Mailman 3 August 2025 - NumPy-Discussion

next NumPy community meeting - Wednesday, April 9th, 2025 at 5 pm UTC - now meeting ONE HOUR EARLIER
by Inessa Pawson Feb. 8, 2026

Feb. 8, 2026

The next NumPy community meeting will be held this Wednesday, April 9th at 17:00 (5 pm) UTC. Please note the TIME CHANGE! Join us via Zoom: https://numfocus-org.zoom.us/j/83278611437?pwd=ekhoLzlHRjdWc0NOY2FQM0NPemdk… Everyone is welcome and encouraged to attend. To add to the meeting agenda the topics you’d like to discuss, follow the link: https://hackmd.io/76o-IxCjQX2mOXO_wwkcpg?both For the notes from the previous meetings, visit: https://github.com/numpy/archive/tree/main/community_meetings -- Cheers, Inessa Inessa Pawson GitHub: inessapawson

1 22

next NumPy Optimization Team meeting - Monday, January 20th, 2025 at 5 pm UTC
by Inessa Pawson Jan. 31, 2026

Jan. 31, 2026

Hi, all! The next NumPy Optimization Team meeting will be held on Monday, January 20th at 17:00 UTC. Join us via Zoom: https://numfocus-org.zoom.us/j/81261288210?pwd=iwV99tGSjR61RTGEERKM4QKxe46g… Everyone is welcome and encouraged to attend. To add to the meeting agenda the topics you’d like to discuss, follow the link: https://hackmd.io/dVdSlQ0TThWkOk0OkmGsmw?both For the notes from the previous meetings, visit: https://github.com/numpy/archive/tree/main/optim_team_meetings -- Cheers, Inessa Inessa Pawson GitHub: inessapawson

1 17

next NumPy triage meeting - Wednesday, May 28th, 2025 at 5 pm UTC
by Inessa Pawson Jan. 4, 2026

Jan. 4, 2026

The next NumPy triage meeting will be held this Wednesday, May 28th at 17:00 (5 pm) UTC. This is a meeting where we synchronously triage prioritized PRs and issues. Everyone is welcome to attend and contribute to a conversation. Join us via Zoom: https://numfocus-org.zoom.us/j/82096749952?pwd=MW9oUmtKQ1c3a2gydGk1RTdYUUVX… Please notify us of issues or PRs that you’d like to have reviewed by adding a GitHub link to them in the meeting agenda: https://hackmd.io/68i_JvOYQfy9ERiHgXMPvg -- Cheers, Inessa Inessa Pawson GitHub: inessapawson

1 14

ENH: Add a non-circular shift function
by Carlos Martin Nov. 23, 2025

Nov. 23, 2025

Feature request: Add a non-circular shift function that is similar to [`numpy.roll`](https://numpy.org/doc/stable/reference/generated/numpy.roll.html) but, instead of wrapping elements around, replaces unoccupied entries with a specified fill value. Examples: ``` >>> import numpy as np >>> a = np.arange(1, 6) >>> a array([1, 2, 3, 4, 5]) >>> np.shifted(a, 2) array([0, 0, 1, 2, 3]) >>> np.shifted(a, -2) array([3, 4, 5, 0, 0]) >>> a = np.arange(1, 10).reshape((3, 3)) >>> a array([[1, 2, 3], [4, 5, 6], [7, 8, 9]]) >>> np.shifted(a, (1, -2), (0, 1)) array([[0, 0, 0], [3, 0, 0], [6, 0, 0]]) ``` Prior discussion: https://github.com/numpy/numpy/issues/26220 Pull request: https://github.com/numpy/numpy/pull/29389

1 1

next NumPy Newcomers' Hour - June 27th, 2024 at 10 pm UTC
by Inessa Pawson Sept. 12, 2025

Sept. 12, 2025

Our next Newcomers' Hour will be held TOMORROW, June 27th at 10 pm UTC. Stop by to ask questions, share your progress, celebrate success, or just to say hi. To add to the meeting agenda the topics you’d like to discuss, follow the link: https://hackmd.io/3f3otyyuTte3FU9y3QzsLg?both. Join the meeting via Zoom: https://us06web.zoom.us/j/82563808729?pwd=ZFU3Z2dMcXBGb05YemRsaGE1OW5nQT09. -- Cheers, Inessa Inessa Pawson GitHub: inessapawson

1 11

Feature proposal: weighted gram matrix / sandwich product
by Christian Lorentzen Sept. 2, 2025

Sept. 2, 2025

Dear numpy community I would like to propose a mew feature in np.linalg to compute weighted gram matrix (sandwich product), see https://github.com/numpy/numpy/issues/29559. Proposed new feature or change: The solvers for many important statistical models like Generalized Linear Models often need to compute |X.T @ W @ X| where |X| is a 2-dimensional array of features (features in columns, observations in rows) and |W| is very often a diagonal array of (current) weights. This computation is usually the main computational bottleneck. It would be great if numpy could provide an efficient implementation of it, e.g. |np.linalg.sandwicht_product(X, weight=w)|. Why numpy? It seems even more unrealistic to me, to get it into BLAS implementations. Numpy has support for SIMD via Highways. Computational alternatives Drawbacks of |(X.T * diag(W)) @ X|: * Additional memory allocation to compute |X.T @ diag(W)| , same size as (usually large) |X|. * The result is symmetric, but this fact is not used. So at least a factor of 2 is possible. Note that without weights, |X.T @ X| uses BLAS syrk. But the weights are crucial. Drawback of |Z = np.sqrt(diag(W))[:, None] * X| and then |Z @ Z|: * Additional memory allocation for |Z|, same size as (usually large) |X|. * Taking square roots for possibly large |diag(W)| Additional information https://github.com/Quantco/tabmat has an implementation of it with XSIMD. Best Christian Lorentzen

3 4

same_value casting in ndarray.astype
by matti picus Aug. 27, 2025

Aug. 27, 2025

In PR 29129 <https://github.com/numpy/numpy/pull/29129> I propose to add a new casting option to ndarray.astype(dtype, casting='XXX') called same_value. It changes the inner casting loop to check that the value round-trips, and raises a value error on overflow or float-to-int truncation. While the PR adds infrastructure for further same_value casting in user-defined dtypes and elsewhere in NumPy, the current work is only exposed in ndarray.astype (and not in np.astype, since there is no casting option in that function). This was motivated by user desire to have some kind of value-based casting. See also the discussion of saturation-based ufuncs <https://mail.python.org/archives/list/numpy-discussion@python.org/thread/B6…> What do you think? Matti

2 1

new time slot for community calls
by Inessa Pawson Aug. 18, 2025

Aug. 18, 2025

Hi, all! Some of our active community members are unable to attend Wednesday meetings due to scheduling conflicts. To support broader participation, I’ve set up a poll for us to find a new time slot: https://crab.fit/numpy-community-calls-582036 Please indicate your availability by Sunday, August 18th EOD. -- Cheers, Inessa Inessa Pawson GitHub: inessapawson

1 1

Feature proposal
by Андрей Рыбаков Aug. 7, 2025

Aug. 7, 2025

Dear all, I would like to discuss a «cosmetic» feature for numpy. I find it useful in my day-to-day work to inspect the values of 1D and 2D numpy arrays. Typically I was saving the data in the .txt format and then used the next best tool to analyse the numbers. However, I would prefer to be able to just print an array directly in the interactive console (python console for instance) and have the printed array formatted in such a way that it can be easily read by human eyes. Now, the type of formatting is subjective. Nevertheless, I would like to propose the one that I have converged to over the years. The proposed implementation is a standalone function, written in pure python (as this «visual» representation will be suitable only for the reasonable-sized array, then there is no need to worry about the performance), for example, def decorate_array( object: array_like, fmt: str, header: list, footer: list, column_header: list, footer column: list, …, ) If the numpy community agrees that this function fits in the package, then I would like to adapt the implementation that I am using now. I would open a PR, but I saw that it is recommended to ask for an opinion on this mailing list first. Below are the examples of how this function would behave (again, the style is subjective): 1. Plain real-valued array decorate_array( object = np.random.random((10,3)), fmt = ">.4f", ) ┌────────┬────────┬────────┐ │ 0.0410 │ 0.2844 │ 0.4251 │ ├────────┼────────┼────────┤ │ 0.9270 │ 0.7509 │ 0.9892 │ ├────────┼────────┼────────┤ │ 0.7380 │ 0.1245 │ 0.1855 │ ├────────┼────────┼────────┤ │ 0.4738 │ 0.6671 │ 0.8455 │ ├────────┼────────┼────────┤ │ 0.4442 │ 0.5946 │ 0.5874 │ ├────────┼────────┼────────┤ │ 0.0789 │ 0.5374 │ 0.6505 │ ├────────┼────────┼────────┤ │ 0.8006 │ 0.0970 │ 0.4456 │ ├────────┼────────┼────────┤ │ 0.4884 │ 0.4162 │ 0.8283 │ ├────────┼────────┼────────┤ │ 0.5082 │ 0.5625 │ 0.8812 │ ├────────┼────────┼────────┤ │ 0.2013 │ 0.9937 │ 0.3731 │ └────────┴────────┴────────┘ 2. Headers and footers decorate_array( object = np.random.random((2,3)), fmt = ">.5f", header = ["l-h-corner", "h1", "h2", "h3"], footer = ["l-f-corner", "f1", "f2", "f3"], column_header = ["ch1", "ch2", "ch3"], footer column = ["cf1", "cf2", "cf3"], ) ┌────────────┬─────────┬─────────┬─────────┬─────┐ │ l-h-corner │ h1 │ h2 │ h3 │ │ ├────────────┼─────────┼─────────┼─────────┼─────┤ │ ch1 │ 0.84675 │ 0.80002 │ 0.79360 │ cf1 │ ├────────────┼─────────┼─────────┼─────────┼─────┤ │ ch2 │ 0.93482 │ 0.28542 │ 0.28057 │ cf2 │ ├────────────┼─────────┼─────────┼─────────┼─────┤ │ l-f-corner │ f1 │ f2 │ f3 │ │ └────────────┴─────────┴─────────┴─────────┴─────┘ 3. Complex numbers ┌──────────────┬──────────────┬──────────────┐ │ 0.60 - i0.12 │ 0.66 + i0.18 │ 0.31 - i0.22 │ ├──────────────┼──────────────┼──────────────┤ │ 0.52 - i0.33 │ 0.05 - i0.03 │ 0.23 + i0.37 │ └──────────────┴──────────────┴──────────────┘ There are more verbose examples written in the docs page of my sandbox python package, but I am not sure if posting the link to it here would be appropriate. The three examples above should be enough to grasp the idea. Best, Andrey

3 3

Re: Feature proposal
by Андрей Рыбаков Aug. 7, 2025

Aug. 7, 2025

I see that the examples were misformatted, here how they were intended to look like. Best, Andrey

2 1