Mailman 3 March 2026 - NumPy-Discussion

NumPy 2.4.3 Release
by Charles R Harris March 9, 2026

March 9, 2026

Hi All, On behalf of the NumPy team, I'm pleased to announce the release of NumPy 2.4.3. The NumPy 2.4.3 is a patch release that fixes bugs discovered after the 2.4.2 release. The most user visible fix may be a threading fix for OpenBLAS on ARM, closing issue #30816. This release supports Python versions 3.11-3.14. Wheels can be downloaded from PyPI <https://pypi.org/project/numpy/2.4.3>; release notes and the changelog are available on Github <https://github.com/numpy/numpy/releases/tag/v2.4.3>. *Contributors* A total of 11 people contributed to this release. People with a "+" by their names contributed a patch for the first time. - Antareep Sarkar + - Charles Harris - Joren Hammudoglu - Matthieu Darbois - Matti Picus - Nathan Goldbaum - Peter Hawkins - Pieter Eendebak - Sebastian Berg - Warren Weckesser - stratakis + *Pull requests merged* A total of 14 pull requests were merged for this release. - #30759: MAINT: Prepare 2.4.x for further development - #30827: BUG: Fix some leaks found via LeakSanitizer (#30756) - #30841: MAINT: Synchronize 2.4.x submodules with main - #30849: TYP: matlib: missing extended precision imports - #30850: BUG: Fix weak hash function in np.isin(). (#30840) - #30921: BUG: fix infinite recursion in np.ma.flatten_structured_array... - #30922: BUG: Fix buffer overrun in CPU baseline validation (#30877) - #30923: BUG: Fix busdaycalendar's handling of a bool array weekmask... - #30924: BUG: Fix reference leaks and NULL pointer dereferences (#30908) - #30925: MAINT: fix two minor issues noticed when touching the C API setup - #30955: ENH: Test .kind not .char in np.testing.assert_equal (#30879) - #30957: BUG: fix type issues in uses if PyDataType macros - #30958: MAINT: Don't use vulture 2.15, it has false positives - #30973: MAINT: update openblas (#30961) Cheers, Charles Harris

1 0

Deprecation of `generic` unit in `timedelta64` dtype
by riku-sakamoto March 7, 2026

March 7, 2026

Dear NumPy developers, Hi all, I am Riku. I currently work on the PR #29619 ( https://github.com/numpy/numpy/pull/29619 ), which deprecates `generic` unit in `timedelta64` dtype. This change is expected to have a large impact on the downstream libraries, so I sent this email to notify it. Let me briefly summarize the motivation and changed behavior. Motivation: `generic` unit in `timedelta64` dtype is allowed to be operated with any other `timedelta64` dtype. So, comparison between several `timedelta64` dtypes including `generic` unit becomes not transitive. It is reported in the issue #28287 ( https://github.com/numpy/numpy/issues/28287 ) ```python x, y, z = np.timedelta64(1, "ms"), np.timedelta64(2), np.timedelta64(5, "ns") print(x < y < z) # True, but x < z is False ``` Changed behavior: After the deprecation, when using `generic` unit in `timedelta64` dtype, `DeprecationWarning` will be raised. Following operations will raise the warning: ```python np.timedelta64(1) np.array([1, 2], dtype="m8") # `+ 1` is treated as adding 1 with generic timedelta np.array([1, 2], dtype="m8[s]") + 1 ``` I want to note that `NaT` is also required to have specific unit. So, the following code will also raise the warning: ```python np.array(["NaT"], dtype="m8") ``` I'd appreciate it if you let me know when you have any questions or concerns about this change. Any feedback is welcome. Best Regards, Riku

3 2

Current policy on AI-generated code in NumPy
by David Cournapeau March 4, 2026

March 4, 2026

Hi, I know there has been discussions in the past on AI-generated contributions. Is there a current policy for NumPy ? E.g. do we request that contributors are the "sole contributor" to the written code, or do we allow code written by AI as long as it follows the usual quality requirements ? Context of my question: ~18 months ago I started in some spare time writing an ARPACK-replacement in numpy/scipy during scipy sprint. At that time, I used ChatGPT for the "research part" only: literature review, explain to me some existing BSD implementation in Julia for points I could not understand. I implemented the python code myself. There is still quite a bit of work needed to be a viable replacement for ARPACK. Seeing the progress of the AI tooling in my team at work, and how I myself use those tools for other hobby projects, I believe I could finish that replacement very quickly with those tools today. But I don't want to "taint" the work if this would risk the chances of integration into scipy proper. Thanks, David

24 120