TYP: Fix reverse operators typing for objects with __array__ and __array_ufunc__ #29591
+37
−4
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
…func__ Fixes numpy#29486 ## Description This PR addresses an issue with static typing of reverse operators in NumPy's `ndarray` class. When using objects that implement both `__array__` and `__array_ufunc__` (like pandas Series) in reverse operations with NumPy arrays, the static typing incorrectly reports the result as an `ndarray` instead of the appropriate type from the left-hand side object. The solution implements a Protocol-based approach that adds overloads to the reverse operators (`__radd__`, `__rsub__`, `__rmul__`, etc.) in the `ndarray` class. These overloads check if the left-hand side operand implements the corresponding forward operator and, if so, delegate to that operator, returning its result type. ## Changes 1. Added a `_CanAdd` Protocol in `numpy/_typing/_callable.pyi` 2. Added new overloads to reverse operators in `numpy/__init__.pyi` that use the Protocol 3. Added necessary imports for the new Protocol and TypeVar ## Testing The changes have been syntax-checked with `python -m compileall` and maintain backward compatibility with existing code. For a detailed explanation of the problem and solution, see ISSUE_29486_SOLUTION.md
This comment has been minimized.
This comment has been minimized.
jorenham
reviewed
Aug 19, 2025
numpy/__init__.pyi
Outdated
| @@ -207,6 +208,7 @@ from typing import ( | |||
| SupportsIndex, | |||
| TypeAlias, | |||
| TypedDict, | |||
| TypeVar, | |||
There was a problem hiding this comment.
The TypeVar should be imported from typing_extensions because we use the PEP 696 default kwarg, which requires Python 3.13+
jorenham
reviewed
Aug 19, 2025
numpy/_typing/_callable.pyi
Outdated
| @@ -42,6 +42,7 @@ from ._scalars import _BoolLike_co, _IntLike_co, _NumberLike_co | |||
|
|
|||
| _T1 = TypeVar("_T1") | |||
| _T2 = TypeVar("_T2") | |||
| _T = TypeVar("_T") | |||
jorenham
requested changes
Aug 19, 2025
jorenham
reviewed
Aug 19, 2025
There was a problem hiding this comment.
I can see how you might want to use _CanAdd for __radd__, but it doesn't make much sense to also use that for other reflected binops like __rpow__
Co-authored-by: Joren Hammudoglu <jhammudoglu@gmail.com>
This comment has been minimized.
This comment has been minimized.
|
Thanks for this PR; it's always good to see interest in improving NumPy's type annotations. But keep in mind that every commit you push notifies us, and that running the CI isn't free. So please run the mypy tests locally first ( Also note that this is a very difficult problem to solve, and I'm still not sure what the right approach is, even though I do this for a living. Don't take this the wrong way, but based on your changes so far, it looks to me like you're not an expert in writing typing stubs. I'm a bit worried that the time that your generously investing in this might go to waste, due to its complexity. So perhaps it would be more efficient (and more fun) to work on a different numpy typing issue? Some options that come to mind are #29595, or annotating one of the untyped |
|
Diff from mypy_primer, showing the effect of this PR on type check results on a corpus of open source code: static-frame (https://github.com/static-frame/static-frame)
+ static_frame/core/rank.py:106: error: Unused "type: ignore" comment [unused-ignore]
xarray (https://github.com/pydata/xarray)
+ xarray/computation/weighted.py: note: In function "_weighted_quantile":
+ xarray/computation/weighted.py:333: error: Incompatible types in assignment (expression has type "int", variable has type "ndarray[tuple[Any, ...], dtype[Any]]") [assignment]
+ xarray/computation/weighted.py:337: error: Incompatible types in assignment (expression has type "float", variable has type "ndarray[tuple[Any, ...], dtype[Any]]") [assignment]
+ xarray/computation/weighted.py:341: error: Incompatible types in assignment (expression has type "float", variable has type "ndarray[tuple[Any, ...], dtype[Any]]") [assignment]
+ xarray/computation/weighted.py:343: error: Incompatible types in assignment (expression has type "float", variable has type "ndarray[tuple[Any, ...], dtype[Any]]") [assignment]
+ xarray/tests/test_namedarray.py: note: In member "test_real_and_imag" of class "TestNamedArray":
+ xarray/tests/test_namedarray.py:274: error: Incompatible types in assignment (expression has type "ndarray[tuple[int], dtype[float64]]", variable has type "ndarray[Any, dtype[complex128]]") [assignment]
+ xarray/tests/test_namedarray.py: note: In class "TestNamedArray":
+ xarray/tests/test_variable.py: note: In member "test_copy" of class "VariableSubclassobjects":
+ xarray/tests/test_variable.py:555: error: "Never" has no attribute "astype" [attr-defined]
+ xarray/tests/test_variable.py: note: At top level:
+ xarray/tests/test_duck_array_ops.py:892: error: Need type annotation for "expected" [var-annotated]
+ xarray/tests/test_duck_array_ops.py:908: error: Need type annotation for "expected2" [var-annotated]
+ xarray/tests/test_duck_array_ops.py: note: In function "test_datetime_to_numeric_potential_overflow":
+ xarray/tests/test_duck_array_ops.py:980: error: Need type annotation for "expected" [var-annotated]
+ xarray/tests/test_dataset.py: note: In member "test_constructor" of class "TestDataset":
+ xarray/tests/test_dataset.py:475: error: Need type annotation for "x1" [var-annotated]
+ xarray/tests/test_cftimeindex.py: note: In function "test_asi8":
+ xarray/tests/test_cftimeindex.py:1380: error: Need type annotation for "expected" [var-annotated]
+ xarray/tests/test_backends.py: note: In function "test_use_cftime_standard_calendar_default_in_range":
+ xarray/tests/test_backends.py:6639: error: Need type annotation for "decoded_x" [var-annotated]
+ xarray/tests/test_backends.py:6640: error: Need type annotation for "decoded_time" [var-annotated]
+ xarray/tests/test_backends.py: note: In function "test_use_cftime_false_standard_calendar_in_range":
+ xarray/tests/test_backends.py:6732: error: Need type annotation for "decoded_x" [var-annotated]
+ xarray/tests/test_backends.py:6733: error: Need type annotation for "decoded_time" [var-annotated]
arviz (https://github.com/arviz-devs/arviz)
+ arviz/stats/ecdf_utils.py:69: error: "Never" has no attribute "astype" [attr-defined]
+ examples/matplotlib/mpl_plot_hdi.py:15: error: Need type annotation for "y_data" [var-annotated]
+ examples/bokeh/bokeh_plot_hdi.py:12: error: Need type annotation for "y_data" [var-annotated]
optuna (https://github.com/optuna/optuna)
+ optuna/samplers/_tpe/_truncnorm.py:75: error: Incompatible return value type (got "float", expected "ndarray[tuple[Any, ...], dtype[Any]]") [return-value]
+ tests/hypervolume_tests/test_wfg.py:22: error: Need type annotation for "r" [var-annotated]
+ tests/hypervolume_tests/test_wfg.py:35: error: Need type annotation for "r" [var-annotated]
+ tests/hypervolume_tests/test_wfg.py:53: error: Need type annotation for "r" [var-annotated]
+ tests/hypervolume_tests/test_wfg.py:63: error: Need type annotation for "r" [var-annotated]
+ tests/hypervolume_tests/test_wfg.py:73: error: Need type annotation for "r" [var-annotated]
+ tests/hypervolume_tests/test_wfg.py:81: error: Need type annotation for "r" [var-annotated]
+ tests/hypervolume_tests/test_wfg.py:91: error: Need type annotation for "r" [var-annotated]
+ tests/hypervolume_tests/test_wfg.py:101: error: Need type annotation for "s" [var-annotated]
+ optuna/importance/_ped_anova/evaluator.py:43: error: Need type annotation for "loss_values" [var-annotated]
scipy (https://github.com/scipy/scipy)
+ scipy/io/matlab/tests/test_mio.py:46: error: Need type annotation for "theta" [var-annotated]
+ scipy/interpolate/_cubic.py:562: error: Unsupported target for indexed assignment ("Never") [index]
+ scipy/_lib/cobyqa/examples/powell2015.py:49: error: Need type annotation for "bub" [var-annotated]
pandas-stubs (https://github.com/pandas-dev/pandas-stubs)
+ tests/test_plotting.py:201: error: Unsupported left operand type for + ("Never") [operator]
+ tests/series/arithmetic/bool/test_sub.py:87: error: Expression is of type "Any", not "Never" [assert-type]
hydpy (https://github.com/hydpy-dev/hydpy)
+ hydpy/auxs/statstools.py:1053: error: Incompatible return value type (got "float", expected "ndarray[tuple[Any, ...], dtype[float64]]") [return-value]
pandas (https://github.com/pandas-dev/pandas)
+ pandas/core/dtypes/missing.py:590: error: Need type annotation for "taker" [var-annotated]
+ pandas/io/formats/format.py:1603: error: "Never" has no attribute "round" [attr-defined]
+ pandas/io/formats/format.py:1611: error: Need type annotation for "unique_pcts" [var-annotated]
+ pandas/io/formats/format.py:1614: error: Value of type "Never" is not indexable [index]
+ pandas/io/formats/format.py:1616: error: Value of type "Never" is not indexable [index]
+ pandas/core/indexing.py:1889: error: Need type annotation for "taker" [var-annotated]
+ pandas/core/internals/managers.py:2484: error: Need type annotation for "empty_arr" [var-annotated]
+ pandas/core/indexes/base.py:6171: error: Need type annotation for "no_matches" [var-annotated]
+ pandas/core/groupby/generic.py:1033: error: Need type annotation for "idchanges" [var-annotated]
+ pandas/core/reshape/reshape.py:996: error: Need type annotation for "idxs" [var-annotated]
colour (https://github.com/colour-science/colour)
+ colour/algebra/interpolation.py:308: error: Unsupported left operand type for + ("Never") [operator]
+ colour/algebra/interpolation.py:309: error: Unsupported left operand type for + ("Never") [operator]
+ colour/temperature/kang2002.py:177: error: Unsupported left operand type for - ("Never") [operator]
+ colour/temperature/kang2002.py:178: error: Unsupported left operand type for - ("Never") [operator]
+ colour/temperature/kang2002.py:179: error: Unsupported left operand type for - ("Never") [operator]
+ colour/models/yrg.py:119: error: Unsupported left operand type for - ("Never") [operator]
+ colour/models/yrg.py:120: error: Unsupported left operand type for + ("Never") [operator]
+ colour/models/rgb/transfer_functions/log.py:321: error: Redundant cast to "float" [redundant-cast]
+ colour/models/rgb/transfer_functions/dcdm.py:174: error: Need type annotation for "XYZ" [var-annotated]
+ colour/difference/delta_e.py:404: error: Need type annotation for "delta_theta" [var-annotated]
+ colour/contrast/barten1999.py:246: error: Need type annotation for "E" [var-annotated]
+ colour/colorimetry/yellowness.py:280: error: Need type annotation for "WI" [var-annotated]
+ colour/appearance/llab.py:555: error: Need type annotation for "z" [var-annotated]
+ colour/appearance/llab.py:559: error: Need type annotation for "a" [var-annotated]
+ colour/appearance/llab.py:560: error: Need type annotation for "b" [var-annotated]
+ colour/appearance/hunt.py:572: error: Incompatible return value type (got "floating[_16Bit] | floating[_32Bit] | float64", expected "ndarray[tuple[Any, ...], dtype[floating[_16Bit] | floating[_32Bit] | float64]]") [return-value]
+ colour/appearance/hunt.py:603: error: Incompatible return value type (got "floating[_16Bit] | floating[_32Bit] | float64", expected "ndarray[tuple[Any, ...], dtype[floating[_16Bit] | floating[_32Bit] | float64]]") [return-value]
+ colour/appearance/hunt.py:766: error: Incompatible return value type (got "float", expected "ndarray[tuple[Any, ...], dtype[floating[_16Bit] | floating[_32Bit] | float64]]") [return-value]
+ colour/appearance/hunt.py:1187: error: Incompatible types in assignment (expression has type "floating[_16Bit] | float64 | floating[_32Bit]", variable has type "float") [assignment]
+ colour/appearance/hunt.py:1245: error: Incompatible return value type (got "floating[_16Bit] | float64 | floating[_32Bit]", expected "ndarray[tuple[Any, ...], dtype[floating[_16Bit] | floating[_32Bit] | float64]]") [return-value]
+ colour/appearance/hke.py:226: error: Need type annotation for "theta_2" [var-annotated]
+ colour/appearance/hke.py:226: error: Need type annotation for "theta_3" [var-annotated]
+ colour/appearance/hke.py:226: error: Need type annotation for "theta_4" [var-annotated]
+ colour/models/rgb/transfer_functions/sony.py:330: error: Incompatible return value type (got "float", expected "ndarray[tuple[Any, ...], dtype[floating[_16Bit] | floating[_32Bit] | float64]]") [return-value]
+ colour/appearance/rlab.py:280: error: Need type annotation for "LR" [var-annotated]
+ colour/colorimetry/luminance.py:206: error: Unsupported left operand type for - ("Never") [operator]
+ colour/models/hdr_cie_lab.py:228: error: Need type annotation for "a_hdr" [var-annotated]
+ colour/models/hdr_cie_lab.py:229: error: Need type annotation for "b_hdr" [var-annotated]
+ colour/models/cie_luv.py:139: error: Unsupported left operand type for - ("Never") [operator]
+ colour/models/cie_luv.py:140: error: Unsupported left operand type for - ("Never") [operator]
+ colour/models/cie_luv.py:216: error: Need type annotation for "b" [var-annotated]
+ colour/models/cie_lab.py:116: error: Need type annotation for "a" [var-annotated]
+ colour/models/cie_lab.py:117: error: Need type annotation for "b" [var-annotated]
+ colour/colorimetry/illuminants.py:360: error: Unsupported left operand type for + ("Never") [operator]
+ colour/temperature/mccamy1992.py:81: error: Unsupported left operand type for + ("Never") [operator]
+ colour/phenomena/rayleigh.py:222: error: Incompatible return value type (got "floating[_16Bit] | floating[_32Bit] | float64", expected "ndarray[tuple[Any, ...], dtype[floating[_16Bit] | floating[_32Bit] | float64]]") [return-value]
+ colour/phenomena/rayleigh.py:368: error: Unsupported left operand type for + ("Never") [operator]
+ colour/phenomena/rayleigh.py:406: error: Unsupported left operand type for + ("Never") [operator]
+ colour/phenomena/rayleigh.py:480: error: Incompatible return value type (got "float", expected "ndarray[tuple[Any, ...], dtype[floating[_16Bit] | floating[_32Bit] | float64]]") [return-value]
+ colour/appearance/scam.py:270: error: Need type annotation for "z" [var-annotated]
+ colour/appearance/scam.py:426: error: Need type annotation for "z" [var-annotated]
+ colour/appearance/scam.py:440: error: Need type annotation for "I" [var-annotated]
+ colour/appearance/ciecam02.py:625: error: Need type annotation for "N_bb" [var-annotated]
+ colour/appearance/ciecam02.py:625: error: Need type annotation for "N_cb" [var-annotated]
- colour/notation/munsell.py:1955: error: Incompatible types in assignment (expression has type "int", variable has type "ndarray[tuple[Any, ...], dtype[floating[_16Bit] | floating[_32Bit] | float64]]") [assignment]
+ colour/notation/munsell.py:440: error: Need type annotation for "V" [var-annotated]
+ colour/appearance/hellwig2022.py:384: error: Need type annotation for "Q_HK" [var-annotated]
+ colour/appearance/hellwig2022.py:762: error: Need type annotation for "_2_h" [var-annotated]
+ colour/appearance/hellwig2022.py:763: error: Need type annotation for "_3_h" [var-annotated]
+ colour/appearance/hellwig2022.py:764: error: Need type annotation for "_4_h" [var-annotated]
+ colour/appearance/hellwig2022.py:961: error: Need type annotation for "P_p_1" [var-annotated]
+ colour/appearance/ciecam16.py:689: error: Unsupported left operand type for / ("Never") [operator]
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Fixes #29486
Description
This PR addresses an issue with static typing of reverse operators in NumPy's
ndarrayclass. When using objects that implement both__array__and__array_ufunc__(like pandas Series) in reverse operations with NumPy arrays, the static typing incorrectly reports the result as anndarrayinstead of the appropriate type from the left-hand side object.The solution implements a Protocol-based approach that adds overloads to the reverse operators (
__radd__,__rsub__,__rmul__, etc.) in thendarrayclass. These overloads check if the left-hand side operand implements the corresponding forward operator and, if so, delegate to that operator, returning its result type.