bug#61072: How to change the length of the separation lines in eldoc, used by eglot?

[Felician Nemeth]

João Távora <joaotavora@gmail.com> writes:

> See docstring of eldoc-documentation-functions.

The docstring is quite easy to follow.  Nice work.

> I stopped short of adding :content-type there.  There's not much point
> in moving rendering from Eglot to ElDoc if the rendering is going to
> suck just as bad.  So I think this has to be analyzed carefully.  Here
> are some ideas:
>
> 1. Use eww to render HTML, as you suggest.  But how to get that HTML?
>    Call an external process?

My idea was to leave the door open for a potential ElDoc backend that
produces raw HTML documentation.  But it is a half-backed idea.  I don't
know how images should be specified.  Just as an IMG tag: <img
src="example.jpg">.  Or inline like in an email.

If I understood correctly, one of your original ideas was to let ElDoc
backends (Eglot) to specify a flag whether the documentation was in
markdown format.  Using terminology of the
eldoc-documentation-functions, I simply suggested a :content-type key to
generalize the markdown flag hoping Emacs had helper functions to help
handling a content-type specification.

Unfortunately, I now see a small problem with the :content-type
approach.  How should Eglot know whether ElDoc can render a markdown
formatted documentation?  If Eglot can request the documentation in
multiple formats, how it should know which one to choose.

> 2. You seem to suggest that eww can render markdown directly.  Sure?

This is a misunderstanding.  I didn't intend to suggest that.

> 3. Lobby for markdown.el to become more render-friendly, (remove hard
>    newlines from paragraphs, remove invisible text, etc.)

Sure, but can a better markdown.el solve the original issue?  Is there a
way to render a separation line independently of the current
window-width?

> 4. Make a new Markdown mode based on a tree-sitter grammar

I think a simple mode just to view a markdown document would be useful.
However, tree-sitter only helps parse a document.  Turning an abstract
syntax tree into a viewable Emacs buffer is still a substantial work, I
suppose.

Felicián