Dotted path relative references

[lmmx]

I have begun to develop the new simplified form of relative cross-reference and think it should default to on, because as I see it this cannot cause backward incompatibility: it was invalid to use a .-prefixed reference in a docstring before now, so surely there is no pre-existing code whose behaviour we need to preserve? Can I confirm that's the case?

• feat(relative-refs): add and document a new relative_crossrefs config setting

This PR will close #27:

• Relative cross-references

[pawamoy]

Can I confirm that's the case?

Sounds right. And in that case, we don't even need a setting, right?

Could you share how you plan to implement this? I don't want to gate-keep eternally but I might have reserves regarding an implementation directly within the Python handler and not the autorefs plugin.

[lmmx]

My thoughts exactly, adding needless configuration is just introducing complexity, I will remove it.

I was just ramping up to consider that question! Apologies as I just came to make this PR because I thought your comment on the autorefs plugin repo was telling me to do that haha, I'd be happy to contribute wherever you think it should go.

Rereading what you said I now see it was supposed to imply a PR should be made to autorefs, apologies.

I still want to implement this, particularly in a generic way, through autorefs

[lmmx]

Closing in favour of a direct modification to autorefs instead

Update - new PR is here:

• Parse .-prefixed reference identifiers (relative identifiers) autorefs#37

[pawamoy]

Sorry for the confusion! And no worries 🙂 Thanks a lot for contributing to mkdocstrings and friends 😄 I'll try to unlock some time soon to elaborate a bit more on what would be need in autorefs (and possibly handlers too) to make this work.