TYP: lib._function_base_impl: many typing improvements
#30246
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Member
Author
|
This turned out to be a lot more work than I had initially anticipated 😅 |
This comment has been minimized.
This comment has been minimized.
Merged
7 tasks
This comment has been minimized.
This comment has been minimized.
Member
Author
|
The primer error for The primer error for The disappeared |
Member
|
Ping @MarcoGorelli - any chance you can give this a once-over? |
Member
Author
|
Yea that would help :). In the meantime, let's see if mr. robot has anything useful to say about this |
Contributor
There was a problem hiding this comment.
Pull Request Overview
This PR significantly improves typing for functions in numpy.lib._function_base_impl, adding shape-typing support, more accurate dtype specifications, validation for callable signatures, and comprehensive overloads for 30+ functions.
Key Changes:
- Enhanced type precision with shape-aware types (e.g.,
_Array1D,_Array2D) and specific dtype returns - Expanded overloads for functions like
average,median,percentile,quantile,gradient, etc. to handle different input shapes and scalar types - Added callable signature validation for
piecewisefunction usingParamSpec
Reviewed Changes
Copilot reviewed 3 out of 3 changed files in this pull request and generated 4 comments.
| File | Description |
|---|---|
| numpy/lib/_function_base_impl.pyi | Extensive typing improvements including new type aliases, overloads for shape-preservation, and more precise return types for 30+ functions |
| numpy/typing/tests/data/reveal/lib_function_base.pyi | Updated type assertion tests to validate new typing improvements with shape-specific and dtype-specific checks |
| numpy/typing/tests/data/fail/lib_function_base.pyi | Updated error type expectations to match new overload signatures |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Member
Author
|
Ok I have to admit that that was a pretty solid review. I think I'll ask copilot for reviews more often now. |
|
Diff from mypy_primer, showing the effect of this PR on type check results on a corpus of open source code: colour (https://github.com/colour-science/colour)
+ colour/models/rgb/cylindrical.py:531: error: List item 0 has incompatible type "ndarray[tuple[Any, ...], dtype[floating[_16Bit] | floating[_32Bit] | float64]]"; expected "_SupportsArray[dtype[numpy.bool[builtins.bool]]] | _NestedSequence[_SupportsArray[dtype[numpy.bool[builtins.bool]]]] | builtins.bool | _NestedSequence[builtins.bool]" [list-item]
+ colour/models/rgb/cylindrical.py:532: error: List item 1 has incompatible type "ndarray[tuple[Any, ...], dtype[floating[_16Bit] | floating[_32Bit] | float64]]"; expected "_SupportsArray[dtype[numpy.bool[builtins.bool]]] | _NestedSequence[_SupportsArray[dtype[numpy.bool[builtins.bool]]]] | builtins.bool | _NestedSequence[builtins.bool]" [list-item]
+ colour/models/rgb/cylindrical.py:533: error: List item 2 has incompatible type "ndarray[tuple[Any, ...], dtype[floating[_16Bit] | floating[_32Bit] | float64]]"; expected "_SupportsArray[dtype[numpy.bool[builtins.bool]]] | _NestedSequence[_SupportsArray[dtype[numpy.bool[builtins.bool]]]] | builtins.bool | _NestedSequence[builtins.bool]" [list-item]
+ colour/models/rgb/cylindrical.py:534: error: List item 3 has incompatible type "ndarray[tuple[Any, ...], dtype[floating[_16Bit] | floating[_32Bit] | float64]]"; expected "_SupportsArray[dtype[numpy.bool[builtins.bool]]] | _NestedSequence[_SupportsArray[dtype[numpy.bool[builtins.bool]]]] | builtins.bool | _NestedSequence[builtins.bool]" [list-item]
+ colour/models/rgb/cylindrical.py:535: error: List item 4 has incompatible type "ndarray[tuple[Any, ...], dtype[floating[_16Bit] | floating[_32Bit] | float64]]"; expected "_SupportsArray[dtype[numpy.bool[builtins.bool]]] | _NestedSequence[_SupportsArray[dtype[numpy.bool[builtins.bool]]]] | builtins.bool | _NestedSequence[builtins.bool]" [list-item]
+ colour/models/rgb/cylindrical.py:536: error: List item 5 has incompatible type "ndarray[tuple[Any, ...], dtype[floating[_16Bit] | floating[_32Bit] | float64]]"; expected "_SupportsArray[dtype[numpy.bool[builtins.bool]]] | _NestedSequence[_SupportsArray[dtype[numpy.bool[builtins.bool]]]] | builtins.bool | _NestedSequence[builtins.bool]" [list-item]
optuna (https://github.com/optuna/optuna)
+ optuna/samplers/_tpe/_truncnorm.py:262: error: No overload variant of "select" matches argument types "list[Any]", "list[object]", "ndarray[tuple[Any, ...], dtype[Any]]" [call-overload]
+ optuna/samplers/_tpe/_truncnorm.py:262: note: Possible overload variants:
+ optuna/samplers/_tpe/_truncnorm.py:262: note: def [_ArrayT: ndarray[tuple[Any, ...], dtype[Any]]] select(condlist: _SizedIterable[_SupportsArray[dtype[numpy.bool[builtins.bool]]] | _NestedSequence[_SupportsArray[dtype[numpy.bool[builtins.bool]]]] | builtins.bool | _NestedSequence[builtins.bool]], choicelist: Sequence[_ArrayT], default: complex | str | bytes | generic[Any] = ...) -> _ArrayT
+ optuna/samplers/_tpe/_truncnorm.py:262: note: def [_ScalarT: generic[Any]] select(condlist: _SizedIterable[_SupportsArray[dtype[numpy.bool[builtins.bool]]] | _NestedSequence[_SupportsArray[dtype[numpy.bool[builtins.bool]]]] | builtins.bool | _NestedSequence[builtins.bool]], choicelist: Sequence[_SupportsArray[dtype[_ScalarT]] | _NestedSequence[_SupportsArray[dtype[_ScalarT]]]] | ndarray[tuple[Any, ...], dtype[_ScalarT]], default: complex | str | bytes | generic[Any] = ...) -> ndarray[tuple[Any, ...], dtype[_ScalarT]]
+ optuna/samplers/_tpe/_truncnorm.py:262: note: def select(condlist: _SizedIterable[_SupportsArray[dtype[numpy.bool[builtins.bool]]] | _NestedSequence[_SupportsArray[dtype[numpy.bool[builtins.bool]]]] | builtins.bool | _NestedSequence[builtins.bool]], choicelist: Sequence[Buffer | _SupportsArray[dtype[Any]] | _NestedSequence[_SupportsArray[dtype[Any]]] | complex | bytes | str | _NestedSequence[complex | bytes | str]], default: complex | str | bytes | generic[Any] = ...) -> ndarray[tuple[Any, ...], dtype[Any]]
arviz (https://github.com/arviz-devs/arviz)
- arviz/stats/ecdf_utils.py:324: error: Incompatible return value type (got "floating[Any]", expected "float") [return-value]
|
Member
|
Thanks Joren. |
cakedev0
pushed a commit
to cakedev0/numpy
that referenced
this pull request
Dec 5, 2025
* TYP: ``lib._function_base_impl``: many typing improvements * TYP: use the any trick when calling ``gradiant`` with arrays of unknown shape * TYP: fix regression in ``append`` caught by mypy_primer * TYP: Apply Copilot's review suggestions (yes, they actually made sense)
IndifferentArea
pushed a commit
to IndifferentArea/numpy
that referenced
this pull request
Dec 7, 2025
* TYP: ``lib._function_base_impl``: many typing improvements * TYP: use the any trick when calling ``gradiant`` with arrays of unknown shape * TYP: fix regression in ``append`` caught by mypy_primer * TYP: Apply Copilot's review suggestions (yes, they actually made sense)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
For most
numpy.lib._function_base_implfunctions, the adds support for shape-typing, improves the accuracy of returned dtypes, adds validation for callable argument signatures, and adds inline overload comments, and fixes overlapping overloads, towards #27032.The following functions are affected:
np.averagenp.asarray_chkfinitenp.piecewisenp.extractnp.selectnp.copynp.gradientnp.diffnp.interpnp.anglenp.unwrapnp.sort_complexnp.covnp.corrcoefnp.blackmannp.bartlettnp.hanningnp.hammingnp.kaisernp.i0np.sincnp.mediannp.percentilenp.quantilenp.trapezoidnp.meshgridnp.placenp.deletenp.insertnp.appendnp.digitize