Skip to content

ENH: add timedelta64 support boolean division and floor - #32555

Merged
ngoldbaum merged 11 commits into
numpy:mainfrom
mhafizazh:timedelta64-div-boolean
Sep 9, 2026
Merged

ENH: add timedelta64 support boolean division and floor#32555
ngoldbaum merged 11 commits into
numpy:mainfrom
mhafizazh:timedelta64-div-boolean

Conversation

@mhafizazh

@mhafizazh mhafizazh commented Sep 9, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

PR summary

This PR based on issue #18445. Initially timedelta64 only accept int##, if timedelta64 is divided by boolean value such as

m = np.array([1], dtype="timedelta64[s]")
b = np.array([True])

output:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
numpy._core._exceptions._UFuncBinaryResolutionError: ufunc 'divide' cannot use operands with types dtype('<m8[s]') and dtype('bool')

I added condition in ufunct_type_resolution.c so PyUFunc_DivisionTypeResolver() bool i s upcast to int64 (true=1, False=0).

new output:

>>> import numpy as np
>>> m = np.array([1], dtype="timedelta64[s]")
>>> b = np.array([True])
>>> m/b
array([1], dtype='timedelta64[s]')

First time committer introduction

I'm new contributors and I joined previous community meeting maybe 1-2 month ago, and suggested to take a look at the old issues and see if the bug still exist. I want to contribute to NumPy to challenge myself to contribute large repository and I use NumPy almost everyday for my internship too.

AI Disclosure

I used opencode to help draft both the code fix, and the test. I reviewed, wrote, and ran the test with myself.

@ngoldbaum

ngoldbaum commented Sep 9, 2026 •
edited
Loading

Copy link
Copy Markdown
Member

Nice minimal fix. The test failures in the CI run are unrelated breakage on canonical's infrastructure.

This needs a release note. I also think this impacts the type stubs (cc @jorenham). It's probably also worth adding test cases for numpy booleans (numpy.True_ is not the same thing as True).

@jorenham

jorenham commented Sep 9, 2026

Copy link
Copy Markdown
Member

This needs a release note. I also think this impacts the type stubs (cc @jorenham). It's probably also worth adding test cases for numpy booleans (numpy.True_ is not the same thing as True).

Right now the stubs already accept builtins.bool as a direct consequence of bool being a subtype of int. If numpy.bool_ is now also accepted, then the corresponding overloads should indeed be updated to also accept bool_:

numpy/numpy/__init__.pyi

Lines 12881 to 12916 in 88ddb98

@overload
def __truediv__(self, b: timedelta64, /) -> float64: ...
@overload
def __truediv__(self: timedelta64[dt.timedelta], b: dt.timedelta, /) -> float: ...
@overload
def __truediv__(self: timedelta64[Never], b: float | floating | integer, /) -> timedelta64: ...
@overload
def __truediv__[AnyItemT: (dt.timedelta, int, None)](
self: timedelta64[AnyItemT], b: int | integer, /
) -> timedelta64[AnyItemT]: ...
@overload
def __truediv__[AnyItemT: (dt.timedelta, int, None)](
self: timedelta64[AnyItemT], b: float | floating, /
) -> timedelta64[AnyItemT | None]: ...
@overload
def __truediv__(self, b: float | floating | integer, /) -> timedelta64: ...
@overload
def __rtruediv__(self, a: timedelta64, /) -> float64: ...
@overload
def __rtruediv__(self: timedelta64[dt.timedelta], a: dt.timedelta, /) -> float: ...
@overload
def __floordiv__(self, b: timedelta64, /) -> int64: ...
@overload
def __floordiv__(self: timedelta64[dt.timedelta], b: dt.timedelta, /) -> int: ...
@overload
def __floordiv__(self: timedelta64[Never], b: float | floating | integer, /) -> timedelta64: ...
@overload
def __floordiv__[AnyItemT: (dt.timedelta, int, None)](
self: timedelta64[AnyItemT], b: int | integer, /
) -> timedelta64[AnyItemT]: ...
@overload
def __floordiv__[AnyItemT: (dt.timedelta, int, None)](
self: timedelta64[AnyItemT], b: float | floating, /
) -> timedelta64[AnyItemT | None]: ...

@jorenham jorenham added the 56 - Needs Release Note. Needs an entry in doc/release/upcoming_changes label Sep 9, 2026
@mhafizazh mhafizazh changed the title fixing so timedelta64 support boolean division and floor BUG: fixing so timedelta64 support boolean division and floor Sep 9, 2026
@mhafizazh

mhafizazh commented Sep 9, 2026

Copy link
Copy Markdown
Contributor Author

@ngoldbaum thanks for your feedback. I just added release note.

ngoldbaum
ngoldbaum reviewed Sep 9, 2026
View reviewed changes
Comment thread numpy/_core/tests/test_datetime.py Resolved
@ngoldbaum ngoldbaum added 01 - Enhancement and removed 56 - Needs Release Note. Needs an entry in doc/release/upcoming_changes labels Sep 9, 2026
@ngoldbaum ngoldbaum changed the title BUG: fixing so timedelta64 support boolean division and floor ENH: add timedelta64 support boolean division and floor Sep 9, 2026
@ngoldbaum ngoldbaum added this to the 2.6.0 Release milestone Sep 9, 2026
ngoldbaum
ngoldbaum approved these changes Sep 9, 2026
View reviewed changes

@ngoldbaum ngoldbaum left a comment

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks! Someone else will need to give this a look and merge it but this looks good to me.

ngoldbaum
ngoldbaum requested changes Sep 9, 2026
View reviewed changes

@ngoldbaum ngoldbaum left a comment

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Oh sorry, I missed that you still need to update the type stubs. Joren can probably help if you get stuck there.

@mhafizazh
mhafizazh requested a review from ngoldbaum September 9, 2026 21:02
@jorenham
jorenham self-requested a review September 9, 2026 21:03
jorenham
jorenham reviewed Sep 9, 2026
View reviewed changes

@jorenham jorenham left a comment

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Stub changes look good.

I have some minor suggestions to simplify things a bit, because these unions now coincide with the existing _FloatLike_co and _IntLike_co type aliases:

numpy/numpy/_typing/_scalars.py

Lines 13 to 14 in ca30acf

type _IntLike_co = int | np.integer | np.bool
type _FloatLike_co = float | np.floating | np.integer | np.bool

Comment thread numpy/__init__.pyi Resolved
Comment thread numpy/__init__.pyi Resolved
Comment thread numpy/__init__.pyi Resolved
Comment thread numpy/__init__.pyi Resolved
Comment thread numpy/__init__.pyi Resolved
mhafizazh and others added 5 commits September 9, 2026 15:25
@mhafizazh @jorenham
Update numpy/__init__.pyi
580f767 
Co-authored-by: Joren Hammudoglu <[email protected]>
@mhafizazh @jorenham
Update numpy/__init__.pyi
be2b5b7 
Co-authored-by: Joren Hammudoglu <[email protected]>
@mhafizazh @jorenham
Update numpy/__init__.pyi
9df2af0 
Co-authored-by: Joren Hammudoglu <[email protected]>
@mhafizazh @jorenham
Update numpy/__init__.pyi
6829e94 
Co-authored-by: Joren Hammudoglu <[email protected]>
@mhafizazh @jorenham
Update numpy/__init__.pyi
0aa3e6a 
Co-authored-by: Joren Hammudoglu <[email protected]>
@mhafizazh

mhafizazh commented Sep 9, 2026

Copy link
Copy Markdown
Contributor Author

Stub changes look good.

I have some minor suggestions to simplify things a bit, because these unions now coincide with the existing _FloatLike_co and _IntLike_co type aliases:

numpy/numpy/_typing/_scalars.py

Lines 13 to 14 in ca30acf

type _IntLike_co = int | np.integer | np.bool
type _FloatLike_co = float | np.floating | np.integer | np.bool

I see. It looks much cleaner now. Thank you!

jorenham
jorenham reviewed Sep 9, 2026
View reviewed changes
Comment thread numpy/__init__.pyi Resolved
@mhafizazh @jorenham
Update numpy/__init__.pyi
eaf55e9 
Co-authored-by: Joren Hammudoglu <[email protected]>
@ngoldbaum
ngoldbaum merged commit ee2a62a into numpy:main Sep 9, 2026
93 checks passed
@ngoldbaum

ngoldbaum commented Sep 9, 2026

Copy link
Copy Markdown
Member

Thank you!

@mhafizazh mhafizazh mentioned this pull request Oct 28, 2026
Open
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@ngoldbaum ngoldbaum ngoldbaum approved these changes
@jorenham jorenham jorenham approved these changes

Assignees

No one assigned

Labels

01 - Enhancement

Projects

None yet

Milestone

2.6.0 Release

Development

Successfully merging this pull request may close these issues.

3 participants

@mhafizazh @ngoldbaum @jorenham