core: Inferencer fold_stmt() on StmtKind::Assign should not force unify value_ty with target_ty for __setitem__ assignments. #460

New Issue

lyken · 2024-07-16T12:19:04+08:00

lyken commented

2024-07-16 12:19:04 +08:00

value_ty does not have to be the same as target_ty in general (note that ndarray subscript assignment depends on #411):

arr = np_zeros((3, 4))
another_arr = np_ones((2,))

arr[:,1:3] = another_arr
# not ok
# LHS type == ndarray[2; float64]
# RHS type == ndarray[1; float64]

arr[:] = 1
# not ok
# LHS type == ndarray[2; float64]
# RHS type = float64

If this is rectified, Generator::gen_assign should add an extra parameter value_ty: Type.

Everything here may go the same for AugAssign.

`value_ty` does not have to be the same as `target_ty` in general (note that ndarray subscript assignment depends on https://git.m-labs.hk/M-Labs/nac3/issues/411): ```python arr = np_zeros((3, 4)) another_arr = np_ones((2,)) arr[:,1:3] = another_arr # not ok # LHS type == ndarray[2; float64] # RHS type == ndarray[1; float64] arr[:] = 1 # not ok # LHS type == ndarray[2; float64] # RHS type = float64 ``` If this is rectified, `Generator::gen_assign` should add an extra parameter `value_ty: Type`. Everything here may go the same for `AugAssign`.

lyken referenced this issue from a commit

2024-07-16 13:05:21 +08:00

WIP: stuck on #460

lyken referenced this issue from a commit

2024-07-20 21:25:02 +08:00

core/typecheck: extend typechecker with fold_assign()

lyken referenced this issue from a commit

2024-07-30 16:09:58 +08:00

core/typecheck: extend typechecker with fold_assign()

lyken changed title from ~~core: Inferencer `fold_stmt()` on `StmtKind::Assign` should not force unify `value_ty` with `target_ty`.~~ to core: Inferencer `fold_stmt()` on `StmtKind::Assign` should not force unify `value_ty` with `target_ty` for `__setitem__` assignments.

2024-07-31 10:48:08 +08:00

~~lyken referenced this issue 2024-07-31 10:48:35 +08:00~~

core: Inferencer should treat item assignments differently than direct assignments #482

lyken commented

2024-07-31 10:51:01 +08:00

#482 is a more general issue than this. Closing.

https://git.m-labs.hk/M-Labs/nac3/issues/482 is a more general issue than this. Closing.

lyken closed this issue

2024-07-31 10:51:02 +08:00

lyken referenced this issue

2024-08-09 17:23:32 +08:00

core: Generalize `__setitem__` and `__getitem__` #487