feat[next]: adding python containers #2232
Conversation
egparedes
force-pushed
the
python-containers
branch
from
2988116 to
5a6ce4f
Compare
egparedes
force-pushed
the
python-containers
branch
from
5a6ce4f to
7ce5f52
Compare
| return extractors if extractors else None | ||
|
|
||
| def __postinit__(self) -> None: | ||
| def __post_init__(self) -> None: |
There was a problem hiding this comment.
TODO: add a pygrep expression in pre-commit to catch this (or check if ruff has a rule for this).
| *self.program_type.definition.pos_or_kw_args.values(), | ||
| ] | ||
| ): | ||
| if isinstance(type_spec, ts.NamedTupleType): |
There was a problem hiding this comment.
If it's a ts.TupleType that contains NamedTupleType, we have to generate an extractor.
src/gt4py/next/containers.py
Outdated
| def extract( | ||
| container, | ||
| ) -> NestedTuple[ | ||
| PythonContainerValue | ||
| ]: # TODO the input is NestedAnyContainer[PythonContainerValue], to be defined |
There was a problem hiding this comment.
possibly remove again once we have the extractor for all args
| def __setitem__(self, index: AnyIndexSpec, value: Field | core_defs.ScalarT) -> None: ... | ||
|
|
||
|
|
||
| NumericValue: TypeAlias = core_defs.Scalar | Field |
There was a problem hiding this comment.
| NumericValue: TypeAlias = core_defs.Scalar | Field | |
| #: Elements of containers: TODO: use more examples | |
| NumericValue: TypeAlias = core_defs.Scalar | Field |
|
cscs-ci run |
| return len(self.types) | ||
|
|
||
|
|
||
| class NamedTupleType(TupleType): |
There was a problem hiding this comment.
Inheriting from TupleType is dangerous as NamedTuples can not be used in all places where tuples are allowed, but we check if something is a tuple by doing isinstance(..., TupleType)
There was a problem hiding this comment.
Maybe call it StructType because due to the original_python_type it is not a just a NamedTuple anymore.
|
|
||
| class NamedTupleType(TupleType): | ||
| keys: list[str] | ||
| original_python_type: str # Format: '__module__:__qualname__' (used by pkgutil.resolve_name()) |
There was a problem hiding this comment.
I would add a comment that described where is value may be used and where not. For example inside of the type_deduction is should never be used.
In case there is a view of a global array, we do not have a problem, because we never modify the strides of globals. The problems start when there is a view of a transient. The SDFG is constructed using C order, which good for CPU. However, for GPU we go to FORTRAN order which is essentially swapping the memory order. But we can not only do that for transients, but must also do it for views, which might be difficult, because views can add or remove certain dimensions. This PR does _not_ solve it, but instead generates an error if such a case is detected.
## Description We found that there was an issue with absolute k-indexing inside the conditional of a while loop. This was due to a missing propagation of the symbol-table into the conditional. This PR fixes that issue. ## Requirements - [x] All fixes and/or new features come with corresponding tests. - [ ] Important design decisions have been documented in the appropriate ADR inside the [docs/development/ADRs/](docs/development/ADRs/README.md) folder. N/A
Adding support for typed namedtuples and Dataclasses as basic containers structures for related Fields. Changes initially prototyped in #2211