Generate doc tables for transplanted methods, continued

[lilyminium]

Fixes #1845

Continues #2176
Changes made in this Pull Request:

• Writes a sorted list of methods and properties without arguments like group or self.

I haven't addressed the duplicate citation issue. Links just go to the original MDAnalysis.core.topologyattrs.Attrname.Method docstring citation.

To be honest, IMO the best solution would be to just move the methods to the core classes and fail gracefully when the required topology attribute isn't there. This approach has many downsides, including:

• I sorted the properties/methods alphabetically, but all of them nevertheless appear before the *Group's actual properties/methods
• You must manually add .. include:: XXX_methods_table.txt and .. include:: XXX_methods_docs.txt to each docstring every time you add a new transplant target
• The Universe transplanted methods are now in a docstring that's largely about its __init__ method and will show up in help(Universe)
• citations confusingly link to an identical entry on a different page
• People looking at the code now have a mismatch between code and documentation
• what happens when you have a method that requires >1 topology attribute?

PR Checklist

• Tests?
• Docs?
• CHANGELOG updated?
• Issue raised/referenced?

[lilyminium]

Looks like the screenshot below or have a look at these built docs. I felt a little blurb was needed to explain the sudden table.

[codecov]

Codecov Report

Merging #2418 into develop will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff            @@
##           develop    #2418   +/-   ##
========================================
  Coverage    90.12%   90.12%           
========================================
  Files          177      177           
  Lines        22511    22511           
  Branches      2913     2913           
========================================
  Hits         20288    20288           
  Misses        1620     1620           
  Partials       603      603

Impacted Files	Coverage Δ
package/MDAnalysis/core/universe.py	95.23% <ø> (ø)	⬆️
package/MDAnalysis/core/groups.py	98.22% <ø> (ø)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 4adbd6c...6ccd4a5. Read the comment docs.

[richardjgowers]

I’m ok with moving methods into the base class. I think the whole transplanted methods thing went a little too far.

[jbarnoud]

To add to @lilyminium argument, having the transplanted method also means that the methods are less discoverable, that they are clearly not IDE friendly, and that error messages are unhelpful when switching topology format.

[jbarnoud]

I’m ok with moving methods into the base class. I think the whole transplanted methods thing went a little too far.

I gave a go at the move in #2425. I am not done with it, but it feels already like a massive regression. The reasons @richardjgowers gave to justify the transplanted methods stand and become more obvious when undoing the change.

From #601:

Reasons for this approach

• Prevents the base AtomGroup class from becoming bloated with every imaginable method
• All code related to a given Attribute lives in the same module
• Can include extra Attributes & methods without installing MDA from source

[orbeckst]

This PR has been superseded by PR #2427 and #3007 so I am closing it.