✨ Write needextend rst from need-id-refs markers

[juiwenchen]

• Added new module needextend_write, which consume the output of analyse and generate the corresponding rst with needextend directive
• Added write rst CLI which use needextend_write
• Added a Sphinx demo project for the whole workflow. Simply run rye run demo:clean to have documentation ready
• added title option to inject the specified section in the generated rst

[ubmarco]

The invocation works! Some points:

Can you change the CLI interface to codelinks write rst --type need-id-refs.
Then it is expandable in future. Think about writing the source code to a json as discussed in the morning, it could be codelinks write sources.
Also with above proposal we could do codelinks write rst --type need-id-refs --type need so we can write multiple things to a single RST.

This looks like a bug to me. The input

leads to

Last one wins or it's an exception in SN, haven't tested.
I think the remote url should be a list in case one need is reference by 2 locations.

Generally I think the output will be pretty ugly because the URLs are so long. Maybe https://sphinx-needs.readthedocs.io/en/latest/configuration.html#needs-string-links would help shorten what is displayed while still being able to have the full link.

This leads to an exception:

 rye run codelinks bridge -r foo output/marked_content.json

KeyError: 'foo'

Is there a use case for rye run codelinks bridge --outdir <somedir>? There is just a single file generated with a pre-given name.
Why not give it a single file as in rye run codelinks bridge --outpath <file> that accepts a file path?

[ubmarco]

outdated

[ubmarco]

same here

[ubmarco]

have you head of https://docs.astral.sh/ruff/rules/blind-except/#blind-except-ble001 ?

[ubmarco]

btw, this can also fail easily - permission issues and stuff