-
Notifications
You must be signed in to change notification settings - Fork 155
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Use oldest-supported-numpy #208
Comments
Hi @henryiii , I do think that's a cool package, and we may use it in the future. However, we do currently test our supported NumPy range to ensure that the build version is compatible with all of our supported versions (not every NumPy minor version increment introduces a binary incompatibility). Have you run into any bugs or specific issues? Otherwise I will mark this as a feature request rather than a bug. Thanks! |
I think you mean patch, not minor. The NumPy team test What I'd recommend, if you use any newer header features that need a recent NumPy, is setting the minimum version manually if you want for the cases in https://github.com/scipy/oldest-supported-numpy/blob/66fc65cfa7bfbd621f0b9d4758ab800afa54c617/setup.cfg#L21-L67, then allow all newer Pythons and rarer architectures use
Nope, just noticed this checking unrelated breakages in conda-forge (related to pip 22.1.0). |
Also, on the I'll go put my nose back in my own business now! Cheers. :) |
I did mean minor. Though I agree/know that in general one needs to treat minor changes as breaking, my point was that we specifically tested the 1.20-1.21 increment to ensure that it was not incompatible (or at least not for the parts we were using). Though I am very aware that it's possible we missed something.
Totally agree. For the majority of our packages it would be good to just abstract the requirement away to the NumPy maintainers. We'll likely start migrating to this approach in our next version update cycle.
Also agree, we're not very consistent with this accross Ocean (e.g. https://github.com/dwavesystems/dimod/blob/7826a91192d547b9618da28a48d9fd067a3c0085/setup.py#L82 does include a cap, but we don't here)
We appreciate folks reaching out! Truly! This is the best part of open source. And I now know of a cool utility package that I didn't before. |
As the person, that (probably) initiated this discussion on conda-forge :) Would be awesome to have a wider range of NumPy available as I currently need to do some manual adjusting to get every package of the SDK to be compatible with each other. (This applies to the whole of the ocean SDK though, not only here). |
I am going to transfer this issue over to our sdk, so that we can track changes across all of Ocean |
Although we need some newer features of NumPy at runtime, we can safely build against the older headers. I was not able to detect any performance regressions from doing so either. See dwavesystems/dwave-ocean-sdk#208
Although we need some newer features of NumPy at runtime, we can safely build against the older headers. I was not able to detect any performance regressions from doing so either. See dwavesystems/dwave-ocean-sdk#208
I just noticed
https://github.com/dwavesystems/dwave-preprocessing/blob/be12b72b1e9c8ef450f83e93b844801df641e7bd/pyproject.toml#L6
This should almost always be oldest-supported-numpy if it's in a pyproject.toml file. https://github.com/scipy/oldest-supported-numpy - see https://scikit-hep.org/developer/packaging#special-additions-numpy or the note at https://cibuildwheel.readthedocs.io/en/stable/options/#before-build for example.
The text was updated successfully, but these errors were encountered: