Running `build.py` to render PEPs leaves residual files in a new `topic/` folder

[pradyunsg]

It seems that the auto-generation of documents that were added in #2579 also leaves the generated files on disk, and they are not git-ignored.

Details

❯ git status 
On branch main
Your branch is up to date with 'origin/main'.

nothing to commit, working tree clean

❯ git show HEAD --stat
commit 0ffc78187c3b33d3ede64156dbf9e79f823ec116 (HEAD -> main, upstream/main, upstream/HEAD, origin/main)
Author: Pradyun Gedam <pradyunsg@gmail.com>
Date: Sat Jun 18 21:53:23 2022 +0100

 PEP 632: Remove `Topic: Packaging` header (#2656)

 pep-0632.rst | 1 -
 1 file changed, 1 deletion(-)

❯ python ./build.py 
Running Sphinx v4.5.0
[snip]
❯ git status 
On branch main
Your branch is up to date with 'origin/main'.

Untracked files:
 (use "git add <file>..." to include in what will be committed)
 topic/

nothing added to commit but untracked files present (use "git add" to track)

❯ ls topic/
index.rst packaging.rst

PS: While I still think it would be nicer to not generate .rst files in-memory (which would avoid this issue + the regeneration of the files on every run), I'm not gonna advocate for it anymore. :)

[pradyunsg]

FWIW, I noticed this because this broke the workflow I had, running sphinx-autobuild -b dirhtml . build/dirhtml. That isn't the only command affected, as showing in the "details" in the issue description above.

[hugovk]

Shall we add the topic directory to .gitignore?

We should have make clean wipe it too.

[CAM-Gerlach]

Seems reasonable to me; not sure if @pradyunsg has other suggestions...

For the record, I generally think @pradyunsg 's idea to to have actual checked-in reST source files with whatever intro paragraph we want for each topic + PEP 0 itself, and have a directive to generate the index tables (parameterized by topic name) instead is a good one.

[hugovk]

Please see PR #2683 to gitignore the generated topic dir and delete it via make clean.

For the record, I generally think @pradyunsg 's idea to to have actual checked-in reST source files with whatever intro paragraph we want for each topic + PEP 0 itself, and have a directive to generate the index tables (parameterized by topic name) instead is a good one.

Yeah, this could be useful for topics needing longer or more complex intro paragraph(s), and could be linted as part of an .rst too.