Sebastian Berg

Results 60 issues of Sebastian Berg

This is a squash of multiple changes (since it was unwieldy anyway) and it is *not* a polished solution, rather *basis for discussion*. To outline the design idea a little...

cat:feature
prio:medium

This PR does three somewhat related things: 1. I fixed the `npy_2_compat.h` so now it should work fine when backported. 2. Add a short segment that some functionality moved (I...

30 - API

Right now, this is a heads-up a start to: * Introduce `_PyArray_LegacyDescr` for the full struct * Introduce some additional access macros for the fields that may not exist now...

03 - Maintenance

The following changes are in progress: * [ ] An overview of these changes and how downstream code should be updated for the transition guide. * [ ] The flags...

component: numpy._core
63 - C API

**Please don't hesitate to edit this issue** The board has some other issues as well, but I would like to list (and add) an issue to track the main C-API...

63 - C API

This adds a dedicated page about promotion rules. --- Please don't hesitate to rip it apart, I haven't tried to really iterate yet, but I think it would be useful...

04 - Documentation

This is a low priority issue, but would be nice to fix: ``` **663466** ********************************************************************** **663466** build/testenv/lib/python3.9/site-packages/numpy/core/tests/test_nditer.py::test_iter_copy_casts[O-e] **663466** ********************************************************************** ==663466== 24,000 (+24,000) bytes in 1,000 (+1,000) blocks are definitely lost...

gh-23275 added broadcasting before the promotion happened, which is incorrect for 0-D objects as they may have special promotion rules (value-based or Python int/float/complex being special). First reported at https://github.com/cupy/cupy/issues/7946...

00 - Bug

**Please don't hesitate to edit this issue to expand/clarify/add/remove items** Adopting NEP 50 is the most important part of a NumPy 2.0 release opinion because it consists of larger changes...

62 - Python API

Quantile/percentile insert the dimensions of `q` at the beginning of the result. Nanquantile seems not to be written to correctly support multi-dimensional `q` and ends up splitting it up: ```python...

00 - Bug
component: numpy.lib