Skip to content

Require numpy >=1.16#7383

Merged
jrbourbeau merged 2 commits intodask:mainfrom
crusaderky:numpy116
Mar 25, 2021
Merged

Require numpy >=1.16#7383
jrbourbeau merged 2 commits intodask:mainfrom
crusaderky:numpy116

Conversation

@crusaderky
Copy link
Collaborator

Needed to work around a mamba bug in numpy 1.15 on MacOSX (#7227)

@jsignell
Copy link
Member

This seems fine to me, especially given the conversation in #7378, but let's get a 👍 from @jrbourbeau as well.

# you modify these, make sure to change the corresponding line there.
extras_require = {
"array": ["numpy >= 1.15.1"],
"array": ["numpy >= 1.16"],
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'd prefer the mins to include the patch unless there is a reason not to.

Suggested change
"array": ["numpy >= 1.16"],
"array": ["numpy >= 1.16.0"],

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

See NEP-29: https://numpy.org/neps/nep-0029-deprecation_policy.html#detailed-description
I understand that it states (albeit in a less than perfectly clear wording) that support for patch versions would overdeliver on the contract stated by the NEP. The idea is that bumping a patch version is (typically) fairly painless.

From personal experience on xarray, pint, and my various personal projects, leaving the patch version free relieves quite a bit of pain from the maintainers while being hardly an issue for any user.

@jakirkham
Copy link
Member

Any thoughts on just going to NumPy 1.17 as a minimum ( #7378 (comment) )? That still older than 18 months and will allow us to drop a lot of compatibility code

cc @pentschev (for vis)

@crusaderky
Copy link
Collaborator Author

@jakirkham I'm strongly in favour for bumping to 1.17 directly

Copy link
Member

@jrbourbeau jrbourbeau left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks @crusaderky! Generally bumping to 1.16 seems okay to me

However I get the sense that this isn't super time sensitive (feel free to correct me if that's not the case). So given the discussion in #7378 and the appetite to bump to 1.17+, what do we think about waiting a few days to get input from the rest of the community on the broader minimum version policy and, if there's no significant pushback, we adopt the policy and update this PR to numpy >= 1.17?

Copy link
Member

@jrbourbeau jrbourbeau left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for your work here @crusaderky! I'm going to merge this now as we're still coming to consensus over in #7378

@jrbourbeau jrbourbeau merged commit 2598af2 into dask:main Mar 25, 2021
@crusaderky crusaderky deleted the numpy116 branch March 29, 2021 15:43
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants