Setup updates 202304

[mkelley]

Drop Python 3.7.

numpy -> 1.18
astropy -> 4.3
synphot -> 1.1
astroquery -> 4.6

Update CI scripts.

[codecov]

Codecov Report

❗ No coverage uploaded for pull request base (main@997d64d). Click here to learn what that means.
The diff coverage is n/a.

@@           Coverage Diff           @@
##             main     #370   +/-   ##
=======================================
  Coverage        ?   75.65%           
=======================================
  Files           ?       78           
  Lines           ?     6795           
  Branches        ?        0           
=======================================
  Hits            ?     5141           
  Misses          ?     1654           
  Partials        ?        0           

[mkelley]

I think this is waiting for old tests to run, but I've replaced those tests. So, ready for review @jianyangli

[jianyangli]

@mkelley Can you remind me the motivation for this update? My philosophy is that we want to support the minimum versions of dependence packages that can satisfy the requirements of sbpy, although this philosophy could be too simplified. Do we have to require the upgrade of those packages you changed the version requirement here? Thanks.

[mkelley]

I should have provided the justification with my PR notes. I understand the need to support a broad set of versions for the users, but that also puts a larger burden on the developers and the testing resources. I recall that we set out to always support astropy's long-term support (LTS) and current versions, which I think are 5.0 and 5.3. However, I don't know where that is documented, and we have to revisit this anyway given that astropy no longer has LTS versions: https://github.com/astropy/astropy-APEs/blob/main/APE21.rst

• Python 3.7 is no longer supported as of this month: https://www.python.org/downloads/ . I'm a little concerned about supporting a version of python that is no longer getting security fixes.
• astropy v4.3... I thought this was the LTS version, but now I see it is 5.0. We can leave this as is.
• synphot 1.1 is needed for compatibility with astropy v4.3
• numpy 1.18... I'm not sure where I picked that up. Looking at astropy, 4.3 needs numpy >=1.17, astropy 5.0 needs 1.18 <= numpy >=1.25.

So, I have some inconsistencies here to fix. Do recommend we stick with astropy 4.3 and set numpy to 1.17, or move to astropy 5.0?

[mkelley]

Ah, numpy 1.17 does not support Python 3.8.

And looking at astroquery's release notes, perhaps we can pin that to 0.4.5?

[jianyangli]

Thanks Mike for the detailed rationales! It's great to document it here. I agree with all of them. I think we can stick to astropy 4.3 and astroquery 0.4.5. I don't see strong reasons for lifting the bar further up for these two packages for now.

[mkelley]

sbpy now requires astroquery>=0.4.5 (down from 0.4.6)

Tests that use from_jplspec now require astroquery 0.4.7, see astropy/astroquery#2717

astropy 5.3 changed how unit strings are represented, which broke some documentation tests. These test now require astropy 5.3, or else they are skipped.

[mkelley]

@jianyangli finally back to successful builds! Let me know if it is ready to go.

[jianyangli]

All looks good to me. Thanks for the updates.

[jianyangli]

Looks like the changes in this file are trivial formatting changes and single quotes to double quotes?

[mkelley]

Oh crud. This was automatically reformatted with Black. I've managed to avoid that most of the time. Anyway, the important change was in one or two doc tests where I modified 1000 to 1000.0 to get the test to pass.