Support `diff` highlighting in Markdown code blocks

[Daeraxa]

Have you checked for existing feature requests?

• Completed

Summary

Pulsar should be able to handle simple diff syntax within Markdown code blocks. e.g.

- line 1
- line 2
+ new line 1 
+ new line 2

Currently it highlights them just as a default gfm text scope:

For reference this is how it looks in Helix:

This is at least supported in markdown-preview (at least with the default Linguist highlighter):

I appreciate that (I think?) we don't support diff highlighting out of the box? (The language-git package suggests installing https://web.pulsar-edit.dev/packages/language-diff) but installing that doesn't seem inject it into the GFM grammar either.

Kind of related but maybe we should integrate diff support directly into language-git by default?

What benefits does this feature provide?

Diff highlighting in markdown code blocks

Any alternatives?

Not that I could find

Other examples:

Helix. Very likely in VSCode too.

[savetheclocktower]

Yeah, this would require a package that's like language-diff but is a Tree-sitter grammar.

The Tree-sitter grammar for Markdown is (in my subjective estimation) about 1,000% better than the TextMate Markdown grammar. (Markdown is complex and needs a thorough grammar.) But the downside is that syntax highlighting can only be injected into code blocks if the grammar doing the highlighting is also a Tree-sitter grammar.

You can replicate this with a fenced code block for, say, less or coffeescript. Both of those will produce syntax highlighting in the markdown-preview view (because we embed an editor for that code block and it runs in isolation, so it can pick either kind of grammar) but not in the buffer itself. That's because we don't have Tree-sitter grammars for Less or CoffeeScript. (One exists for Less but I haven't had the time to look into it. I'd be surprised if one existed for CoffeeScript.)

Anyway, it's a good idea to have a diff grammar, and it should arguably be part of core as well. I bet a Tree-sitter parser exists for diffs already, and even if it didn't, it would not be extremely onerous to write one.

[Daeraxa]

This is the one being used in Helix - https://github.com/tree-sitter-grammars/tree-sitter-diff