Re: Tree-sitter integration on feature/tree-sitter (severe performance issues together with linum-mode)

[Stefan Monnier]

> Yeah! I can do that. Do you have some ideas on the syntax you would use for
> font-lock?

Not really, which is why it hasn't happened yet I guess :-)

`syntax-propertize-rules` could be an inspiration (tho, compared to
font-lock, this one had the added constraint that it had to be applied
in a single scan of an N-element regexp whereas font-lock can use
N scans of simpler regexps).

But maybe it's better if you try to come up with your own plan without
suffering from my biases.  Then I can look at it and complain :-)
The result will probably be better.

> If possible I’d like treesit stuff to be similar to font-lock.

Since font-lock's hasn't happened yet, this should presumably happen by
font-lock mimicking treesit rather than the reverse.

> For example I made treesit to use font-lock’s decoration level system.

That is the kind of bias that would have better been avoided (I think
a better approach is to let the major mode specify features that can be
included/excluded: practice seems to have shown that a "level" is too
meaningless and coarse).


        Stefan