deprecate copy_arrays with warning

[zacharyburnett]

resolves #1710 #1711

Checklist:

• pre-commit checks ran successfully
• tests ran successfully
• for a public change, a changelog entry was added
• for a public change, documentation was updated
• for any new features, unit tests were added

[braingram]

I left a few suggestions. One major one being the warning message. Also, if you can think of other ways we can inform users of the upcoming memmap default change that would be great! It's a rather major change and this switch from "copy_arrays" to "memmap" was somewhat intended to provide a warning to the users.

[braingram]

I also added the Downstream CI label. Some of those packages don't turn warnings into errors (like jwst). Would you check which downstream packages are going to fail (or produce new warnings) with these changes?

[zacharyburnett]

I also added the Downstream CI label. Some of those packages don't turn warnings into errors (like jwst). Would you check which downstream packages are going to fail (or produce new warnings) with these changes?

Sure!

[zacharyburnett]

Also, if you can think of other ways we can inform users of the upcoming memmap default change that would be great! It's a rather major change and this switch from "copy_arrays" to "memmap" was somewhat intended to provide a warning to the users.

I tried adding this warning box but GitHub doesn't render the RST correctly:

[braingram]

Code changes look good to me!

I added a few more suggestions to docs.

The remaining issue is the downstream usage. Of the failed packages most we can update (and perhaps bump asdf to at least 3.1.0 (to use memmap). dkist follows SPEC-0000 and considers asdf a "core" package so we can open a PR which accommodates the older versions (only providing memmap for 3.1.0 and after).

Are there any warnings about this change in the non-failing jobs?

[zacharyburnett]

a few warnings in asdf-astropy and sunpy; I can work on PRs for stdatamodels and roman_datamodels

[braingram]

Thanks!

asdf-astropy is maintained by "us" so that could also use a PR (and gwcs). I also try to make PRs for our other downstream users for stuff like this and am happy to help with PRs. Would you like to divide these up? I think so far we have:

• asdf-zarr: https://github.com/search?q=repo%3Aasdf-format%2Fasdf-zarr%20copy_arrays&type=code
• stdatamodels
• roman_datamodels
• gwcs https://github.com/search?q=repo%3Aspacetelescope%2Fgwcs%20copy_arrays&type=code
• asdf-astropy
• one use in jwst: https://github.com/search?q=repo%3Aspacetelescope%2Fjwst%20copy_arrays&type=code
• a few in romancal: https://github.com/search?q=repo%3Aspacetelescope%2Fromancal+copy_arrays&type=code
• weldx
• sunpy
• dkist

What a kettle of fish! :)

[braingram]

Also one of my suggestions is causing a pre-commit failure.