WiP: A
10000
dded type annotations for the np.core.fromnumeric module
#59
Conversation
See https://github.com/numpy/numpy-stubs/issues/54. Added annotations and tests for: * np.warnings * np.ModuleDeprecationWarning * np.VisibleDeprecationWarning * np.ComplexWarning * np.RankWarning * np.TooHardError * np.AxisError
* Removed ``warnings`` (#57 (comment)) * Deleted a now redundant comment (#57 (comment)) * Created and/or moved a number of tests to fail/ and reveal/ (#57 (comment) and #57 (comment)) * Formatted the main __init__.pyi file with black (#57 (review))
This should be fine, but let me open an issue to give people a chance to object. |
|
No problem, I've just created the pull request at #63. |
* Added new overloads to the ``np.core.fromnumeric`` functions. * Added the ``_ArrayOrScalarCommon.__array__()`` method. * Added the ``_ArrayLikeSeq`` Protocol. * Renamed ``_Scalar`` to ``_ScalarNumeric``. * Wrapped up the reveal tests. * Removed ``typing_extenions`` from the dependencies; see #63
| # Can be used as such `Union[Sequence[T], _ArrayLikeSeq]` as ndarray | ||
| # is not a proper Sequence | ||
| # TODO: Specify the to-be returned array type once ndarray is a Generic | ||
| class _ArrayLikeSeq(Protocol): |
There was a problem hiding this comment.
Union[Sequence[T], _ArrayLikeSeq] can be used for annotating sequences, as ndarray is technically not a proper sequence.
In the future (once ndarray is a generic) this can be changed into Union[Sequence[T], _ArrayLikeSeq[T]]
| # would be more suited here but they have one significant disadvatange: | ||
| # all these methods are also defined available in ndarray, | ||
| # which has no guarantee of being scalar-esque (i.e. a 0D array) | ||
| _ScalarNumeric = Union[int, float, complex, bool, bool_, number] |
There was a problem hiding this comment.
Not the most ideal solution (due to the lack of duck-typing) but it serves its purpose.
|
I see that the diff here has become quite large-I would appreciate it if you could split this into several smaller PRs. (At the end of a work day it's easier to motivate myself to review small pieces of code.) |
|
No problem, I'll create a new (smaller) pull request later today with the first set of changes. |
See https://github.com/numpy/numpy-stubs/issues/54.
Introduced type hints for all (public) functions from the
np.core.fromnumericmodule.Note thatSee #63.typing_extensionshas also been added as a dependency for Python version prior to 3.8.Functions
takereshapechooserepeatputswapaxestransposepartitionargpartitionsortargsortargmaxargminsearchsortedresizesqueezediagonaltraceravelnonzeroshapecompressclipsumallanycumsumptpamaxaminalenprodcumprodndimsizearoundmeanstdvarround_,product,cumproduct,sometrueandalltrueTo do