Refactor the ASTBuilder to get rid of the currentAttr attribute.

[tristanlatr]

Refactor the ASTBuilder to get rid of the currentAttr attribute.

This attribute was used to attach the right docstring node to the right Attribute object. Now it uses AST node navigation (with the .parent attribute) instead for fetching the docstring node for an ast.Assign.

This change removes a attribute being mutated at different places in the code, replaces this kind of "unsafe" state tracking (meaning not with pop() and push()) by some more explicit solution.

The zopeinferface extension needed to be adjusted as well because it relied on the docstring assignment feature in an implicit way, now it's explicit what we're doing.

[codecov]

Codecov Report

Attention: Patch coverage is 87.87879% with 12 lines in your changes missing coverage. Please review.

Project coverage is 92.03%. Comparing base (0d07110) to head (52f6a15).
Report is 1 commits behind head on master.

Files with missing lines	Patch %	Lines
pydoctor/astbuilder.py	92.95%	4 Missing and 1 partial ⚠️
pydoctor/astutils.py	77.27%	3 Missing and 2 partials ⚠️
pydoctor/extensions/zopeinterface.py	66.66%	0 Missing and 2 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #585      +/-   ##
==========================================
- Coverage   92.07%   92.03%   -0.04%     
==========================================
  Files          47       47              
  Lines        8412     8465      +53     
  Branches     1856     1868      +12     
==========================================
+ Hits         7745     7791      +46     
- Misses        391      395       +4     
- Partials      276      279       +3     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[tristanlatr]

After a second though, I believe it’s worth it.

Mainly because:

• it’s now more explicit what happens in the zope interface extension.

• Narrowing down the ASTBuilder and ModuleVistor interface is a good thing since it’s used in customization mechanism.

• Also the parentage feature of the ast tree makes our builder more powerful.

[glyph]

@tristanlatr sorry to punt on this, but can you address the conflicts and then re-request from @twisted/twisted-contributors rather than me personally?

[tristanlatr]

This PR is ready to be merged. It does not worth a changelog entry since it's only refactoring. It also fixes a edge case bug that probably got unnoticed.

[github-actions]

Diff from pydoctor_primer, showing the effect of this PR on open source code:

numpy (https://github.com/numpy/numpy)
- /projects/numpy/numpy/distutils/ccompiler_opt.py:1038: Existing docstring at line 949 is overriden
- /projects/numpy/numpy/distutils/ccompiler_opt.py:1075: Cannot find link target for "bool"
+ /projects/numpy/numpy/distutils/ccompiler_opt.py:986: Cannot find link target for "bool"