Fix broken inheritance-diagram links due to the smart resolver

[ayshih]

sphinx-doc/sphinx#10614 fixes many of the broken inheritance-diagram links (#48). The remaining broken links are due to a bad interaction with sphinx-automodapi's smart resolver.

sphinx.ext.inheritance_diagram generates a list of class full names (e.g., astropy.coordinates.baseframe.BaseCoordinateFrame) and generates a mapping from class full names to documentation URLs. However, the smart resolver causes the generated mapping to be instead from class documented name (e.g., astropy.coordinates.BaseCoordinateFrame) to documentation URLs. The class documented name can be different from the class full name if the class is not documented where it is defined, but rather at some other location where it is imported. In such situations, the code will fail to find the URL that for the class.

This PR monkey-patches the method that receives the mapping and converts the keys from class documented names to class full names.

Also address #173 (but not the part about updating matrix)

[codecov]

Codecov Report

Merging #172 (508fab4) into main (b389ee2) will increase coverage by 0.04%.
The diff coverage is 100.00%.

❗ Current head 508fab4 differs from pull request most recent head 5d6ab81. Consider uploading reports for the commit 5d6ab81 to get more accurate results

@@            Coverage Diff             @@
##             main     #172      +/-   ##
==========================================
+ Coverage   89.83%   89.87%   +0.04%     
==========================================
  Files          27       27              
  Lines        1141     1146       +5     
==========================================
+ Hits         1025     1030       +5     
  Misses        116      116              

Files Changed	Coverage Δ
sphinx_automodapi/automodsumm.py	85.92% <100.00%> (+0.21%)	⬆️

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[ayshih]

This PR also fixes a bug that results from the fact that newer versions of Sphinx use pathlib for paths rather than just strings.

[pllim]

Thanks! Didn't you have a draft PR on astropy to test this? Or was that for something else? I don't see it linked here.

[ayshih]

Sure. I've repurposed astropy/astropy#15104 to test against sphinx dev (not my fork) and my sphinx-automodapi fork

[pllim]

Sigh, I guess it is unfair to you if pydata/pydata-sphinx-theme#1404 blocks this. I guess if you have another draft PR to Sunpy or something that can show the diagram is fixed, that would be also acceptable.

[ayshih]

Confirmed to work for sunpy (sunpy/sunpy#6267 (comment)), although we'd be blocked from using it immediately because our normal theme pins us to Sphinx 5.3. Without sphinx-doc/sphinx#10614, this PR has no effect.

[pllim]

Thanks for all your work and patience in fixing this long standing bug!

I guess the upstream fix will be some release after sphinx 7.1.2. We can make a release here once they release, so feel free to ping us again if we are too slow.

😸

[bsipocz]

I guess the upstream fix will be some release after sphinx 7.1.2

Would you release, or shall I?

[pllim]

@bsipocz , I need to take a lunch break now, so if you can do it faster, please go ahead!

[bsipocz]

I need to take a lunch break now, so if you can do it faster, please go ahead!

not faster, but I can do it in the evening.

[pllim]

In that case, let me just do it. I am back from break. Thanks for the ping!