The third-generation ARTIQ compiler";
inputs.nixpkgs.url = github:NixOS/nixpkgs/nixos-21.11;
outputs = { self, nixpkgs }:
let
pkgs = import nixpkgs { system = "x86_64-linux"; };
in rec {
packages.x86_64-linux = rec {
llvm-nac3 = pkgs.callPackage ./nix/llvm {};
nac3artiq = pkgs.python3Packages.toPythonModule (
pkgs.rustPlatform.buildRustPackage {
name = "nac3artiq";
outputs = [ "out" "runkernel" "standalone" ];
src = self;
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 ])) ];
checkPhase =
''
echo "Checking nac3standalone demos..."
pushd nac3standalone/demo
patchShebangs .
./check_demos.sh
popd
echo "Running Cargo tests..."
cargoCheckHook
'';
installPhase =
''
PYTHON_SITEPACKAGES=$out/${pkgs.python3Packages.python.sitePackages}
mkdir -p $PYTHON_SITEPACKAGES
cp target/x86_64-unknown-linux-gnu/release/libnac3artiq.so $PYTHON_SITEPACKAGES/nac3artiq.so
mkdir -p $runkernel/bin
cp target/x86_64-unknown-linux-gnu/release/runkernel $runkernel/bin
mkdir -p $standalone/bin
cp target/x86_64-unknown-linux-gnu/release/nac3standalone $standalone/bin
'';
}
);
python3-mimalloc = pkgs.python3 // rec {
withMimalloc = pkgs.python3.buildEnv.override({ makeWrapperArgs = [ "--set LD_PRELOAD ${pkgs.mimalloc}/lib/libmimalloc.so" ]; });
withPackages = f: let packages = f pkgs.python3.pkgs; in withMimalloc.override { extraLibs = packages; };
};
# LLVM PGO support
llvm-nac3-instrumented = pkgs.callPackage ./nix/llvm {
stdenv = pkgs.llvmPackages_13.stdenv;
extraCmakeFlags = [ "-DLLVM_BUILD_INSTRUMENTED=IR" ];
};
nac3artiq-instrumented = pkgs.python3Packages.toPythonModule (
pkgs.rustPlatform.buildRustPackage {
name = "nac3artiq-instrumented";
src = self;
cargoLock = { lockFile = ./Cargo.lock; };
nativeBuildInputs = [ pkgs.python3 pkgs.llvmPackages_13.clang-unwrapped llvm-nac3-instrumented ];
buildInputs = [ pkgs.python3 llvm-nac3-instrumented ];
cargoBuildFlags = [ "--package" "nac3artiq" "--features" "init-llvm-profile" ];
doCheck = false;
configurePhase =
''
export CARGO_TARGET_X86_64_UNKNOWN_LINUX_GNU_RUSTFLAGS="-C link-arg=-L${pkgs.llvmPackages_13.compiler-rt}/lib/linux -C link-arg=-lclang_rt.profile-x86_64"
'';
installPhase =
''
TARGET_DIR=$out/${pkgs.python3Packages.python.sitePackages}
mkdir -p $TARGET_DIR
cp target/x86_64-unknown-linux-gnu/release/libnac3artiq.so $TARGET_DIR/nac3artiq.so
'';
}
);
nac3artiq-profile = pkgs.stdenvNoCC.mkDerivation {
name = "nac3artiq-profile";
src = self;
buildInputs = [ (python3-mimalloc.withPackages(ps: [ ps.numpy nac3artiq-instrumented ])) pkgs.lld_13 pkgs.llvmPackages_13.libllvm ];
phases = [ "buildPhase" "installPhase" ];
# TODO: get more representative code.
buildPhase = "python $src/nac3artiq/demo/demo.py";
installPhase =
''
mkdir $out
llvm-profdata merge -o $out/llvm.profdata /build/llvm/build/profiles/*
'';
};
llvm-nac3-pgo = pkgs.callPackage ./nix/llvm {
stdenv = pkgs.llvmPackages_13.stdenv;
extraCmakeFlags = [ "-DLLVM_PROFDATA_FILE=${nac3artiq-profile}/llvm.profdata" ];
};
nac3artiq-pgo = pkgs.python3Packages.toPythonModule (
pkgs.rustPlatform.buildRustPackage {
name = "nac3artiq-pgo";
src = self;
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" ];
installPhase =
''
TARGET_DIR=$out/${pkgs.python3Packages.python.sitePackages}
mkdir -p $TARGET_DIR
cp target/x86_64-unknown-linux-gnu/release/libnac3artiq.so $TARGET_DIR/nac3artiq.so
'';
}
);
};
packages.x86_64-w64-mingw32 = import ./nix/windows { inherit pkgs; };
devShell.x86_64-linux = pkgs.mkShell {
name = "nac3-dev-shell";
buildInputs = with pkgs; [
# build dependencies
packages.x86_64-linux.llvm-nac3
llvmPackages_13.clang-unwrapped # IRRT
cargo
rustc
# runtime dependencies
lld_13
(packages.x86_64-linux.python3-mimalloc.withPackages(ps: [ ps.numpy ]))
# development tools
cargo-insta
clippy
rustfmt
];
};
devShells.x86_64-linux.msys2 = pkgs.mkShell {
name = "nac3-dev-shell-msys2";
buildInputs = with pkgs; [
curl
pacman
fakeroot
packages.x86_64-w64-mingw32.wine-msys2
];
};
hydraJobs = {
inherit (packages.x86_64-linux) llvm-nac3 nac3artiq;
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 = {
binaryCachePublicKeys = ["nixbld.m-labs.hk-1:5aSRVA5b320xbNvu30tqxVPXpld73bhtOeH6uAjRyHc="];
binaryCaches = ["https://nixbld.m-labs.hk" "https://cache.nixos.org"];
};
}

24
nac3artiq/Cargo.toml Normal file
View File

@ -0,0 +1,24 @@
[package]
name = "nac3artiq"
version = "0.1.0"
authors = ["M-Labs"]
edition = "2018"
[lib]
name = "nac3artiq"
crate-type = ["cdylib"]
[dependencies]
pyo3 = { version = "0.14", features = ["extension-module"] }
parking_lot = "0.11"
tempfile = "3"
nac3parser = { path = "../nac3parser" }
nac3core = { path = "../nac3core" }
[dependencies.inkwell]
version = "0.1.0-beta.4"
default-features = false
features = ["llvm13-0", "target-x86", "target-arm", "target-riscv", "no-libffi-linking"]
[features]
init-llvm-profile = []

26
nac3artiq/demo/demo.py Normal file
View File

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

View File

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

View File

@ -0,0 +1,59 @@
class EmbeddingMap:
def __init__(self):
self.object_inverse_map = {}
self.object_map = {}
self.string_map = {}
self.string_reverse_map = {}
self.function_map = {}
self.attributes_writeback = []
# preallocate exception names
self.preallocate_runtime_exception_names(["RuntimeError",
"RTIOUnderflow",
"RTIOOverflow",
"RTIODestinationUnreachable",
"DMAError",
"I2CError",
"CacheError",
"SPIError",
"0:ZeroDivisionError",
"0:IndexError",
"0:UnwrapNoneError"])
def preallocate_runtime_exception_names(self, names):
for i, name in enumerate(names):
if ":" not in name:
name = "0:artiq.coredevice.exceptions." + name
exn_id = self.store_str(name)
assert exn_id == i
def store_function(self, key, fun):
self.function_map[key] = fun
return key
def store_object(self, obj):
obj_id = id(obj)
if obj_id in self.object_inverse_map:
return self.object_inverse_map[obj_id]
key = len(self.object_map) + 1
self.object_map[key] = obj
self.object_inverse_map[obj_id] = key
return key
def store_str(self, s):
if s in self.string_reverse_map:
return self.string_reverse_map[s]
key = len(self.string_map)
self.string_map[key] = s
self.string_reverse_map[s] = key
return key
def retrieve_function(self, key):
return self.function_map[key]
def retrieve_object(self, key):
return self.object_map[key]
def retrieve_str(self, key):
return self.string_map[key]

283
nac3artiq/demo/min_artiq.py Normal file
View File

@ -0,0 +1,283 @@
from inspect import getfullargspec
from functools import wraps
from types import SimpleNamespace
from numpy import int32, int64
from typing import Generic, TypeVar
from math import floor, ceil
import nac3artiq
from embedding_map import EmbeddingMap
__all__ = [
"Kernel", "KernelInvariant", "virtual",
"Option", "Some", "none", "UnwrapNoneError",
"round64", "floor64", "ceil64",
"extern", "kernel", "portable", "nac3",
"rpc", "ms", "us", "ns",
"print_int32", "print_int64",
"Core", "TTLOut",
"parallel", "sequential"
]
T = TypeVar('T')
class Kernel(Generic[T]):
pass
class KernelInvariant(Generic[T]):
pass
# The virtual class must exist before nac3artiq.NAC3 is created.
class virtual(Generic[T]):
pass
class Option(Generic[T]):
_nac3_option: T
def __init__(self, v: T):
self._nac3_option = v
def is_none(self):
return self._nac3_option is None
def is_some(self):
return not self.is_none()
def unwrap(self):
if self.is_none():
raise UnwrapNoneError()
return self._nac3_option
def __repr__(self) -> str:
if self.is_none():
return "none"
else:
return "Some({})".format(repr(self._nac3_option))
def __str__(self) -> str:
if self.is_none():
return "none"
else:
return "Some({})".format(str(self._nac3_option))
def Some(v: T) -> Option[T]:
return Option(v)
none = Option(None)
def round64(x):
return round(x)
def floor64(x):
return floor(x)
def ceil64(x):
return ceil(x)
import device_db
core_arguments = device_db.device_db["core"]["arguments"]
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()
registered_classes = set()
def register_function(fun):
assert allow_registration
registered_functions.add(fun)
def register_class(cls):
assert allow_registration
registered_classes.add(cls)
def extern(function):
"""Decorates a function declaration defined by the core device runtime."""
register_function(function)
return function
def rpc(function):
"""Decorates a function declaration defined by the core device runtime."""
register_function(function)
return function
def kernel(function_or_method):
"""Decorates a function or method to be executed on the core device."""
register_function(function_or_method)
argspec = getfullargspec(function_or_method)
if argspec.args and argspec.args[0] == "self":
@wraps(function_or_method)
def run_on_core(self, *args, **kwargs):
fake_method = SimpleNamespace(__self__=self, __name__=function_or_method.__name__)
self.core.run(fake_method, *args, **kwargs)
else:
@wraps(function_or_method)
def run_on_core(*args, **kwargs):
raise RuntimeError("Kernel functions need explicit core.run()")
return run_on_core
def portable(function):
"""Decorates a function or method to be executed on the same device (host/core device) as the caller."""
register_function(function)
return function
def nac3(cls):
"""
Decorates a class to be analyzed by NAC3.
All classes containing kernels or portable methods must use this decorator.
"""
register_class(cls)
return cls
ms = 1e-3
us = 1e-6
ns = 1e-9
@extern
def rtio_init():
raise NotImplementedError("syscall not simulated")
@extern
def rtio_get_counter() -> int64:
raise NotImplementedError("syscall not simulated")
@extern
def rtio_output(target: int32, data: int32):
raise NotImplementedError("syscall not simulated")
@extern
def rtio_input_timestamp(timeout_mu: int64, channel: int32) -> int64:
raise NotImplementedError("syscall not simulated")
@extern
def rtio_input_data(channel: int32) -> int32:
raise NotImplementedError("syscall not simulated")
# These is not part of ARTIQ and only available in runkernel. Defined here for convenience.
@extern
def print_int32(x: int32):
raise NotImplementedError("syscall not simulated")
@extern
def print_int64(x: int64):
raise NotImplementedError("syscall not simulated")
@nac3
class Core:
ref_period: KernelInvariant[float]
def __init__(self):
self.ref_period = core_arguments["ref_period"]
def run(self, method, *args, **kwargs):
global allow_registration
embedding = EmbeddingMap()
if allow_registration:
compiler.analyze(registered_functions, registered_classes)
allow_registration = False
if hasattr(method, "__self__"):
obj = method.__self__
name = method.__name__
else:
obj = method
name = ""
compiler.compile_method_to_file(obj, name, args, "module.elf", embedding)
@kernel
def reset(self):
rtio_init()
at_mu(rtio_get_counter() + int64(125000))
@kernel
def break_realtime(self):
min_now = rtio_get_counter() + int64(125000)
if now_mu() < min_now:
at_mu(min_now)
@portable
def seconds_to_mu(self, seconds: float) -> int64:
return int64(round(seconds/self.ref_period))
@portable
def mu_to_seconds(self, mu: int64) -> float:
return float(mu)*self.ref_period
@kernel
def delay(self, dt: float):
delay_mu(self.seconds_to_mu(dt))
@nac3
class TTLOut:
core: KernelInvariant[Core]
channel: KernelInvariant[int32]
target_o: KernelInvariant[int32]
def __init__(self, core: Core, channel: int32):
self.core = core
self.channel = channel
self.target_o = channel << 8
@kernel
def output(self):
pass
@kernel
def set_o(self, o: bool):
rtio_output(self.target_o, 1 if o else 0)
@kernel
def on(self):
self.set_o(True)
@kernel
def off(self):
self.set_o(False)
@kernel
def pulse_mu(self, duration: int64):
self.on()
delay_mu(duration)
self.off()
@kernel
def pulse(self, duration: float):
self.on()
self.core.delay(duration)
self.off()
@nac3
class KernelContextManager:
@kernel
def __enter__(self):
pass
@kernel
def __exit__(self):
pass
@nac3
class UnwrapNoneError(Exception):
"""raised when unwrapping a none value"""
artiq_builtin = True
parallel = KernelContextManager()
sequential = KernelContextManager()

1
nac3artiq/demo/nac3artiq.so Symbolic link
View File

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

575
nac3artiq/src/codegen.rs Normal file
View File

@ -0,0 +1,575 @@
use nac3core::{
codegen::{
expr::gen_call,
stmt::{gen_block, gen_with},
CodeGenContext, CodeGenerator,
},
symbol_resolver::ValueEnum,
toplevel::{DefinitionId, GenCall},
typecheck::typedef::{FunSignature, FuncArg, Type, TypeEnum}
};
use nac3parser::ast::{Expr, ExprKind, Located, Stmt, StmtKind, StrRef};
use inkwell::{
context::Context, module::Linkage, types::IntType, values::BasicValueEnum, AddressSpace,
};
use pyo3::{PyObject, PyResult, Python, types::{PyDict, PyList}};
use crate::{symbol_resolver::InnerResolver, timeline::TimeFns};
use std::{
collections::hash_map::DefaultHasher,
collections::HashMap,
hash::{Hash, Hasher},
sync::Arc,
};
pub struct ArtiqCodeGenerator<'a> {
name: String,
size_t: u32,
name_counter: u32,
start: Option<Expr<Option<Type>>>,
end: Option<Expr<Option<Type>>>,
timeline: &'a (dyn TimeFns + Sync),
}
impl<'a> ArtiqCodeGenerator<'a> {
pub fn new(
name: String,
size_t: u32,
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 }
}
}
impl<'b> CodeGenerator for ArtiqCodeGenerator<'b> {
fn get_name(&self) -> &str {
&self.name
}
fn get_size_type<'ctx>(&self, ctx: &'ctx Context) -> IntType<'ctx> {
if self.size_t == 32 {
ctx.i32_type()
} else {
ctx.i64_type()
}
}
fn gen_call<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
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)?;
if let Some(end) = self.end.clone() {
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 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, start.custom.unwrap())?;
self.timeline.emit_at_mu(ctx, start_val);
}
Ok(result)
}
fn gen_with<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
stmt: &Stmt<Option<Type>>,
) -> Result<(), String> {
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.
// Each function call directly inside a parallel block will reset the timeline after
// execution. A parallel block within a sequential block (or not within any block) will
// set the timeline to the max now_mu within the block (and the outer max now_mu will also
// be updated).
//
// Implementation: We track the start and end separately.
// - If there is a start variable, it indicates that we are directly inside a
// parallel block and we have to reset the timeline after every function call.
// - 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() {
let old_start = self.start.take();
let old_end = self.end.take();
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())?
} 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
// associative type is used by limiting the lifetime of CodeGenerator to
// the LLVM Context.
// The name is guaranteed to be unique as users cannot use this as variable
// name.
self.start = old_start.clone().map_or_else(
|| {
let start = format!("with-{}-start", self.name_counter).into();
let start_expr = Located {
// location does not matter at this point
location: stmt.location,
node: ExprKind::Name { id: start, ctx: name_ctx.clone() },
custom: Some(ctx.primitives.int64),
};
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)),
)?;
let end = format!("with-{}-end", self.name_counter).into();
let end_expr = Located {
// location does not matter at this point
location: stmt.location,
node: ExprKind::Name { id: end, ctx: name_ctx.clone() },
custom: Some(ctx.primitives.int64),
};
let end = self.gen_store_target(ctx, &end_expr)?;
ctx.builder.build_store(end, now);
self.end = Some(end_expr);
self.name_counter += 1;
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
// block... e.g.
// if ...:
// return
// Perhaps we can fix this by using actual with block?
let reset_position = if let Some(terminator) = current.get_terminator() {
ctx.builder.position_before(&terminator);
true
} 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())?;
// 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
if let Some(old_end) = &old_end {
let outer_end_val = self
.gen_expr(ctx, old_end)?
.unwrap()
.to_basic_value_enum(ctx, self, old_end.custom.unwrap())?;
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() {
let start = self.start.take();
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, 'a>(
ctx: &mut CodeGenContext<'ctx, 'a>,
ty: Type,
buffer: &mut Vec<u8>,
) -> Result<(), String> {
use nac3core::typecheck::typedef::TypeEnum::*;
let int32 = ctx.primitives.int32;
let int64 = ctx.primitives.int64;
let float = ctx.primitives.float;
let bool = ctx.primitives.bool;
let str = ctx.primitives.str;
let none = ctx.primitives.none;
if ctx.unifier.unioned(ty, int32) {
buffer.push(b'i');
} else if ctx.unifier.unioned(ty, int64) {
buffer.push(b'I');
} else if ctx.unifier.unioned(ty, float) {
buffer.push(b'f');
} else if ctx.unifier.unioned(ty, bool) {
buffer.push(b'b');
} else if ctx.unifier.unioned(ty, str) {
buffer.push(b's');
} else if ctx.unifier.unioned(ty, none) {
buffer.push(b'n');
} else {
let ty_enum = ctx.unifier.get_ty(ty);
match &*ty_enum {
TTuple { ty } => {
buffer.push(b't');
buffer.push(ty.len() as u8);
for ty in ty {
gen_rpc_tag(ctx, *ty, buffer)?;
}
}
TList { ty } => {
buffer.push(b'l');
gen_rpc_tag(ctx, *ty, buffer)?;
}
_ => return Err(format!("Unsupported type: {:?}", ctx.unifier.stringify(ty))),
}
}
Ok(())
}
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(inkwell::AddressSpace::Generic);
let size_type = generator.get_size_type(ctx.ctx);
let int8 = ctx.ctx.i8_type();
let int32 = ctx.ctx.i32_type();
let tag_ptr_type = ctx.ctx.struct_type(&[ptr_type.into(), size_type.into()], false);
let service_id = int32.const_int(fun.1.0 as u64, false);
// -- setup rpc tags
let mut tag = Vec::new();
if obj.is_some() {
tag.push(b'O');
}
for arg in fun.0.args.iter() {
gen_rpc_tag(ctx, arg.ty, &mut tag)?;
}
tag.push(b':');
gen_rpc_tag(ctx, fun.0.ret, &mut tag)?;
let mut hasher = DefaultHasher::new();
tag.hash(&mut hasher);
let hash = format!("{}", hasher.finish());
let tag_ptr = ctx
.module
.get_global(hash.as_str())
.unwrap_or_else(|| {
let tag_arr_ptr = ctx.module.add_global(
int8.array_type(tag.len() as u32),
None,
format!("tagptr{}", fun.1 .0).as_str(),
);
tag_arr_ptr.set_initializer(&int8.const_array(
&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);
tag_ptr.set_linkage(Linkage::Private);
tag_ptr.set_initializer(&ctx.ctx.const_struct(
&[
tag_arr_ptr.as_pointer_value().const_cast(ptr_type).into(),
size_type.const_int(tag.len() as u64, false).into(),
],
false,
));
tag_ptr
})
.as_pointer_value();
let arg_length = args.len() + if obj.is_some() { 1 } else { 0 };
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",
);
// -- rpc args handling
let mut keys = fun.0.args.clone();
let mut mapping = HashMap::new();
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.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))
.collect::<Result<Vec<_>, _>>()?;
if let Some(obj) = obj {
if let ValueEnum::Static(obj) = obj.1 {
real_params.insert(0, obj.get_const_obj(ctx, generator));
} else {
// should be an error here...
panic!("only host object is allowed");
}
}
for (i, arg) in real_params.iter().enumerate() {
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),
)
};
ctx.builder.build_store(arg_ptr, arg_slot);
}
// call
let rpc_send = ctx.module.get_function("rpc_send").unwrap_or_else(|| {
ctx.module.add_function(
"rpc_send",
ctx.ctx.void_type().fn_type(
&[
int32.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",
);
// reclaim stack space used by arguments
ctx.builder.build_call(
stackrestore,
&[stackptr.try_as_basic_value().unwrap_left().into()],
"rpc.stackrestore",
);
// -- receive value:
// T result = {
// void *ret_ptr = alloca(sizeof(T));
// void *ptr = ret_ptr;
// loop: int size = rpc_recv(ptr);
// // Non-zero: Provide `size` bytes of extra storage for variable-length data.
// if(size) { ptr = alloca(size); goto loop; }
// else *(T*)ret_ptr
// }
let rpc_recv = ctx.module.get_function("rpc_recv").unwrap_or_else(|| {
ctx.module.add_function("rpc_recv", int32.fn_type(&[ptr_type.into()], false), None)
});
if ctx.unifier.unioned(fun.0.ret, ctx.primitives.none) {
ctx.build_call_or_invoke(rpc_recv, &[ptr_type.const_null().into()], "rpc_recv");
return Ok(None);
}
let prehead_bb = ctx.builder.get_insert_block().unwrap();
let current_function = prehead_bb.get_parent().unwrap();
let head_bb = ctx.ctx.append_basic_block(current_function, "rpc.head");
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_type(generator, fun.0.ret);
let need_load = !ret_ty.is_pointer_type();
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");
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",
);
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");
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);
ctx.builder.position_at_end(tail_bb);
let result = ctx.builder.build_load(slot, "rpc.result");
if need_load {
ctx.builder.build_call(
stackrestore,
&[stackptr.try_as_basic_value().unwrap_left().into()],
"rpc.stackrestore",
);
}
Ok(Some(result))
}
pub fn attributes_writeback<'ctx, 'a>(
ctx: &mut CodeGenContext<'ctx, 'a>,
generator: &mut dyn CodeGenerator,
inner_resolver: &InnerResolver,
host_attributes: PyObject,
) -> Result<(), String> {
Python::with_gil(|py| -> PyResult<Result<(), String>> {
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.iter() {
let val = val.as_ref(py);
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))
}
let ty = ty.unwrap();
match &*ctx.unifier.get_ty(ty) {
TypeEnum::TObj { fields, obj_id, .. }
if *obj_id != ctx.primitives.option.get_obj_id(&ctx.unifier) =>
{
// 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.iter() {
if !is_mutable {
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(
obj.into_pointer_value(),
&[zero, int32.const_int(index as u64, false)])));
}
}
if !attributes.is_empty() {
let pydict = PyDict::new(py);
pydict.set_item("obj", val)?;
pydict.set_item("fields", attributes)?;
host_attributes.append(pydict)?;
}
},
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()));
}
},
_ => {}
}
}
let fun = FunSignature {
args: values.iter().enumerate().map(|(i, (ty, _))| FuncArg {
name: i.to_string().into(),
ty: *ty,
default_value: None
}).collect(),
ret: ctx.primitives.none,
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, DefinitionId(0)), args, generator) {
return Ok(Err(e));
}
Ok(Ok(()))
}).unwrap()?;
Ok(())
}
pub fn rpc_codegen_callback() -> Arc<GenCall> {
Arc::new(GenCall::new(Box::new(|ctx, obj, fun, args, generator| {
rpc_codegen_callback_fn(ctx, obj, fun, args, generator)
})))
}

56
nac3artiq/src/kernel.ld Normal file
View File

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

922
nac3artiq/src/lib.rs Normal file
View File

@ -0,0 +1,922 @@
use std::collections::{HashMap, HashSet};
use std::fs;
use std::process::Command;
use std::rc::Rc;
use std::sync::Arc;
use inkwell::{
memory_buffer::MemoryBuffer,
passes::{PassManager, PassManagerBuilder},
targets::*,
OptimizationLevel,
};
use nac3core::codegen::gen_func_impl;
use nac3core::toplevel::builtins::get_exn_constructor;
use nac3core::typecheck::typedef::{TypeEnum, Unifier};
use nac3parser::{
ast::{self, ExprKind, Stmt, StmtKind, StrRef},
parser::{self, parse_program},
};
use pyo3::prelude::*;
use pyo3::{exceptions, types::PyBytes, types::PyDict, types::PySet};
use pyo3::create_exception;
use parking_lot::{Mutex, RwLock};
use nac3core::{
codegen::irrt::load_irrt,
codegen::{concrete_type::ConcreteTypeStore, CodeGenTask, WithCall, WorkerRegistry},
symbol_resolver::SymbolResolver,
toplevel::{
composer::{ComposerConfig, TopLevelComposer},
DefinitionId, GenCall, TopLevelDef,
},
typecheck::typedef::{FunSignature, FuncArg},
typecheck::{type_inferencer::PrimitiveStore, typedef::Type},
};
use tempfile::{self, TempDir};
use crate::codegen::attributes_writeback;
use crate::{
codegen::{rpc_codegen_callback, ArtiqCodeGenerator},
symbol_resolver::{InnerResolver, PythonHelper, Resolver, DeferredEvaluationStore},
};
mod codegen;
mod symbol_resolver;
mod timeline;
use timeline::TimeFns;
#[derive(PartialEq, Clone, Copy)]
enum Isa {
Host,
RiscV32G,
RiscV32IMA,
CortexA9,
}
#[derive(Clone)]
pub struct PrimitivePythonId {
int: u64,
int32: u64,
int64: u64,
uint32: u64,
uint64: u64,
float: u64,
float64: u64,
bool: u64,
list: u64,
tuple: u64,
typevar: u64,
none: u64,
exception: u64,
generic_alias: (u64, u64),
virtual_id: u64,
option: u64,
}
type TopLevelComponent = (Stmt, String, PyObject);
// TopLevelComposer is unsendable as it holds the unification table, which is
// unsendable due to Rc. Arc would cause a performance hit.
#[pyclass(unsendable, name = "NAC3")]
struct Nac3 {
isa: Isa,
time_fns: &'static (dyn TimeFns + Sync),
primitive: PrimitiveStore,
builtins: Vec<(StrRef, FunSignature, Arc<GenCall>)>,
pyid_to_def: Arc<RwLock<HashMap<u64, DefinitionId>>>,
primitive_ids: PrimitivePythonId,
working_directory: TempDir,
top_levels: Vec<TopLevelComponent>,
string_store: Arc<RwLock<HashMap<String, i32>>>,
exception_ids: Arc<RwLock<HashMap<usize, usize>>>,
deferred_eval_store: DeferredEvaluationStore
}
create_exception!(nac3artiq, CompileError, exceptions::PyException);
impl Nac3 {
fn register_module(
&mut self,
module: PyObject,
registered_class_ids: &HashSet<u64>,
) -> PyResult<()> {
let (module_name, source_file) = Python::with_gil(|py| -> PyResult<(String, String)> {
let module: &PyAny = module.extract(py)?;
Ok((module.getattr("__name__")?.extract()?, module.getattr("__file__")?.extract()?))
})?;
let source = fs::read_to_string(&source_file).map_err(|e| {
exceptions::PyIOError::new_err(format!("failed to read input file: {}", e))
})?;
let parser_result = parser::parse_program(&source, source_file.into())
.map_err(|e| exceptions::PySyntaxError::new_err(format!("parse error: {}", e)))?;
for mut stmt in parser_result.into_iter() {
let include = match stmt.node {
ast::StmtKind::ClassDef {
ref decorator_list, ref mut body, ref mut bases, ..
} => {
let nac3_class = decorator_list.iter().any(|decorator| {
if let ast::ExprKind::Name { id, .. } = decorator.node {
id.to_string() == "nac3"
} else {
false
}
});
if !nac3_class {
continue;
}
// Drop unregistered (i.e. host-only) base classes.
bases.retain(|base| {
Python::with_gil(|py| -> PyResult<bool> {
let id_fn = PyModule::import(py, "builtins")?.getattr("id")?;
match &base.node {
ast::ExprKind::Name { id, .. } => {
if *id == "Exception".into() {
Ok(true)
} else {
let base_obj = module.getattr(py, id.to_string())?;
let base_id = id_fn.call1((base_obj,))?.extract()?;
Ok(registered_class_ids.contains(&base_id))
}
}
_ => Ok(true),
}
})
.unwrap()
});
body.retain(|stmt| {
if let ast::StmtKind::FunctionDef { ref decorator_list, .. } = stmt.node {
decorator_list.iter().any(|decorator| {
if let ast::ExprKind::Name { id, .. } = decorator.node {
id.to_string() == "kernel"
|| id.to_string() == "portable"
|| id.to_string() == "rpc"
} else {
false
}
})
} else {
true
}
});
true
}
ast::StmtKind::FunctionDef { ref decorator_list, .. } => {
decorator_list.iter().any(|decorator| {
if let ast::ExprKind::Name { id, .. } = decorator.node {
let id = id.to_string();
id == "extern" || id == "portable" || id == "kernel" || id == "rpc"
} else {
false
}
})
}
_ => false,
};
if include {
self.top_levels.push((stmt, module_name.clone(), module.clone()));
}
}
Ok(())
}
fn report_modinit(
arg_names: &[String],
method_name: &str,
resolver: Arc<dyn SymbolResolver + Send + Sync>,
top_level_defs: &[Arc<RwLock<TopLevelDef>>],
unifier: &mut Unifier,
primitives: &PrimitiveStore,
) -> Option<String> {
let base_ty =
match resolver.get_symbol_type(unifier, top_level_defs, primitives, "base".into()) {
Ok(ty) => ty,
Err(e) => return Some(format!("type error inside object launching kernel: {}", e)),
};
let fun_ty = if method_name.is_empty() {
base_ty
} else if let TypeEnum::TObj { fields, .. } = &*unifier.get_ty(base_ty) {
match fields.get(&(*method_name).into()) {
Some(t) => t.0,
None => {
return Some(format!(
"object launching kernel does not have method `{}`",
method_name
))
}
}
} else {
return Some("cannot launch kernel by calling a non-callable".into());
};
if let TypeEnum::TFunc(FunSignature { args, .. }) = &*unifier.get_ty(fun_ty) {
if arg_names.len() > args.len() {
return Some(format!(
"launching kernel function with too many arguments (expect {}, found {})",
args.len(),
arg_names.len(),
));
}
for (i, FuncArg { ty, default_value, name }) in args.iter().enumerate() {
let in_name = match arg_names.get(i) {
Some(n) => n,
None if default_value.is_none() => {
return Some(format!(
"argument `{}` not provided when launching kernel function",
name
))
}
_ => break,
};
let in_ty = match resolver.get_symbol_type(
unifier,
top_level_defs,
primitives,
in_name.clone().into(),
) {
Ok(t) => t,
Err(e) => {
return Some(format!(
"type error ({}) at parameter #{} when calling kernel function",
e, i
))
}
};
if let Err(e) = unifier.unify(in_ty, *ty) {
return Some(format!(
"type error ({}) at parameter #{} when calling kernel function",
e.to_display(unifier).to_string(),
i
));
}
}
} else {
return Some("cannot launch kernel by calling a non-callable".into());
}
None
}
}
fn add_exceptions(
composer: &mut TopLevelComposer,
builtin_def: &mut HashMap<StrRef, DefinitionId>,
builtin_ty: &mut HashMap<StrRef, Type>,
error_names: &[&str]
) -> Vec<Type> {
let mut types = Vec::new();
// note: this is only for builtin exceptions, i.e. the exception name is "0:{exn}"
for name in error_names {
let def_id = composer.definition_ast_list.len();
let (exception_fn, exception_class, exception_cons, exception_type) = get_exn_constructor(
name,
// class id
def_id,
// constructor id
def_id + 1,
&mut composer.unifier,
&composer.primitives_ty
);
composer.definition_ast_list.push((Arc::new(RwLock::new(exception_class)), None));
composer.definition_ast_list.push((Arc::new(RwLock::new(exception_fn)), None));
builtin_ty.insert((*name).into(), exception_cons);
builtin_def.insert((*name).into(), DefinitionId(def_id));
types.push(exception_type);
}
types
}
#[pymethods]
impl Nac3 {
#[new]
fn new(isa: &str, py: Python) -> PyResult<Self> {
let isa = match isa {
"host" => Isa::Host,
"rv32g" => Isa::RiscV32G,
"rv32ima" => Isa::RiscV32IMA,
"cortexa9" => Isa::CortexA9,
_ => return Err(exceptions::PyValueError::new_err("invalid ISA")),
};
let time_fns: &(dyn TimeFns + Sync) = match isa {
Isa::Host => &timeline::EXTERN_TIME_FNS,
Isa::RiscV32G => &timeline::NOW_PINNING_TIME_FNS_64,
Isa::RiscV32IMA => &timeline::NOW_PINNING_TIME_FNS,
Isa::CortexA9 => &timeline::EXTERN_TIME_FNS,
};
let primitive: PrimitiveStore = TopLevelComposer::make_primitives().0;
let builtins = vec![
(
"now_mu".into(),
FunSignature { args: vec![], ret: primitive.int64, vars: HashMap::new() },
Arc::new(GenCall::new(Box::new(move |ctx, _, _, _, _| {
Ok(Some(time_fns.emit_now_mu(ctx)))
}))),
),
(
"at_mu".into(),
FunSignature {
args: vec![FuncArg {
name: "t".into(),
ty: primitive.int64,
default_value: None,
}],
ret: primitive.none,
vars: HashMap::new(),
},
Arc::new(GenCall::new(Box::new(move |ctx, _, fun, args, generator| {
let arg_ty = fun.0.args[0].ty;
let arg = args[0].1.clone().to_basic_value_enum(ctx, generator, arg_ty).unwrap();
time_fns.emit_at_mu(ctx, arg);
Ok(None)
}))),
),
(
"delay_mu".into(),
FunSignature {
args: vec![FuncArg {
name: "dt".into(),
ty: primitive.int64,
default_value: None,
}],
ret: primitive.none,
vars: HashMap::new(),
},
Arc::new(GenCall::new(Box::new(move |ctx, _, fun, args, generator| {
let arg_ty = fun.0.args[0].ty;
let arg = args[0].1.clone().to_basic_value_enum(ctx, generator, arg_ty).unwrap();
time_fns.emit_delay_mu(ctx, arg);
Ok(None)
}))),
),
];
let builtins_mod = PyModule::import(py, "builtins").unwrap();
let id_fn = builtins_mod.getattr("id").unwrap();
let numpy_mod = PyModule::import(py, "numpy").unwrap();
let typing_mod = PyModule::import(py, "typing").unwrap();
let types_mod = PyModule::import(py, "types").unwrap();
let get_id = |x| id_fn.call1((x,)).unwrap().extract().unwrap();
let get_attr_id = |obj: &PyModule, attr| id_fn.call1((obj.getattr(attr).unwrap(),))
.unwrap().extract().unwrap();
let primitive_ids = PrimitivePythonId {
virtual_id: get_id(
builtins_mod
.getattr("globals")
.unwrap()
.call0()
.unwrap()
.get_item("virtual")
.unwrap(
)),
generic_alias: (
get_attr_id(typing_mod, "_GenericAlias"),
get_attr_id(types_mod, "GenericAlias"),
),
none: id_fn
.call1((builtins_mod
.getattr("globals")
.unwrap()
.call0()
.unwrap()
.get_item("none")
.unwrap(),))
.unwrap()
.extract()
.unwrap(),
typevar: get_attr_id(typing_mod, "TypeVar"),
int: get_attr_id(builtins_mod, "int"),
int32: get_attr_id(numpy_mod, "int32"),
int64: get_attr_id(numpy_mod, "int64"),
uint32: get_attr_id(numpy_mod, "uint32"),
uint64: get_attr_id(numpy_mod, "uint64"),
bool: get_attr_id(builtins_mod, "bool"),
float: get_attr_id(builtins_mod, "float"),
float64: get_attr_id(numpy_mod, "float64"),
list: get_attr_id(builtins_mod, "list"),
tuple: get_attr_id(builtins_mod, "tuple"),
exception: get_attr_id(builtins_mod, "Exception"),
option: id_fn
.call1((builtins_mod
.getattr("globals")
.unwrap()
.call0()
.unwrap()
.get_item("Option")
.unwrap(),))
.unwrap()
.extract()
.unwrap(),
};
let working_directory = tempfile::Builder::new().prefix("nac3-").tempdir().unwrap();
fs::write(working_directory.path().join("kernel.ld"), include_bytes!("kernel.ld")).unwrap();
Ok(Nac3 {
isa,
time_fns,
primitive,
builtins,
primitive_ids,
top_levels: Default::default(),
pyid_to_def: Default::default(),
working_directory,
string_store: Default::default(),
exception_ids: Default::default(),
deferred_eval_store: DeferredEvaluationStore::new(),
})
}
fn analyze(&mut self, functions: &PySet, classes: &PySet) -> PyResult<()> {
let (modules, class_ids) =
Python::with_gil(|py| -> PyResult<(HashMap<u64, PyObject>, HashSet<u64>)> {
let mut modules: HashMap<u64, PyObject> = HashMap::new();
let mut class_ids: HashSet<u64> = HashSet::new();
let id_fn = PyModule::import(py, "builtins")?.getattr("id")?;
let getmodule_fn = PyModule::import(py, "inspect")?.getattr("getmodule")?;
for function in functions.iter() {
let module = getmodule_fn.call1((function,))?.extract()?;
modules.insert(id_fn.call1((&module,))?.extract()?, module);
}
for class in classes.iter() {
let module = getmodule_fn.call1((class,))?.extract()?;
modules.insert(id_fn.call1((&module,))?.extract()?, module);
class_ids.insert(id_fn.call1((class,))?.extract()?);
}
Ok((modules, class_ids))
})?;
for module in modules.into_values() {
self.register_module(module, &class_ids)?;
}
Ok(())
}
fn compile_method_to_file(
&mut self,
obj: &PyAny,
method_name: &str,
args: Vec<&PyAny>,
filename: &str,
embedding_map: &PyAny,
py: Python,
) -> PyResult<()> {
let (mut composer, mut builtins_def, mut builtins_ty) = TopLevelComposer::new(
self.builtins.clone(),
ComposerConfig { kernel_ann: Some("Kernel"), kernel_invariant_ann: "KernelInvariant" },
);
let builtins = PyModule::import(py, "builtins")?;
let typings = PyModule::import(py, "typing")?;
let id_fn = builtins.getattr("id")?;
let issubclass = builtins.getattr("issubclass")?;
let exn_class = builtins.getattr("Exception")?;
let store_obj = embedding_map.getattr("store_object").unwrap().to_object(py);
let store_str = embedding_map.getattr("store_str").unwrap().to_object(py);
let store_fun = embedding_map.getattr("store_function").unwrap().to_object(py);
let host_attributes = embedding_map.getattr("attributes_writeback").unwrap().to_object(py);
let global_value_ids: Arc<RwLock<HashMap<_, _>>> = Arc::new(RwLock::new(HashMap::new()));
let helper = PythonHelper {
id_fn: builtins.getattr("id").unwrap().to_object(py),
len_fn: builtins.getattr("len").unwrap().to_object(py),
type_fn: builtins.getattr("type").unwrap().to_object(py),
origin_ty_fn: typings.getattr("get_origin").unwrap().to_object(py),
args_ty_fn: typings.getattr("get_args").unwrap().to_object(py),
store_obj: store_obj.clone(),
store_str,
};
let pyid_to_type = Arc::new(RwLock::new(HashMap::<u64, Type>::new()));
let exception_names = [
"ZeroDivisionError",
"IndexError",
"ValueError",
"RuntimeError",
"AssertionError",
"KeyError",
"NotImplementedError",
"OverflowError",
"IOError",
"UnwrapNoneError",
];
add_exceptions(&mut composer, &mut builtins_def, &mut builtins_ty, &exception_names);
let mut module_to_resolver_cache: HashMap<u64, _> = HashMap::new();
let mut rpc_ids = vec![];
for (stmt, path, module) in self.top_levels.iter() {
let py_module: &PyAny = module.extract(py)?;
let module_id: u64 = id_fn.call1((py_module,))?.extract()?;
let helper = helper.clone();
let class_obj;
if let StmtKind::ClassDef { name, .. } = &stmt.node {
let class = py_module.getattr(name.to_string()).unwrap();
if issubclass.call1((class, exn_class)).unwrap().extract().unwrap() &&
class.getattr("artiq_builtin").is_err() {
class_obj = Some(class);
} else {
class_obj = None;
}
} else {
class_obj = None;
}
let (name_to_pyid, resolver) =
module_to_resolver_cache.get(&module_id).cloned().unwrap_or_else(|| {
let mut name_to_pyid: HashMap<StrRef, u64> = HashMap::new();
let members: &PyDict =
py_module.getattr("__dict__").unwrap().cast_as().unwrap();
for (key, val) in members.iter() {
let key: &str = key.extract().unwrap();
let val = id_fn.call1((val,)).unwrap().extract().unwrap();
name_to_pyid.insert(key.into(), val);
}
let resolver = Arc::new(Resolver(Arc::new(InnerResolver {
id_to_type: builtins_ty.clone().into(),
id_to_def: builtins_def.clone().into(),
pyid_to_def: self.pyid_to_def.clone(),
pyid_to_type: pyid_to_type.clone(),
primitive_ids: self.primitive_ids.clone(),
global_value_ids: global_value_ids.clone(),
class_names: Default::default(),
name_to_pyid: name_to_pyid.clone(),
module: module.clone(),
id_to_pyval: Default::default(),
id_to_primitive: Default::default(),
field_to_val: Default::default(),
helper,
string_store: self.string_store.clone(),
exception_ids: self.exception_ids.clone(),
deferred_eval_store: self.deferred_eval_store.clone(),
})))
as Arc<dyn SymbolResolver + Send + Sync>;
let name_to_pyid = Rc::new(name_to_pyid);
module_to_resolver_cache
.insert(module_id, (name_to_pyid.clone(), resolver.clone()));
(name_to_pyid, resolver)
});
let (name, def_id, ty) = composer
.register_top_level(stmt.clone(), Some(resolver.clone()), path.clone())
.map_err(|e| {
CompileError::new_err(format!(
"compilation failed\n----------\n{}",
e
))
})?;
if let Some(class_obj) = class_obj {
self.exception_ids.write().insert(def_id.0, store_obj.call1(py, (class_obj, ))?.extract(py)?);
}
match &stmt.node {
StmtKind::FunctionDef { decorator_list, .. } => {
if decorator_list.iter().any(|decorator| matches!(decorator.node, ExprKind::Name { id, .. } if id == "rpc".into())) {
store_fun.call1(py, (def_id.0.into_py(py), module.getattr(py, name.to_string()).unwrap())).unwrap();
rpc_ids.push((None, def_id));
}
}
StmtKind::ClassDef { name, body, .. } => {
let class_obj = module.getattr(py, name.to_string()).unwrap();
for stmt in body.iter() {
if let StmtKind::FunctionDef { name, decorator_list, .. } = &stmt.node {
if decorator_list.iter().any(|decorator| matches!(decorator.node, ExprKind::Name { id, .. } if id == "rpc".into())) {
rpc_ids.push((Some((class_obj.clone(), *name)), def_id));
}
}
}
}
_ => ()
}
let id = *name_to_pyid.get(&name).unwrap();
self.pyid_to_def.write().insert(id, def_id);
{
let mut pyid_to_ty = pyid_to_type.write();
if let Some(ty) = ty {
pyid_to_ty.insert(id, ty);
}
}
}
let id_fun = PyModule::import(py, "builtins")?.getattr("id")?;
let mut name_to_pyid: HashMap<StrRef, u64> = HashMap::new();
let module = PyModule::new(py, "tmp")?;
module.add("base", obj)?;
name_to_pyid.insert("base".into(), id_fun.call1((obj,))?.extract()?);
let mut arg_names = vec![];
for (i, arg) in args.into_iter().enumerate() {
let name = format!("tmp{}", i);
module.add(&name, arg)?;
name_to_pyid.insert(name.clone().into(), id_fun.call1((arg,))?.extract()?);
arg_names.push(name);
}
let synthesized = if method_name.is_empty() {
format!("def __modinit__():\n base({})", arg_names.join(", "))
} else {
format!("def __modinit__():\n base.{}({})", method_name, arg_names.join(", "))
};
let mut synthesized =
parse_program(&synthesized, "__nac3_synthesized_modinit__".to_string().into()).unwrap();
let inner_resolver = Arc::new(InnerResolver {
id_to_type: builtins_ty.clone().into(),
id_to_def: builtins_def.clone().into(),
pyid_to_def: self.pyid_to_def.clone(),
pyid_to_type: pyid_to_type.clone(),
primitive_ids: self.primitive_ids.clone(),
global_value_ids: global_value_ids.clone(),
class_names: Default::default(),
id_to_pyval: Default::default(),
id_to_primitive: Default::default(),
field_to_val: Default::default(),
name_to_pyid,
module: module.to_object(py),
helper,
string_store: self.string_store.clone(),
exception_ids: self.exception_ids.clone(),
deferred_eval_store: self.deferred_eval_store.clone(),
});
let resolver = Arc::new(Resolver(inner_resolver.clone())) as Arc<dyn SymbolResolver + Send + Sync>;
let (_, def_id, _) = composer
.register_top_level(synthesized.pop().unwrap(), Some(resolver.clone()), "".into())
.unwrap();
let fun_signature =
FunSignature { args: vec![], ret: self.primitive.none, vars: HashMap::new() };
let mut store = ConcreteTypeStore::new();
let mut cache = HashMap::new();
let signature =
store.from_signature(&mut composer.unifier, &self.primitive, &fun_signature, &mut cache);
let signature = store.add_cty(signature);
if let Err(e) = composer.start_analysis(true) {
// report error of __modinit__ separately
if !e.contains("__nac3_synthesized_modinit__") {
return Err(CompileError::new_err(format!(
"compilation failed\n----------\n{}",
e
)));
} else {
let msg = Self::report_modinit(
&arg_names,
method_name,
resolver.clone(),
&composer.extract_def_list(),
&mut composer.unifier,
&self.primitive,
);
return Err(CompileError::new_err(msg.unwrap_or(e)));
}
}
let top_level = Arc::new(composer.make_top_level_context());
{
let rpc_codegen = rpc_codegen_callback();
let defs = top_level.definitions.read();
for (class_data, id) in rpc_ids.iter() {
let mut def = defs[id.0].write();
match &mut *def {
TopLevelDef::Function { codegen_callback, .. } => {
*codegen_callback = Some(rpc_codegen.clone());
}
TopLevelDef::Class { methods, .. } => {
let (class_def, method_name) = class_data.as_ref().unwrap();
for (name, _, id) in methods.iter() {
if name != method_name {
continue;
}
if let TopLevelDef::Function { codegen_callback, .. } =
&mut *defs[id.0].write()
{
*codegen_callback = Some(rpc_codegen.clone());
store_fun
.call1(
py,
(
id.0.into_py(py),
class_def.getattr(py, name.to_string()).unwrap(),
),
)
.unwrap();
}
}
}
}
}
}
let instance = {
let defs = top_level.definitions.read();
let mut definition = defs[def_id.0].write();
if let TopLevelDef::Function { instance_to_stmt, instance_to_symbol, .. } =
&mut *definition
{
instance_to_symbol.insert("".to_string(), "__modinit__".into());
instance_to_stmt[""].clone()
} else {
unreachable!()
}
};
let task = CodeGenTask {
subst: Default::default(),
symbol_name: "__modinit__".to_string(),
body: instance.body,
signature,
resolver: resolver.clone(),
store,
unifier_index: instance.unifier_id,
calls: instance.calls,
id: 0,
};
let mut store = ConcreteTypeStore::new();
let mut cache = HashMap::new();
let signature =
store.from_signature(&mut composer.unifier, &self.primitive, &fun_signature, &mut cache);
let signature = store.add_cty(signature);
let attributes_writeback_task = CodeGenTask {
subst: Default::default(),
symbol_name: "attributes_writeback".to_string(),
body: Arc::new(Default::default()),
signature,
resolver,
store,
unifier_index: instance.unifier_id,
calls: Arc::new(Default::default()),
id: 0,
};
let isa = self.isa;
let working_directory = self.working_directory.path().to_owned();
let membuffers: Arc<Mutex<Vec<Vec<u8>>>> = Default::default();
let membuffer = membuffers.clone();
let f = Arc::new(WithCall::new(Box::new(move |module| {
let buffer = module.write_bitcode_to_memory();
let buffer = buffer.as_slice().into();
membuffer.lock().push(buffer);
})));
let size_t = if self.isa == Isa::Host { 64 } else { 32 };
let thread_names: Vec<String> = (0..4).map(|_| "main".to_string()).collect();
let threads: Vec<_> = thread_names
.iter()
.map(|s| Box::new(ArtiqCodeGenerator::new(s.to_string(), size_t, self.time_fns)))
.collect();
let membuffer = membuffers.clone();
py.allow_threads(|| {
let (registry, handles) = WorkerRegistry::create_workers(threads, top_level.clone(), f);
registry.add_task(task);
registry.wait_tasks_complete(handles);
let mut generator = ArtiqCodeGenerator::new("attributes_writeback".to_string(), size_t, self.time_fns);
let context = inkwell::context::Context::create();
let module = context.create_module("attributes_writeback");
let builder = context.create_builder();
let (_, module, _) = gen_func_impl(&context, &mut generator, &registry, builder, module,
attributes_writeback_task, |generator, ctx| {
attributes_writeback(ctx, generator, inner_resolver.as_ref(), host_attributes)
}).unwrap();
let buffer = module.write_bitcode_to_memory();
let buffer = buffer.as_slice().into();
membuffer.lock().push(buffer);
});
let context = inkwell::context::Context::create();
let buffers = membuffers.lock();
let main = context
.create_module_from_ir(MemoryBuffer::create_from_memory_range(&buffers[0], "main"))
.unwrap();
for buffer in buffers.iter().skip(1) {
let other = context
.create_module_from_ir(MemoryBuffer::create_from_memory_range(buffer, "main"))
.unwrap();
main.link_in_module(other)
.map_err(|err| CompileError::new_err(err.to_string()))?;
}
let builder = context.create_builder();
let modinit_return = main.get_function("__modinit__").unwrap().get_last_basic_block().unwrap().get_terminator().unwrap();
builder.position_before(&modinit_return);
builder.build_call(main.get_function("attributes_writeback").unwrap(), &[], "attributes_writeback");
main.link_in_module(load_irrt(&context))
.map_err(|err| CompileError::new_err(err.to_string()))?;
let mut function_iter = main.get_first_function();
while let Some(func) = function_iter {
if func.count_basic_blocks() > 0 && func.get_name().to_str().unwrap() != "__modinit__" {
func.set_linkage(inkwell::module::Linkage::Private);
}
function_iter = func.get_next_function();
}
let builder = PassManagerBuilder::create();
builder.set_optimization_level(OptimizationLevel::Aggressive);
let passes = PassManager::create(());
builder.set_inliner_with_threshold(255);
builder.populate_module_pass_manager(&passes);
passes.run_on(&main);
let (triple, features) = match isa {
Isa::Host => (
TargetMachine::get_default_triple(),
TargetMachine::get_host_cpu_features().to_string(),
),
Isa::RiscV32G => {
(TargetTriple::create("riscv32-unknown-linux"), "+a,+m,+f,+d".to_string())
}
Isa::RiscV32IMA => (TargetTriple::create("riscv32-unknown-linux"), "+a,+m".to_string()),
Isa::CortexA9 => (
TargetTriple::create("armv7-unknown-linux-gnueabihf"),
"+dsp,+fp16,+neon,+vfp3".to_string(),
),
};
let target =
Target::from_triple(&triple).expect("couldn't create target from target triple");
let target_machine = target
.create_target_machine(
&triple,
"",
&features,
OptimizationLevel::Default,
RelocMode::PIC,
CodeModel::Default,
)
.expect("couldn't create target machine");
target_machine
.write_to_file(&main, FileType::Object, &working_directory.join("module.o"))
.expect("couldn't write module to file");
let mut linker_args = vec![
"-shared".to_string(),
"--eh-frame-hdr".to_string(),
"-x".to_string(),
"-o".to_string(),
filename.to_string(),
working_directory.join("module.o").to_string_lossy().to_string(),
];
if isa != Isa::Host {
linker_args.push(
"-T".to_string()
+ self.working_directory.path().join("kernel.ld").to_str().unwrap(),
);
}
#[cfg(not(windows))]
let lld_command = "ld.lld";
#[cfg(windows)]
let lld_command = "ld.lld.exe";
if let Ok(linker_status) = Command::new(lld_command).args(linker_args).status() {
if !linker_status.success() {
return Err(CompileError::new_err("failed to start linker"));
}
} else {
return Err(CompileError::new_err(
"linker returned non-zero status code",
));
}
Ok(())
}
fn compile_method_to_mem(
&mut self,
obj: &PyAny,
method_name: &str,
args: Vec<&PyAny>,
embedding_map: &PyAny,
py: Python,
) -> PyResult<PyObject> {
let filename_path = self.working_directory.path().join("module.elf");
let filename = filename_path.to_str().unwrap();
self.compile_method_to_file(obj, method_name, args, filename, embedding_map, py)?;
Ok(PyBytes::new(py, &fs::read(filename).unwrap()).into())
}
}
#[cfg(feature = "init-llvm-profile")]
extern "C" {
fn __llvm_profile_initialize();
}
#[pymodule]
fn nac3artiq(py: Python, m: &PyModule) -> PyResult<()> {
#[cfg(feature = "init-llvm-profile")]
unsafe {
__llvm_profile_initialize();
}
Target::initialize_all(&InitializationConfig::default());
m.add("CompileError", py.get_type::<CompileError>())?;
m.add_class::<Nac3>()?;
Ok(())
}

File diff suppressed because it is too large Load Diff

304
nac3artiq/src/timeline.rs Normal file
View File

@ -0,0 +1,304 @@
use inkwell::{values::BasicValueEnum, AddressSpace, AtomicOrdering};
use nac3core::codegen::CodeGenContext;
pub trait TimeFns {
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 {}
// 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, '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::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_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!();
}
}
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);
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",
);
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(2, 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!();
}
}
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::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_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 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,
"now_trunc",
);
let time_lo = ctx.builder.build_int_truncate(time, i32_type, "now_trunc");
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!();
};
}
}
pub static NOW_PINNING_TIME_FNS_64: NowPinningTimeFns64 = NowPinningTimeFns64 {};
pub struct NowPinningTimeFns {}
impl TimeFns for NowPinningTimeFns {
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");
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_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, '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);
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",
);
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_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!();
}
}
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!();
}
}
}
pub static NOW_PINNING_TIME_FNS: NowPinningTimeFns = NowPinningTimeFns {};
pub struct ExternTimeFns {}
impl TimeFns for ExternTimeFns {
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").try_as_basic_value().left().unwrap()
}
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",
ctx.ctx.void_type().fn_type(&[ctx.ctx.i64_type().into()], false),
None,
)
});
ctx.builder.build_call(at_mu, &[t.into()], "at_mu");
}
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",
ctx.ctx.void_type().fn_type(&[ctx.ctx.i64_type().into()], false),
None,
)
});
ctx.builder.build_call(delay_mu, &[dt.into()], "delay_mu");
}
}
pub static EXTERN_TIME_FNS: ExternTimeFns = ExternTimeFns {};

16
nac3ast/Cargo.toml Normal file
View File

@ -0,0 +1,16 @@
[package]
name = "nac3ast"
version = "0.1.0"
authors = ["RustPython Team", "M-Labs"]
edition = "2018"
[features]
default = ["constant-optimization", "fold"]
constant-optimization = ["fold"]
fold = []
[dependencies]
lazy_static = "1.4.0"
parking_lot = "0.11.1"
string-interner = "0.13.0"
fxhash = "0.2.1"

127
nac3ast/Python.asdl Normal file
View File

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

385
nac3ast/asdl.py Normal file
View File

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

609
nac3ast/asdl_rs.py Executable file
View File

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

1265
nac3ast/src/ast_gen.rs Normal file

File diff suppressed because it is too large Load Diff

213
nac3ast/src/constant.rs Normal file
View File

@ -0,0 +1,213 @@
#[derive(Clone, Debug, PartialEq)]
pub enum Constant {
None,
Bool(bool),
Str(String),
Bytes(Vec<u8>),
Int(i128),
Tuple(Vec<Constant>),
Float(f64),
Complex { real: f64, imag: f64 },
Ellipsis,
}
impl From<String> for Constant {
fn from(s: String) -> Constant {
Self::Str(s)
}
}
impl From<Vec<u8>> for Constant {
fn from(b: Vec<u8>) -> Constant {
Self::Bytes(b)
}
}
impl From<bool> for Constant {
fn from(b: bool) -> Constant {
Self::Bool(b)
}
}
impl From<i32> for Constant {
fn from(i: i32) -> Constant {
Self::Int(i as i128)
}
}
impl From<i64> for Constant {
fn from(i: i64) -> Constant {
Self::Int(i as i128)
}
}
/// Transforms a value prior to formatting it.
#[derive(Copy, Clone, Debug, PartialEq)]
#[repr(u8)]
pub enum ConversionFlag {
/// Converts by calling `str(<value>)`.
Str = b's',
/// Converts by calling `ascii(<value>)`.
Ascii = b'a',
/// Converts by calling `repr(<value>)`.
Repr = b'r',
}
impl ConversionFlag {
pub fn try_from_byte(b: u8) -> Option<Self> {
match b {
b's' => Some(Self::Str),
b'a' => Some(Self::Ascii),
b'r' => Some(Self::Repr),
_ => None,
}
}
}
#[cfg(feature = "constant-optimization")]
#[derive(Default)]
pub struct ConstantOptimizer {
_priv: (),
}
#[cfg(feature = "constant-optimization")]
impl ConstantOptimizer {
#[inline]
pub fn new() -> Self {
Self { _priv: () }
}
}
#[cfg(feature = "constant-optimization")]
impl<U> crate::fold::Fold<U> for ConstantOptimizer {
type TargetU = U;
type Error = std::convert::Infallible;
#[inline]
fn map_user(&mut self, user: U) -> Result<Self::TargetU, Self::Error> {
Ok(user)
}
fn fold_expr(&mut self, node: crate::Expr<U>) -> Result<crate::Expr<U>, Self::Error> {
match node.node {
crate::ExprKind::Tuple { elts, ctx } => {
let elts = elts
.into_iter()
.map(|x| self.fold_expr(x))
.collect::<Result<Vec<_>, _>>()?;
let expr = if elts
.iter()
.all(|e| matches!(e.node, crate::ExprKind::Constant { .. }))
{
let tuple = elts
.into_iter()
.map(|e| match e.node {
crate::ExprKind::Constant { value, .. } => value,
_ => unreachable!(),
})
.collect();
crate::ExprKind::Constant {
value: Constant::Tuple(tuple),
kind: None,
}
} else {
crate::ExprKind::Tuple { elts, ctx }
};
Ok(crate::Expr {
node: expr,
custom: node.custom,
location: node.location,
})
}
_ => crate::fold::fold_expr(self, node),
}
}
}
#[cfg(test)]
mod tests {
#[cfg(feature = "constant-optimization")]
#[test]
fn test_constant_opt() {
use super::*;
use crate::fold::Fold;
use crate::*;
let location = Location::new(0, 0, Default::default());
let custom = ();
let ast = Located {
location,
custom,
node: ExprKind::Tuple {
ctx: ExprContext::Load,
elts: vec![
Located {
location,
custom,
node: ExprKind::Constant {
value: 1.into(),
kind: None,
},
},
Located {
location,
custom,
node: ExprKind::Constant {
value: 2.into(),
kind: None,
},
},
Located {
location,
custom,
node: ExprKind::Tuple {
ctx: ExprContext::Load,
elts: vec![
Located {
location,
custom,
node: ExprKind::Constant {
value: 3.into(),
kind: None,
},
},
Located {
location,
custom,
node: ExprKind::Constant {
value: 4.into(),
kind: None,
},
},
Located {
location,
custom,
node: ExprKind::Constant {
value: 5.into(),
kind: None,
},
},
],
},
},
],
},
};
let new_ast = ConstantOptimizer::new()
.fold_expr(ast)
.unwrap_or_else(|e| match e {});
assert_eq!(
new_ast,
Located {
location,
custom,
node: ExprKind::Constant {
value: Constant::Tuple(vec![
1.into(),
2.into(),
Constant::Tuple(vec![
3.into(),
4.into(),
5.into(),
])
]),
kind: None
},
}
);
}
}

View File

@ -0,0 +1,74 @@
use crate::constant;
use crate::fold::Fold;
use crate::StrRef;
pub(crate) trait Foldable<T, U> {
type Mapped;
fn fold<F: Fold<T, TargetU = U> + ?Sized>(
self,
folder: &mut F,
) -> Result<Self::Mapped, F::Error>;
}
impl<T, U, X> Foldable<T, U> for Vec<X>
where
X: Foldable<T, U>,
{
type Mapped = Vec<X::Mapped>;
fn fold<F: Fold<T, TargetU = U> + ?Sized>(
self,
folder: &mut F,
) -> Result<Self::Mapped, F::Error> {
self.into_iter().map(|x| x.fold(folder)).collect()
}
}
impl<T, U, X> Foldable<T, U> for Option<X>
where
X: Foldable<T, U>,
{
type Mapped = Option<X::Mapped>;
fn fold<F: Fold<T, TargetU = U> + ?Sized>(
self,
folder: &mut F,
) -> Result<Self::Mapped, F::Error> {
self.map(|x| x.fold(folder)).transpose()
}
}
impl<T, U, X> Foldable<T, U> for Box<X>
where
X: Foldable<T, U>,
{
type Mapped = Box<X::Mapped>;
fn fold<F: Fold<T, TargetU = U> + ?Sized>(
self,
folder: &mut F,
) -> Result<Self::Mapped, F::Error> {
(*self).fold(folder).map(Box::new)
}
}
macro_rules! simple_fold {
($($t:ty),+$(,)?) => {
$(impl<T, U> $crate::fold_helpers::Foldable<T, U> for $t {
type Mapped = Self;
#[inline]
fn fold<F: Fold<T, TargetU = U> + ?Sized>(
self,
_folder: &mut F,
) -> Result<Self::Mapped, F::Error> {
Ok(self)
}
})+
};
}
simple_fold!(
usize,
String,
bool,
StrRef,
constant::Constant,
constant::ConversionFlag
);

53
nac3ast/src/impls.rs Normal file
View File

@ -0,0 +1,53 @@
use crate::{Constant, ExprKind};
impl<U> ExprKind<U> {
/// Returns a short name for the node suitable for use in error messages.
pub fn name(&self) -> &'static str {
match self {
ExprKind::BoolOp { .. } | ExprKind::BinOp { .. } | ExprKind::UnaryOp { .. } => {
"operator"
}
ExprKind::Subscript { .. } => "subscript",
ExprKind::Await { .. } => "await expression",
ExprKind::Yield { .. } | ExprKind::YieldFrom { .. } => "yield expression",
ExprKind::Compare { .. } => "comparison",
ExprKind::Attribute { .. } => "attribute",
ExprKind::Call { .. } => "function call",
ExprKind::Constant { value, .. } => match value {
Constant::Str(_)
| Constant::Int(_)
| Constant::Float(_)
| Constant::Complex { .. }
| Constant::Bytes(_) => "literal",
Constant::Tuple(_) => "tuple",
Constant::Bool(_) | Constant::None => "keyword",
Constant::Ellipsis => "ellipsis",
},
ExprKind::List { .. } => "list",
ExprKind::Tuple { .. } => "tuple",
ExprKind::Dict { .. } => "dict display",
ExprKind::Set { .. } => "set display",
ExprKind::ListComp { .. } => "list comprehension",
ExprKind::DictComp { .. } => "dict comprehension",
ExprKind::SetComp { .. } => "set comprehension",
ExprKind::GeneratorExp { .. } => "generator expression",
ExprKind::Starred { .. } => "starred",
ExprKind::Slice { .. } => "slice",
ExprKind::JoinedStr { values } => {
if values
.iter()
.any(|e| matches!(e.node, ExprKind::JoinedStr { .. }))
{
"f-string expression"
} else {
"literal"
}
}
ExprKind::FormattedValue { .. } => "f-string expression",
ExprKind::Name { .. } => "name",
ExprKind::Lambda { .. } => "lambda",
ExprKind::IfExp { .. } => "conditional expression",
ExprKind::NamedExpr { .. } => "named expression",
}
}
}

14
nac3ast/src/lib.rs Normal file
View File

@ -0,0 +1,14 @@
#[macro_use]
extern crate lazy_static;
mod ast_gen;
mod constant;
#[cfg(feature = "fold")]
mod fold_helpers;
mod impls;
mod location;
pub use ast_gen::*;
pub use location::{Location, FileName};
pub type Suite<U = ()> = Vec<Stmt<U>>;

94
nac3ast/src/location.rs Normal file
View File

@ -0,0 +1,94 @@
//! Datatypes to support source location information.
use crate::ast_gen::StrRef;
use std::fmt;
#[derive(Clone, Copy, Debug, PartialEq)]
pub struct FileName(pub StrRef);
impl Default for FileName {
fn default() -> Self {
FileName("unknown".into())
}
}
impl From<String> for FileName {
fn from(s: String) -> Self {
FileName(s.into())
}
}
/// A location somewhere in the sourcecode.
#[derive(Clone, Copy, Debug, Default, PartialEq)]
pub struct Location {
pub row: usize,
pub column: usize,
pub file: FileName
}
impl fmt::Display for Location {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
write!(f, "{}: line {} column {}", self.file.0, self.row, self.column)
}
}
impl Location {
pub fn visualize<'a>(
&self,
line: &'a str,
desc: impl fmt::Display + 'a,
) -> impl fmt::Display + 'a {
struct Visualize<'a, D: fmt::Display> {
loc: Location,
line: &'a str,
desc: D,
}
impl<D: fmt::Display> fmt::Display for Visualize<'_, D> {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
write!(
f,
"{}\n{}\n{arrow:>pad$}",
self.desc,
self.line,
pad = self.loc.column,
arrow = "",
)
}
}
Visualize {
loc: *self,
line,
desc,
}
}
}
impl Location {
pub fn new(row: usize, column: usize, file: FileName) -> Self {
Location { row, column, file }
}
pub fn row(&self) -> usize {
self.row
}
pub fn column(&self) -> usize {
self.column
}
pub fn reset(&mut self) {
self.row = 1;
self.column = 1;
}
pub fn go_right(&mut self) {
self.column += 1;
}
pub fn go_left(&mut self) {
self.column -= 1;
}
pub fn newline(&mut self) {
self.row += 1;
self.column = 1;
}
}

View File

@ -5,12 +5,21 @@ authors = ["M-Labs"]
edition = "2018" edition = "2018"
[dependencies] [dependencies]
num-bigint = "0.3" itertools = "0.10.1"
num-traits = "0.2" crossbeam = "0.8.1"
inkwell = { git = "https://github.com/TheDan64/inkwell", branch = "master", features = ["llvm10-0"] } parking_lot = "0.11.1"
rustpython-parser = { git = "https://github.com/RustPython/RustPython", branch = "master" } rayon = "1.5.1"
indoc = "1.0" nac3parser = { path = "../nac3parser" }
[dependencies.inkwell]
version = "0.1.0-beta.4"
default-features = false
features = ["llvm13-0", "target-x86", "target-arm", "target-riscv", "no-libffi-linking"]
[dev-dependencies] [dev-dependencies]
test-case = "1.2.0" test-case = "1.2.0"
indoc = "1.0" indoc = "1.0"
insta = "=1.11.0"
[build-dependencies]
regex = "1"

72
nac3core/build.rs Normal file
View File

@ -0,0 +1,72 @@
use regex::Regex;
use std::{
env,
fs::File,
io::Write,
path::Path,
process::{Command, Stdio},
};
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.
*/
const FLAG: &[&str] = &[
"--target=wasm32",
FILE,
"-O3",
"-emit-llvm",
"-S",
"-Wall",
"-Wextra",
"-o",
"-",
];
let output = Command::new("clang")
.args(FLAG)
.output()
.map(|o| {
assert!(o.status.success(), "{}", std::str::from_utf8(&o.stderr).unwrap());
o
})
.unwrap();
// https://github.com/rust-lang/regex/issues/244
let output = std::str::from_utf8(&output.stdout).unwrap().replace("\r\n", "\n");
let mut filtered_output = String::with_capacity(output.len());
let regex_filter = regex::Regex::new(r"(?ms:^define.*?\}$)|(?m:^declare.*?$)").unwrap();
for f in regex_filter.captures_iter(&output) {
assert!(f.len() == 1);
filtered_output.push_str(&f[0]);
filtered_output.push('\n');
}
let filtered_output = Regex::new("(#\\d+)|(, *![0-9A-Za-z.]+)|(![0-9A-Za-z.]+)|(!\".*?\")")
.unwrap()
.replace_all(&filtered_output, "");
println!("cargo:rerun-if-env-changed=DEBUG_DUMP_IRRT");
if env::var("DEBUG_DUMP_IRRT").is_ok() {
let mut file = File::create(out_path.join("irrt.ll")).unwrap();
file.write_all(output.as_bytes()).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")
.stdin(Stdio::piped())
.arg("-o")
.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())
}

View File

@ -0,0 +1,305 @@
use crate::{
symbol_resolver::SymbolValue,
toplevel::DefinitionId,
typecheck::{
type_inferencer::PrimitiveStore,
typedef::{FunSignature, FuncArg, Type, TypeEnum, Unifier},
},
};
use nac3parser::ast::StrRef;
use std::collections::HashMap;
pub struct ConcreteTypeStore {
store: Vec<ConcreteTypeEnum>,
}
#[derive(Copy, Clone, Eq, PartialEq, Hash, Debug)]
pub struct ConcreteType(usize);
#[derive(Clone, Debug)]
pub struct ConcreteFuncArg {
pub name: StrRef,
pub ty: ConcreteType,
pub default_value: Option<SymbolValue>,
}
#[derive(Clone, Debug)]
pub enum Primitive {
Int32,
Int64,
UInt32,
UInt64,
Float,
Bool,
None,
Range,
Str,
Exception,
}
#[derive(Debug)]
pub enum ConcreteTypeEnum {
TPrimitive(Primitive),
TTuple {
ty: Vec<ConcreteType>,
},
TList {
ty: ConcreteType,
},
TObj {
obj_id: DefinitionId,
fields: HashMap<StrRef, (ConcreteType, bool)>,
params: HashMap<u32, ConcreteType>,
},
TVirtual {
ty: ConcreteType,
},
TFunc {
args: Vec<ConcreteFuncArg>,
ret: ConcreteType,
vars: HashMap<u32, ConcreteType>,
},
}
impl ConcreteTypeStore {
pub fn new() -> ConcreteTypeStore {
ConcreteTypeStore {
store: vec![
ConcreteTypeEnum::TPrimitive(Primitive::Int32),
ConcreteTypeEnum::TPrimitive(Primitive::Int64),
ConcreteTypeEnum::TPrimitive(Primitive::Float),
ConcreteTypeEnum::TPrimitive(Primitive::Bool),
ConcreteTypeEnum::TPrimitive(Primitive::None),
ConcreteTypeEnum::TPrimitive(Primitive::Range),
ConcreteTypeEnum::TPrimitive(Primitive::Str),
ConcreteTypeEnum::TPrimitive(Primitive::Exception),
ConcreteTypeEnum::TPrimitive(Primitive::UInt32),
ConcreteTypeEnum::TPrimitive(Primitive::UInt64),
],
}
}
pub fn get(&self, cty: ConcreteType) -> &ConcreteTypeEnum {
&self.store[cty.0]
}
pub fn from_signature(
&mut self,
unifier: &mut Unifier,
primitives: &PrimitiveStore,
signature: &FunSignature,
cache: &mut HashMap<Type, Option<ConcreteType>>,
) -> ConcreteTypeEnum {
ConcreteTypeEnum::TFunc {
args: signature
.args
.iter()
.map(|arg| ConcreteFuncArg {
name: arg.name,
ty: self.from_unifier_type(unifier, primitives, arg.ty, cache),
default_value: arg.default_value.clone(),
})
.collect(),
ret: self.from_unifier_type(unifier, primitives, signature.ret, cache),
vars: signature
.vars
.iter()
.map(|(id, ty)| (*id, self.from_unifier_type(unifier, primitives, *ty, cache)))
.collect(),
}
}
pub fn from_unifier_type(
&mut self,
unifier: &mut Unifier,
primitives: &PrimitiveStore,
ty: Type,
cache: &mut HashMap<Type, Option<ConcreteType>>,
) -> ConcreteType {
let ty = unifier.get_representative(ty);
if unifier.unioned(ty, primitives.int32) {
ConcreteType(0)
} else if unifier.unioned(ty, primitives.int64) {
ConcreteType(1)
} else if unifier.unioned(ty, primitives.float) {
ConcreteType(2)
} else if unifier.unioned(ty, primitives.bool) {
ConcreteType(3)
} else if unifier.unioned(ty, primitives.none) {
ConcreteType(4)
} else if unifier.unioned(ty, primitives.range) {
ConcreteType(5)
} else if unifier.unioned(ty, primitives.str) {
ConcreteType(6)
} else if unifier.unioned(ty, primitives.exception) {
ConcreteType(7)
} else if unifier.unioned(ty, primitives.uint32) {
ConcreteType(8)
} else if unifier.unioned(ty, primitives.uint64) {
ConcreteType(9)
} else if let Some(cty) = cache.get(&ty) {
if let Some(cty) = cty {
*cty
} else {
let index = self.store.len();
// placeholder
self.store.push(ConcreteTypeEnum::TPrimitive(Primitive::Int32));
let result = ConcreteType(index);
cache.insert(ty, Some(result));
result
}
} else {
cache.insert(ty, None);
let ty_enum = unifier.get_ty(ty);
let result = match &*ty_enum {
TypeEnum::TTuple { ty } => ConcreteTypeEnum::TTuple {
ty: ty
.iter()
.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
.iter()
.filter_map(|(name, ty)| {
// here we should not have type vars, but some partial instantiated
// class methods can still have uninstantiated type vars, so
// filter out all the methods, as this will not affect codegen
if let TypeEnum::TFunc(..) = &*unifier.get_ty(ty.0) {
None
} else {
Some((
*name,
(
self.from_unifier_type(unifier, primitives, ty.0, cache),
ty.1,
),
))
}
})
.collect(),
params: params
.iter()
.map(|(id, ty)| {
(*id, self.from_unifier_type(unifier, primitives, *ty, cache))
})
.collect(),
},
TypeEnum::TVirtual { ty } => ConcreteTypeEnum::TVirtual {
ty: self.from_unifier_type(unifier, primitives, *ty, cache),
},
TypeEnum::TFunc(signature) => {
self.from_signature(unifier, primitives, &*signature, cache)
}
_ => unreachable!(),
};
let index = if let Some(ConcreteType(index)) = cache.get(&ty).unwrap() {
self.store[*index] = result;
*index
} else {
self.store.push(result);
self.store.len() - 1
};
cache.insert(ty, Some(ConcreteType(index)));
ConcreteType(index)
}
}
pub fn to_unifier_type(
&self,
unifier: &mut Unifier,
primitives: &PrimitiveStore,
cty: ConcreteType,
cache: &mut HashMap<ConcreteType, Option<Type>>,
) -> Type {
if let Some(ty) = cache.get_mut(&cty) {
return if let Some(ty) = ty {
*ty
} else {
*ty = Some(unifier.get_dummy_var().0);
ty.unwrap()
};
}
cache.insert(cty, None);
let result = match &self.store[cty.0] {
ConcreteTypeEnum::TPrimitive(primitive) => {
let ty = match primitive {
Primitive::Int32 => primitives.int32,
Primitive::Int64 => primitives.int64,
Primitive::UInt32 => primitives.uint32,
Primitive::UInt64 => primitives.uint64,
Primitive::Float => primitives.float,
Primitive::Bool => primitives.bool,
Primitive::None => primitives.none,
Primitive::Range => primitives.range,
Primitive::Str => primitives.str,
Primitive::Exception => primitives.exception,
};
*cache.get_mut(&cty).unwrap() = Some(ty);
return ty;
}
ConcreteTypeEnum::TTuple { ty } => TypeEnum::TTuple {
ty: ty
.iter()
.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) }
}
ConcreteTypeEnum::TObj { obj_id, fields, params } => TypeEnum::TObj {
obj_id: *obj_id,
fields: fields
.iter()
.map(|(name, cty)| {
(*name, (self.to_unifier_type(unifier, primitives, cty.0, cache), cty.1))
})
.collect::<HashMap<_, _>>(),
params: 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
.iter()
.map(|arg| FuncArg {
name: arg.name,
ty: self.to_unifier_type(unifier, primitives, arg.ty, cache),
default_value: arg.default_value.clone(),
})
.collect(),
ret: self.to_unifier_type(unifier, primitives, *ret, cache),
vars: vars
.iter()
.map(|(id, cty)| (*id, self.to_unifier_type(unifier, primitives, *cty, cache)))
.collect::<HashMap<_, _>>(),
}),
};
let result = unifier.add_ty(result);
if let Some(ty) = cache.get(&cty).unwrap() {
unifier.unify(*ty, result).unwrap();
}
cache.insert(cty, Some(result));
result
}
pub fn add_cty(&mut self, cty: ConcreteTypeEnum) -> ConcreteType {
self.store.push(cty);
ConcreteType(self.store.len() - 1)
}
}
impl Default for ConcreteTypeStore {
fn default() -> Self {
Self::new()
}
}

1554
nac3core/src/codegen/expr.rs Normal file

File diff suppressed because it is too large Load Diff

View File

@ -0,0 +1,209 @@
use crate::{
codegen::{expr::*, stmt::*, CodeGenContext},
symbol_resolver::ValueEnum,
toplevel::{DefinitionId, TopLevelDef},
typecheck::typedef::{FunSignature, Type},
};
use inkwell::{
context::Context,
types::{BasicTypeEnum, IntType},
values::{BasicValueEnum, PointerValue},
};
use nac3parser::ast::{Expr, Stmt, StrRef};
pub trait CodeGenerator {
/// Return the module name for the code generator.
fn get_name(&self) -> &str;
fn get_size_type<'ctx>(&self, ctx: &'ctx Context) -> IntType<'ctx>;
/// Generate function call and returns the function return value.
/// - obj: Optional object for method call.
/// - fun: Function signature and definition ID.
/// - params: Function parameters. Note that this does not include the object even if the
/// function is a class method.
fn gen_call<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
obj: Option<(Type, ValueEnum<'ctx>)>,
fun: (&FunSignature, DefinitionId),
params: Vec<(Option<StrRef>, ValueEnum<'ctx>)>,
) -> Result<Option<BasicValueEnum<'ctx>>, String>
where
Self: Sized,
{
gen_call(self, ctx, obj, fun, params)
}
/// Generate object constructor and returns the constructed object.
/// - signature: Function signature of the constructor.
/// - def: Class definition for the constructor class.
/// - params: Function parameters.
fn gen_constructor<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
signature: &FunSignature,
def: &TopLevelDef,
params: Vec<(Option<StrRef>, ValueEnum<'ctx>)>,
) -> Result<BasicValueEnum<'ctx>, String>
where
Self: Sized,
{
gen_constructor(self, ctx, signature, def, params)
}
/// Generate a function instance.
/// - obj: Optional object for method call.
/// - fun: Function signature, definition ID and the substitution key.
/// - params: Function parameters. Note that this does not include the object even if the
/// function is a class method.
/// Note that this function should check if the function is generated in another thread (due to
/// possible race condition), see the default implementation for an example.
fn gen_func_instance<'ctx, 'a>(
&mut self,
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)
}
/// Generate the code for an expression.
fn gen_expr<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
expr: &Expr<Option<Type>>,
) -> Result<Option<ValueEnum<'ctx>>, String>
where
Self: Sized,
{
gen_expr(self, ctx, expr)
}
/// Allocate memory for a variable and return a pointer pointing to it.
/// The default implementation places the allocations at the start of the function.
fn gen_var_alloc<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
ty: BasicTypeEnum<'ctx>,
) -> Result<PointerValue<'ctx>, String> {
gen_var(ctx, ty)
}
/// Return a pointer pointing to the target of the expression.
fn gen_store_target<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
pattern: &Expr<Option<Type>>,
) -> Result<PointerValue<'ctx>, String>
where
Self: Sized,
{
gen_store_target(self, ctx, pattern)
}
/// Generate code for an assignment expression.
fn gen_assign<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
target: &Expr<Option<Type>>,
value: ValueEnum<'ctx>,
) -> Result<(), String>
where
Self: Sized,
{
gen_assign(self, ctx, target, value)
}
/// Generate code for a while expression.
/// Return true if the while loop must early return
fn gen_while<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
stmt: &Stmt<Option<Type>>,
) -> Result<(), String>
where
Self: Sized,
{
gen_while(self, ctx, stmt)
}
/// 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, 'a>,
stmt: &Stmt<Option<Type>>,
) -> Result<(), String>
where
Self: Sized,
{
gen_for(self, ctx, stmt)
}
/// Generate code for an if expression.
/// Return true if the statement must early return
fn gen_if<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
stmt: &Stmt<Option<Type>>,
) -> Result<(), String>
where
Self: Sized,
{
gen_if(self, ctx, stmt)
}
fn gen_with<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
stmt: &Stmt<Option<Type>>,
) -> Result<(), String>
where
Self: Sized,
{
gen_with(self, ctx, stmt)
}
/// Generate code for a statement
/// Return true if the statement must early return
fn gen_stmt<'ctx, 'a>(
&mut self,
ctx: &mut CodeGenContext<'ctx, 'a>,
stmt: &Stmt<Option<Type>>,
) -> Result<(), String>
where
Self: Sized,
{
gen_stmt(self, ctx, stmt)
}
}
pub struct DefaultCodeGenerator {
name: String,
size_t: u32,
}
impl DefaultCodeGenerator {
pub fn new(name: String, size_t: u32) -> DefaultCodeGenerator {
assert!(size_t == 32 || size_t == 64);
DefaultCodeGenerator { name, size_t }
}
}
impl CodeGenerator for DefaultCodeGenerator {
fn get_name(&self) -> &str {
&self.name
}
fn get_size_type<'ctx>(&self, ctx: &'ctx Context) -> IntType<'ctx> {
// it should be unsigned, but we don't really need unsigned and this could save us from
// having to do a bit cast...
if self.size_t == 32 {
ctx.i32_type()
} else {
ctx.i64_type()
}
}
}

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;
}

View File

@ -0,0 +1,434 @@
use crate::typecheck::typedef::Type;
use super::{CodeGenContext, CodeGenerator};
use inkwell::{
attributes::{Attribute, AttributeLoc},
context::Context,
memory_buffer::MemoryBuffer,
module::Module,
types::BasicTypeEnum,
values::{IntValue, PointerValue},
AddressSpace, IntPredicate,
};
use nac3parser::ast::Expr;
pub fn load_irrt(ctx: &Context) -> Module {
let bitcode_buf = MemoryBuffer::create_from_memory_range(
include_bytes!(concat!(env!("OUT_DIR"), "/irrt.bc")),
"irrt_bitcode_buffer",
);
let irrt_mod = Module::parse_bitcode_from_buffer(&bitcode_buf, ctx).unwrap();
let inline_attr = Attribute::get_named_enum_kind_id("alwaysinline");
for symbol in &[
"__nac3_int_exp_int32_t",
"__nac3_int_exp_int64_t",
"__nac3_range_slice_len",
"__nac3_slice_index_bound",
] {
let function = irrt_mod.get_function(symbol).unwrap();
function.add_attribute(AttributeLoc::Function, ctx.create_enum_attribute(inline_attr, 0));
}
irrt_mod
}
// repeated squaring method adapted from GNU Scientific Library:
// https://git.savannah.gnu.org/cgit/gsl.git/tree/sys/pow_int.c
pub fn integer_power<'ctx, 'a>(
generator: &mut dyn CodeGenerator,
ctx: &mut CodeGenContext<'ctx, 'a>,
base: IntValue<'ctx>,
exp: IntValue<'ctx>,
signed: bool,
) -> IntValue<'ctx> {
let symbol = match (base.get_type().get_bit_width(), exp.get_type().get_bit_width(), signed) {
(32, 32, true) => "__nac3_int_exp_int32_t",
(64, 64, true) => "__nac3_int_exp_int64_t",
(32, 32, false) => "__nac3_int_exp_uint32_t",
(64, 64, false) => "__nac3_int_exp_uint64_t",
_ => unreachable!(),
};
let base_type = base.get_type();
let pow_fun = ctx.module.get_function(symbol).unwrap_or_else(|| {
let fn_type = base_type.fn_type(&[base_type.into(), base_type.into()], false);
ctx.module.add_function(symbol, fn_type, None)
});
// throw exception when exp < 0
let ge_zero = ctx.builder.build_int_compare(
IntPredicate::SGE,
exp,
exp.get_type().const_zero(),
"assert_int_pow_ge_0",
);
ctx.make_assert(
generator,
ge_zero,
"0:ValueError",
"integer power must be positive or zero",
[None, None, None],
ctx.current_loc,
);
ctx.builder
.build_call(pow_fun, &[base.into(), exp.into()], "call_int_pow")
.try_as_basic_value()
.unwrap_left()
.into_int_value()
}
pub fn calculate_len_for_slice_range<'ctx, 'a>(
generator: &mut dyn CodeGenerator,
ctx: &mut CodeGenContext<'ctx, 'a>,
start: IntValue<'ctx>,
end: IntValue<'ctx>,
step: IntValue<'ctx>,
) -> IntValue<'ctx> {
const SYMBOL: &str = "__nac3_range_slice_len";
let len_func = ctx.module.get_function(SYMBOL).unwrap_or_else(|| {
let i32_t = ctx.ctx.i32_type();
let fn_t = i32_t.fn_type(&[i32_t.into(), i32_t.into(), i32_t.into()], false);
ctx.module.add_function(SYMBOL, fn_t, None)
});
// assert step != 0, throw exception if not
let not_zero = ctx.builder.build_int_compare(
IntPredicate::NE,
step,
step.get_type().const_zero(),
"range_step_ne",
);
ctx.make_assert(
generator,
not_zero,
"0:ValueError",
"step must not be zero",
[None, None, None],
ctx.current_loc,
);
ctx.builder
.build_call(len_func, &[start.into(), end.into(), step.into()], "calc_len")
.try_as_basic_value()
.left()
.unwrap()
.into_int_value()
}
/// NOTE: the output value of the end index of this function should be compared ***inclusively***,
/// because python allows `a[2::-1]`, whose semantic is `[a[2], a[1], a[0]]`, which is equivalent to
/// NO numeric slice in python.
///
/// equivalent code:
/// ```pseudo_code
/// match (start, end, step):
/// case (s, e, None | Some(step)) if step > 0:
/// return (
/// match s:
/// case None:
/// 0
/// case Some(s):
/// handle_in_bound(s)
/// ,match e:
/// case None:
/// length - 1
/// case Some(e):
/// handle_in_bound(e) - 1
/// ,step == None ? 1 : step
/// )
/// case (s, e, Some(step)) if step < 0:
/// return (
/// match s:
/// case None:
/// length - 1
/// case Some(s):
/// s = handle_in_bound(s)
/// if s == length:
/// s - 1
/// else:
/// s
/// ,match e:
/// case None:
/// 0
/// case Some(e):
/// handle_in_bound(e) + 1
/// ,step
/// )
/// ```
pub fn handle_slice_indices<'a, 'ctx, G: CodeGenerator>(
start: &Option<Box<Expr<Option<Type>>>>,
end: &Option<Box<Expr<Option<Type>>>>,
step: &Option<Box<Expr<Option<Type>>>>,
ctx: &mut CodeGenContext<'ctx, 'a>,
generator: &mut G,
list: PointerValue<'ctx>,
) -> Result<(IntValue<'ctx>, IntValue<'ctx>, IntValue<'ctx>), String> {
let int32 = ctx.ctx.i32_type();
let zero = int32.const_zero();
let one = int32.const_int(1, false);
let length = ctx.build_gep_and_load(list, &[zero, one]).into_int_value();
let length = ctx.builder.build_int_truncate_or_bit_cast(length, int32, "leni32");
Ok(match (start, end, step) {
(s, e, None) => (
s.as_ref().map_or_else(
|| Ok(int32.const_zero()),
|s| handle_slice_index_bound(s, ctx, generator, length),
)?,
{
let e = e.as_ref().map_or_else(
|| Ok(length),
|e| handle_slice_index_bound(e, ctx, generator, length),
)?;
ctx.builder.build_int_sub(e, one, "final_end")
},
one,
),
(s, e, Some(step)) => {
let step = generator
.gen_expr(ctx, step)?
.unwrap()
.to_basic_value_enum(ctx, generator, ctx.primitives.int32)?
.into_int_value();
// assert step != 0, throw exception if not
let not_zero = ctx.builder.build_int_compare(
IntPredicate::NE,
step,
step.get_type().const_zero(),
"range_step_ne",
);
ctx.make_assert(
generator,
not_zero,
"0:ValueError",
"slice step cannot be zero",
[None, None, None],
ctx.current_loc,
);
let len_id = ctx.builder.build_int_sub(length, one, "lenmin1");
let neg = ctx.builder.build_int_compare(IntPredicate::SLT, step, zero, "step_is_neg");
(
match s {
Some(s) => {
let s = handle_slice_index_bound(s, ctx, generator, length)?;
ctx.builder
.build_select(
ctx.builder.build_and(
ctx.builder.build_int_compare(
IntPredicate::EQ,
s,
length,
"s_eq_len",
),
neg,
"should_minus_one",
),
ctx.builder.build_int_sub(s, one, "s_min"),
s,
"final_start",
)
.into_int_value()
}
None => ctx.builder.build_select(neg, len_id, zero, "stt").into_int_value(),
},
match e {
Some(e) => {
let e = handle_slice_index_bound(e, ctx, generator, length)?;
ctx.builder
.build_select(
neg,
ctx.builder.build_int_add(e, one, "end_add_one"),
ctx.builder.build_int_sub(e, one, "end_sub_one"),
"final_end",
)
.into_int_value()
}
None => ctx.builder.build_select(neg, zero, len_id, "end").into_int_value(),
},
step,
)
}
})
}
/// this function allows index out of range, since python
/// allows index out of range in slice (`a = [1,2,3]; a[1:10] == [2,3]`).
pub fn handle_slice_index_bound<'a, 'ctx, G: CodeGenerator>(
i: &Expr<Option<Type>>,
ctx: &mut CodeGenContext<'ctx, 'a>,
generator: &mut G,
length: IntValue<'ctx>,
) -> Result<IntValue<'ctx>, String> {
const SYMBOL: &str = "__nac3_slice_index_bound";
let func = ctx.module.get_function(SYMBOL).unwrap_or_else(|| {
let i32_t = ctx.ctx.i32_type();
let fn_t = i32_t.fn_type(&[i32_t.into(), i32_t.into()], false);
ctx.module.add_function(SYMBOL, fn_t, None)
});
let i = generator.gen_expr(ctx, i)?.unwrap().to_basic_value_enum(ctx, generator, i.custom.unwrap())?;
Ok(ctx
.builder
.build_call(func, &[i.into(), length.into()], "bounded_ind")
.try_as_basic_value()
.left()
.unwrap()
.into_int_value())
}
/// This function handles 'end' **inclusively**.
/// Order of tuples assign_idx and value_idx is ('start', 'end', 'step').
/// Negative index should be handled before entering this function
pub fn list_slice_assignment<'ctx, 'a>(
generator: &mut dyn CodeGenerator,
ctx: &mut CodeGenContext<'ctx, 'a>,
ty: BasicTypeEnum<'ctx>,
dest_arr: PointerValue<'ctx>,
dest_idx: (IntValue<'ctx>, IntValue<'ctx>, IntValue<'ctx>),
src_arr: PointerValue<'ctx>,
src_idx: (IntValue<'ctx>, IntValue<'ctx>, IntValue<'ctx>),
) {
let size_ty = generator.get_size_type(ctx.ctx);
let int8_ptr = ctx.ctx.i8_type().ptr_type(AddressSpace::Generic);
let int32 = ctx.ctx.i32_type();
let (fun_symbol, elem_ptr_type) = ("__nac3_list_slice_assign_var_size", int8_ptr);
let slice_assign_fun = {
let ty_vec = vec![
int32.into(), // dest start idx
int32.into(), // dest end idx
int32.into(), // dest step
elem_ptr_type.into(), // dest arr ptr
int32.into(), // dest arr len
int32.into(), // src start idx
int32.into(), // src end idx
int32.into(), // src step
elem_ptr_type.into(), // src arr ptr
int32.into(), // src arr len
int32.into(), // size
];
ctx.module.get_function(fun_symbol).unwrap_or_else(|| {
let fn_t = int32.fn_type(ty_vec.as_slice(), false);
ctx.module.add_function(fun_symbol, fn_t, None)
})
};
let zero = int32.const_zero();
let one = int32.const_int(1, false);
let dest_arr_ptr = ctx.build_gep_and_load(dest_arr, &[zero, zero]);
let dest_arr_ptr = ctx.builder.build_pointer_cast(
dest_arr_ptr.into_pointer_value(),
elem_ptr_type,
"dest_arr_ptr_cast",
);
let dest_len = ctx.build_gep_and_load(dest_arr, &[zero, one]).into_int_value();
let dest_len = ctx.builder.build_int_truncate_or_bit_cast(dest_len, int32, "srclen32");
let src_arr_ptr = ctx.build_gep_and_load(src_arr, &[zero, zero]);
let src_arr_ptr = ctx.builder.build_pointer_cast(
src_arr_ptr.into_pointer_value(),
elem_ptr_type,
"src_arr_ptr_cast",
);
let src_len = ctx.build_gep_and_load(src_arr, &[zero, one]).into_int_value();
let src_len = ctx.builder.build_int_truncate_or_bit_cast(src_len, int32, "srclen32");
// index in bound and positive should be done
// assert if dest.step == 1 then len(src) <= len(dest) else len(src) == len(dest), and
// throw exception if not satisfied
let src_end = ctx.builder
.build_select(
ctx.builder.build_int_compare(
inkwell::IntPredicate::SLT,
src_idx.2,
zero,
"is_neg",
),
ctx.builder.build_int_sub(src_idx.1, one, "e_min_one"),
ctx.builder.build_int_add(src_idx.1, one, "e_add_one"),
"final_e",
)
.into_int_value();
let dest_end = ctx.builder
.build_select(
ctx.builder.build_int_compare(
inkwell::IntPredicate::SLT,
dest_idx.2,
zero,
"is_neg",
),
ctx.builder.build_int_sub(dest_idx.1, one, "e_min_one"),
ctx.builder.build_int_add(dest_idx.1, one, "e_add_one"),
"final_e",
)
.into_int_value();
let src_slice_len =
calculate_len_for_slice_range(generator, ctx, src_idx.0, src_end, src_idx.2);
let dest_slice_len =
calculate_len_for_slice_range(generator, ctx, dest_idx.0, dest_end, dest_idx.2);
let src_eq_dest = ctx.builder.build_int_compare(
IntPredicate::EQ,
src_slice_len,
dest_slice_len,
"slice_src_eq_dest",
);
let src_slt_dest = ctx.builder.build_int_compare(
IntPredicate::SLT,
src_slice_len,
dest_slice_len,
"slice_src_slt_dest",
);
let dest_step_eq_one = ctx.builder.build_int_compare(
IntPredicate::EQ,
dest_idx.2,
dest_idx.2.get_type().const_int(1, false),
"slice_dest_step_eq_one",
);
let cond_1 = ctx.builder.build_and(dest_step_eq_one, src_slt_dest, "slice_cond_1");
let cond = ctx.builder.build_or(src_eq_dest, cond_1, "slice_cond");
ctx.make_assert(
generator,
cond,
"0:ValueError",
"attempt to assign sequence of size {0} to slice of size {1} with step size {2}",
[Some(src_slice_len), Some(dest_slice_len), Some(dest_idx.2)],
ctx.current_loc,
);
let new_len = {
let args = vec![
dest_idx.0.into(), // dest start idx
dest_idx.1.into(), // dest end idx
dest_idx.2.into(), // dest step
dest_arr_ptr.into(), // dest arr ptr
dest_len.into(), // dest arr len
src_idx.0.into(), // src start idx
src_idx.1.into(), // src end idx
src_idx.2.into(), // src step
src_arr_ptr.into(), // src arr ptr
src_len.into(), // src arr len
{
let s = match ty {
BasicTypeEnum::FloatType(t) => t.size_of(),
BasicTypeEnum::IntType(t) => t.size_of(),
BasicTypeEnum::PointerType(t) => t.size_of(),
BasicTypeEnum::StructType(t) => t.size_of().unwrap(),
_ => unreachable!(),
};
ctx.builder.build_int_truncate_or_bit_cast(s, int32, "size")
}
.into(),
];
ctx.builder
.build_call(slice_assign_fun, args.as_slice(), "slice_assign")
.try_as_basic_value()
.unwrap_left()
.into_int_value()
};
// update length
let need_update =
ctx.builder.build_int_compare(IntPredicate::NE, new_len, dest_len, "need_update");
let current = ctx.builder.get_insert_block().unwrap().get_parent().unwrap();
let update_bb = ctx.ctx.append_basic_block(current, "update");
let cont_bb = ctx.ctx.append_basic_block(current, "cont");
ctx.builder.build_conditional_branch(need_update, update_bb, cont_bb);
ctx.builder.position_at_end(update_bb);
let dest_len_ptr = unsafe { ctx.builder.build_gep(dest_arr, &[zero, one], "dest_len_ptr") };
let new_len = ctx.builder.build_int_z_extend_or_bit_cast(new_len, size_ty, "new_len");
ctx.builder.build_store(dest_len_ptr, new_len);
ctx.builder.build_unconditional_branch(cont_bb);
ctx.builder.position_at_end(cont_bb);
}

608
nac3core/src/codegen/mod.rs Normal file
View File

@ -0,0 +1,608 @@
use crate::{
symbol_resolver::{StaticValue, SymbolResolver},
toplevel::{TopLevelContext, TopLevelDef},
typecheck::{
type_inferencer::{CodeLocation, PrimitiveStore},
typedef::{CallId, FuncArg, Type, TypeEnum, Unifier},
},
};
use crossbeam::channel::{unbounded, Receiver, Sender};
use inkwell::{
AddressSpace,
OptimizationLevel,
attributes::{Attribute, AttributeLoc},
basic_block::BasicBlock,
builder::Builder,
context::Context,
module::Module,
passes::{PassManager, PassManagerBuilder},
types::{AnyType, BasicType, BasicTypeEnum},
values::{BasicValueEnum, FunctionValue, PhiValue, PointerValue}
};
use itertools::Itertools;
use nac3parser::ast::{Stmt, StrRef, Location};
use parking_lot::{Condvar, Mutex};
use std::collections::{HashMap, HashSet};
use std::sync::{
atomic::{AtomicBool, Ordering},
Arc,
};
use std::thread;
pub mod concrete_type;
pub mod expr;
mod generator;
pub mod irrt;
pub mod stmt;
#[cfg(test)]
mod test;
use concrete_type::{ConcreteType, ConcreteTypeEnum, ConcreteTypeStore};
pub use generator::{CodeGenerator, DefaultCodeGenerator};
#[derive(Default)]
pub struct StaticValueStore {
pub lookup: HashMap<Vec<(usize, u64)>, usize>,
pub store: Vec<HashMap<usize, Arc<dyn StaticValue + Send + Sync>>>,
}
pub type VarValue<'ctx> = (PointerValue<'ctx>, Option<Arc<dyn StaticValue + Send + Sync>>, i64);
pub struct CodeGenContext<'ctx, 'a> {
pub ctx: &'ctx Context,
pub builder: Builder<'ctx>,
pub module: Module<'ctx>,
pub top_level: &'a TopLevelContext,
pub unifier: Unifier,
pub resolver: Arc<dyn SymbolResolver + Send + Sync>,
pub static_value_store: Arc<Mutex<StaticValueStore>>,
pub var_assignment: HashMap<StrRef, VarValue<'ctx>>,
pub type_cache: HashMap<Type, BasicTypeEnum<'ctx>>,
pub primitives: PrimitiveStore,
pub calls: Arc<HashMap<CodeLocation, CallId>>,
pub registry: &'a WorkerRegistry,
// const string cache
pub const_strings: HashMap<String, BasicValueEnum<'ctx>>,
// stores the alloca for variables
pub init_bb: BasicBlock<'ctx>,
// the first one is the test_bb, and the second one is bb after the loop
pub loop_target: Option<(BasicBlock<'ctx>, BasicBlock<'ctx>)>,
// unwind target bb
pub unwind_target: Option<BasicBlock<'ctx>>,
// return target bb, just emit ret if no such target
pub return_target: Option<BasicBlock<'ctx>>,
pub return_buffer: Option<PointerValue<'ctx>>,
// outer catch clauses
pub outer_catch_clauses:
Option<(Vec<Option<BasicValueEnum<'ctx>>>, BasicBlock<'ctx>, PhiValue<'ctx>)>,
pub need_sret: bool,
pub current_loc: Location,
}
impl<'ctx, 'a> CodeGenContext<'ctx, 'a> {
pub fn is_terminated(&self) -> bool {
self.builder.get_insert_block().unwrap().get_terminator().is_some()
}
}
type Fp = Box<dyn Fn(&Module) + Send + Sync>;
pub struct WithCall {
fp: Fp,
}
impl WithCall {
pub fn new(fp: Fp) -> WithCall {
WithCall { fp }
}
pub fn run<'ctx>(&self, m: &Module<'ctx>) {
(self.fp)(m)
}
}
pub struct WorkerRegistry {
sender: Arc<Sender<Option<CodeGenTask>>>,
receiver: Arc<Receiver<Option<CodeGenTask>>>,
panicked: AtomicBool,
task_count: Mutex<usize>,
thread_count: usize,
wait_condvar: Condvar,
top_level_ctx: Arc<TopLevelContext>,
static_value_store: Arc<Mutex<StaticValueStore>>,
}
impl WorkerRegistry {
pub fn create_workers<G: CodeGenerator + Send + 'static>(
generators: Vec<Box<G>>,
top_level_ctx: Arc<TopLevelContext>,
f: Arc<WithCall>,
) -> (Arc<WorkerRegistry>, Vec<thread::JoinHandle<()>>) {
let (sender, receiver) = unbounded();
let task_count = Mutex::new(0);
let wait_condvar = Condvar::new();
// init: 0 to be empty
let mut static_value_store: StaticValueStore = Default::default();
static_value_store.lookup.insert(Default::default(), 0);
static_value_store.store.push(Default::default());
let registry = Arc::new(WorkerRegistry {
sender: Arc::new(sender),
receiver: Arc::new(receiver),
thread_count: generators.len(),
panicked: AtomicBool::new(false),
static_value_store: Arc::new(Mutex::new(static_value_store)),
task_count,
wait_condvar,
top_level_ctx,
});
let mut handles = Vec::new();
for mut generator in generators.into_iter() {
let registry = registry.clone();
let registry2 = registry.clone();
let f = f.clone();
let handle = thread::spawn(move || {
registry.worker_thread(generator.as_mut(), f);
});
let handle = thread::spawn(move || {
if let Err(e) = handle.join() {
if let Some(e) = e.downcast_ref::<&'static str>() {
eprintln!("Got an error: {}", e);
} else {
eprintln!("Got an unknown error: {:?}", e);
}
registry2.panicked.store(true, Ordering::SeqCst);
registry2.wait_condvar.notify_all();
}
});
handles.push(handle);
}
(registry, handles)
}
pub fn wait_tasks_complete(&self, handles: Vec<thread::JoinHandle<()>>) {
{
let mut count = self.task_count.lock();
while *count != 0 {
if self.panicked.load(Ordering::SeqCst) {
break;
}
self.wait_condvar.wait(&mut count);
}
}
for _ in 0..self.thread_count {
self.sender.send(None).unwrap();
}
{
let mut count = self.task_count.lock();
while *count != self.thread_count {
if self.panicked.load(Ordering::SeqCst) {
break;
}
self.wait_condvar.wait(&mut count);
}
}
for handle in handles {
handle.join().unwrap();
}
if self.panicked.load(Ordering::SeqCst) {
panic!("tasks panicked");
}
}
pub fn add_task(&self, task: CodeGenTask) {
*self.task_count.lock() += 1;
self.sender.send(Some(task)).unwrap();
}
fn worker_thread<G: CodeGenerator>(&self, generator: &mut G, f: Arc<WithCall>) {
let context = Context::create();
let mut builder = context.create_builder();
let module = context.create_module(generator.get_name());
let pass_builder = PassManagerBuilder::create();
pass_builder.set_optimization_level(OptimizationLevel::Default);
let passes = PassManager::create(&module);
pass_builder.populate_function_pass_manager(&passes);
let mut errors = HashSet::new();
while let Some(task) = self.receiver.recv().unwrap() {
let tmp_module = context.create_module("tmp");
match gen_func(&context, generator, self, builder, tmp_module, task) {
Ok(result) => {
builder = result.0;
passes.run_on(&result.2);
module.link_in_module(result.1).unwrap();
}
Err((old_builder, e)) => {
builder = old_builder;
errors.insert(e);
}
}
*self.task_count.lock() -= 1;
self.wait_condvar.notify_all();
}
if !errors.is_empty() {
panic!("Codegen error: {}", errors.into_iter().sorted().join("\n----------\n"));
}
let result = module.verify();
if let Err(err) = result {
println!("{}", module.print_to_string().to_str().unwrap());
println!("{}", err);
panic!()
}
f.run(&module);
let mut lock = self.task_count.lock();
*lock += 1;
self.wait_condvar.notify_all();
}
}
pub struct CodeGenTask {
pub subst: Vec<(Type, ConcreteType)>,
pub store: ConcreteTypeStore,
pub symbol_name: String,
pub signature: ConcreteType,
pub body: Arc<Vec<Stmt<Option<Type>>>>,
pub calls: Arc<HashMap<CodeLocation, CallId>>,
pub unifier_index: usize,
pub resolver: Arc<dyn SymbolResolver + Send + Sync>,
pub id: usize,
}
fn get_llvm_type<'ctx>(
ctx: &'ctx Context,
generator: &mut dyn CodeGenerator,
unifier: &mut Unifier,
top_level: &TopLevelContext,
type_cache: &mut HashMap<Type, BasicTypeEnum<'ctx>>,
primitives: &PrimitiveStore,
ty: Type,
) -> BasicTypeEnum<'ctx> {
use TypeEnum::*;
// we assume the type cache should already contain primitive types,
// and they should be passed by value instead of passing as pointer.
type_cache.get(&unifier.get_representative(ty)).cloned().unwrap_or_else(|| {
let ty_enum = unifier.get_ty(ty);
let result = match &*ty_enum {
TObj { obj_id, fields, .. } => {
// check to avoid treating primitives other than Option as classes
if obj_id.0 <= 10 {
match (unifier.get_ty(ty).as_ref(), unifier.get_ty(primitives.option).as_ref())
{
(
TypeEnum::TObj { obj_id, params, .. },
TypeEnum::TObj { obj_id: opt_id, .. },
) if *obj_id == *opt_id => {
return get_llvm_type(
ctx,
generator,
unifier,
top_level,
type_cache,
primitives,
*params.iter().next().unwrap().1,
)
.ptr_type(AddressSpace::Generic)
.into();
}
_ => unreachable!("must be option type"),
}
}
// a struct with fields in the order of declaration
let top_level_defs = top_level.definitions.read();
let definition = top_level_defs.get(obj_id.0).unwrap();
let ty = if let TopLevelDef::Class { name, fields: fields_list, .. } =
&*definition.read()
{
let struct_type = ctx.opaque_struct_type(&name.to_string());
type_cache.insert(unifier.get_representative(ty), struct_type.ptr_type(AddressSpace::Generic).into());
let fields = fields_list
.iter()
.map(|f| {
get_llvm_type(
ctx,
generator,
unifier,
top_level,
type_cache,
primitives,
fields[&f.0].0,
)
})
.collect_vec();
struct_type.set_body(&fields, false);
struct_type.ptr_type(AddressSpace::Generic).into()
} else {
unreachable!()
};
return ty;
}
TTuple { ty } => {
// a struct with fields in the order present in the tuple
let fields = ty
.iter()
.map(|ty| get_llvm_type(ctx, generator, unifier, top_level, type_cache, primitives, *ty))
.collect_vec();
ctx.struct_type(&fields, false).into()
}
TList { ty } => {
// a struct with an integer and a pointer to an array
let element_type =
get_llvm_type(ctx, generator, unifier, top_level, type_cache, primitives, *ty);
let fields = [
element_type.ptr_type(AddressSpace::Generic).into(),
generator.get_size_type(ctx).into(),
];
ctx.struct_type(&fields, false).ptr_type(AddressSpace::Generic).into()
}
TVirtual { .. } => unimplemented!(),
_ => unreachable!("{}", ty_enum.get_type_name()),
};
type_cache.insert(unifier.get_representative(ty), result);
result
})
}
fn need_sret<'ctx>(ctx: &'ctx Context, ty: BasicTypeEnum<'ctx>) -> bool {
fn need_sret_impl<'ctx>(ctx: &'ctx Context, ty: BasicTypeEnum<'ctx>, maybe_large: bool) -> bool {
match ty {
BasicTypeEnum::IntType(_) | BasicTypeEnum::PointerType(_) => false,
BasicTypeEnum::FloatType(_) if maybe_large => false,
BasicTypeEnum::StructType(ty) if maybe_large && ty.count_fields() <= 2 =>
ty.get_field_types().iter().any(|ty| need_sret_impl(ctx, *ty, false)),
_ => true,
}
}
need_sret_impl(ctx, ty, true)
}
pub fn gen_func_impl<'ctx, G: CodeGenerator, F: FnOnce(&mut G, &mut CodeGenContext) -> Result<(), String>> (
context: &'ctx Context,
generator: &mut G,
registry: &WorkerRegistry,
builder: Builder<'ctx>,
module: Module<'ctx>,
task: CodeGenTask,
codegen_function: F
) -> Result<(Builder<'ctx>, Module<'ctx>, FunctionValue<'ctx>), (Builder<'ctx>, String)> {
let top_level_ctx = registry.top_level_ctx.clone();
let static_value_store = registry.static_value_store.clone();
let (mut unifier, primitives) = {
let (unifier, primitives) = &top_level_ctx.unifiers.read()[task.unifier_index];
(Unifier::from_shared_unifier(unifier), *primitives)
};
unifier.top_level = Some(top_level_ctx.clone());
let mut cache = HashMap::new();
for (a, b) in task.subst.iter() {
// this should be unification between variables and concrete types
// and should not cause any problem...
let b = task.store.to_unifier_type(&mut unifier, &primitives, *b, &mut cache);
unifier
.unify(*a, b)
.or_else(|err| {
if matches!(&*unifier.get_ty(*a), TypeEnum::TRigidVar { .. }) {
unifier.replace_rigid_var(*a, b);
Ok(())
} else {
Err(err)
}
})
.unwrap()
}
// rebuild primitive store with unique representatives
let primitives = PrimitiveStore {
int32: unifier.get_representative(primitives.int32),
int64: unifier.get_representative(primitives.int64),
uint32: unifier.get_representative(primitives.uint32),
uint64: unifier.get_representative(primitives.uint64),
float: unifier.get_representative(primitives.float),
bool: unifier.get_representative(primitives.bool),
none: unifier.get_representative(primitives.none),
range: unifier.get_representative(primitives.range),
str: unifier.get_representative(primitives.str),
exception: unifier.get_representative(primitives.exception),
option: unifier.get_representative(primitives.option),
};
let mut type_cache: HashMap<_, _> = [
(primitives.int32, context.i32_type().into()),
(primitives.int64, context.i64_type().into()),
(primitives.uint32, context.i32_type().into()),
(primitives.uint64, context.i64_type().into()),
(primitives.float, context.f64_type().into()),
(primitives.bool, context.bool_type().into()),
(primitives.str, {
let str_type = context.opaque_struct_type("str");
let fields = [
context.i8_type().ptr_type(AddressSpace::Generic).into(),
generator.get_size_type(context).into(),
];
str_type.set_body(&fields, false);
str_type.into()
}),
(primitives.range, context.i32_type().array_type(3).ptr_type(AddressSpace::Generic).into()),
]
.iter()
.cloned()
.collect();
type_cache.insert(primitives.exception, {
let exception = context.opaque_struct_type("Exception");
let int32 = context.i32_type().into();
let int64 = context.i64_type().into();
let str_ty = *type_cache.get(&primitives.str).unwrap();
let fields = [int32, str_ty, int32, int32, str_ty, str_ty, int64, int64, int64];
exception.set_body(&fields, false);
exception.ptr_type(AddressSpace::Generic).into()
});
// NOTE: special handling of option cannot use this type cache since it contains type var,
// handled inside get_llvm_type instead
let (args, ret) = if let ConcreteTypeEnum::TFunc { args, ret, .. } =
task.store.get(task.signature)
{
(
args.iter()
.map(|arg| FuncArg {
name: arg.name,
ty: task.store.to_unifier_type(&mut unifier, &primitives, arg.ty, &mut cache),
default_value: arg.default_value.clone(),
})
.collect_vec(),
task.store.to_unifier_type(&mut unifier, &primitives, *ret, &mut cache),
)
} else {
unreachable!()
};
let ret_type = if unifier.unioned(ret, primitives.none) {
None
} else {
Some(get_llvm_type(context, generator, &mut unifier, top_level_ctx.as_ref(), &mut type_cache, &primitives, ret))
};
let has_sret = ret_type.map_or(false, |ty| need_sret(context, ty));
let mut params = args
.iter()
.map(|arg| {
get_llvm_type(
context,
generator,
&mut unifier,
top_level_ctx.as_ref(),
&mut type_cache,
&primitives,
arg.ty,
)
.into()
})
.collect_vec();
if has_sret {
params.insert(0, ret_type.unwrap().ptr_type(AddressSpace::Generic).into());
}
let fn_type = match ret_type {
Some(ret_type) if !has_sret => ret_type.fn_type(&params, false),
_ => context.void_type().fn_type(&params, false)
};
let symbol = &task.symbol_name;
let fn_val =
module.get_function(symbol).unwrap_or_else(|| module.add_function(symbol, fn_type, None));
if let Some(personality) = &top_level_ctx.personality_symbol {
let personality = module.get_function(personality).unwrap_or_else(|| {
let ty = context.i32_type().fn_type(&[], true);
module.add_function(personality, ty, None)
});
fn_val.set_personality_function(personality);
}
if has_sret {
fn_val.add_attribute(AttributeLoc::Param(0),
context.create_type_attribute(Attribute::get_named_enum_kind_id("sret"),
ret_type.unwrap().as_any_type_enum()));
}
let init_bb = context.append_basic_block(fn_val, "init");
builder.position_at_end(init_bb);
let body_bb = context.append_basic_block(fn_val, "body");
let mut var_assignment = HashMap::new();
let offset = if has_sret { 1 } else { 0 };
for (n, arg) in args.iter().enumerate() {
let param = fn_val.get_nth_param((n as u32) + offset).unwrap();
let alloca = builder.build_alloca(
get_llvm_type(
context,
generator,
&mut unifier,
top_level_ctx.as_ref(),
&mut type_cache,
&primitives,
arg.ty,
),
&arg.name.to_string(),
);
builder.build_store(alloca, param);
var_assignment.insert(arg.name, (alloca, None, 0));
}
let return_buffer = if has_sret {
Some(fn_val.get_nth_param(0).unwrap().into_pointer_value())
} else {
fn_type.get_return_type().map(|v| builder.build_alloca(v, "$ret"))
};
let static_values = {
let store = registry.static_value_store.lock();
store.store[task.id].clone()
};
for (k, v) in static_values.into_iter() {
let (_, static_val, _) = var_assignment.get_mut(&args[k].name).unwrap();
*static_val = Some(v);
}
builder.build_unconditional_branch(body_bb);
builder.position_at_end(body_bb);
let mut code_gen_context = CodeGenContext {
ctx: context,
resolver: task.resolver,
top_level: top_level_ctx.as_ref(),
calls: task.calls,
loop_target: None,
return_target: None,
return_buffer,
unwind_target: None,
outer_catch_clauses: None,
const_strings: Default::default(),
registry,
var_assignment,
type_cache,
primitives,
init_bb,
builder,
module,
unifier,
static_value_store,
need_sret: has_sret,
current_loc: Default::default(),
};
let result = codegen_function(generator, &mut code_gen_context);
// after static analysis, only void functions can have no return at the end.
if !code_gen_context.is_terminated() {
code_gen_context.builder.build_return(None);
}
let CodeGenContext { builder, module, .. } = code_gen_context;
if let Err(e) = result {
return Err((builder, e));
}
Ok((builder, module, fn_val))
}
pub fn gen_func<'ctx, G: CodeGenerator>(
context: &'ctx Context,
generator: &mut G,
registry: &WorkerRegistry,
builder: Builder<'ctx>,
module: Module<'ctx>,
task: CodeGenTask,
) -> Result<(Builder<'ctx>, Module<'ctx>, FunctionValue<'ctx>), (Builder<'ctx>, String)> {
let body = task.body.clone();
gen_func_impl(context, generator, registry, builder, module, task, |generator, ctx| {
for stmt in body.iter() {
generator.gen_stmt(ctx, stmt)?;
}
Ok(())
})
}

1079
nac3core/src/codegen/stmt.rs Normal file

File diff suppressed because it is too large Load Diff

View File

@ -0,0 +1,364 @@
use crate::{
codegen::{
concrete_type::ConcreteTypeStore, CodeGenContext, CodeGenTask, DefaultCodeGenerator,
WithCall, WorkerRegistry,
},
symbol_resolver::{SymbolResolver, ValueEnum},
toplevel::{
composer::TopLevelComposer, DefinitionId, FunInstance, TopLevelContext, TopLevelDef,
},
typecheck::{
type_inferencer::{FunctionData, Inferencer, PrimitiveStore},
typedef::{FunSignature, FuncArg, Type, TypeEnum, Unifier},
},
};
use indoc::indoc;
use nac3parser::{
ast::{fold::Fold, StrRef},
parser::parse_program,
};
use parking_lot::RwLock;
use std::collections::{HashMap, HashSet};
use std::sync::Arc;
struct Resolver {
id_to_type: HashMap<StrRef, Type>,
id_to_def: RwLock<HashMap<StrRef, DefinitionId>>,
class_names: HashMap<StrRef, Type>,
}
impl Resolver {
pub fn add_id_def(&self, id: StrRef, def: DefinitionId) {
self.id_to_def.write().insert(id, def);
}
}
impl SymbolResolver for Resolver {
fn get_default_param_value(
&self,
_: &nac3parser::ast::Expr,
) -> Option<crate::symbol_resolver::SymbolValue> {
unimplemented!()
}
fn get_symbol_type(
&self,
_: &mut Unifier,
_: &[Arc<RwLock<TopLevelDef>>],
_: &PrimitiveStore,
str: StrRef,
) -> Result<Type, String> {
self.id_to_type.get(&str).cloned().ok_or_else(|| format!("cannot find symbol `{}`", str))
}
fn get_symbol_value<'ctx, 'a>(
&self,
_: StrRef,
_: &mut CodeGenContext<'ctx, 'a>,
) -> Option<ValueEnum<'ctx>> {
unimplemented!()
}
fn get_identifier_def(&self, id: StrRef) -> Result<DefinitionId, String> {
self.id_to_def
.read()
.get(&id)
.cloned()
.ok_or_else(|| format!("cannot find symbol `{}`", id))
}
fn get_string_id(&self, _: &str) -> i32 {
unimplemented!()
}
fn get_exception_id(&self, _tyid: usize) -> usize {
unimplemented!()
}
}
#[test]
fn test_primitives() {
let source = indoc! { "
c = a + b
d = a if c == 1 else 0
return d
"};
let statements = parse_program(source, Default::default()).unwrap();
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());
unifier.top_level = Some(top_level.clone());
let resolver = Arc::new(Resolver {
id_to_type: HashMap::new(),
id_to_def: RwLock::new(HashMap::new()),
class_names: Default::default(),
}) as Arc<dyn SymbolResolver + Send + Sync>;
let threads = vec![DefaultCodeGenerator::new("test".into(), 32).into()];
let signature = FunSignature {
args: vec![
FuncArg { name: "a".into(), ty: primitives.int32, default_value: None },
FuncArg { name: "b".into(), ty: primitives.int32, default_value: None },
],
ret: primitives.int32,
vars: HashMap::new(),
};
let mut store = ConcreteTypeStore::new();
let mut cache = HashMap::new();
let signature = store.from_signature(&mut unifier, &primitives, &signature, &mut cache);
let signature = store.add_cty(signature);
let mut function_data = FunctionData {
resolver: resolver.clone(),
bound_variables: Vec::new(),
return_type: Some(primitives.int32),
};
let mut virtual_checks = Vec::new();
let mut calls = HashMap::new();
let mut identifiers: 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: Default::default(),
primitives: &primitives,
virtual_checks: &mut virtual_checks,
calls: &mut calls,
defined_identifiers: identifiers.clone(),
in_handler: false,
};
inferencer.variable_mapping.insert("a".into(), inferencer.primitives.int32);
inferencer.variable_mapping.insert("b".into(), inferencer.primitives.int32);
let statements = statements
.into_iter()
.map(|v| inferencer.fold_stmt(v))
.collect::<Result<Vec<_>, _>>()
.unwrap();
inferencer.check_block(&statements, &mut identifiers).unwrap();
let top_level = Arc::new(TopLevelContext {
definitions: Arc::new(RwLock::new(std::mem::take(&mut *top_level.definitions.write()))),
unifiers: Arc::new(RwLock::new(vec![(unifier.get_shared_unifier(), primitives)])),
personality_symbol: None,
});
let task = CodeGenTask {
subst: Default::default(),
symbol_name: "testing".into(),
body: Arc::new(statements),
unifier_index: 0,
calls: Arc::new(calls),
resolver,
store,
signature,
id: 0,
};
let f = Arc::new(WithCall::new(Box::new(|module| {
// the following IR is equivalent to
// ```
// ; ModuleID = 'test.ll'
// source_filename = "test"
//
// ; Function Attrs: norecurse nounwind readnone
// define i32 @testing(i32 %0, i32 %1) local_unnamed_addr #0 {
// init:
// %add = add i32 %1, %0
// %cmp = icmp eq i32 %add, 1
// %ifexpr = select i1 %cmp, i32 %0, i32 0
// ret i32 %ifexpr
// }
//
// attributes #0 = { norecurse nounwind readnone }
// ```
// after O2 optimization
let expected = indoc! {"
; ModuleID = 'test'
source_filename = \"test\"
define i32 @testing(i32 %0, i32 %1) {
init:
%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
}
"}
.trim();
assert_eq!(expected, module.print_to_string().to_str().unwrap().trim());
})));
let (registry, handles) = WorkerRegistry::create_workers(threads, top_level, f);
registry.add_task(task);
registry.wait_tasks_complete(handles);
}
#[test]
fn test_simple_call() {
let source_1 = indoc! { "
a = foo(a)
return a * 2
"};
let statements_1 = parse_program(source_1, Default::default()).unwrap();
let source_2 = indoc! { "
return a + 1
"};
let statements_2 = parse_program(source_2, Default::default()).unwrap();
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());
unifier.top_level = Some(top_level.clone());
let signature = FunSignature {
args: vec![FuncArg { name: "a".into(), ty: primitives.int32, default_value: None }],
ret: primitives.int32,
vars: HashMap::new(),
};
let fun_ty = unifier.add_ty(TypeEnum::TFunc(signature.clone()));
let mut store = ConcreteTypeStore::new();
let mut cache = HashMap::new();
let signature = store.from_signature(&mut unifier, &primitives, &signature, &mut cache);
let signature = store.add_cty(signature);
let foo_id = top_level.definitions.read().len();
top_level.definitions.write().push(Arc::new(RwLock::new(TopLevelDef::Function {
name: "foo".to_string(),
simple_name: "foo".into(),
signature: fun_ty,
var_id: vec![],
instance_to_stmt: HashMap::new(),
instance_to_symbol: HashMap::new(),
resolver: None,
codegen_callback: None,
loc: None,
})));
let resolver = Resolver {
id_to_type: HashMap::new(),
id_to_def: RwLock::new(HashMap::new()),
class_names: Default::default(),
};
resolver.add_id_def("foo".into(), DefinitionId(foo_id));
let resolver = Arc::new(resolver) as Arc<dyn SymbolResolver + Send + Sync>;
if let TopLevelDef::Function { resolver: r, .. } =
&mut *top_level.definitions.read()[foo_id].write()
{
*r = Some(resolver.clone());
} else {
unreachable!()
}
let threads = vec![DefaultCodeGenerator::new("test".into(), 32).into()];
let mut function_data = FunctionData {
resolver: resolver.clone(),
bound_variables: Vec::new(),
return_type: Some(primitives.int32),
};
let mut virtual_checks = Vec::new();
let mut calls = HashMap::new();
let mut identifiers: 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: Default::default(),
primitives: &primitives,
virtual_checks: &mut virtual_checks,
calls: &mut calls,
defined_identifiers: identifiers.clone(),
in_handler: false,
};
inferencer.variable_mapping.insert("a".into(), inferencer.primitives.int32);
inferencer.variable_mapping.insert("foo".into(), fun_ty);
let statements_1 = statements_1
.into_iter()
.map(|v| inferencer.fold_stmt(v))
.collect::<Result<Vec<_>, _>>()
.unwrap();
let calls1 = inferencer.calls.clone();
inferencer.calls.clear();
let statements_2 = statements_2
.into_iter()
.map(|v| inferencer.fold_stmt(v))
.collect::<Result<Vec<_>, _>>()
.unwrap();
if let TopLevelDef::Function { instance_to_stmt, .. } =
&mut *top_level.definitions.read()[foo_id].write()
{
instance_to_stmt.insert(
"".to_string(),
FunInstance {
body: Arc::new(statements_2),
calls: Arc::new(inferencer.calls.clone()),
subst: Default::default(),
unifier_id: 0,
},
);
} else {
unreachable!()
}
inferencer.check_block(&statements_1, &mut identifiers).unwrap();
let top_level = Arc::new(TopLevelContext {
definitions: Arc::new(RwLock::new(std::mem::take(&mut *top_level.definitions.write()))),
unifiers: Arc::new(RwLock::new(vec![(unifier.get_shared_unifier(), primitives)])),
personality_symbol: None,
});
let task = CodeGenTask {
subst: Default::default(),
symbol_name: "testing".to_string(),
body: Arc::new(statements_1),
calls: Arc::new(calls1),
unifier_index: 0,
resolver,
signature,
store,
id: 0,
};
let f = Arc::new(WithCall::new(Box::new(|module| {
let expected = indoc! {"
; ModuleID = 'test'
source_filename = \"test\"
define i32 @testing(i32 %0) {
init:
%call = call i32 @foo.0(i32 %0)
%mul = mul i32 %call, 2
ret i32 %mul
}
define i32 @foo.0(i32 %0) {
init:
%add = add i32 %0, 1
ret i32 %add
}
"}
.trim();
assert_eq!(expected, module.print_to_string().to_str().unwrap().trim());
})));
let (registry, handles) = WorkerRegistry::create_workers(threads, top_level, f);
registry.add_task(task);
registry.wait_tasks_complete(handles);
}

View File

@ -1,13 +1,7 @@
#![warn(clippy::all)] #![warn(clippy::all)]
#![allow(clippy::clone_double_ref)] #![allow(dead_code)]
#[cfg(test)]
extern crate test_case;
extern crate num_bigint;
extern crate inkwell;
extern crate rustpython_parser;
extern crate indoc;
mod typecheck;
pub mod codegen;
pub mod symbol_resolver;
pub mod toplevel;
pub mod typecheck;

View File

@ -0,0 +1,380 @@
use std::fmt::Debug;
use std::sync::Arc;
use std::{collections::HashMap, fmt::Display};
use crate::typecheck::typedef::TypeEnum;
use crate::{
codegen::CodeGenContext,
toplevel::{DefinitionId, TopLevelDef},
};
use crate::{
codegen::CodeGenerator,
typecheck::{
type_inferencer::PrimitiveStore,
typedef::{Type, Unifier},
},
};
use inkwell::values::{BasicValueEnum, FloatValue, IntValue, PointerValue, StructValue};
use itertools::{chain, izip};
use nac3parser::ast::{Expr, Location, StrRef};
use parking_lot::RwLock;
#[derive(Clone, PartialEq, Debug)]
pub enum SymbolValue {
I32(i32),
I64(i64),
U32(u32),
U64(u64),
Str(String),
Double(f64),
Bool(bool),
Tuple(Vec<SymbolValue>),
OptionSome(Box<SymbolValue>),
OptionNone,
}
impl Display for SymbolValue {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
match self {
SymbolValue::I32(i) => write!(f, "{}", i),
SymbolValue::I64(i) => write!(f, "int64({})", i),
SymbolValue::U32(i) => write!(f, "uint32({})", i),
SymbolValue::U64(i) => write!(f, "uint64({})", i),
SymbolValue::Str(s) => write!(f, "\"{}\"", s),
SymbolValue::Double(d) => write!(f, "{}", d),
SymbolValue::Bool(b) => {
if *b {
write!(f, "True")
} else {
write!(f, "False")
}
}
SymbolValue::Tuple(t) => {
write!(f, "({})", t.iter().map(|v| format!("{}", v)).collect::<Vec<_>>().join(", "))
}
SymbolValue::OptionSome(v) => write!(f, "Some({})", v),
SymbolValue::OptionNone => write!(f, "none"),
}
}
}
pub trait StaticValue {
fn get_unique_identifier(&self) -> u64;
fn get_const_obj<'ctx, 'a>(
&self,
ctx: &mut CodeGenContext<'ctx, 'a>,
generator: &mut dyn CodeGenerator,
) -> BasicValueEnum<'ctx>;
fn to_basic_value_enum<'ctx, 'a>(
&self,
ctx: &mut CodeGenContext<'ctx, 'a>,
generator: &mut dyn CodeGenerator,
expected_ty: Type,
) -> Result<BasicValueEnum<'ctx>, String>;
fn get_field<'ctx, 'a>(
&self,
name: StrRef,
ctx: &mut CodeGenContext<'ctx, 'a>,
) -> Option<ValueEnum<'ctx>>;
fn get_tuple_element<'ctx>(&self, index: u32) -> Option<ValueEnum<'ctx>>;
}
#[derive(Clone)]
pub enum ValueEnum<'ctx> {
Static(Arc<dyn StaticValue + Send + Sync>),
Dynamic(BasicValueEnum<'ctx>),
}
impl<'ctx> From<BasicValueEnum<'ctx>> for ValueEnum<'ctx> {
fn from(v: BasicValueEnum<'ctx>) -> Self {
ValueEnum::Dynamic(v)
}
}
impl<'ctx> From<PointerValue<'ctx>> for ValueEnum<'ctx> {
fn from(v: PointerValue<'ctx>) -> Self {
ValueEnum::Dynamic(v.into())
}
}
impl<'ctx> From<IntValue<'ctx>> for ValueEnum<'ctx> {
fn from(v: IntValue<'ctx>) -> Self {
ValueEnum::Dynamic(v.into())
}
}
impl<'ctx> From<FloatValue<'ctx>> for ValueEnum<'ctx> {
fn from(v: FloatValue<'ctx>) -> Self {
ValueEnum::Dynamic(v.into())
}
}
impl<'ctx> From<StructValue<'ctx>> for ValueEnum<'ctx> {
fn from(v: StructValue<'ctx>) -> Self {
ValueEnum::Dynamic(v.into())
}
}
impl<'ctx> ValueEnum<'ctx> {
pub fn to_basic_value_enum<'a>(
self,
ctx: &mut CodeGenContext<'ctx, 'a>,
generator: &mut dyn CodeGenerator,
expected_ty: Type,
) -> Result<BasicValueEnum<'ctx>, String> {
match self {
ValueEnum::Static(v) => v.to_basic_value_enum(ctx, generator, expected_ty),
ValueEnum::Dynamic(v) => Ok(v),
}
}
}
pub trait SymbolResolver {
// get type of type variable identifier or top-level function type
fn get_symbol_type(
&self,
unifier: &mut Unifier,
top_level_defs: &[Arc<RwLock<TopLevelDef>>],
primitives: &PrimitiveStore,
str: StrRef,
) -> Result<Type, String>;
// get the top-level definition of identifiers
fn get_identifier_def(&self, str: StrRef) -> Result<DefinitionId, String>;
fn get_symbol_value<'ctx, 'a>(
&self,
str: StrRef,
ctx: &mut CodeGenContext<'ctx, 'a>,
) -> Option<ValueEnum<'ctx>>;
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;
fn handle_deferred_eval(
&self,
_unifier: &mut Unifier,
_top_level_defs: &[Arc<RwLock<TopLevelDef>>],
_primitives: &PrimitiveStore
) -> Result<(), String> {
Ok(())
}
}
thread_local! {
static IDENTIFIER_ID: [StrRef; 11] = [
"int32".into(),
"int64".into(),
"float".into(),
"bool".into(),
"virtual".into(),
"list".into(),
"tuple".into(),
"str".into(),
"Exception".into(),
"uint32".into(),
"uint64".into(),
];
}
// 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, String> {
use nac3parser::ast::ExprKind::*;
let ids = IDENTIFIER_ID.with(|ids| *ids);
let int32_id = ids[0];
let int64_id = ids[1];
let float_id = ids[2];
let bool_id = ids[3];
let virtual_id = ids[4];
let 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 {
Ok(primitives.int32)
} else if *id == int64_id {
Ok(primitives.int64)
} else if *id == uint32_id {
Ok(primitives.uint32)
} else if *id == uint64_id {
Ok(primitives.uint64)
} else if *id == float_id {
Ok(primitives.float)
} else if *id == bool_id {
Ok(primitives.bool)
} else if *id == str_id {
Ok(primitives.str)
} else if *id == exn_id {
Ok(primitives.exception)
} else {
let obj_id = resolver.get_identifier_def(*id);
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(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: Default::default(),
}))
} else {
Err(format!("Cannot use function name as type at {}", loc))
}
}
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(format!("Unknown type annotation {} at {}", id, loc))
}
}
}
}
};
let subscript_name_handle = |id: &StrRef, slice: &Expr<T>, unifier: &mut Unifier| {
if *id == virtual_id {
let ty = parse_type_annotation(resolver, top_level_defs, unifier, primitives, slice)?;
Ok(unifier.add_ty(TypeEnum::TVirtual { ty }))
} else if *id == 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
.iter()
.map(|elt| {
parse_type_annotation(resolver, top_level_defs, unifier, primitives, elt)
})
.collect::<Result<Vec<_>, _>>()?;
Ok(unifier.add_ty(TypeEnum::TTuple { ty }))
} else {
Err("Expected multiple elements for tuple".into())
}
} else {
let types = if let Tuple { elts, .. } = &slice.node {
elts.iter()
.map(|v| {
parse_type_annotation(resolver, top_level_defs, unifier, primitives, v)
})
.collect::<Result<Vec<_>, _>>()?
} else {
vec![parse_type_annotation(resolver, top_level_defs, unifier, primitives, slice)?]
};
let obj_id = resolver.get_identifier_def(*id)?;
let def = top_level_defs[obj_id.0].read();
if let TopLevelDef::Class { fields, methods, type_vars, .. } = &*def {
if types.len() != type_vars.len() {
return Err(format!(
"Unexpected number of type parameters: expected {} but got {}",
type_vars.len(),
types.len()
));
}
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
} else {
unreachable!()
};
subst.insert(id, *ty);
}
let mut fields = fields
.iter()
.map(|(attr, ty, is_mutable)| {
let ty = unifier.subst(*ty, &subst).unwrap_or(*ty);
(*attr, (ty, *is_mutable))
})
.collect::<HashMap<_, _>>();
fields.extend(methods.iter().map(|(attr, ty, _)| {
let ty = unifier.subst(*ty, &subst).unwrap_or(*ty);
(*attr, (ty, false))
}));
Ok(unifier.add_ty(TypeEnum::TObj { obj_id, fields, params: subst }))
} else {
Err("Cannot use function name as type".into())
}
}
};
match &expr.node {
Name { id, .. } => name_handling(id, expr.location, unifier),
Subscript { value, slice, .. } => {
if let Name { id, .. } = &value.node {
subscript_name_handle(id, slice, unifier)
} else {
Err(format!("unsupported type expression at {}", expr.location))
}
}
_ => Err(format!("unsupported type expression at {}", expr.location)),
}
}
impl dyn SymbolResolver + Send + Sync {
pub fn parse_type_annotation<T>(
&self,
top_level_defs: &[Arc<RwLock<TopLevelDef>>],
unifier: &mut Unifier,
primitives: &PrimitiveStore,
expr: &Expr<T>,
) -> Result<Type, String> {
parse_type_annotation(self, top_level_defs, unifier, primitives, expr)
}
pub fn get_type_name(
&self,
top_level_defs: &[Arc<RwLock<TopLevelDef>>],
unifier: &mut Unifier,
ty: Type,
) -> String {
unifier.internal_stringify(
ty,
&mut |id| {
if let TopLevelDef::Class { name, .. } = &*top_level_defs[id].read() {
name.to_string()
} else {
unreachable!("expected class definition")
}
},
&mut |id| format!("typevar{}", id),
&mut None,
)
}
}
impl Debug for dyn SymbolResolver + Send + Sync {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
write!(f, "")
}
}

File diff suppressed because it is too large Load Diff

File diff suppressed because it is too large Load Diff

View File

@ -0,0 +1,585 @@
use std::convert::TryInto;
use crate::symbol_resolver::SymbolValue;
use nac3parser::ast::{Constant, Location};
use super::*;
impl TopLevelDef {
pub fn to_string(&self, unifier: &mut Unifier) -> String {
match self {
TopLevelDef::Class { name, ancestors, fields, methods, type_vars, .. } => {
let fields_str = fields
.iter()
.map(|(n, ty, _)| (n.to_string(), unifier.stringify(*ty)))
.collect_vec();
let methods_str = methods
.iter()
.map(|(n, ty, id)| (n.to_string(), unifier.stringify(*ty), *id))
.collect_vec();
format!(
"Class {{\nname: {:?},\nancestors: {:?},\nfields: {:?},\nmethods: {:?},\ntype_vars: {:?}\n}}",
name,
ancestors.iter().map(|ancestor| ancestor.stringify(unifier)).collect_vec(),
fields_str.iter().map(|(a, _)| a).collect_vec(),
methods_str.iter().map(|(a, b, _)| (a, b)).collect_vec(),
type_vars.iter().map(|id| unifier.stringify(*id)).collect_vec(),
)
}
TopLevelDef::Function { name, signature, var_id, .. } => format!(
"Function {{\nname: {:?},\nsig: {:?},\nvar_id: {:?}\n}}",
name,
unifier.stringify(*signature),
{
// preserve the order for debug output and test
let mut r = var_id.clone();
r.sort_unstable();
r
}
),
}
}
}
impl TopLevelComposer {
pub fn make_primitives() -> (PrimitiveStore, Unifier) {
let mut unifier = Unifier::new();
let int32 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(0),
fields: HashMap::new(),
params: HashMap::new(),
});
let int64 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(1),
fields: HashMap::new(),
params: HashMap::new(),
});
let float = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(2),
fields: HashMap::new(),
params: HashMap::new(),
});
let bool = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(3),
fields: HashMap::new(),
params: HashMap::new(),
});
let none = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(4),
fields: HashMap::new(),
params: HashMap::new(),
});
let range = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(5),
fields: HashMap::new(),
params: HashMap::new(),
});
let str = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(6),
fields: HashMap::new(),
params: HashMap::new(),
});
let exception = unifier.add_ty(TypeEnum::TObj {
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()
.collect::<HashMap<_, _>>(),
params: HashMap::new(),
});
let uint32 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(8),
fields: HashMap::new(),
params: HashMap::new(),
});
let uint64 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(9),
fields: HashMap::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: 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.0,
vars: HashMap::from([(option_type_var.1, option_type_var.0)]),
}));
let option = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(10),
fields: vec![
("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: HashMap::from([(option_type_var.1, option_type_var.0)]),
});
let primitives = PrimitiveStore {
int32,
int64,
float,
bool,
none,
range,
str,
exception,
uint32,
uint64,
option,
};
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
pub fn make_top_level_class_def(
index: usize,
resolver: Option<Arc<dyn SymbolResolver + Send + Sync>>,
name: StrRef,
constructor: Option<Type>,
loc: Option<Location>,
) -> TopLevelDef {
TopLevelDef::Class {
name,
object_id: DefinitionId(index),
type_vars: Default::default(),
fields: Default::default(),
methods: Default::default(),
ancestors: Default::default(),
constructor,
resolver,
loc,
}
}
/// when first registering, the type is a invalid value
pub fn make_top_level_function_def(
name: String,
simple_name: StrRef,
ty: Type,
resolver: Option<Arc<dyn SymbolResolver + Send + Sync>>,
loc: Option<Location>,
) -> TopLevelDef {
TopLevelDef::Function {
name,
simple_name,
signature: ty,
var_id: Default::default(),
instance_to_symbol: Default::default(),
instance_to_stmt: Default::default(),
resolver,
codegen_callback: None,
loc,
}
}
pub fn make_class_method_name(mut class_name: String, method_name: &str) -> String {
class_name.push('.');
class_name.push_str(method_name);
class_name
}
pub fn get_class_method_def_info(
class_methods_def: &[(StrRef, Type, DefinitionId)],
method_name: StrRef,
) -> Result<(Type, DefinitionId), String> {
for (name, ty, def_id) in class_methods_def {
if name == &method_name {
return Ok((*ty, *def_id));
}
}
Err(format!("no method {} in the current class", method_name))
}
/// get all base class def id of a class, excluding itself. \
/// this function should called only after the direct parent is set
/// and before all the ancestors are set
/// and when we allow single inheritance \
/// the order of the returned list is from the child to the deepest ancestor
pub fn get_all_ancestors_helper(
child: &TypeAnnotation,
temp_def_list: &[Arc<RwLock<TopLevelDef>>],
) -> 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 {
parent = Self::get_parent(&p, temp_def_list);
let p_id = if let TypeAnnotation::CustomClass { id, .. } = &p {
*id
} else {
unreachable!("must be class kind annotation")
};
// check cycle
let no_cycle = result.iter().all(|x| {
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("cyclic inheritance detected".into());
}
}
Ok(result)
}
/// should only be called when finding all ancestors, so panic when wrong
fn get_parent(
child: &TypeAnnotation,
temp_def_list: &[Arc<RwLock<TopLevelDef>>],
) -> Option<TypeAnnotation> {
let child_id = if let TypeAnnotation::CustomClass { id, .. } = child {
*id
} else {
unreachable!("should be class type annotation")
};
let child_def = temp_def_list.get(child_id.0).unwrap();
let child_def = child_def.read();
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<u32, String> {
if let TypeEnum::TVar { id, .. } = unifier.get_ty(var_ty).as_ref() {
Ok(*id)
} else {
Err("not type var".to_string())
}
}
pub fn check_overload_function_type(
this: Type,
other: Type,
unifier: &mut Unifier,
type_var_to_concrete_def: &HashMap<Type, TypeAnnotation>,
) -> bool {
let this = unifier.get_ty(this);
let this = this.as_ref();
let other = unifier.get_ty(other);
let other = other.as_ref();
if let (
TypeEnum::TFunc(FunSignature { args: this_args, ret: this_ret, .. }),
TypeEnum::TFunc(FunSignature { args: other_args, ret: other_ret, .. }),
) = (this, other)
{
// check 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, .. }| {
(name, type_var_to_concrete_def.get(ty).unwrap())
}))
.all(|(this, other)| {
if this.0 == &"self".into() && this.0 == other.0 {
true
} else {
this.0 == other.0
&& check_overload_type_annotation_compatible(this.1, other.1, unifier)
}
});
// check rets
let ret_ok = check_overload_type_annotation_compatible(
type_var_to_concrete_def.get(this_ret).unwrap(),
type_var_to_concrete_def.get(other_ret).unwrap(),
unifier,
);
// return
args_ok && ret_ok
} else {
unreachable!("this function must be called with function type")
}
}
pub fn check_overload_field_type(
this: Type,
other: Type,
unifier: &mut Unifier,
type_var_to_concrete_def: &HashMap<Type, TypeAnnotation>,
) -> bool {
check_overload_type_annotation_compatible(
type_var_to_concrete_def.get(&this).unwrap(),
type_var_to_concrete_def.get(&other).unwrap(),
unifier,
)
}
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 {
ast::StmtKind::AnnAssign { target, .. }
if {
if let ast::ExprKind::Attribute { value, .. } = &target.node {
if let ast::ExprKind::Name { id, .. } = &value.node {
id == &"self".into()
} else {
false
}
} else {
false
}
} =>
{
return Err(format!(
"redundant type annotation for class fields at {}",
s.location
))
}
ast::StmtKind::Assign { targets, .. } => {
for t in targets {
if let ast::ExprKind::Attribute { value, attr, .. } = &t.node {
if let ast::ExprKind::Name { id, .. } = &value.node {
if id == &"self".into() {
result.insert(*attr);
}
}
}
}
}
// TODO: do not check for For and While?
ast::StmtKind::For { body, orelse, .. }
| ast::StmtKind::While { body, orelse, .. } => {
result.extend(Self::get_all_assigned_field(body.as_slice())?);
result.extend(Self::get_all_assigned_field(orelse.as_slice())?);
}
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())?)
.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())?)
.cloned()
.collect::<HashSet<_>>();
result.extend(inited_for_sure);
result.extend(Self::get_all_assigned_field(finalbody.as_slice())?);
}
ast::StmtKind::With { body, .. } => {
result.extend(Self::get_all_assigned_field(body.as_slice())?);
}
ast::StmtKind::Pass { .. } => {}
ast::StmtKind::Assert { .. } => {}
ast::StmtKind::Expr { .. } => {}
_ => {
unimplemented!()
}
}
}
Ok(result)
}
pub fn parse_parameter_default_value(
default: &ast::Expr,
resolver: &(dyn SymbolResolver + Send + Sync),
) -> Result<SymbolValue, String> {
parse_parameter_default_value(default, resolver)
}
pub fn check_default_param_type(
val: &SymbolValue,
ty: &TypeAnnotation,
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,
unifier: &mut Unifier,
primitive: &PrimitiveStore,
) -> bool {
match (found, expect) {
(TypeAnnotation::Primitive(f), TypeAnnotation::Primitive(e)) => {
unifier.unioned(*f, *e)
}
(
TypeAnnotation::CustomClass { id: f_id, params: f_param },
TypeAnnotation::CustomClass { id: e_id, params: e_param },
) => {
*f_id == *e_id
&& *f_id == primitive.option.get_obj_id(unifier)
&& (f_param.is_empty()
|| (f_param.len() == 1
&& e_param.len() == 1
&& is_compatible(&f_param[0], &e_param[0], unifier, primitive)))
}
(TypeAnnotation::Tuple(f), TypeAnnotation::Tuple(e)) => {
f.len() == e.len()
&& f.iter()
.zip(e.iter())
.all(|(f, e)| is_compatible(f, e, unifier, primitive))
}
_ => false,
}
}
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(())
}
}
}
pub fn parse_parameter_default_value(
default: &ast::Expr,
resolver: &(dyn SymbolResolver + Send + Sync),
) -> 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(format!("integer value out of range at {}", loc))
}
}
Constant::Float(v) => Ok(SymbolValue::Double(*v)),
Constant::Bool(v) => Ok(SymbolValue::Bool(*v)),
Constant::Tuple(tuple) => Ok(SymbolValue::Tuple(
tuple.iter().map(|x| handle_constant(x, loc)).collect::<Result<Vec<_>, _>>()?,
)),
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::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(format!("default param value out of range 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(format!("default param value out of range 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(format!("default param value out of range at {}", default.location)),
}
}
_ => 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<_>, _>>()?
)),
ast::ExprKind::Name { id, .. } if id == &"none".into() => Ok(SymbolValue::OptionNone),
ast::ExprKind::Name { id, .. } => {
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
)
)
}
_ => Err(format!(
"unsupported default parameter (not primitive type, option or tuple) at {}",
default.location
))
}
}

View File

@ -0,0 +1,138 @@
use std::{
borrow::BorrowMut,
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};
use crate::{
codegen::CodeGenerator,
symbol_resolver::{SymbolResolver, ValueEnum},
typecheck::{type_inferencer::CodeLocation, typedef::CallId},
};
use inkwell::values::BasicValueEnum;
use itertools::{izip, Itertools};
use nac3parser::ast::{self, Location, Stmt, StrRef};
use parking_lot::RwLock;
#[derive(PartialEq, Eq, PartialOrd, Ord, Clone, Copy, Hash, Debug)]
pub struct DefinitionId(pub usize);
pub mod builtins;
pub mod composer;
pub mod helper;
pub mod type_annotation;
use composer::*;
use type_annotation::*;
#[cfg(test)]
mod test;
type GenCallCallback = Box<
dyn for<'ctx, 'a> Fn(
&mut CodeGenContext<'ctx, 'a>,
Option<(Type, ValueEnum<'ctx>)>,
(&FunSignature, DefinitionId),
Vec<(Option<StrRef>, ValueEnum<'ctx>)>,
&mut dyn CodeGenerator,
) -> Result<Option<BasicValueEnum<'ctx>>, String>
+ Send
+ Sync,
>;
pub struct GenCall {
fp: GenCallCallback,
}
impl GenCall {
pub fn new(fp: GenCallCallback) -> GenCall {
GenCall { fp }
}
pub fn run<'ctx, 'a>(
&self,
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> {
(self.fp)(ctx, obj, fun, args, generator)
}
}
impl Debug for GenCall {
fn fmt(&self, _: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
Ok(())
}
}
#[derive(Clone, Debug)]
pub struct FunInstance {
pub body: Arc<Vec<Stmt<Option<Type>>>>,
pub calls: Arc<HashMap<CodeLocation, CallId>>,
pub subst: HashMap<u32, Type>,
pub unifier_id: usize,
}
#[derive(Debug, Clone)]
pub enum TopLevelDef {
Class {
// name for error messages and symbols
name: StrRef,
// object ID used for TypeEnum
object_id: DefinitionId,
/// type variables bounded to the class.
type_vars: Vec<Type>,
// class fields
// name, type, is mutable
fields: Vec<(StrRef, Type, bool)>,
// class methods, pointing to the corresponding function definition.
methods: Vec<(StrRef, Type, DefinitionId)>,
// ancestor classes, including itself.
ancestors: Vec<TypeAnnotation>,
// symbol resolver of the module defined the class, none if it is built-in type
resolver: Option<Arc<dyn SymbolResolver + Send + Sync>>,
// constructor type
constructor: Option<Type>,
// definition location
loc: Option<Location>,
},
Function {
// prefix for symbol, should be unique globally
name: String,
// simple name, the same as in method/function definition
simple_name: StrRef,
// function signature.
signature: Type,
// instantiated type variable IDs
var_id: Vec<u32>,
/// Function instance to symbol mapping
/// Key: string representation of type variable values, sorted by variable ID in ascending
/// order, including type variables associated with the class.
/// Value: function symbol name.
instance_to_symbol: HashMap<String, String>,
/// Function instances to annotated AST mapping
/// Key: string representation of type variable values, sorted by variable ID in ascending
/// order, including type variables associated with the class. Excluding rigid type
/// variables.
/// rigid type variables that would be substituted when the function is instantiated.
instance_to_stmt: HashMap<String, FunInstance>,
// symbol resolver of the module defined the class
resolver: Option<Arc<dyn SymbolResolver + Send + Sync>>,
// custom codegen callback
codegen_callback: Option<Arc<GenCall>>,
// definition location
loc: Option<Location>,
},
}
pub struct TopLevelContext {
pub definitions: Arc<RwLock<Vec<Arc<RwLock<TopLevelDef>>>>>,
pub unifiers: Arc<RwLock<Vec<(SharedUnifier, PrimitiveStore)>>>,
pub personality_symbol: Option<String>,
}

View File

@ -0,0 +1,14 @@
---
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: [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",
]

View File

@ -0,0 +1,17 @@
---
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[typevar7]\", \"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: [\"typevar7\"]\n}\n",
"Function {\nname: \"B.__init__\",\nsig: \"fn[[], none]\",\nvar_id: []\n}\n",
"Function {\nname: \"B.fun\",\nsig: \"fn[[a:int32, b:T], list[virtual[B[bool]]]]\",\nvar_id: []\n}\n",
"Class {\nname: \"C\",\nancestors: [\"C\", \"B[bool]\", \"A[float]\"],\nfields: [\"a\", \"b\", \"c\", \"d\", \"e\"],\nmethods: [(\"__init__\", \"fn[[], none]\"), (\"fun\", \"fn[[a:int32, b:T], list[virtual[B[bool]]]]\"), (\"foo\", \"fn[[c:C], none]\")],\ntype_vars: []\n}\n",
"Function {\nname: \"C.__init__\",\nsig: \"fn[[], none]\",\nvar_id: []\n}\n",
]

View File

@ -0,0 +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: [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",
]

View File

@ -0,0 +1,15 @@
---
source: nac3core/src/toplevel/test.rs
assertion_line: 549
expression: res_vec
---
[
"Class {\nname: \"A\",\nancestors: [\"A[typevar6, typevar7]\"],\nfields: [\"a\", \"b\"],\nmethods: [(\"__init__\", \"fn[[a:A[bool, float], b:B], none]\"), (\"fun\", \"fn[[a:A[bool, float]], A[bool, int32]]\")],\ntype_vars: [\"typevar6\", \"typevar7\"]\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[int32, list[B]]], tuple[A[bool, virtual[A[B, int32]]], B]]\",\nvar_id: []\n}\n",
]

View File

@ -0,0 +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: [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: [34]\n}\n",
]

View File

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

View File

@ -0,0 +1,801 @@
use crate::{
codegen::CodeGenContext,
symbol_resolver::{SymbolResolver, ValueEnum},
toplevel::DefinitionId,
typecheck::{
type_inferencer::PrimitiveStore,
typedef::{Type, Unifier},
},
};
use indoc::indoc;
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>>,
class_names: Mutex<HashMap<StrRef, Type>>,
}
impl ResolverInternal {
fn add_id_def(&self, id: StrRef, def: DefinitionId) {
self.id_to_def.lock().insert(id, def);
}
fn add_id_type(&self, id: StrRef, ty: Type) {
self.id_to_type.lock().insert(id, ty);
}
}
struct Resolver(Arc<ResolverInternal>);
impl SymbolResolver for Resolver {
fn get_default_param_value(
&self,
_: &nac3parser::ast::Expr,
) -> Option<crate::symbol_resolver::SymbolValue> {
unimplemented!()
}
fn get_symbol_type(
&self,
_: &mut Unifier,
_: &[Arc<RwLock<TopLevelDef>>],
_: &PrimitiveStore,
str: StrRef,
) -> Result<Type, String> {
self.0
.id_to_type
.lock()
.get(&str)
.cloned()
.ok_or_else(|| format!("cannot find symbol `{}`", str))
}
fn get_symbol_value<'ctx, 'a>(
&self,
_: StrRef,
_: &mut CodeGenContext<'ctx, 'a>,
) -> Option<ValueEnum<'ctx>> {
unimplemented!()
}
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 {
unimplemented!()
}
fn get_exception_id(&self, _tyid: usize) -> usize {
unimplemented!()
}
}
#[test_case(
vec![
indoc! {"
def fun(a: int32) -> int32:
return a
"},
indoc! {"
class A:
def __init__(self):
self.a: int32 = 3
"},
indoc! {"
class B:
def __init__(self):
self.b: float = 4.3
def fun(self):
self.b = self.b + 3.0
"},
indoc! {"
def foo(a: float):
a + 1.0
"},
indoc! {"
class C(B):
def __init__(self):
self.c: int32 = 4
self.a: bool = True
"}
];
"register"
)]
fn test_simple_register(source: Vec<&str>) {
let mut composer: TopLevelComposer = Default::default();
for s in source {
let ast = parse_program(s, Default::default()).unwrap();
let ast = ast[0].clone();
composer.register_top_level(ast, None, "".into()).unwrap();
}
}
#[test_case(
vec![
indoc! {"
def fun(a: int32) -> int32:
return a
"},
indoc! {"
def foo(a: float):
a + 1.0
"},
indoc! {"
def f(b: int64) -> int32:
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: Vec<&str>, tys: Vec<&str>, names: Vec<&str>) {
let mut composer: TopLevelComposer = Default::default();
let internal_resolver = Arc::new(ResolverInternal {
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, Default::default()).unwrap();
let ast = ast[0].clone();
let (id, def_id, ty) =
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);
}
}
composer.start_analysis(true).unwrap();
for (i, (def, _)) in composer.definition_ast_list.iter().skip(composer.builtin_num).enumerate()
{
let def = &*def.read();
if let TopLevelDef::Function { signature, name, .. } = def {
let ty_str = composer.unifier.internal_stringify(
*signature,
&mut |id| id.to_string(),
&mut |id| id.to_string(),
&mut None,
);
assert_eq!(ty_str, tys[i]);
assert_eq!(name, names[i]);
}
}
}
#[test_case(
vec![
indoc! {"
class A():
a: int32
def __init__(self):
self.a = 3
def fun(self, b: B):
pass
def foo(self, a: T, b: V):
pass
"},
indoc! {"
class B(C):
def __init__(self):
pass
"},
indoc! {"
class C(A):
def __init__(self):
pass
def fun(self, b: B):
a = 1
pass
"},
indoc! {"
def foo(a: A):
pass
"},
indoc! {"
def ff(a: T) -> V:
pass
"}
],
vec![];
"simple class compose"
)]
#[test_case(
vec![
indoc! {"
class Generic_A(Generic[V], B):
a: int64
def __init__(self):
self.a = 123123123123
def fun(self, a: int32) -> V:
pass
"},
indoc! {"
class B:
aa: bool
def __init__(self):
self.aa = False
def foo(self, b: T):
pass
"}
],
vec![];
"generic class"
)]
#[test_case(
vec![
indoc! {"
def foo(a: list[int32], b: tuple[T, float]) -> A[B, bool]:
pass
"},
indoc! {"
class A(Generic[T, V]):
a: T
b: V
def __init__(self, v: V):
self.a = 1
self.b = v
def fun(self, a: T) -> V:
pass
"},
indoc! {"
def gfun(a: A[list[float], int32]):
pass
"},
indoc! {"
class B:
def __init__(self):
pass
"}
],
vec![];
"list tuple generic"
)]
#[test_case(
vec![
indoc! {"
class A(Generic[T, V]):
a: A[float, bool]
b: B
def __init__(self, a: A[float, bool], b: B):
self.a = a
self.b = b
def fun(self, a: A[float, bool]) -> A[bool, int32]:
pass
"},
indoc! {"
class B(A[int64, bool]):
def __init__(self):
pass
def foo(self, b: B) -> B:
pass
def bar(self, a: A[list[B], int32]) -> tuple[A[virtual[A[B, int32]], bool], B]:
pass
"}
],
vec![];
"self1"
)]
#[test_case(
vec![
indoc! {"
class A(Generic[T]):
a: int32
b: T
c: A[int64]
def __init__(self, t: T):
self.a = 3
self.b = T
def fun(self, a: int32, b: T) -> list[virtual[B[bool]]]:
pass
def foo(self, c: C):
pass
"},
indoc! {"
class B(Generic[V], A[float]):
d: C
def __init__(self):
pass
def fun(self, a: int32, b: T) -> list[virtual[B[bool]]]:
# override
pass
"},
indoc! {"
class C(B[bool]):
e: int64
def __init__(self):
pass
"}
],
vec![];
"inheritance_override"
)]
#[test_case(
vec![
indoc! {"
class A(Generic[T]):
def __init__(self):
pass
def fun(self, a: A[T]) -> A[T]:
pass
"}
],
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):
pass
"},
indoc! {"
class B(A):
def __init__(self):
pass
"}
],
vec!["cyclic inheritance detected"];
"cyclic1"
)]
#[test_case(
vec![
indoc! {"
class A(B[bool, int64]):
def __init__(self):
pass
"},
indoc! {"
class B(Generic[V, T], C[int32]):
def __init__(self):
pass
"},
indoc! {"
class C(Generic[T], A):
def __init__(self):
pass
"},
],
vec!["cyclic inheritance detected"];
"cyclic2"
)]
#[test_case(
vec![
indoc! {"
class A:
pass
"}
],
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(
vec![indoc! {"
class A:
def __init__():
pass
"}],
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):
pass
"},
indoc! {"
class B:
def __init__(self):
pass
"},
indoc! {"
class C:
def __init__(self):
pass
"}
],
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
b: T
c: A[int64]
def __init__(self, t: T):
self.a = 3
self.b = T
def fun(self, a: int32, b: T) -> list[virtual[B[bool]]]:
pass
"},
indoc! {"
class B(Generic[V], A[float]):
def __init__(self):
pass
def fun(self, a: int32, b: T) -> list[virtual[B[int32]]]:
# override
pass
"}
],
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
b: T
c: A[int64]
def __init__(self, t: T):
self.a = 3
self.b = T
def fun(self, a: int32, b: T) -> list[virtual[B[bool]]]:
pass
"},
indoc! {"
class B(Generic[V], A[float]):
a: int32
def __init__(self):
pass
def fun(self, a: int32, b: T) -> list[virtual[B[bool]]]:
# override
pass
"}
],
vec!["field `a` has already declared in the ancestor classes"];
"err_incompatible_inheritance_field"
)]
#[test_case(
vec![
indoc! {"
class A:
def __init__(self):
pass
"},
indoc! {"
class A:
a: int32
def __init__(self):
pass
"}
],
vec!["duplicate definition of class `A` (at unknown: line 1 column 1)"];
"class same name"
)]
fn test_analyze(source: Vec<&str>, res: Vec<&str>) {
let print = false;
let mut composer: TopLevelComposer = Default::default();
let internal_resolver = make_internal_resolver_with_tvar(
vec![
("T".into(), vec![]),
("V".into(), vec![composer.primitives_ty.bool, composer.primitives_ty.int32]),
("G".into(), vec![composer.primitives_ty.bool, composer.primitives_ty.int64]),
],
&mut composer.unifier,
print,
);
let resolver =
Arc::new(Resolver(internal_resolver.clone())) as Arc<dyn SymbolResolver + Send + Sync>;
for s in source {
let ast = parse_program(s, Default::default()).unwrap();
let ast = ast[0].clone();
let (id, def_id, ty) = {
match composer.register_top_level(ast, Some(resolver.clone()), "".into()) {
Ok(x) => x,
Err(msg) => {
if print {
println!("{}", msg);
} else {
assert_eq!(res[0], msg);
}
return;
}
}
};
internal_resolver.add_id_def(id, def_id);
if let Some(ty) = ty {
internal_resolver.add_id_type(id, ty);
}
}
if let Err(msg) = composer.start_analysis(false) {
if print {
println!("{}", msg);
} else {
assert_eq!(res[0], msg);
}
} else {
// skip 5 to skip primitives
let mut res_vec: Vec<String> = Vec::new();
for (def, _) in composer.definition_ast_list.iter().skip(composer.builtin_num) {
let def = &*def.read();
res_vec.push(format!("{}\n", def.to_string(composer.unifier.borrow_mut())));
}
insta::assert_debug_snapshot!(res_vec);
}
}
#[test_case(
vec![
indoc! {"
def fun(a: int32, b: int32) -> int32:
return a + b
"},
indoc! {"
def fib(n: int32) -> int32:
if n <= 2:
return 1
a = fib(n - 1)
b = fib(n - 2)
return fib(n - 1)
"}
],
vec![];
"simple function"
)]
#[test_case(
vec![
indoc! {"
class A:
a: int32
def __init__(self):
self.a = 3
def fun(self) -> int32:
b = self.a + 3
return b * self.a
def clone(self) -> A:
SELF = self
return SELF
def sum(self) -> int32:
if self.a == 0:
return self.a
else:
a = self.a
self.a = self.a - 1
return a + self.sum()
def fib(self, a: int32) -> int32:
if a <= 2:
return 1
return self.fib(a - 1) + self.fib(a - 2)
"},
indoc! {"
def fun(a: A) -> int32:
return a.fun() + 2
"}
],
vec![];
"simple class body"
)]
#[test_case(
vec![
indoc! {"
def fun(a: V, c: G, t: T) -> V:
b = a
cc = c
ret = fun(b, cc, t)
return ret * ret
"},
indoc! {"
def sum_three(l: list[V]) -> V:
return l[0] + l[1] + l[2]
"},
indoc! {"
def sum_sq_pair(p: tuple[V, V]) -> list[V]:
a = p[0]
b = p[1]
a = a**a
b = b**b
return [a, b]
"}
],
vec![];
"type var fun"
)]
#[test_case(
vec![
indoc! {"
class A(Generic[G]):
a: G
b: bool
def __init__(self, aa: G):
self.a = aa
if 2 > 1:
self.b = True
else:
# self.b = False
pass
def fun(self, a: G) -> list[G]:
ret = [a, self.a]
return ret if self.b else self.fun(self.a)
"}
],
vec![];
"type var class"
)]
#[test_case(
vec![
indoc! {"
class A:
def fun(self):
pass
"},
indoc!{"
class B:
a: int32
b: bool
def __init__(self):
# self.b = False
if 3 > 2:
self.a = 3
self.b = False
else:
self.a = 4
self.b = True
"}
],
vec![];
"no_init_inst_check"
)]
fn test_inference(source: Vec<&str>, res: Vec<&str>) {
let print = true;
let mut composer: TopLevelComposer = Default::default();
let internal_resolver = make_internal_resolver_with_tvar(
vec![
("T".into(), vec![]),
(
"V".into(),
vec![
composer.primitives_ty.float,
composer.primitives_ty.int32,
composer.primitives_ty.int64,
],
),
("G".into(), vec![composer.primitives_ty.bool, composer.primitives_ty.int64]),
],
&mut composer.unifier,
print,
);
let resolver =
Arc::new(Resolver(internal_resolver.clone())) as Arc<dyn SymbolResolver + Send + Sync>;
for s in source {
let ast = parse_program(s, Default::default()).unwrap();
let ast = ast[0].clone();
let (id, def_id, ty) = {
match composer.register_top_level(ast, Some(resolver.clone()), "".into()) {
Ok(x) => x,
Err(msg) => {
if print {
println!("{}", msg);
} else {
assert_eq!(res[0], msg);
}
return;
}
}
};
internal_resolver.add_id_def(id, def_id);
if let Some(ty) = ty {
internal_resolver.add_id_type(id, ty);
}
}
if let Err(msg) = composer.start_analysis(true) {
if print {
println!("{}", msg);
} else {
assert_eq!(res[0], msg);
}
} else {
// skip 5 to skip primitives
let mut stringify_folder = TypeToStringFolder { unifier: &mut composer.unifier };
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 {
println!(
"=========`{}`: number of instances: {}===========",
name,
instance_to_stmt.len()
);
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());
println!("--------------------");
}
println!("\n");
}
}
}
}
}
fn make_internal_resolver_with_tvar(
tvars: Vec<(StrRef, Vec<Type>)>,
unifier: &mut Unifier,
print: bool,
) -> Arc<ResolverInternal> {
let res: Arc<ResolverInternal> = ResolverInternal {
id_to_def: Default::default(),
id_to_type: tvars
.into_iter()
.map(|(name, range)| {
(name, {
let (ty, id) = unifier.get_fresh_var_with_range(range.as_slice(), None, None);
if print {
println!("{}: {:?}, typevar{}", name, ty, id);
}
ty
})
})
.collect::<HashMap<_, _>>()
.into(),
class_names: Default::default(),
}
.into();
if print {
println!();
}
res
}
struct TypeToStringFolder<'a> {
unifier: &'a mut Unifier,
}
impl<'a> Fold<Option<Type>> for TypeToStringFolder<'a> {
type TargetU = String;
type Error = String;
fn map_user(&mut self, user: Option<Type>) -> Result<Self::TargetU, Self::Error> {
Ok(if let Some(ty) = user {
self.unifier.internal_stringify(
ty,
&mut |id| format!("class{}", id.to_string()),
&mut |id| format!("typevar{}", id.to_string()),
&mut None,
)
} else {
"None".into()
})
}
}

View File

@ -0,0 +1,529 @@
use super::*;
#[derive(Clone, Debug)]
pub enum TypeAnnotation {
Primitive(Type),
// we use type vars kind at params to represent self type
CustomClass {
id: DefinitionId,
// params can also be type var
params: Vec<TypeAnnotation>,
},
// can only be CustomClassKind
Virtual(Box<TypeAnnotation>),
TypeVar(Type),
List(Box<TypeAnnotation>),
Tuple(Vec<TypeAnnotation>),
}
impl TypeAnnotation {
pub fn stringify(&self, unifier: &mut Unifier) -> String {
use TypeAnnotation::*;
match self {
Primitive(ty) | TypeVar(ty) => unifier.stringify(*ty),
CustomClass { id, params } => {
let class_name = match unifier.top_level {
Some(ref top) => {
if let TopLevelDef::Class { name, .. } =
&*top.definitions.read()[id.0].read()
{
(*name).into()
} else {
unreachable!()
}
}
None => format!("class_def_{}", id.0),
};
format!(
"{}{}",
class_name,
{
let param_list = params.iter().map(|p| p.stringify(unifier)).collect_vec().join(", ");
if param_list.is_empty() {
"".into()
} else {
format!("[{}]", param_list)
}
}
)
}
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(", "))
}
}
}
}
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>>,
) -> Result<TypeAnnotation, String> {
let name_handle = |id: &StrRef,
unifier: &mut Unifier,
locked: HashMap<DefinitionId, Vec<Type>>| {
if id == &"int32".into() {
Ok(TypeAnnotation::Primitive(primitives.int32))
} else if id == &"int64".into() {
Ok(TypeAnnotation::Primitive(primitives.int64))
} else if id == &"uint32".into() {
Ok(TypeAnnotation::Primitive(primitives.uint32))
} else if id == &"uint64".into() {
Ok(TypeAnnotation::Primitive(primitives.uint64))
} else if id == &"float".into() {
Ok(TypeAnnotation::Primitive(primitives.float))
} else if id == &"bool".into() {
Ok(TypeAnnotation::Primitive(primitives.bool))
} else if id == &"str".into() {
Ok(TypeAnnotation::Primitive(primitives.str))
} else if id == &"Exception".into() {
Ok(TypeAnnotation::CustomClass { id: 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();
if let Some(def_read) = def_read {
if let TopLevelDef::Class { type_vars, .. } = &*def_read {
type_vars.clone()
} else {
return Err(format!(
"function cannot be used as a type (at {})",
expr.location
));
}
} else {
locked.get(&obj_id).unwrap().clone()
}
};
// check param number here
if !type_vars.is_empty() {
return Err(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)).0;
unifier.unify(var, ty).unwrap();
Ok(TypeAnnotation::TypeVar(ty))
} else {
Err(format!("`{}` is not a valid type annotation (at {})", id, expr.location))
}
} else {
Err(format!("`{}` is not a valid type annotation (at {})", id, expr.location))
}
};
let class_name_handle =
|id: &StrRef,
slice: &ast::Expr<T>,
unifier: &mut Unifier,
mut locked: HashMap<DefinitionId, Vec<Type>>| {
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 {
if let TopLevelDef::Class { type_vars, .. } = &*def_read {
type_vars.clone()
} else {
unreachable!("must be class here")
}
} else {
locked.get(&obj_id).unwrap().clone()
}
};
// we do not check whether the application of type variables are compatible here
let param_type_infos = {
let params_ast = if let ast::ExprKind::Tuple { elts, .. } = &slice.node {
elts.iter().collect_vec()
} else {
vec![slice]
};
if type_vars.len() != params_ast.len() {
return Err(format!(
"expect {} type parameters but got {} (at {})",
type_vars.len(),
params_ast.len(),
params_ast[0].location,
));
}
let result = params_ast
.iter()
.map(|x| {
parse_ast_to_type_annotation_kinds(
resolver,
top_level_defs,
unifier,
primitives,
x,
{
locked.insert(obj_id, type_vars.clone());
locked.clone()
},
)
})
.collect::<Result<Vec<_>, _>>()?;
// make sure the result do not contain any type vars
let no_type_var =
result.iter().all(|x| get_type_var_contained_in_type_annotation(x).is_empty());
if no_type_var {
result
} else {
return Err(format!(
"application of type vars to generic class \
is not currently supported (at {})",
params_ast[0].location
));
}
};
Ok(TypeAnnotation::CustomClass { id: obj_id, params: param_type_infos })
};
match &expr.node {
ast::ExprKind::Name { id, .. } => name_handle(id, unifier, locked),
// virtual
ast::ExprKind::Subscript { value, slice, .. }
if {
matches!(&value.node, ast::ExprKind::Name { id, .. } if id == &"virtual".into())
} =>
{
let def = parse_ast_to_type_annotation_kinds(
resolver,
top_level_defs,
unifier,
primitives,
slice.as_ref(),
locked,
)?;
if !matches!(def, TypeAnnotation::CustomClass { .. }) {
unreachable!("must be concretized custom class kind in the virtual")
}
Ok(TypeAnnotation::Virtual(def.into()))
}
// 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 {
matches!(&value.node, ast::ExprKind::Name { id, .. } if id == &"Option".into())
} =>
{
let def_ann = parse_ast_to_type_annotation_kinds(
resolver,
top_level_defs,
unifier,
primitives,
slice.as_ref(),
locked,
)?;
let id =
if let TypeEnum::TObj { obj_id, .. } = unifier.get_ty(primitives.option).as_ref() {
*obj_id
} else {
unreachable!()
};
Ok(TypeAnnotation::CustomClass { id, params: vec![def_ann] })
}
// tuple
ast::ExprKind::Subscript { value, slice, .. }
if {
matches!(&value.node, ast::ExprKind::Name { id, .. } if id == &"tuple".into())
} =>
{
let tup_elts = {
if let ast::ExprKind::Tuple { elts, .. } = &slice.node {
elts.as_slice()
} else {
std::slice::from_ref(slice.as_ref())
}
};
let type_annotations = tup_elts
.iter()
.map(|e| {
parse_ast_to_type_annotation_kinds(
resolver,
top_level_defs,
unifier,
primitives,
e,
locked.clone(),
)
})
.collect::<Result<Vec<_>, _>>()?;
Ok(TypeAnnotation::Tuple(type_annotations))
}
// custom class
ast::ExprKind::Subscript { value, slice, .. } => {
if let ast::ExprKind::Name { id, .. } = &value.node {
class_name_handle(id, slice, unifier, locked)
} else {
Err(format!("unsupported expression type for class name (at {})", value.location))
}
}
_ => Err(format!("unsupported expression for type annotation (at {})", expr.location)),
}
}
// no need to have the `locked` parameter, unlike the `parse_ast_to_type_annotation_kinds`, since
// when calling this function, there should be no topleveldef::class being write, and this function
// also only read the toplevedefs
pub fn get_type_from_type_annotation_kinds(
top_level_defs: &[Arc<RwLock<TopLevelDef>>],
unifier: &mut Unifier,
primitives: &PrimitiveStore,
ann: &TypeAnnotation,
subst_list: &mut Option<Vec<Type>>
) -> Result<Type, 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.deref();
if let TopLevelDef::Class { fields, methods, type_vars, .. } = class_def {
if type_vars.len() != params.len() {
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,
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: HashMap<u32, Type> = HashMap::new();
for (tvar, p) in type_vars.iter().zip(param_ty) {
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 || {
let temp = unifier.get_fresh_var_with_range(
range.as_slice(),
*name,
*loc,
);
unifier.unify(temp.0, p).is_ok()
}
};
if ok {
result.insert(*id, p);
} else {
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 None
),
*id
));
}
} else {
unreachable!("must be generic type var")
}
}
result
};
let mut tobj_fields = methods
.iter()
.map(|(name, ty, _)| {
let subst_ty = unifier.subst(*ty, &subst).unwrap_or(*ty);
// methods are immutable
(*name, (subst_ty, false))
})
.collect::<HashMap<_, _>>();
tobj_fields.extend(fields.iter().map(|(name, ty, mutability)| {
let subst_ty = unifier.subst(*ty, &subst).unwrap_or(*ty);
(*name, (subst_ty, *mutability))
}));
let need_subst = !subst.is_empty();
let ty = unifier.add_ty(TypeEnum::TObj {
obj_id: *obj_id,
fields: tobj_fields,
params: subst,
});
if need_subst {
subst_list.as_mut().map(|wl| wl.push(ty));
}
Ok(ty)
}
} 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
)?;
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, primitives, x, subst_list)
})
.collect::<Result<Vec<_>, _>>()?;
Ok(unifier.add_ty(TypeEnum::TTuple { ty: tys }))
}
}
}
/// given an def id, return a type annotation of self \
/// ```python
/// class A(Generic[T, V]):
/// def fun(self):
/// ```
/// the type of `self` should be similar to `A[T, V]`, where `T`, `V`
/// considered to be type variables associated with the class \
/// \
/// But note that here we do not make a duplication of `T`, `V`, we directly
/// use them as they are in the TopLevelDef::Class since those in the
/// TopLevelDef::Class.type_vars will be substitute later when seeing applications/instantiations
/// the Type of their fields and methods will also be subst when application/instantiation
pub fn make_self_type_annotation(type_vars: &[Type], object_id: DefinitionId) -> TypeAnnotation {
TypeAnnotation::CustomClass {
id: object_id,
params: type_vars.iter().map(|ty| TypeAnnotation::TypeVar(*ty)).collect_vec(),
}
}
/// get all the occurences of type vars contained in a type annotation
/// e.g. `A[int, B[T], V, virtual[C[G]]]` => [T, V, G]
/// this function will not make a duplicate of type var
pub fn get_type_var_contained_in_type_annotation(ann: &TypeAnnotation) -> Vec<TypeAnnotation> {
let mut result: Vec<TypeAnnotation> = Vec::new();
match ann {
TypeAnnotation::TypeVar(..) => result.push(ann.clone()),
TypeAnnotation::Virtual(ann) => {
result.extend(get_type_var_contained_in_type_annotation(ann.as_ref()))
}
TypeAnnotation::CustomClass { params, .. } => {
for p in params {
result.extend(get_type_var_contained_in_type_annotation(p));
}
}
TypeAnnotation::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(..) => {}
}
result
}
/// check the type compatibility for overload
pub fn check_overload_type_annotation_compatible(
this: &TypeAnnotation,
other: &TypeAnnotation,
unifier: &mut Unifier,
) -> bool {
match (this, other) {
(TypeAnnotation::Primitive(a), TypeAnnotation::Primitive(b)) => a == b,
(TypeAnnotation::TypeVar(a), TypeAnnotation::TypeVar(b)) => {
let a = unifier.get_ty(*a);
let a = a.deref();
let b = unifier.get_ty(*b);
let b = b.deref();
if let (
TypeEnum::TVar { id: a, fields: None, .. },
TypeEnum::TVar { id: b, fields: None, .. },
) = (a, b)
{
a == b
} else {
unreachable!("must be type var")
}
}
(TypeAnnotation::Virtual(a), TypeAnnotation::Virtual(b))
| (TypeAnnotation::List(a), TypeAnnotation::List(b)) => {
check_overload_type_annotation_compatible(a.as_ref(), b.as_ref(), unifier)
}
(TypeAnnotation::Tuple(a), TypeAnnotation::Tuple(b)) => {
a.len() == b.len() && {
a.iter()
.zip(b)
.all(|(a, b)| check_overload_type_annotation_compatible(a, b, unifier))
}
}
(
TypeAnnotation::CustomClass { id: a, params: a_p },
TypeAnnotation::CustomClass { id: b, params: b_p },
) => {
a.0 == b.0 && {
a_p.len() == b_p.len() && {
a_p.iter()
.zip(b_p)
.all(|(a, b)| check_overload_type_annotation_compatible(a, b, unifier))
}
}
}
_ => false,
}
}

View File

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

View File

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

View File

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

View File

@ -0,0 +1,306 @@
use crate::typecheck::typedef::TypeEnum;
use super::type_inferencer::Inferencer;
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<(), String> {
if matches!(expr.custom, Some(ty) if self.unifier.unioned(ty, self.primitives.none)) {
Err(format!("Error at {}: cannot have value none", expr.location))
} else {
Ok(())
}
}
fn check_pattern(
&mut self,
pattern: &Expr<Option<Type>>,
defined_identifiers: &mut HashSet<StrRef>,
) -> Result<(), String> {
match &pattern.node {
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);
}
self.should_have_value(pattern)?;
Ok(())
}
ExprKind::Tuple { elts, .. } => {
for elt in elts.iter() {
self.check_pattern(elt, defined_identifiers)?;
self.should_have_value(elt)?;
}
Ok(())
}
ExprKind::Subscript { value, slice, .. } => {
self.check_expr(value, defined_identifiers)?;
self.should_have_value(value)?;
self.check_expr(slice, defined_identifiers)?;
if let TypeEnum::TTuple { .. } = &*self.unifier.get_ty(value.custom.unwrap()) {
return Err(format!(
"Error at {}: cannot assign to tuple element",
value.location
));
}
Ok(())
}
ExprKind::Constant { .. } => {
Err(format!("cannot assign to a constant (at {})", pattern.location))
}
_ => self.check_expr(pattern, defined_identifiers),
}
}
fn check_expr(
&mut self,
expr: &Expr<Option<Type>>,
defined_identifiers: &mut HashSet<StrRef>,
) -> Result<(), String> {
// there are some cases where the custom field is None
if let Some(ty) = &expr.custom {
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 {
ExprKind::Name { id, .. } => {
if id == &"none".into() {
return Ok(());
}
self.should_have_value(expr)?;
if !defined_identifiers.contains(id) {
match self.function_data.resolver.get_symbol_type(
self.unifier,
&self.top_level.definitions.read(),
self.primitives,
*id,
) {
Ok(_) => {
self.defined_identifiers.insert(*id);
}
Err(e) => {
return Err(format!(
"type error at identifier `{}` ({}) at {}",
id, e, expr.location
));
}
}
}
}
ExprKind::List { elts, .. }
| ExprKind::Tuple { elts, .. }
| ExprKind::BoolOp { values: elts, .. } => {
for elt in elts.iter() {
self.check_expr(elt, defined_identifiers)?;
self.should_have_value(elt)?;
}
}
ExprKind::Attribute { value, .. } => {
self.check_expr(value, defined_identifiers)?;
self.should_have_value(value)?;
}
ExprKind::BinOp { left, right, .. } => {
self.check_expr(left, defined_identifiers)?;
self.check_expr(right, defined_identifiers)?;
self.should_have_value(left)?;
self.should_have_value(right)?;
}
ExprKind::UnaryOp { operand, .. } => {
self.check_expr(operand, defined_identifiers)?;
self.should_have_value(operand)?;
}
ExprKind::Compare { left, comparators, .. } => {
for elt in once(left.as_ref()).chain(comparators.iter()) {
self.check_expr(elt, defined_identifiers)?;
self.should_have_value(elt)?;
}
}
ExprKind::Subscript { value, slice, .. } => {
self.should_have_value(value)?;
self.check_expr(value, defined_identifiers)?;
self.check_expr(slice, defined_identifiers)?;
}
ExprKind::IfExp { test, body, orelse } => {
self.check_expr(test, defined_identifiers)?;
self.check_expr(body, defined_identifiers)?;
self.check_expr(orelse, defined_identifiers)?;
}
ExprKind::Slice { lower, upper, step } => {
for elt in [lower.as_ref(), upper.as_ref(), step.as_ref()].iter().flatten() {
self.should_have_value(elt)?;
self.check_expr(elt, defined_identifiers)?;
}
}
ExprKind::Lambda { args, body } => {
let mut defined_identifiers = defined_identifiers.clone();
for arg in args.args.iter() {
// TODO: should we check the types here?
if !defined_identifiers.contains(&arg.node.arg) {
defined_identifiers.insert(arg.node.arg);
}
}
self.check_expr(body, &mut defined_identifiers)?;
}
ExprKind::ListComp { elt, generators, .. } => {
// in our type inference stage, we already make sure that there is only 1 generator
let ast::Comprehension { target, iter, ifs, .. } = &generators[0];
self.check_expr(iter, defined_identifiers)?;
self.should_have_value(iter)?;
let mut defined_identifiers = defined_identifiers.clone();
self.check_pattern(target, &mut defined_identifiers)?;
self.should_have_value(target)?;
for term in once(elt.as_ref()).chain(ifs.iter()) {
self.check_expr(term, &mut defined_identifiers)?;
self.should_have_value(term)?;
}
}
ExprKind::Call { func, args, keywords } => {
for expr in once(func.as_ref())
.chain(args.iter())
.chain(keywords.iter().map(|v| v.node.value.as_ref()))
{
self.check_expr(expr, defined_identifiers)?;
self.should_have_value(expr)?;
}
}
ExprKind::Constant { .. } => {}
_ => {
unimplemented!()
}
}
Ok(())
}
// check 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, 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.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.iter() {
self.check_stmt(stmt, &mut local_defined_identifiers)?;
}
Ok(false)
}
StmtKind::If { test, body, orelse, .. } => {
self.check_expr(test, defined_identifiers)?;
self.should_have_value(test)?;
let mut body_identifiers = defined_identifiers.clone();
let mut orelse_identifiers = defined_identifiers.clone();
let body_returned = self.check_block(body, &mut body_identifiers)?;
let orelse_returned = self.check_block(orelse, &mut orelse_identifiers)?;
for ident in body_identifiers.iter() {
if !defined_identifiers.contains(ident) && orelse_identifiers.contains(ident) {
defined_identifiers.insert(*ident);
}
}
Ok(body_returned && orelse_returned)
}
StmtKind::While { test, body, orelse, .. } => {
self.check_expr(test, defined_identifiers)?;
self.should_have_value(test)?;
let mut defined_identifiers = defined_identifiers.clone();
self.check_block(body, &mut defined_identifiers)?;
self.check_block(orelse, &mut defined_identifiers)?;
Ok(false)
}
StmtKind::With { items, body, .. } => {
let mut new_defined_identifiers = defined_identifiers.clone();
for item in items.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)?;
}
}
self.check_block(body, &mut new_defined_identifiers)?;
Ok(false)
}
StmtKind::Try { body, handlers, orelse, finalbody, .. } => {
self.check_block(body, &mut defined_identifiers.clone())?;
self.check_block(orelse, &mut defined_identifiers.clone())?;
for handler in handlers.iter() {
let mut defined_identifiers = defined_identifiers.clone();
let ast::ExcepthandlerKind::ExceptHandler { name, body, .. } = &handler.node;
if let Some(name) = name {
defined_identifiers.insert(*name);
}
self.check_block(body, &mut defined_identifiers)?;
}
self.check_block(finalbody, defined_identifiers)?;
Ok(false)
}
StmtKind::Expr { value, .. } => {
self.check_expr(value, defined_identifiers)?;
Ok(false)
}
StmtKind::Assign { targets, value, .. } => {
self.check_expr(value, defined_identifiers)?;
self.should_have_value(value)?;
for target in targets {
self.check_pattern(target, defined_identifiers)?;
}
Ok(false)
}
StmtKind::AnnAssign { target, value, .. } => {
if let Some(value) = value {
self.check_expr(value, defined_identifiers)?;
self.should_have_value(value)?;
self.check_pattern(target, defined_identifiers)?;
}
Ok(false)
}
StmtKind::Return { value, .. } => {
if let Some(value) = value {
self.check_expr(value, defined_identifiers)?;
self.should_have_value(value)?;
}
Ok(true)
}
StmtKind::Raise { exc, .. } => {
if let Some(value) = exc {
self.check_expr(value, defined_identifiers)?;
}
Ok(true)
}
// break, raise, etc.
_ => Ok(false),
}
}
pub fn check_block(
&mut self,
block: &[Stmt<Option<Type>>],
defined_identifiers: &mut HashSet<StrRef>,
) -> Result<bool, String> {
let mut ret = false;
for stmt in block {
if ret {
return Err(format!("dead code at {:?}", stmt.location));
}
if self.check_stmt(stmt, defined_identifiers)? {
ret = true;
}
}
Ok(ret)
}
}

View File

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

View File

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

View File

@ -1,4 +1,11 @@
use rustpython_parser::ast::{Cmpop, Operator, Unaryop}; use crate::typecheck::{
type_inferencer::*,
typedef::{FunSignature, FuncArg, Type, TypeEnum, Unifier},
};
use nac3parser::ast::{self, StrRef};
use nac3parser::ast::{Cmpop, Operator, Unaryop};
use std::collections::HashMap;
use std::rc::Rc;
pub fn binop_name(op: &Operator) -> &'static str { pub fn binop_name(op: &Operator) -> &'static str {
match op { match op {
@ -38,21 +45,286 @@ pub fn binop_assign_name(op: &Operator) -> &'static str {
pub fn unaryop_name(op: &Unaryop) -> &'static str { pub fn unaryop_name(op: &Unaryop) -> &'static str {
match op { match op {
Unaryop::UAdd => "__pos__", Unaryop::UAdd => "__pos__",
Unaryop::USub => "__neg__", Unaryop::USub => "__neg__",
Unaryop::Not => "__not__", Unaryop::Not => "__not__",
Unaryop::Invert => "__inv__", Unaryop::Invert => "__inv__",
} }
} }
pub fn comparison_name(op: &Cmpop) -> Option<&'static str> { pub fn comparison_name(op: &Cmpop) -> Option<&'static str> {
match op { match op {
Cmpop::Lt => Some("__lt__"), Cmpop::Lt => Some("__lt__"),
Cmpop::LtE => Some("__le__"), Cmpop::LtE => Some("__le__"),
Cmpop::Gt => Some("__gt__"), Cmpop::Gt => Some("__gt__"),
Cmpop::GtE => Some("__ge__"), Cmpop::GtE => Some("__ge__"),
Cmpop::Eq => Some("__eq__"), Cmpop::Eq => Some("__eq__"),
Cmpop::NotEq => Some("__ne__"), Cmpop::NotEq => Some("__ne__"),
_ => None, _ => None,
} }
} }
pub(super) fn with_fields<F>(unifier: &mut Unifier, ty: Type, f: F)
where
F: FnOnce(&mut Unifier, &mut HashMap<StrRef, (Type, bool)>),
{
let (id, mut fields, params) =
if let TypeEnum::TObj { obj_id, fields, params } = &*unifier.get_ty(ty) {
(*obj_id, fields.clone(), params.clone())
} else {
unreachable!()
};
f(unifier, &mut fields);
unsafe {
let unification_table = unifier.get_unification_table();
unification_table.set_value(ty, Rc::new(TypeEnum::TObj { obj_id: id, fields, params }));
}
}
pub fn impl_binop(
unifier: &mut Unifier,
store: &PrimitiveStore,
ty: Type,
other_ty: &[Type],
ret_ty: 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 (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::<HashMap<_, _>>()
} else {
HashMap::new()
};
for op in ops {
fields.insert(binop_name(op).into(), {
(
unifier.add_ty(TypeEnum::TFunc(FunSignature {
ret: ret_ty,
vars: function_vars.clone(),
args: vec![FuncArg {
ty: other_ty,
default_value: None,
name: "other".into(),
}],
})),
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: Type, ops: &[ast::Unaryop]) {
with_fields(unifier, ty, |unifier, fields| {
for op in ops {
fields.insert(
unaryop_name(op).into(),
(
unifier.add_ty(TypeEnum::TFunc(FunSignature {
ret: ret_ty,
vars: HashMap::new(),
args: vec![],
})),
false,
),
);
}
});
}
pub fn impl_cmpop(
unifier: &mut Unifier,
store: &PrimitiveStore,
ty: Type,
other_ty: Type,
ops: &[ast::Cmpop],
) {
with_fields(unifier, ty, |unifier, fields| {
for op in ops {
fields.insert(
comparison_name(op).unwrap().into(),
(
unifier.add_ty(TypeEnum::TFunc(FunSignature {
ret: store.bool,
vars: HashMap::new(),
args: vec![FuncArg {
ty: other_ty,
default_value: None,
name: "other".into(),
}],
})),
false,
),
);
}
});
}
/// Add, Sub, Mult
pub fn impl_basic_arithmetic(
unifier: &mut Unifier,
store: &PrimitiveStore,
ty: Type,
other_ty: &[Type],
ret_ty: Type,
) {
impl_binop(
unifier,
store,
ty,
other_ty,
ret_ty,
&[ast::Operator::Add, ast::Operator::Sub, ast::Operator::Mult],
)
}
/// Pow
pub fn impl_pow(
unifier: &mut Unifier,
store: &PrimitiveStore,
ty: Type,
other_ty: &[Type],
ret_ty: Type,
) {
impl_binop(unifier, store, ty, other_ty, ret_ty, &[ast::Operator::Pow])
}
/// BitOr, BitXor, BitAnd
pub fn impl_bitwise_arithmetic(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type) {
impl_binop(
unifier,
store,
ty,
&[ty],
ty,
&[ast::Operator::BitAnd, ast::Operator::BitOr, ast::Operator::BitXor],
)
}
/// 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])
}
/// 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: Type,
) {
impl_binop(unifier, store, ty, other_ty, ret_ty, &[ast::Operator::FloorDiv])
}
/// Mod
pub fn impl_mod(
unifier: &mut Unifier,
store: &PrimitiveStore,
ty: Type,
other_ty: &[Type],
ret_ty: Type,
) {
impl_binop(unifier, store, ty, other_ty, ret_ty, &[ast::Operator::Mod])
}
/// UAdd, USub
pub fn impl_sign(unifier: &mut Unifier, _store: &PrimitiveStore, ty: Type) {
impl_unaryop(unifier, ty, ty, &[ast::Unaryop::UAdd, ast::Unaryop::USub])
}
/// Invert
pub fn impl_invert(unifier: &mut Unifier, _store: &PrimitiveStore, ty: Type) {
impl_unaryop(unifier, ty, ty, &[ast::Unaryop::Invert])
}
/// Not
pub fn impl_not(unifier: &mut Unifier, store: &PrimitiveStore, ty: Type) {
impl_unaryop(unifier, ty, store.bool, &[ast::Unaryop::Not])
}
/// 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,
&[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) {
impl_cmpop(unifier, store, ty, ty, &[ast::Cmpop::Eq, ast::Cmpop::NotEq])
}
pub fn set_primitives_magic_methods(store: &PrimitiveStore, unifier: &mut Unifier) {
let PrimitiveStore {
int32: int32_t,
int64: int64_t,
float: float_t,
bool: bool_t,
uint32: uint32_t,
uint64: uint64_t,
..
} = *store;
/* int ======== */
for t in [int32_t, int64_t, uint32_t, uint64_t] {
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]);
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);
}
/* float ======== */
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 ======== */
impl_not(unifier, store, bool_t);
impl_eq(unifier, store, bool_t);
}

View File

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

View File

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

View File

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

View File

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

View File

@ -0,0 +1,186 @@
use std::collections::HashMap;
use std::fmt::Display;
use crate::typecheck::typedef::TypeEnum;
use super::typedef::{RecordKey, Type, Unifier};
use nac3parser::ast::{Location, StrRef};
#[derive(Debug, Clone)]
pub enum TypeErrorKind {
TooManyArguments {
expected: usize,
got: usize,
},
MissingArgs(String),
UnknownArgName(StrRef),
IncorrectArgType {
name: StrRef,
expected: Type,
got: Type,
},
FieldUnificationError {
field: RecordKey,
types: (Type, Type),
loc: (Option<Location>, Option<Location>),
},
IncompatibleRange(Type, Vec<Type>),
IncompatibleTypes(Type, Type),
MutationError(RecordKey, Type),
NoSuchField(RecordKey, Type),
TupleIndexOutOfBounds {
index: i32,
len: i32,
},
RequiresTypeAnn,
PolymorphicFunctionPointer,
}
#[derive(Debug, Clone)]
pub struct TypeError {
pub kind: TypeErrorKind,
pub loc: Option<Location>,
}
impl TypeError {
pub fn new(kind: TypeErrorKind, loc: Option<Location>) -> TypeError {
TypeError { kind, loc }
}
pub fn at(mut self, loc: Option<Location>) -> TypeError {
self.loc = self.loc.or(loc);
self
}
pub fn to_display(self, unifier: &Unifier) -> DisplayTypeError {
DisplayTypeError { err: self, unifier }
}
}
pub struct DisplayTypeError<'a> {
pub err: TypeError,
pub unifier: &'a Unifier,
}
fn loc_to_str(loc: Option<Location>) -> String {
match loc {
Some(loc) => format!("(in {})", loc),
None => "".to_string(),
}
}
impl<'a> Display for DisplayTypeError<'a> {
fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
use TypeErrorKind::*;
let mut notes = Some(HashMap::new());
match &self.err.kind {
TooManyArguments { expected, got } => {
write!(f, "Too many arguments. Expected {} but got {}", expected, got)
}
MissingArgs(args) => {
write!(f, "Missing arguments: {}", args)
}
UnknownArgName(name) => {
write!(f, "Unknown argument name: {}", name)
}
IncorrectArgType { name, expected, got } => {
let expected = self.unifier.stringify_with_notes(*expected, &mut notes);
let got = self.unifier.stringify_with_notes(*got, &mut notes);
write!(
f,
"Incorrect argument type for {}. Expected {}, but got {}",
name, expected, got
)
}
FieldUnificationError { field, types, loc } => {
let lhs = self.unifier.stringify_with_notes(types.0, &mut notes);
let rhs = self.unifier.stringify_with_notes(types.1, &mut notes);
write!(
f,
"Unable to unify field {}: Got types {}{} and {}{}",
field,
lhs,
loc_to_str(loc.0),
rhs,
loc_to_str(loc.1)
)
}
IncompatibleRange(t, ts) => {
let t = self.unifier.stringify_with_notes(*t, &mut notes);
let ts = ts
.iter()
.map(|t| self.unifier.stringify_with_notes(*t, &mut notes))
.collect::<Vec<_>>();
write!(f, "Expected any one of these types: {}, but got {}", ts.join(", "), t)
}
IncompatibleTypes(t1, t2) => {
let type1 = self.unifier.get_ty_immutable(*t1);
let type2 = self.unifier.get_ty_immutable(*t2);
match (&*type1, &*type2) {
(TypeEnum::TCall(calls), _) => {
let loc = self.unifier.calls[calls[0].0].loc;
let result = write!(
f,
"{} is not callable",
self.unifier.stringify_with_notes(*t2, &mut notes)
);
if let Some(loc) = loc {
result?;
write!(f, " (in {})", loc)?;
return Ok(());
}
result
}
(TypeEnum::TTuple { ty: ty1 }, TypeEnum::TTuple { ty: ty2 })
if ty1.len() != ty2.len() =>
{
let t1 = self.unifier.stringify_with_notes(*t1, &mut notes);
let t2 = self.unifier.stringify_with_notes(*t2, &mut notes);
write!(f, "Tuple length mismatch: got {} 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: {} and {}", t1, t2)
}
}
}
MutationError(name, t) => {
if let TypeEnum::TTuple { .. } = &*self.unifier.get_ty_immutable(*t) {
write!(f, "Cannot assign to an element of a tuple")
} else {
let t = self.unifier.stringify_with_notes(*t, &mut notes);
write!(f, "Cannot assign to field {} of {}, which is immutable", name, t)
}
}
NoSuchField(name, t) => {
let t = self.unifier.stringify_with_notes(*t, &mut notes);
write!(f, "`{}::{}` field/method does not exist", t, name)
}
TupleIndexOutOfBounds { index, len } => {
write!(
f,
"Tuple index out of bounds. Got {} but tuple has only {} elements",
index, len
)
}
RequiresTypeAnn => {
write!(f, "Unable to infer virtual object type: Type annotation required")
}
PolymorphicFunctionPointer => {
write!(f, "Polymorphic function pointers is not supported")
}
}?;
if let Some(loc) = self.err.loc {
write!(f, " at {}", loc)?;
}
let notes = notes.unwrap();
if !notes.is_empty() {
write!(f, "\n\nNotes:")?;
for line in notes.values() {
write!(f, "\n {}", line)?;
}
}
Ok(())
}
}

File diff suppressed because it is too large Load Diff

View File

@ -0,0 +1,713 @@
use super::super::{magic_methods::with_fields, typedef::*};
use super::*;
use crate::{
codegen::CodeGenContext,
symbol_resolver::ValueEnum,
toplevel::{DefinitionId, TopLevelDef},
};
use indoc::indoc;
use itertools::zip;
use nac3parser::parser::parse_program;
use parking_lot::RwLock;
use test_case::test_case;
struct Resolver {
id_to_type: HashMap<StrRef, Type>,
id_to_def: HashMap<StrRef, DefinitionId>,
class_names: HashMap<StrRef, Type>,
}
impl SymbolResolver for Resolver {
fn get_default_param_value(
&self,
_: &nac3parser::ast::Expr,
) -> Option<crate::symbol_resolver::SymbolValue> {
unimplemented!()
}
fn get_symbol_type(
&self,
_: &mut Unifier,
_: &[Arc<RwLock<TopLevelDef>>],
_: &PrimitiveStore,
str: StrRef,
) -> Result<Type, String> {
self.id_to_type.get(&str).cloned().ok_or_else(|| format!("cannot find symbol `{}`", str))
}
fn get_symbol_value<'ctx, 'a>(
&self,
_: StrRef,
_: &mut CodeGenContext<'ctx, 'a>,
) -> Option<ValueEnum<'ctx>> {
unimplemented!()
}
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 {
unimplemented!()
}
fn get_exception_id(&self, _tyid: usize) -> usize {
unimplemented!()
}
}
struct TestEnvironment {
pub unifier: Unifier,
pub function_data: FunctionData,
pub primitives: PrimitiveStore,
pub id_to_name: HashMap<usize, StrRef>,
pub identifier_mapping: HashMap<StrRef, Type>,
pub virtual_checks: Vec<(Type, Type, nac3parser::ast::Location)>,
pub calls: HashMap<CodeLocation, CallId>,
pub top_level: TopLevelContext,
}
impl TestEnvironment {
pub fn basic_test_env() -> TestEnvironment {
let mut unifier = Unifier::new();
let int32 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(0),
fields: HashMap::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: HashMap::new(),
}));
fields.insert("__add__".into(), (add_ty, false));
});
let int64 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(1),
fields: HashMap::new(),
params: HashMap::new(),
});
let float = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(2),
fields: HashMap::new(),
params: HashMap::new(),
});
let bool = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(3),
fields: HashMap::new(),
params: HashMap::new(),
});
let none = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(4),
fields: HashMap::new(),
params: HashMap::new(),
});
let range = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(5),
fields: HashMap::new(),
params: HashMap::new(),
});
let str = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(6),
fields: HashMap::new(),
params: HashMap::new(),
});
let exception = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(7),
fields: HashMap::new(),
params: HashMap::new(),
});
let uint32 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(8),
fields: HashMap::new(),
params: HashMap::new(),
});
let uint64 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(9),
fields: HashMap::new(),
params: HashMap::new(),
});
let option = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(10),
fields: HashMap::new(),
params: HashMap::new(),
});
let primitives = PrimitiveStore {
int32,
int64,
float,
bool,
none,
range,
str,
exception,
uint32,
uint64,
option,
};
set_primitives_magic_methods(&primitives, &mut unifier);
let id_to_name = [
(0, "int32".into()),
(1, "int64".into()),
(2, "float".into()),
(3, "bool".into()),
(4, "none".into()),
(5, "range".into()),
(6, "str".into()),
(7, "exception".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: Default::default(),
class_names: Default::default(),
}) as Arc<dyn SymbolResolver + Send + Sync>;
TestEnvironment {
top_level: TopLevelContext {
definitions: Default::default(),
unifiers: Default::default(),
personality_symbol: None,
},
unifier,
function_data: FunctionData {
resolver,
bound_variables: Vec::new(),
return_type: None,
},
primitives,
id_to_name,
identifier_mapping,
virtual_checks: Vec::new(),
calls: HashMap::new(),
}
}
fn new() -> TestEnvironment {
let mut unifier = Unifier::new();
let mut identifier_mapping = HashMap::new();
let mut top_level_defs: Vec<Arc<RwLock<TopLevelDef>>> = Vec::new();
let int32 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(0),
fields: HashMap::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: HashMap::new(),
}));
fields.insert("__add__".into(), (add_ty, false));
});
let int64 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(1),
fields: HashMap::new(),
params: HashMap::new(),
});
let float = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(2),
fields: HashMap::new(),
params: HashMap::new(),
});
let bool = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(3),
fields: HashMap::new(),
params: HashMap::new(),
});
let none = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(4),
fields: HashMap::new(),
params: HashMap::new(),
});
let range = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(5),
fields: HashMap::new(),
params: HashMap::new(),
});
let str = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(6),
fields: HashMap::new(),
params: HashMap::new(),
});
let exception = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(7),
fields: HashMap::new(),
params: HashMap::new(),
});
let uint32 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(8),
fields: HashMap::new(),
params: HashMap::new(),
});
let uint64 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(9),
fields: HashMap::new(),
params: HashMap::new(),
});
let option = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(10),
fields: HashMap::new(),
params: HashMap::new(),
});
identifier_mapping.insert("None".into(), none);
for (i, name) in ["int32", "int64", "float", "bool", "none", "range", "str", "Exception"]
.iter()
.enumerate()
{
top_level_defs.push(
RwLock::new(TopLevelDef::Class {
name: (*name).into(),
object_id: DefinitionId(i),
type_vars: Default::default(),
fields: Default::default(),
methods: Default::default(),
ancestors: Default::default(),
resolver: None,
constructor: None,
loc: None,
})
.into(),
);
}
let defs = 7;
let primitives = PrimitiveStore {
int32,
int64,
float,
bool,
none,
range,
str,
exception,
uint32,
uint64,
option,
};
let (v0, id) = unifier.get_dummy_var();
let foo_ty = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(defs + 1),
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![v0],
fields: [("a".into(), v0, true)].into(),
methods: Default::default(),
ancestors: Default::default(),
resolver: None,
constructor: None,
loc: None,
})
.into(),
);
identifier_mapping.insert(
"Foo".into(),
unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![],
ret: foo_ty,
vars: [(id, v0)].iter().cloned().collect(),
})),
);
let fun = unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![],
ret: int32,
vars: Default::default(),
}));
let bar = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(defs + 2),
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: Default::default(),
fields: [("a".into(), int32, true), ("b".into(), fun, true)].into(),
methods: Default::default(),
ancestors: Default::default(),
resolver: None,
constructor: None,
loc: None,
})
.into(),
);
identifier_mapping.insert(
"Bar".into(),
unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![],
ret: bar,
vars: Default::default(),
})),
);
let bar2 = unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(defs + 3),
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: Default::default(),
fields: [("a".into(), bool, true), ("b".into(), fun, false)].into(),
methods: Default::default(),
ancestors: Default::default(),
resolver: None,
constructor: None,
loc: None,
})
.into(),
);
identifier_mapping.insert(
"Bar2".into(),
unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![],
ret: bar2,
vars: Default::default(),
})),
);
let class_names = [("Bar".into(), bar), ("Bar2".into(), bar2)].iter().cloned().collect();
let id_to_name = [
"int32".into(),
"int64".into(),
"float".into(),
"bool".into(),
"none".into(),
"range".into(),
"str".into(),
"exception".into(),
"Foo".into(),
"Bar".into(),
"Bar2".into(),
]
.iter()
.enumerate()
.map(|(a, b)| (a, *b))
.collect();
let top_level = TopLevelContext {
definitions: Arc::new(top_level_defs.into()),
unifiers: Default::default(),
personality_symbol: None,
};
let resolver = Arc::new(Resolver {
id_to_type: identifier_mapping.clone(),
id_to_def: [
("Foo".into(), DefinitionId(defs + 1)),
("Bar".into(), DefinitionId(defs + 2)),
("Bar2".into(), DefinitionId(defs + 3)),
]
.iter()
.cloned()
.collect(),
class_names,
}) as Arc<dyn SymbolResolver + Send + Sync>;
TestEnvironment {
unifier,
top_level,
function_data: FunctionData {
resolver,
bound_variables: Vec::new(),
return_type: None,
},
primitives,
id_to_name,
identifier_mapping,
virtual_checks: Vec::new(),
calls: HashMap::new(),
}
}
fn get_inferencer(&mut self) -> Inferencer {
Inferencer {
top_level: &self.top_level,
function_data: &mut self.function_data,
unifier: &mut self.unifier,
variable_mapping: Default::default(),
primitives: &mut self.primitives,
virtual_checks: &mut self.virtual_checks,
calls: &mut self.calls,
defined_identifiers: Default::default(),
in_handler: false,
}
}
}
#[test_case(indoc! {"
a = 1234
b = int64(2147483648)
c = 1.234
d = True
"},
[("a", "int32"), ("b", "int64"), ("c", "float"), ("d", "bool")].iter().cloned().collect(),
&[]
; "primitives test")]
#[test_case(indoc! {"
a = lambda x, y: x
b = lambda x: a(x, x)
c = 1.234
d = b(c)
"},
[("a", "fn[[x:float, y:float], float]"), ("b", "fn[[x:float], float]"), ("c", "float"), ("d", "float")].iter().cloned().collect(),
&[]
; "lambda test")]
#[test_case(indoc! {"
a = lambda x: x + x
b = lambda x: a(x) + x
a = b
c = b(1)
"},
[("a", "fn[[x:int32], int32]"), ("b", "fn[[x:int32], int32]"), ("c", "int32")].iter().cloned().collect(),
&[]
; "lambda test 2")]
#[test_case(indoc! {"
a = lambda x: x
b = lambda x: x
foo1 = Foo()
foo2 = Foo()
c = a(foo1.a)
d = b(foo2.a)
a(True)
b(123)
"},
[("a", "fn[[x:bool], bool]"), ("b", "fn[[x:int32], int32]"), ("c", "bool"),
("d", "int32"), ("foo1", "Foo[bool]"), ("foo2", "Foo[int32]")].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]")].iter().cloned().collect(),
&[]
; "listcomp test")]
#[test_case(indoc! {"
a = virtual(Bar(), Bar)
b = a.b()
a = virtual(Bar2())
"},
[("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]")].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);
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().cloned().collect();
defined_identifiers.insert("virtual".into());
let mut inferencer = env.get_inferencer();
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))
.collect::<Result<Vec<_>, _>>()
.unwrap();
inferencer.check_block(&statements, &mut defined_identifiers).unwrap();
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 None,
);
println!("{}: {}", k, name);
}
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 None,
);
assert_eq!(format!("{}: {}", k, v), format!("{}: {}", k, name));
}
assert_eq!(inferencer.virtual_checks.len(), virtuals.len());
for ((a, b, _), (x, y)) in zip(inferencer.virtual_checks.iter(), virtuals) {
let a = inferencer.unifier.internal_stringify(
*a,
&mut |v| (*id_to_name.get(&v).unwrap()).into(),
&mut |v| format!("v{}", v),
&mut None,
);
let b = inferencer.unifier.internal_stringify(
*b,
&mut |v| (*id_to_name.get(&v).unwrap()).into(),
&mut |v| format!("v{}", v),
&mut None,
);
assert_eq!(&a, x);
assert_eq!(&b, y);
}
}
#[test_case(indoc! {"
a = 2
b = 2
c = a + b
d = a - b
e = a * b
f = a / b
g = a // b
h = a % b
"},
[("a", "int32"),
("b", "int32"),
("c", "int32"),
("d", "int32"),
("e", "int32"),
("f", "float"),
("g", "int32"),
("h", "int32")].iter().cloned().collect()
; "int32")]
#[test_case(
indoc! {"
a = 2.4
b = 3.6
c = a + b
d = a - b
e = a * b
f = a / b
g = a // b
h = a % b
i = a ** b
ii = 3
j = a ** b
"},
[("a", "float"),
("b", "float"),
("c", "float"),
("d", "float"),
("e", "float"),
("f", "float"),
("g", "float"),
("h", "float"),
("i", "float"),
("ii", "int32"),
("j", "float")].iter().cloned().collect()
; "float"
)]
#[test_case(
indoc! {"
a = int64(12312312312)
b = int64(24242424424)
c = a + b
d = a - b
e = a * b
f = a / b
g = a // b
h = a % b
i = a == b
j = a > b
k = a < b
l = a != b
"},
[("a", "int64"),
("b", "int64"),
("c", "int64"),
("d", "int64"),
("e", "int64"),
("f", "float"),
("g", "int64"),
("h", "int64"),
("i", "bool"),
("j", "bool"),
("k", "bool"),
("l", "bool")].iter().cloned().collect()
; "int64"
)]
#[test_case(
indoc! {"
a = True
b = False
c = a == b
d = not a
e = a != b
"},
[("a", "bool"),
("b", "bool"),
("c", "bool"),
("d", "bool"),
("e", "bool")].iter().cloned().collect()
; "boolean"
)]
fn test_primitive_magic_methods(source: &str, mapping: HashMap<&str, &str>) {
println!("source:\n{}", source);
let mut env = TestEnvironment::basic_test_env();
let id_to_name = std::mem::take(&mut env.id_to_name);
let mut defined_identifiers: HashSet<_> = env.identifier_mapping.keys().cloned().collect();
defined_identifiers.insert("virtual".into());
let mut inferencer = env.get_inferencer();
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))
.collect::<Result<Vec<_>, _>>()
.unwrap();
inferencer.check_block(&statements, &mut defined_identifiers).unwrap();
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 None,
);
println!("{}: {}", k, name);
}
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 None,
);
assert_eq!(format!("{}: {}", k, v), format!("{}: {}", k, name));
}
}

View File

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

File diff suppressed because it is too large Load Diff

View File

@ -0,0 +1,571 @@
use super::super::magic_methods::with_fields;
use super::*;
use indoc::indoc;
use itertools::Itertools;
use std::collections::HashMap;
use test_case::test_case;
impl Unifier {
/// Check whether two types are equal.
fn eq(&mut self, a: Type, b: Type) -> bool {
if a == b {
return true;
}
let (ty_a, ty_b) = {
let table = &mut self.unification_table;
if table.unioned(a, b) {
return true;
}
(table.probe_value(a).clone(), table.probe_value(b).clone())
};
match (&*ty_a, &*ty_b) {
(
TypeEnum::TVar { fields: None, id: id1, .. },
TypeEnum::TVar { fields: None, id: id2, .. },
) => id1 == id2,
(
TypeEnum::TVar { fields: Some(map1), .. },
TypeEnum::TVar { fields: Some(map2), .. },
) => self.map_eq2(map1, map2),
(TypeEnum::TTuple { ty: ty1 }, TypeEnum::TTuple { ty: ty2 }) => {
ty1.len() == ty2.len()
&& ty1.iter().zip(ty2.iter()).all(|(t1, t2)| self.eq(*t1, *t2))
}
(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),
// TCall and TFunc are not yet implemented
_ => false,
}
}
fn map_eq<K>(&mut self, map1: &Mapping<K>, map2: &Mapping<K>) -> bool
where
K: std::hash::Hash + std::cmp::Eq + std::clone::Clone,
{
if map1.len() != map2.len() {
return false;
}
for (k, v) in map1.iter() {
if !map2.get(k).map(|v1| self.eq(*v, *v1)).unwrap_or(false) {
return false;
}
}
true
}
fn map_eq2<K>(&mut self, map1: &Mapping<K, RecordField>, map2: &Mapping<K, RecordField>) -> bool
where
K: std::hash::Hash + std::cmp::Eq + std::clone::Clone,
{
if map1.len() != map2.len() {
return false;
}
for (k, v) in map1.iter() {
if !map2.get(k).map(|v1| self.eq(v.ty, v1.ty)).unwrap_or(false) {
return false;
}
}
true
}
}
struct TestEnvironment {
pub unifier: Unifier,
pub type_mapping: HashMap<String, Type>,
}
impl TestEnvironment {
fn new() -> TestEnvironment {
let mut unifier = Unifier::new();
let mut type_mapping = HashMap::new();
type_mapping.insert(
"int".into(),
unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(0),
fields: HashMap::new(),
params: HashMap::new(),
}),
);
type_mapping.insert(
"float".into(),
unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(1),
fields: HashMap::new(),
params: HashMap::new(),
}),
);
type_mapping.insert(
"bool".into(),
unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(2),
fields: HashMap::new(),
params: HashMap::new(),
}),
);
let (v0, id) = unifier.get_dummy_var();
type_mapping.insert(
"Foo".into(),
unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(3),
fields: [("a".into(), (v0, true))].iter().cloned().collect::<HashMap<_, _>>(),
params: [(id, v0)].iter().cloned().collect::<HashMap<_, _>>(),
}),
);
TestEnvironment { unifier, type_mapping }
}
fn parse(&mut self, typ: &str, mapping: &Mapping<String>) -> Type {
let result = self.internal_parse(typ, mapping);
assert!(result.1.is_empty());
result.0
}
fn internal_parse<'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_else(|| typ.len());
match &typ[..end] {
"tuple" => {
let mut s = &typ[end..];
assert!(&s[0..1] == "[");
let mut ty = Vec::new();
while &s[0..1] != "]" {
let result = self.internal_parse(&s[1..], mapping);
ty.push(result.0);
s = result.1;
}
(self.unifier.add_ty(TypeEnum::TTuple { ty }), &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!(&s[0..1] == "[");
let mut fields = HashMap::new();
while &s[0..1] != "]" {
let eq = s.find('=').unwrap();
let key = s[1..eq].into();
let result = self.internal_parse(&s[eq + 1..], mapping);
fields.insert(key, RecordField::new(result.0, true, None));
s = result.1;
}
(self.unifier.add_record(fields), &s[1..])
}
x => {
let mut s = &typ[end..];
let ty = mapping.get(x).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.as_ref() {
if !params.is_empty() {
assert!(&s[0..1] == "[");
let mut p = Vec::new();
while &s[0..1] != "]" {
let result = self.internal_parse(&s[1..], mapping);
p.push(result.0);
s = result.1;
}
s = &s[1..];
ty = self
.unifier
.subst(ty, &params.keys().cloned().zip(p.into_iter()).collect())
.unwrap_or(ty);
}
}
ty
});
(ty, s)
}
}
}
fn unify(&mut self, typ1: Type, typ2: Type) -> Result<(), String> {
self.unifier.unify(typ1, typ2).map_err(|e| e.to_display(&self.unifier).to_string())
}
}
#[test_case(2,
&[("v1", "v2"), ("v2", "float")],
&[("v1", "float"), ("v2", "float")]
; "simple variable"
)]
#[test_case(2,
&[("v1", "list[v2]"), ("v1", "list[float]")],
&[("v1", "list[float]"), ("v2", "float")]
; "list element"
)]
#[test_case(3,
&[
("v1", "Record[a=v3,b=v3]"),
("v2", "Record[b=float,c=v3]"),
("v1", "v2")
],
&[
("v1", "Record[a=float,b=float,c=float]"),
("v2", "Record[a=float,b=float,c=float]"),
("v3", "float")
]
; "record merge"
)]
#[test_case(3,
&[
("v1", "Record[a=float]"),
("v2", "Foo[v3]"),
("v1", "v2")
],
&[
("v1", "Foo[float]"),
("v3", "float")
]
; "record obj merge"
)]
/// Test cases for valid unifications.
fn test_unify(
variable_count: u32,
unify_pairs: &[(&'static str, &'static str)],
verify_pairs: &[(&'static str, &'static str)],
) {
let unify_count = unify_pairs.len();
// test all permutations...
for perm in unify_pairs.iter().permutations(unify_count) {
let mut env = TestEnvironment::new();
let mut mapping = HashMap::new();
for i in 1..=variable_count {
let v = env.unifier.get_dummy_var();
mapping.insert(format!("v{}", i), v.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.iter() {
let t1 = env.parse(a, &mapping);
let t2 = env.parse(b, &mapping);
pairs.push((t1, t2));
}
for (t1, t2) in pairs {
env.unifier.unify(t1, t2).unwrap();
}
for (a, b) in verify_pairs.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));
assert!(env.unifier.eq(t1, t2));
}
}
}
#[test_case(2,
&[
("v1", "tuple[int]"),
("v2", "list[int]"),
],
(("v1", "v2"), "Incompatible types: list[0] and tuple[0]")
; "type mismatch"
)]
#[test_case(2,
&[
("v1", "tuple[int]"),
("v2", "tuple[float]"),
],
(("v1", "v2"), "Incompatible types: 0 and 1")
; "tuple parameter mismatch"
)]
#[test_case(2,
&[
("v1", "tuple[int,int]"),
("v2", "tuple[int]"),
],
(("v1", "v2"), "Tuple length mismatch: got tuple[0, 0] and tuple[0]")
; "tuple length mismatch"
)]
#[test_case(3,
&[
("v1", "Record[a=float,b=int]"),
("v2", "Foo[v3]"),
],
(("v1", "v2"), "`3[typevar4]::b` field/method does not exist")
; "record obj merge"
)]
/// Test cases for invalid unifications.
fn test_invalid_unification(
variable_count: u32,
unify_pairs: &[(&'static str, &'static str)],
erroneous_pair: ((&'static str, &'static str), &'static str),
) {
let mut env = TestEnvironment::new();
let mut mapping = HashMap::new();
for i in 1..=variable_count {
let v = env.unifier.get_dummy_var();
mapping.insert(format!("v{}", i), v.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.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));
for (a, b) in pairs {
env.unifier.unify(a, b).unwrap();
}
assert_eq!(env.unify(t1, t2), Err(erroneous_pair.1.to_string()));
}
#[test]
fn test_recursive_subst() {
let mut env = TestEnvironment::new();
let int = *env.type_mapping.get("int").unwrap();
let foo_id = *env.type_mapping.get("Foo").unwrap();
let foo_ty = env.unifier.get_ty(foo_id);
let mapping: HashMap<_, _>;
with_fields(&mut env.unifier, foo_id, |_unifier, fields| {
fields.insert("rec".into(), (foo_id, true));
});
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);
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]
fn test_virtual() {
let mut env = TestEnvironment::new();
let int = env.parse("int", &HashMap::new());
let fun = env.unifier.add_ty(TypeEnum::TFunc(FunSignature {
args: vec![],
ret: int,
vars: HashMap::new(),
}));
let bar = env.unifier.add_ty(TypeEnum::TObj {
obj_id: DefinitionId(5),
fields: [("f".into(), (fun, false)), ("a".into(), (int, false))]
.iter()
.cloned()
.collect::<HashMap<StrRef, _>>(),
params: HashMap::new(),
});
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))].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))].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))].iter().cloned().collect());
assert_eq!(env.unify(b, d), Err("`virtual[5]::b` field/method does not exist".to_string()));
}
#[test]
fn test_typevar_range() {
let mut env = TestEnvironment::new();
let int = env.parse("int", &HashMap::new());
let boolean = env.parse("bool", &HashMap::new());
let float = env.parse("float", &HashMap::new());
let int_list = env.parse("list[int]", &HashMap::new());
let float_list = env.parse("list[float]", &HashMap::new());
// unification between v and int
// where v in (int, bool)
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).0;
assert_eq!(
env.unify(int_list, v),
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).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).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).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).0;
// unification between v and list[float]
// where v in (int, list[v1]), v1 in (int, bool)
assert_eq!(
env.unify(float_list, v),
Err("Expected any one of these types: 0, list[typevar5], but got list[1]\n\nNotes:\n typevar5 ∈ {0, 2}".to_string())
);
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).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).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::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).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::TList { ty: int });
assert_eq!(
env.unify(a_list, int_list),
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).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),
Err("Expected any one of these types: 0, 1, but got 2".into())
);
}
#[test]
fn test_rigid_var() {
let mut env = TestEnvironment::new();
let a = env.unifier.get_fresh_rigid_var(None, None).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: typevar3 and typevar2".to_string()));
env.unifier.unify(list_a, list_x).unwrap();
assert_eq!(env.unify(list_x, list_int), Err("Incompatible types: 0 and typevar2".to_string()));
env.unifier.replace_rigid_var(a, int);
env.unifier.unify(list_x, list_int).unwrap();
}
#[test]
fn test_instantiation() {
let mut env = TestEnvironment::new();
let int = env.parse("int", &HashMap::new());
let boolean = env.parse("bool", &HashMap::new());
let float = env.parse("float", &HashMap::new());
let list_int = env.parse("list[int]", &HashMap::new());
let obj_map: HashMap<_, _> =
[(0usize, "int"), (1, "float"), (2, "bool")].iter().cloned().collect();
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).0;
// t = TypeVar('t')
// v = TypeVar('v', int, bool)
// v1 = TypeVar('v1', 'list[v]', int)
// v2 = TypeVar('v2', 'list[int]', float)
// v3 = TypeVar('v3', tuple[v, v1, v2], t)
// what values can v3 take?
let types = env.unifier.get_instantiations(v3).unwrap();
let expected_types = indoc! {"
tuple[bool, int, float]
tuple[bool, int, list[int]]
tuple[bool, list[bool], float]
tuple[bool, list[bool], list[int]]
tuple[bool, list[int], float]
tuple[bool, list[int], list[int]]
tuple[int, int, float]
tuple[int, int, list[int]]
tuple[int, list[bool], float]
tuple[int, list[bool], list[int]]
tuple[int, list[int], float]
tuple[int, list[int], list[int]]
v5"
}
.split('\n')
.collect_vec();
let types = types
.iter()
.map(|ty| {
env.unifier.internal_stringify(
*ty,
&mut |i| obj_map.get(&i).unwrap().to_string(),
&mut |i| format!("v{}", i),
&mut None,
)
})
.sorted()
.collect_vec();
assert_eq!(expected_types, types);
}

View File

@ -0,0 +1,169 @@
use std::rc::Rc;
use itertools::izip;
#[derive(Copy, Clone, PartialEq, Eq, Debug, Hash)]
pub struct UnificationKey(usize);
#[derive(Clone)]
pub struct UnificationTable<V> {
parents: Vec<usize>,
ranks: Vec<u32>,
values: Vec<Option<V>>,
log: Vec<Action<V>>,
generation: u32,
}
#[derive(Clone, Debug)]
enum Action<V> {
Parent {
key: usize,
original_parent: usize,
},
Value {
key: usize,
original_value: Option<V>,
},
Rank {
key: usize,
original_rank: u32,
},
Marker {
generation: u32,
}
}
impl<V> Default for UnificationTable<V> {
fn default() -> Self {
Self::new()
}
}
impl<V> UnificationTable<V> {
pub fn new() -> UnificationTable<V> {
UnificationTable { parents: Vec::new(), ranks: Vec::new(), values: Vec::new(), log: Vec::new(), generation: 0 }
}
pub fn new_key(&mut self, v: V) -> UnificationKey {
let index = self.parents.len();
self.parents.push(index);
self.ranks.push(0);
self.values.push(Some(v));
UnificationKey(index)
}
pub fn unify(&mut self, a: UnificationKey, b: UnificationKey) {
let mut a = self.find(a);
let mut b = self.find(b);
if a == b {
return;
}
if self.ranks[a] < self.ranks[b] {
std::mem::swap(&mut a, &mut b);
}
self.log.push(Action::Parent { key: b, original_parent: self.parents[b] });
self.parents[b] = a;
if self.ranks[a] == self.ranks[b] {
self.log.push(Action::Rank { key: a, original_rank: self.ranks[a] });
self.ranks[a] += 1;
}
}
pub fn probe_value_immutable(&self, key: UnificationKey) -> &V {
let mut root = key.0;
let mut parent = self.parents[root];
while root != parent {
root = parent;
// parent = root.parent
parent = self.parents[parent];
}
self.values[parent].as_ref().unwrap()
}
pub fn probe_value(&mut self, a: UnificationKey) -> &V {
let index = self.find(a);
self.values[index].as_ref().unwrap()
}
pub fn set_value(&mut self, a: UnificationKey, v: V) {
let index = self.find(a);
let original_value = self.values[index].replace(v);
self.log.push(Action::Value { key: index, original_value });
}
pub fn unioned(&mut self, a: UnificationKey, b: UnificationKey) -> bool {
self.find(a) == self.find(b)
}
pub fn get_representative(&mut self, key: UnificationKey) -> UnificationKey {
UnificationKey(self.find(key))
}
fn find(&mut self, key: UnificationKey) -> usize {
let mut root = key.0;
let mut parent = self.parents[root];
while root != parent {
// a = parent.parent
let a = self.parents[parent];
// root.parent = parent.parent
self.log.push(Action::Parent { key: root, original_parent: self.parents[root] });
self.parents[root] = a;
root = parent;
// parent = root.parent
parent = a;
}
parent
}
pub fn get_snapshot(&mut self) -> (usize, u32) {
let generation = self.generation;
self.log.push(Action::Marker { generation });
self.generation += 1;
(self.log.len(), generation)
}
pub fn restore_snapshot(&mut self, snapshot: (usize, u32)) {
let (log_len, generation) = snapshot;
assert!(self.log.len() >= log_len, "snapshot restoration error");
assert!(matches!(self.log[log_len - 1], Action::Marker { generation: gen } if gen == generation), "snapshot restoration error");
for action in self.log.drain(log_len - 1..).rev() {
match action {
Action::Parent { key, original_parent } => {
self.parents[key] = original_parent;
}
Action::Value { key, original_value } => {
self.values[key] = original_value;
}
Action::Rank { key, original_rank } => {
self.ranks[key] = original_rank;
}
Action::Marker { .. } => {}
}
}
}
pub fn discard_snapshot(&mut self, snapshot: (usize, u32)) {
let (log_len, generation) = snapshot;
assert!(self.log.len() >= log_len, "snapshot discard error");
assert!(matches!(self.log[log_len - 1], Action::Marker { generation: gen } if gen == generation), "snapshot discard error");
self.log.clear();
}
}
impl<V> UnificationTable<Rc<V>>
where
V: Clone,
{
pub fn get_send(&self) -> UnificationTable<V> {
let values = izip!(self.values.iter(), self.parents.iter())
.enumerate()
.map(|(i, (v, p))| if *p == i { v.as_ref().map(|v| v.as_ref().clone()) } else { None })
.collect();
UnificationTable { parents: self.parents.clone(), ranks: self.ranks.clone(), values, log: Vec::new(), generation: 0 }
}
pub fn from_send(table: &UnificationTable<V>) -> UnificationTable<Rc<V>> {
let values = table.values.iter().cloned().map(|v| v.map(Rc::new)).collect();
UnificationTable { parents: table.parents.clone(), ranks: table.ranks.clone(), values, log: Vec::new(), generation: 0 }
}
}

View File

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

View File

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

View File

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

View File

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

View File

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

24
nac3parser/Cargo.toml Normal file
View File

@ -0,0 +1,24 @@
[package]
name = "nac3parser"
version = "0.1.2"
description = "Parser for python code."
authors = [ "RustPython Team", "M-Labs" ]
build = "build.rs"
license = "MIT"
edition = "2018"
[build-dependencies]
lalrpop = "0.19.6"
[dependencies]
nac3ast = { path = "../nac3ast" }
lalrpop-util = "0.19.6"
log = "0.4.1"
unic-emoji-char = "0.9"
unic-ucd-ident = "0.9"
unicode_names2 = "0.4"
phf = { version = "0.9", features = ["macros"] }
ahash = "0.7.2"
[dev-dependencies]
insta = "=1.11.0"

56
nac3parser/README.md Normal file
View File

@ -0,0 +1,56 @@
# nac3parser
This directory has the code for python lexing, parsing and generating Abstract Syntax Trees (AST).
This is the RustPython parser with modifications for NAC3.
The steps are:
- Lexical analysis: splits the source code into tokens.
- Parsing and generating the AST: transforms those tokens into an AST. Uses `LALRPOP`, a Rust parser generator framework.
The RustPython team wrote [a blog post](https://rustpython.github.io/2020/04/02/thing-explainer-parser.html) with screenshots and an explanation to help you understand the steps by seeing them in action.
For more information on LALRPOP, here is a link to the [LALRPOP book](https://github.com/lalrpop/lalrpop).
There is a readme in the `src` folder with the details of each file.
## Directory content
`build.rs`: The build script.
`Cargo.toml`: The config file.
The `src` directory has:
**lib.rs**
This is the crate's root.
**lexer.rs**
This module takes care of lexing python source text. This means source code is translated into separate tokens.
**parser.rs**
A python parsing module. Use this module to parse python code into an AST. There are three ways to parse python code. You could parse a whole program, a single statement, or a single expression.
**ast.rs**
Implements abstract syntax tree (AST) nodes for the python language. Roughly equivalent to [the python AST](https://docs.python.org/3/library/ast.html).
**python.lalrpop**
Python grammar.
**token.rs**
Different token definitions. Loosely based on token.h from CPython source.
**errors.rs**
Define internal parse error types. The goal is to provide a matching and a safe error API, masking errors from LALR.
**fstring.rs**
Format strings.
**function.rs**
Collection of functions for parsing parameters, arguments.
**location.rs**
Datatypes to support source location information.
**mode.rs**
Execution mode check. Allowed modes are `exec`, `eval` or `single`.

3
nac3parser/build.rs Normal file
View File

@ -0,0 +1,3 @@
fn main() {
lalrpop::process_root().unwrap()
}

View File

@ -0,0 +1,85 @@
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>
) -> Result<Vec<Ident>, ParseError<Location, Tok, LexicalError>> {
if com_loc.column() != stmt_loc.column() && !nac3com_above.is_empty() {
return Err(ParseError::User {
error: LexicalError {
location: com_loc,
error: LexicalErrorType::OtherError(
format!(
"config comment at top must have the same indentation with what it applies (comment at {}, statement at {})",
com_loc,
stmt_loc,
)
)
}
})
};
Ok(
nac3com_above
.into_iter()
.map(|(com, _)| com)
.chain(nac3com_end.map_or_else(|| vec![].into_iter(), |com| vec![com].into_iter()))
.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>> {
if com_above_loc.column() != stmts[0].location.column() && !nac3com_above.is_empty() {
return Err(ParseError::User {
error: LexicalError {
location: com_above_loc,
error: LexicalErrorType::OtherError(
format!(
"config comment at top must have the same indentation with what it applies (comment at {}, statement at {})",
com_above_loc,
stmts[0].location,
)
)
}
})
}
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])
);
Ok(())
}
fn apply_config_comments<U>(stmt: &mut Stmt<U>, comments: Vec<Ident>) {
match &mut stmt.node {
StmtKind::Pass { config_comment, .. }
| StmtKind::Delete { config_comment, .. }
| StmtKind::Expr { config_comment, .. }
| StmtKind::Assign { config_comment, .. }
| StmtKind::AugAssign { config_comment, .. }
| StmtKind::AnnAssign { config_comment, .. }
| StmtKind::Break { config_comment, .. }
| StmtKind::Continue { config_comment, .. }
| StmtKind::Return { config_comment, .. }
| StmtKind::Raise { config_comment, .. }
| StmtKind::Import { config_comment, .. }
| StmtKind::ImportFrom { config_comment, .. }
| StmtKind::Global { config_comment, .. }
| StmtKind::Nonlocal { config_comment, .. }
| StmtKind::Assert { config_comment, .. } => config_comment.extend(comments),
_ => { unreachable!("only small statements should call this function") }
}
}

239
nac3parser/src/error.rs Normal file
View File

@ -0,0 +1,239 @@
//! Define internal parse error types
//! The goal is to provide a matching and a safe error API, maksing errors from LALR
use lalrpop_util::ParseError as LalrpopError;
use crate::ast::Location;
use crate::token::Tok;
use std::error::Error;
use std::fmt;
/// Represents an error during lexical scanning.
#[derive(Debug, PartialEq)]
pub struct LexicalError {
pub error: LexicalErrorType,
pub location: Location,
}
#[derive(Debug, PartialEq)]
pub enum LexicalErrorType {
StringError,
UnicodeError,
NestingError,
IndentationError,
TabError,
TabsAfterSpaces,
DefaultArgumentError,
PositionalArgumentError,
DuplicateKeywordArgumentError,
UnrecognizedToken { tok: char },
FStringError(FStringErrorType),
LineContinuationError,
Eof,
OtherError(String),
}
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::UnicodeError => write!(f, "Got unexpected unicode"),
LexicalErrorType::NestingError => write!(f, "Got unexpected nesting"),
LexicalErrorType::IndentationError => {
write!(f, "unindent does not match any outer indentation level")
}
LexicalErrorType::TabError => {
write!(f, "inconsistent use of tabs and spaces in indentation")
}
LexicalErrorType::TabsAfterSpaces => {
write!(f, "Tabs not allowed as part of indentation after spaces")
}
LexicalErrorType::DefaultArgumentError => {
write!(f, "non-default argument follows default argument")
}
LexicalErrorType::DuplicateKeywordArgumentError => {
write!(f, "keyword argument repeated")
}
LexicalErrorType::PositionalArgumentError => {
write!(f, "positional argument follows keyword argument")
}
LexicalErrorType::UnrecognizedToken { 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),
}
}
}
// TODO: consolidate these with ParseError
#[derive(Debug, PartialEq)]
pub struct FStringError {
pub error: FStringErrorType,
pub location: Location,
}
#[derive(Debug, PartialEq)]
pub enum FStringErrorType {
UnclosedLbrace,
UnopenedRbrace,
ExpectedRbrace,
InvalidExpression(Box<ParseErrorType>),
InvalidConversionFlag,
EmptyExpression,
MismatchedDelimiter,
ExpressionNestedTooDeeply,
}
impl fmt::Display for FStringErrorType {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
match self {
FStringErrorType::UnclosedLbrace => write!(f, "Unclosed '{{'"),
FStringErrorType::UnopenedRbrace => write!(f, "Unopened '}}'"),
FStringErrorType::ExpectedRbrace => write!(f, "Expected '}}' after conversion flag."),
FStringErrorType::InvalidExpression(error) => {
write!(f, "Invalid expression: {}", error)
}
FStringErrorType::InvalidConversionFlag => write!(f, "Invalid conversion flag"),
FStringErrorType::EmptyExpression => write!(f, "Empty expression"),
FStringErrorType::MismatchedDelimiter => write!(f, "Mismatched delimiter"),
FStringErrorType::ExpressionNestedTooDeeply => {
write!(f, "expressions nested too deeply")
}
}
}
}
impl From<FStringError> for LalrpopError<Location, Tok, LexicalError> {
fn from(err: FStringError) -> Self {
lalrpop_util::ParseError::User {
error: LexicalError {
error: LexicalErrorType::FStringError(err.error),
location: err.location,
},
}
}
}
/// Represents an error during parsing
#[derive(Debug, PartialEq)]
pub struct ParseError {
pub error: ParseErrorType,
pub location: Location,
}
#[derive(Debug, PartialEq)]
pub enum ParseErrorType {
/// Parser encountered an unexpected end of input
Eof,
/// Parser encountered an extra token
ExtraToken(Tok),
/// Parser encountered an invalid token
InvalidToken,
/// Parser encountered an unexpected token
UnrecognizedToken(Tok, Option<String>),
/// Maps to `User` type from `lalrpop-util`
Lexical(LexicalErrorType),
}
/// Convert `lalrpop_util::ParseError` to our internal type
impl From<LalrpopError<Location, Tok, LexicalError>> for ParseError {
fn from(err: LalrpopError<Location, Tok, LexicalError>) -> Self {
match err {
// 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
};
ParseError {
error: ParseErrorType::UnrecognizedToken(token.1, expected),
location: token.0,
}
}
LalrpopError::UnrecognizedEOF { location, .. } => ParseError {
error: ParseErrorType::Eof,
location,
},
}
}
}
impl fmt::Display for ParseError {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
write!(f, "{} at {}", self.error, self.location)
}
}
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::InvalidToken => write!(f, "Got invalid token"),
ParseErrorType::UnrecognizedToken(ref tok, ref expected) => {
if *tok == Tok::Indent {
write!(f, "unexpected indent")
} else if expected.as_deref() == Some("Indent") {
write!(f, "expected an indented block")
} else {
write!(f, "Got unexpected token {}", tok)
}
}
ParseErrorType::Lexical(ref error) => write!(f, "{}", error),
}
}
}
impl Error for ParseErrorType {}
impl ParseErrorType {
pub fn is_indentation_error(&self) -> bool {
match self {
ParseErrorType::Lexical(LexicalErrorType::IndentationError) => true,
ParseErrorType::UnrecognizedToken(token, expected) => {
*token == Tok::Indent || expected.clone() == Some("Indent".to_owned())
}
_ => false,
}
}
pub fn is_tab_error(&self) -> bool {
matches!(
self,
ParseErrorType::Lexical(LexicalErrorType::TabError)
| ParseErrorType::Lexical(LexicalErrorType::TabsAfterSpaces)
)
}
}
impl std::ops::Deref for ParseError {
type Target = ParseErrorType;
fn deref(&self) -> &Self::Target {
&self.error
}
}
impl Error for ParseError {
fn source(&self) -> Option<&(dyn Error + 'static)> {
None
}
}

405
nac3parser/src/fstring.rs Normal file
View File

@ -0,0 +1,405 @@
use std::iter;
use std::mem;
use std::str;
use crate::ast::{Constant, ConversionFlag, Expr, ExprKind, Location};
use crate::error::{FStringError, FStringErrorType, ParseError};
use crate::parser::parse_expression;
use self::FStringErrorType::*;
struct FStringParser<'a> {
chars: iter::Peekable<str::Chars<'a>>,
str_location: Location,
}
impl<'a> FStringParser<'a> {
fn new(source: &'a str, str_location: Location) -> Self {
Self {
chars: source.chars().peekable(),
str_location,
}
}
#[inline]
fn expr(&self, node: ExprKind) -> Expr {
Expr::new(self.str_location, node)
}
fn parse_formatted_value(&mut self) -> Result<Vec<Expr>, FStringErrorType> {
let mut expression = String::new();
let mut spec = None;
let mut delims = Vec::new();
let mut conversion = None;
let mut pred_expression_text = String::new();
let mut trailing_seq = String::new();
while let Some(ch) = self.chars.next() {
match ch {
// can be integrated better with the remainign code, but as a starting point ok
// in general I would do here a tokenizing of the fstrings to omit this peeking.
'!' if self.chars.peek() == Some(&'=') => {
expression.push_str("!=");
self.chars.next();
}
'=' if self.chars.peek() == Some(&'=') => {
expression.push_str("==");
self.chars.next();
}
'>' if self.chars.peek() == Some(&'=') => {
expression.push_str(">=");
self.chars.next();
}
'<' if self.chars.peek() == Some(&'=') => {
expression.push_str("<=");
self.chars.next();
}
'!' if delims.is_empty() && self.chars.peek() != Some(&'=') => {
if expression.trim().is_empty() {
return Err(EmptyExpression);
}
conversion = Some(match self.chars.next() {
Some('s') => ConversionFlag::Str,
Some('a') => ConversionFlag::Ascii,
Some('r') => ConversionFlag::Repr,
Some(_) => {
return Err(InvalidConversionFlag);
}
None => {
return Err(ExpectedRbrace);
}
});
if let Some(&peek) = self.chars.peek() {
if peek != '}' && peek != ':' {
return Err(ExpectedRbrace);
}
} else {
return Err(ExpectedRbrace);
}
}
// match a python 3.8 self documenting expression
// format '{' PYTHON_EXPRESSION '=' FORMAT_SPECIFIER? '}'
'=' if self.chars.peek() != Some(&'=') && delims.is_empty() => {
pred_expression_text = expression.to_string(); // safe expression before = to print it
}
':' if delims.is_empty() => {
let mut nested = false;
let mut in_nested = false;
let mut spec_expression = String::new();
while let Some(&next) = self.chars.peek() {
match next {
'{' => {
if in_nested {
return Err(ExpressionNestedTooDeeply);
}
in_nested = true;
nested = true;
self.chars.next();
continue;
}
'}' => {
if in_nested {
in_nested = false;
self.chars.next();
}
break;
}
_ => (),
}
spec_expression.push(next);
self.chars.next();
}
if in_nested {
return Err(UnclosedLbrace);
}
spec = Some(if nested {
Box::new(
self.expr(ExprKind::FormattedValue {
value: Box::new(
parse_fstring_expr(&spec_expression)
.map_err(|e| InvalidExpression(Box::new(e.error)))?,
),
conversion: None,
format_spec: None,
}),
)
} else {
Box::new(self.expr(ExprKind::Constant {
value: spec_expression.to_owned().into(),
kind: None,
}))
})
}
'(' | '{' | '[' => {
expression.push(ch);
delims.push(ch);
}
')' => {
if delims.pop() != Some('(') {
return Err(MismatchedDelimiter);
}
expression.push(ch);
}
']' => {
if delims.pop() != Some('[') {
return Err(MismatchedDelimiter);
}
expression.push(ch);
}
'}' if !delims.is_empty() => {
if delims.pop() != Some('{') {
return Err(MismatchedDelimiter);
}
expression.push(ch);
}
'}' => {
if expression.is_empty() {
return Err(EmptyExpression);
}
let ret = if pred_expression_text.is_empty() {
vec![self.expr(ExprKind::FormattedValue {
value: Box::new(
parse_fstring_expr(&expression)
.map_err(|e| InvalidExpression(Box::new(e.error)))?,
),
conversion,
format_spec: spec,
})]
} else {
vec![
self.expr(ExprKind::Constant {
value: Constant::Str(pred_expression_text + "="),
kind: None,
}),
self.expr(ExprKind::Constant {
value: trailing_seq.into(),
kind: None,
}),
self.expr(ExprKind::FormattedValue {
value: Box::new(
parse_fstring_expr(&expression)
.map_err(|e| InvalidExpression(Box::new(e.error)))?,
),
conversion,
format_spec: spec,
}),
]
};
return Ok(ret);
}
'"' | '\'' => {
expression.push(ch);
for next in &mut self.chars {
expression.push(next);
if next == ch {
break;
}
}
}
' ' if !pred_expression_text.is_empty() => {
trailing_seq.push(ch);
}
_ => {
expression.push(ch);
}
}
}
Err(UnclosedLbrace)
}
fn parse(mut self) -> Result<Expr, FStringErrorType> {
let mut content = String::new();
let mut values = vec![];
while let Some(ch) = self.chars.next() {
match ch {
'{' => {
if let Some('{') = self.chars.peek() {
self.chars.next();
content.push('{');
} else {
if !content.is_empty() {
values.push(self.expr(ExprKind::Constant {
value: mem::take(&mut content).into(),
kind: None,
}));
}
values.extend(self.parse_formatted_value()?);
}
}
'}' => {
if let Some('}') = self.chars.peek() {
self.chars.next();
content.push('}');
} else {
return Err(UnopenedRbrace);
}
}
_ => {
content.push(ch);
}
}
}
if !content.is_empty() {
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,
}),
1 => values.into_iter().next().unwrap(),
_ => self.expr(ExprKind::JoinedStr { values }),
};
Ok(s)
}
}
fn parse_fstring_expr(source: &str) -> Result<Expr, ParseError> {
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 })
}
#[cfg(test)]
mod tests {
use super::*;
fn parse_fstring(source: &str) -> Result<Expr, FStringErrorType> {
FStringParser::new(source, Location::default()).parse()
}
#[test]
fn test_parse_fstring() {
let source = "{a}{ b }{{foo}}";
let parse_ast = parse_fstring(&source).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_fstring_nested_spec() {
let source = "{foo:{spec}}";
let parse_ast = parse_fstring(&source).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_fstring_not_nested_spec() {
let source = "{foo:spec}";
let parse_ast = parse_fstring(&source).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_empty_fstring() {
insta::assert_debug_snapshot!(parse_fstring("").unwrap());
}
#[test]
fn test_fstring_parse_selfdocumenting_base() {
let src = "{user=}";
let parse_ast = parse_fstring(&src).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_fstring_parse_selfdocumenting_base_more() {
let src = "mix {user=} with text and {second=}";
let parse_ast = parse_fstring(&src).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_fstring_parse_selfdocumenting_format() {
let src = "{user=:>10}";
let parse_ast = parse_fstring(&src).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_invalid_fstring() {
assert_eq!(parse_fstring("{5!a"), Err(ExpectedRbrace));
assert_eq!(parse_fstring("{5!a1}"), Err(ExpectedRbrace));
assert_eq!(parse_fstring("{5!"), Err(ExpectedRbrace));
assert_eq!(parse_fstring("abc{!a 'cat'}"), Err(EmptyExpression));
assert_eq!(parse_fstring("{!a"), Err(EmptyExpression));
assert_eq!(parse_fstring("{ !a}"), Err(EmptyExpression));
assert_eq!(parse_fstring("{5!}"), Err(InvalidConversionFlag));
assert_eq!(parse_fstring("{5!x}"), Err(InvalidConversionFlag));
assert_eq!(parse_fstring("{a:{a:{b}}"), Err(ExpressionNestedTooDeeply));
assert_eq!(parse_fstring("{a:b}}"), Err(UnopenedRbrace));
assert_eq!(parse_fstring("}"), Err(UnopenedRbrace));
assert_eq!(parse_fstring("{a:{b}"), Err(UnclosedLbrace));
assert_eq!(parse_fstring("{"), Err(UnclosedLbrace));
assert_eq!(parse_fstring("{}"), Err(EmptyExpression));
// TODO: check for InvalidExpression enum?
assert!(parse_fstring("{class}").is_err());
}
#[test]
fn test_parse_fstring_not_equals() {
let source = "{1 != 2}";
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();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_fstring_selfdoc_prec_space() {
let source = "{x =}";
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();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_fstring_yield_expr() {
let source = "{yield}";
let parse_ast = parse_fstring(&source).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
}

View File

@ -0,0 +1,96 @@
use ahash::RandomState;
use std::collections::HashSet;
use crate::ast;
use crate::error::{LexicalError, LexicalErrorType};
pub struct ArgumentList {
pub args: Vec<ast::Expr>,
pub keywords: Vec<ast::Keyword>,
}
type ParameterDefs = (Vec<ast::Arg>, Vec<ast::Arg>, Vec<ast::Expr>);
type ParameterDef = (ast::Arg, Option<ast::Expr>);
pub fn parse_params(
params: (Vec<ParameterDef>, Vec<ParameterDef>),
) -> Result<ParameterDefs, LexicalError> {
let mut posonly = Vec::with_capacity(params.0.len());
let mut names = Vec::with_capacity(params.1.len());
let mut defaults = vec![];
let mut try_default = |name: &ast::Arg, default| {
if let Some(default) = default {
defaults.push(default);
} else if !defaults.is_empty() {
// Once we have started with defaults, all remaining arguments must
// have defaults
return Err(LexicalError {
error: LexicalErrorType::DefaultArgumentError,
location: name.location,
});
}
Ok(())
};
for (name, default) in params.0 {
try_default(&name, default)?;
posonly.push(name);
}
for (name, default) in params.1 {
try_default(&name, default)?;
names.push(name);
}
Ok((posonly, names, defaults))
}
type FunctionArgument = (Option<(ast::Location, Option<String>)>, ast::Expr);
pub fn parse_args(func_args: Vec<FunctionArgument>) -> Result<ArgumentList, LexicalError> {
let mut args = vec![];
let mut keywords = vec![];
let mut keyword_names = HashSet::with_capacity_and_hasher(func_args.len(), RandomState::new());
for (name, value) in func_args {
match name {
Some((location, name)) => {
if let Some(keyword_name) = &name {
if keyword_names.contains(keyword_name) {
return Err(LexicalError {
error: LexicalErrorType::DuplicateKeywordArgumentError,
location,
});
}
keyword_names.insert(keyword_name.clone());
}
keywords.push(ast::Keyword::new(
location,
ast::KeywordData {
arg: name.map(|name| name.into()),
value: Box::new(value),
},
));
}
None => {
// Allow starred args after keyword arguments.
if !keywords.is_empty() && !is_starred(&value) {
return Err(LexicalError {
error: LexicalErrorType::PositionalArgumentError,
location: value.location,
});
}
args.push(value);
}
}
}
Ok(ArgumentList { args, keywords })
}
fn is_starred(exp: &ast::Expr) -> bool {
matches!(exp.node, ast::ExprKind::Starred { .. })
}

1895
nac3parser/src/lexer.rs Normal file

File diff suppressed because it is too large Load Diff

35
nac3parser/src/lib.rs Normal file
View File

@ -0,0 +1,35 @@
//! This crate can be used to parse python sourcecode into a so
//! called AST (abstract syntax tree).
//!
//! The stages involved in this process are lexical analysis and
//! parsing. The lexical analysis splits the sourcecode into
//! tokens, and the parsing transforms those tokens into an AST.
//!
//! For example, one could do this:
//!
//! ```
//! use nac3parser::{parser, ast};
//!
//! let python_source = "print('Hello world')";
//! let python_ast = parser::parse_expression(python_source).unwrap();
//!
//! ```
#[macro_use]
extern crate log;
use lalrpop_util::lalrpop_mod;
pub use nac3ast as ast;
pub mod error;
mod fstring;
mod function;
pub mod lexer;
pub mod mode;
pub mod parser;
lalrpop_mod!(
#[allow(clippy::all)]
#[allow(unused)]
python
);
pub mod token;
pub mod config_comment_helper;

40
nac3parser/src/mode.rs Normal file
View File

@ -0,0 +1,40 @@
use crate::token::Tok;
#[derive(Clone, Copy)]
pub enum Mode {
Module,
Interactive,
Expression,
}
impl Mode {
pub(crate) fn to_marker(self) -> Tok {
match self {
Self::Module => Tok::StartModule,
Self::Interactive => Tok::StartInteractive,
Self::Expression => Tok::StartExpression,
}
}
}
impl std::str::FromStr for Mode {
type Err = ModeParseError;
fn from_str(s: &str) -> Result<Self, ModeParseError> {
match s {
"exec" | "single" => Ok(Mode::Module),
"eval" => Ok(Mode::Expression),
_ => Err(ModeParseError { _priv: () }),
}
}
}
#[derive(Debug)]
pub struct ModeParseError {
_priv: (),
}
impl std::fmt::Display for ModeParseError {
fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
write!(f, r#"mode should be "exec", "eval", or "single""#)
}
}

229
nac3parser/src/parser.rs Normal file
View File

@ -0,0 +1,229 @@
//! Python parsing.
//!
//! Use this module to parse python code into an AST.
//! There are three ways to parse python code. You could
//! parse a whole program, a single statement, or a single
//! expression.
use std::iter;
use crate::ast::{self, FileName};
use crate::error::ParseError;
use crate::lexer;
pub use crate::mode::Mode;
use crate::python;
/*
* Parse python code.
* Grammar may be inspired by antlr grammar for python:
* https://github.com/antlr/grammars-v4/tree/master/python3
*/
/// Parse a full python program, containing usually multiple lines.
pub fn parse_program(source: &str, file: FileName) -> Result<ast::Suite, ParseError> {
parse(source, Mode::Module, file).map(|top| match top {
ast::Mod::Module { body, .. } => body,
_ => unreachable!(),
})
}
/// Parses a python expression
///
/// # Example
/// ```
/// use nac3parser::{parser, ast};
/// let expr = parser::parse_expression("1 + 2").unwrap();
///
/// assert_eq!(
/// expr,
/// ast::Expr {
/// location: ast::Location::new(1, 3, Default::default()),
/// custom: (),
/// node: ast::ExprKind::BinOp {
/// left: Box::new(ast::Expr {
/// location: ast::Location::new(1, 1, Default::default()),
/// custom: (),
/// node: ast::ExprKind::Constant {
/// value: ast::Constant::Int(1.into()),
/// kind: None,
/// }
/// }),
/// op: ast::Operator::Add,
/// right: Box::new(ast::Expr {
/// location: ast::Location::new(1, 5, Default::default()),
/// custom: (),
/// node: ast::ExprKind::Constant {
/// value: ast::Constant::Int(2.into()),
/// kind: None,
/// }
/// })
/// }
/// },
/// );
///
/// ```
pub fn parse_expression(source: &str) -> Result<ast::Expr, ParseError> {
parse(source, Mode::Expression, Default::default()).map(|top| match top {
ast::Mod::Expression { body } => *body,
_ => unreachable!(),
})
}
// 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 = (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)
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn test_parse_empty() {
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, 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, 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, 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, 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, Default::default()).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_tuples() {
let source = "a, b = 4, 5";
insta::assert_debug_snapshot!(parse_program(source, Default::default()).unwrap());
}
#[test]
fn test_parse_class() {
let source = "\
class Foo(A, B):
def __init__(self):
pass
def method_with_default(self, arg='default'):
pass";
insta::assert_debug_snapshot!(parse_program(source, Default::default()).unwrap());
}
#[test]
fn test_parse_dict_comprehension() {
let source = String::from("{x1: x2 for y in z}");
let parse_ast = parse_expression(&source).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_list_comprehension() {
let source = String::from("[x for y in z]");
let parse_ast = parse_expression(&source).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_parse_double_list_comprehension() {
let source = String::from("[x for y, y2 in z for a in b if a < 5 if a > 10]");
let parse_ast = parse_expression(&source).unwrap();
insta::assert_debug_snapshot!(parse_ast);
}
#[test]
fn test_more_comment() {
let source = "\
a: int # nac3: sf1
# nac3: sdf4
for i in (1, '12'): # nac3: sf2
a: int
# nac3: 3
# nac3: 5
while i < 2: # nac3: 4
# nac3: real pass
pass
# nac3: expr1
# nac3: expr3
1 + 2 # nac3: expr2
# nac3: if3
# nac3: if1
if 1: # nac3: if2
3";
insta::assert_debug_snapshot!(parse_program(source, Default::default()).unwrap());
}
#[test]
fn test_sample_comment() {
let source = "\
# nac3: while1
# nac3: while2
# normal comment
while test: # nac3: while3
# nac3: simple assign0
a = 3 # nac3: simple assign1
";
insta::assert_debug_snapshot!(parse_program(source, Default::default()).unwrap());
}
#[test]
fn test_comment_ambiguity() {
let source = "\
if a: d; # nac3: for d
if b: c # nac3: for c
if d: # nac3: for if d
b; b + 3; # nac3: for b + 3
a = 3; a + 3; b = a; # nac3: notif
# nac3: smallsingle1
# nac3: smallsingle3
aa = 3 # nac3: smallsingle2
if a: # nac3: small2
a
for i in a: # nac3: for1
pass
";
insta::assert_debug_snapshot!(parse_program(source, Default::default()).unwrap());
}
#[test]
fn test_comment_should_fail() {
let source = "\
if a: # nac3: something
a = 3
";
assert!(parse_program(source, Default::default()).is_err());
}
}

File diff suppressed because it is too large Load Diff

View File

@ -0,0 +1,80 @@
---
source: nac3parser/src/fstring.rs
assertion_line: 327
expression: parse_ast
---
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: JoinedStr {
values: [
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Str(
"user=",
),
kind: None,
},
},
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Str(
"",
),
kind: None,
},
},
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: FormattedValue {
value: Located {
location: Location {
row: 1,
column: 2,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "user",
ctx: Load,
},
},
conversion: None,
format_spec: None,
},
},
],
},
}

View File

@ -0,0 +1,172 @@
---
source: nac3parser/src/fstring.rs
assertion_line: 335
expression: parse_ast
---
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: JoinedStr {
values: [
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Str(
"mix ",
),
kind: None,
},
},
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Str(
"user=",
),
kind: None,
},
},
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Str(
"",
),
kind: None,
},
},
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: FormattedValue {
value: Located {
location: Location {
row: 1,
column: 2,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "user",
ctx: Load,
},
},
conversion: None,
format_spec: None,
},
},
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Str(
" with text and ",
),
kind: None,
},
},
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Str(
"second=",
),
kind: None,
},
},
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Str(
"",
),
kind: None,
},
},
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: FormattedValue {
value: Located {
location: Location {
row: 1,
column: 2,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "second",
ctx: Load,
},
},
conversion: None,
format_spec: None,
},
},
],
},
}

View File

@ -0,0 +1,97 @@
---
source: nac3parser/src/fstring.rs
assertion_line: 343
expression: parse_ast
---
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: JoinedStr {
values: [
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Str(
"user=",
),
kind: None,
},
},
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Str(
"",
),
kind: None,
},
},
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: FormattedValue {
value: Located {
location: Location {
row: 1,
column: 2,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "user",
ctx: Load,
},
},
conversion: None,
format_spec: Some(
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Str(
">10",
),
kind: None,
},
},
),
},
},
],
},
}

View File

@ -0,0 +1,22 @@
---
source: nac3parser/src/fstring.rs
assertion_line: 319
expression: "parse_fstring(\"\").unwrap()"
---
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Str(
"",
),
kind: None,
},
}

View File

@ -0,0 +1,92 @@
---
source: nac3parser/src/fstring.rs
assertion_line: 298
expression: parse_ast
---
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: JoinedStr {
values: [
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: FormattedValue {
value: Located {
location: Location {
row: 1,
column: 2,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "a",
ctx: Load,
},
},
conversion: None,
format_spec: None,
},
},
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: FormattedValue {
value: Located {
location: Location {
row: 1,
column: 3,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "b",
ctx: Load,
},
},
conversion: None,
format_spec: None,
},
},
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Str(
"{foo}",
),
kind: None,
},
},
],
},
}

View File

@ -0,0 +1,69 @@
---
source: nac3parser/src/fstring.rs
assertion_line: 382
expression: parse_ast
---
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: FormattedValue {
value: Located {
location: Location {
row: 1,
column: 5,
file: FileName(
"unknown",
),
},
custom: (),
node: Compare {
left: Located {
location: Location {
row: 1,
column: 2,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Int(
42,
),
kind: None,
},
},
ops: [
Eq,
],
comparators: [
Located {
location: Location {
row: 1,
column: 8,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Int(
42,
),
kind: None,
},
},
],
},
},
conversion: None,
format_spec: None,
},
}

View File

@ -0,0 +1,63 @@
---
source: nac3parser/src/fstring.rs
assertion_line: 306
expression: parse_ast
---
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: FormattedValue {
value: Located {
location: Location {
row: 1,
column: 2,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "foo",
ctx: Load,
},
},
conversion: None,
format_spec: Some(
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: FormattedValue {
value: Located {
location: Location {
row: 1,
column: 2,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "spec",
ctx: Load,
},
},
conversion: None,
format_spec: None,
},
},
),
},
}

View File

@ -0,0 +1,69 @@
---
source: nac3parser/src/fstring.rs
assertion_line: 375
expression: parse_ast
---
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: FormattedValue {
value: Located {
location: Location {
row: 1,
column: 4,
file: FileName(
"unknown",
),
},
custom: (),
node: Compare {
left: Located {
location: Location {
row: 1,
column: 2,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Int(
1,
),
kind: None,
},
},
ops: [
NotEq,
],
comparators: [
Located {
location: Location {
row: 1,
column: 7,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Int(
2,
),
kind: None,
},
},
],
},
},
conversion: None,
format_spec: None,
},
}

View File

@ -0,0 +1,51 @@
---
source: nac3parser/src/fstring.rs
assertion_line: 314
expression: parse_ast
---
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: FormattedValue {
value: Located {
location: Location {
row: 1,
column: 2,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "foo",
ctx: Load,
},
},
conversion: None,
format_spec: Some(
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Str(
"spec",
),
kind: None,
},
},
),
},
}

View File

@ -0,0 +1,80 @@
---
source: nac3parser/src/fstring.rs
assertion_line: 389
expression: parse_ast
---
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: JoinedStr {
values: [
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Str(
"x =",
),
kind: None,
},
},
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Str(
"",
),
kind: None,
},
},
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: FormattedValue {
value: Located {
location: Location {
row: 1,
column: 2,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "x",
ctx: Load,
},
},
conversion: None,
format_spec: None,
},
},
],
},
}

View File

@ -0,0 +1,80 @@
---
source: nac3parser/src/fstring.rs
assertion_line: 396
expression: parse_ast
---
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: JoinedStr {
values: [
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Str(
"x=",
),
kind: None,
},
},
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Str(
" ",
),
kind: None,
},
},
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: FormattedValue {
value: Located {
location: Location {
row: 1,
column: 2,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "x",
ctx: Load,
},
},
conversion: None,
format_spec: None,
},
},
],
},
}

View File

@ -0,0 +1,33 @@
---
source: nac3parser/src/fstring.rs
assertion_line: 403
expression: parse_ast
---
Located {
location: Location {
row: 0,
column: 0,
file: FileName(
"unknown",
),
},
custom: (),
node: FormattedValue {
value: Located {
location: Location {
row: 1,
column: 2,
file: FileName(
"unknown",
),
},
custom: (),
node: Yield {
value: None,
},
},
conversion: None,
format_spec: None,
},
}

View File

@ -0,0 +1,560 @@
---
source: nac3parser/src/parser.rs
assertion_line: 218
expression: "parse_program(&source, Default::default()).unwrap()"
---
[
Located {
location: Location {
row: 1,
column: 1,
file: FileName(
"unknown",
),
},
custom: (),
node: If {
test: Located {
location: Location {
row: 1,
column: 4,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "a",
ctx: Load,
},
},
body: [
Located {
location: Location {
row: 1,
column: 7,
file: FileName(
"unknown",
),
},
custom: (),
node: Expr {
value: Located {
location: Location {
row: 1,
column: 7,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "d",
ctx: Load,
},
},
config_comment: [
"for d",
],
},
},
],
orelse: [],
config_comment: [],
},
},
Located {
location: Location {
row: 2,
column: 1,
file: FileName(
"unknown",
),
},
custom: (),
node: If {
test: Located {
location: Location {
row: 2,
column: 4,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "b",
ctx: Load,
},
},
body: [
Located {
location: Location {
row: 2,
column: 7,
file: FileName(
"unknown",
),
},
custom: (),
node: Expr {
value: Located {
location: Location {
row: 2,
column: 7,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "c",
ctx: Load,
},
},
config_comment: [
"for c",
],
},
},
],
orelse: [],
config_comment: [],
},
},
Located {
location: Location {
row: 3,
column: 1,
file: FileName(
"unknown",
),
},
custom: (),
node: If {
test: Located {
location: Location {
row: 3,
column: 4,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "d",
ctx: Load,
},
},
body: [
Located {
location: Location {
row: 4,
column: 5,
file: FileName(
"unknown",
),
},
custom: (),
node: Expr {
value: Located {
location: Location {
row: 4,
column: 5,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "b",
ctx: Load,
},
},
config_comment: [],
},
},
Located {
location: Location {
row: 4,
column: 8,
file: FileName(
"unknown",
),
},
custom: (),
node: Expr {
value: Located {
location: Location {
row: 4,
column: 10,
file: FileName(
"unknown",
),
},
custom: (),
node: BinOp {
left: Located {
location: Location {
row: 4,
column: 8,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "b",
ctx: Load,
},
},
op: Add,
right: Located {
location: Location {
row: 4,
column: 12,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Int(
3,
),
kind: None,
},
},
},
},
config_comment: [
"for b + 3",
],
},
},
],
orelse: [],
config_comment: [
"for if d",
],
},
},
Located {
location: Location {
row: 5,
column: 1,
file: FileName(
"unknown",
),
},
custom: (),
node: Assign {
targets: [
Located {
location: Location {
row: 5,
column: 1,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "a",
ctx: Load,
},
},
],
value: Located {
location: Location {
row: 5,
column: 5,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Int(
3,
),
kind: None,
},
},
type_comment: None,
config_comment: [],
},
},
Located {
location: Location {
row: 5,
column: 8,
file: FileName(
"unknown",
),
},
custom: (),
node: Expr {
value: Located {
location: Location {
row: 5,
column: 10,
file: FileName(
"unknown",
),
},
custom: (),
node: BinOp {
left: Located {
location: Location {
row: 5,
column: 8,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "a",
ctx: Load,
},
},
op: Add,
right: Located {
location: Location {
row: 5,
column: 12,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Int(
3,
),
kind: None,
},
},
},
},
config_comment: [],
},
},
Located {
location: Location {
row: 5,
column: 15,
file: FileName(
"unknown",
),
},
custom: (),
node: Assign {
targets: [
Located {
location: Location {
row: 5,
column: 15,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "b",
ctx: Load,
},
},
],
value: Located {
location: Location {
row: 5,
column: 19,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "a",
ctx: Load,
},
},
type_comment: None,
config_comment: [
"notif",
],
},
},
Located {
location: Location {
row: 8,
column: 1,
file: FileName(
"unknown",
),
},
custom: (),
node: Assign {
targets: [
Located {
location: Location {
row: 8,
column: 1,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "aa",
ctx: Load,
},
},
],
value: Located {
location: Location {
row: 8,
column: 6,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Int(
3,
),
kind: None,
},
},
type_comment: None,
config_comment: [
"smallsingle1",
"smallsingle3",
"smallsingle2",
],
},
},
Located {
location: Location {
row: 9,
column: 1,
file: FileName(
"unknown",
),
},
custom: (),
node: If {
test: Located {
location: Location {
row: 9,
column: 4,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "a",
ctx: Load,
},
},
body: [
Located {
location: Location {
row: 10,
column: 5,
file: FileName(
"unknown",
),
},
custom: (),
node: Expr {
value: Located {
location: Location {
row: 10,
column: 5,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "a",
ctx: Load,
},
},
config_comment: [],
},
},
],
orelse: [],
config_comment: [
"small2",
],
},
},
Located {
location: Location {
row: 11,
column: 1,
file: FileName(
"unknown",
),
},
custom: (),
node: For {
target: Located {
location: Location {
row: 11,
column: 5,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "i",
ctx: Load,
},
},
iter: Located {
location: Location {
row: 11,
column: 10,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "a",
ctx: Load,
},
},
body: [
Located {
location: Location {
row: 12,
column: 5,
file: FileName(
"unknown",
),
},
custom: (),
node: Pass {
config_comment: [],
},
},
],
orelse: [],
type_comment: None,
config_comment: [
"for1",
],
},
},
]

View File

@ -0,0 +1,386 @@
---
source: nac3parser/src/parser.rs
assertion_line: 186
expression: "parse_program(&source, Default::default()).unwrap()"
---
[
Located {
location: Location {
row: 1,
column: 1,
file: FileName(
"unknown",
),
},
custom: (),
node: AnnAssign {
target: Located {
location: Location {
row: 1,
column: 1,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "a",
ctx: Load,
},
},
annotation: Located {
location: Location {
row: 1,
column: 4,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "int",
ctx: Load,
},
},
value: None,
simple: true,
config_comment: [
"sf1",
],
},
},
Located {
location: Location {
row: 2,
column: 1,
file: FileName(
"unknown",
),
},
custom: (),
node: For {
target: Located {
location: Location {
row: 3,
column: 5,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "i",
ctx: Load,
},
},
iter: Located {
location: Location {
row: 3,
column: 11,
file: FileName(
"unknown",
),
},
custom: (),
node: Tuple {
elts: [
Located {
location: Location {
row: 3,
column: 11,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Int(
1,
),
kind: None,
},
},
Located {
location: Location {
row: 3,
column: 15,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Str(
"12",
),
kind: None,
},
},
],
ctx: Load,
},
},
body: [
Located {
location: Location {
row: 4,
column: 5,
file: FileName(
"unknown",
),
},
custom: (),
node: AnnAssign {
target: Located {
location: Location {
row: 4,
column: 5,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "a",
ctx: Load,
},
},
annotation: Located {
location: Location {
row: 4,
column: 8,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "int",
ctx: Load,
},
},
value: None,
simple: true,
config_comment: [],
},
},
],
orelse: [],
type_comment: None,
config_comment: [
"sdf4",
"sf2",
],
},
},
Located {
location: Location {
row: 5,
column: 1,
file: FileName(
"unknown",
),
},
custom: (),
node: While {
test: Located {
location: Location {
row: 7,
column: 9,
file: FileName(
"unknown",
),
},
custom: (),
node: Compare {
left: Located {
location: Location {
row: 7,
column: 7,
file: FileName(
"unknown",
),
},
custom: (),
node: Name {
id: "i",
ctx: Load,
},
},
ops: [
Lt,
],
comparators: [
Located {
location: Location {
row: 7,
column: 11,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Int(
2,
),
kind: None,
},
},
],
},
},
body: [
Located {
location: Location {
row: 9,
column: 5,
file: FileName(
"unknown",
),
},
custom: (),
node: Pass {
config_comment: [
"real pass",
],
},
},
Located {
location: Location {
row: 12,
column: 5,
file: FileName(
"unknown",
),
},
custom: (),
node: Expr {
value: Located {
location: Location {
row: 12,
column: 7,
file: FileName(
"unknown",
),
},
custom: (),
node: BinOp {
left: Located {
location: Location {
row: 12,
column: 5,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Int(
1,
),
kind: None,
},
},
op: Add,
right: Located {
location: Location {
row: 12,
column: 9,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Int(
2,
),
kind: None,
},
},
},
},
config_comment: [
"expr1",
"expr3",
"expr2",
],
},
},
Located {
location: Location {
row: 13,
column: 5,
file: FileName(
"unknown",
),
},
custom: (),
node: If {
test: Located {
location: Location {
row: 15,
column: 8,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Int(
1,
),
kind: None,
},
},
body: [
Located {
location: Location {
row: 16,
column: 9,
file: FileName(
"unknown",
),
},
custom: (),
node: Expr {
value: Located {
location: Location {
row: 16,
column: 9,
file: FileName(
"unknown",
),
},
custom: (),
node: Constant {
value: Int(
3,
),
kind: None,
},
},
config_comment: [],
},
},
],
orelse: [],
config_comment: [
"if3",
"if1",
"if2",
],
},
},
],
orelse: [],
config_comment: [
"3",
"5",
"4",
],
},
},
]

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