[Python] Updates for NumPy 2.0 support in the sdist builder#1713
[Python] Updates for NumPy 2.0 support in the sdist builder#1713
Conversation
0b97cde to
5f293f9
Compare
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #1713 +/- ##
=======================================
Coverage 72.84% 72.84%
=======================================
Files 379 379
Lines 53824 53824
Branches 9182 9182
=======================================
Hits 39208 39208
Misses 11642 11642
Partials 2974 2974 ☔ View full report in Codecov by Sentry. |
|
Related to #1675 |
faf2540 to
12b2529
Compare
|
I saw #1675 was closed 🎉 @bryanwweber Is the plan to do some sort of 3.0.0post1 once this work is complete so that a fresh |
I'll probably re-open #1675 until the Conda packages are fixed as well. Then I'll see about backporting 😄 I would like to backport, and see if we can preemptively support Python 3.13 as well |
019e0d4 to
36c7ff3
Compare
speth
left a comment
There was a problem hiding this comment.
Thanks, @bryanwweber, all these changes look good to me. Just one request regarding the documentation.
| - name: Trigger Conda builds | ||
| run: > | ||
| gh workflow run -R cantera/conda-recipes | ||
| main.yml |
There was a problem hiding this comment.
This is the change dropping the builds for the cantera conda channel, correct? I think this would be the right time to update the installation instructions (doc/sphinx/install/conda.md) to reflect conda-forge being the sole source of conda packages.
There was a problem hiding this comment.
They aren't uploaded from this job, but yes this is the one that builds when pushes happen on main. I'll update the docs! 🎉
b52902d to
8488b30
Compare
This can be configured by the top-level example_data option
When building the sdist via scons, the cython step was running twice because of the conditional. In addition, we were getting a warning about building the extension from multiple source files. These changes resolve both problems by cythonizing once if the files aren't present and only including the C++ sources in the extension.
NumPy 2.0 does not support Python 3.8, so we need to special case that version.
…th Pint Update dependencies for Cython/Sdist packages
We're deprecating these builds for the 3.1 release because we don't have a Matlab package in this release.
This will work for Anaconda or miniforge.
8488b30 to
8908f4e
Compare
Co-authored-by: Ray Speth <[email protected]>
This can be configured by the top-level example_data option
When building the sdist via scons, the cython step was running twice because of the conditional. In addition, we were getting a warning about building the extension from multiple source files. These changes resolve both problems by cythonizing once if the files aren't present and only including the C++ sources in the extension.
NumPy 2.0 does not support Python 3.8, so we need to special case that version. Also special case for Pint, which doesn't support NumPy 2.0 on Python 3.9.