feat: Add feature to distinguish attribute directives & modifiers

[AlbertMarashi]

I've added support for enabling the independent styling of directives & modifiers in order to migrate Zed's svelte support to this tree-sitter-svelte instead of https://github.com/Himujjal/tree-sitter-svelte

This will allow us to close some long-standing svelte issues within zed-industries/zed#17529

Supports parsing:

• class:foo={...}
  • (attribute_directive) (attribute_identifier)
• on:click|stopPropagation|preventDefault={...}
  • (attribute_directive) (attribute_identifier) (modifier) (modifier)
• style:--bar={...}
  • (attribute_directive) (attribute_identifier)

All are still wrapped within (attribute_name) node. I'm only familiar with Zed's syntax highlighting conventions, so I left highlights.scm and etc as they currently are

[AlbertMarashi]

Also, I was very confused about the addition of all the new files when I merged from main, and how the package.json no longer contains all of the tree-sitter commands?

Was this supposed to be committed?

[clason]

tree-sitter generate --no-bindings if you only wish to update the parser without touching the bindings (which indeed got more numerous with tree-sitter 0.23.0).

[AlbertMarashi]

I did format all the tests with https://github.com/Sarrus1/tree-sitter-tests-formatter cli

Hopefully it isn't a problem 🙏🏼

edit: I added some contribution docs to the README.MD so others can know how to run the tests and regenerate the parser

[AlbertMarashi]

@amaanq @savetheclocktower I'd love to get this merged so we can get it integrated into Zed's downstream extension!

[savetheclocktower]

I don't have the commit bit on this repo, but @amaanq has enough responsibilities already, so I'll try to reduce the friction on this one. :-)

This is a soft rejection that would change to an approval pretty easily once the feedback is addressed.

@amaanq, there are new nodes introduced — attribute_identifier, attribute_directive, and attribute_modifier (if my feedback is agreeable). For most ordinary attributes, the only effect will be to make attribute_name have a single attribute_identifier child. That doesn't bother me, but please weigh in if you have other suggestions for how this could be handled.

[clason]

The main question is whether the grammar change breaks existing queries. (The repo contains them for a reason.) Adding new nodes is fine, but moving nodes usually isn't. ("grammar semver")

[AlbertMarashi]

The main question is whether the grammar change breaks existing queries. (The repo contains them for a reason.) Adding new nodes is fine, but moving nodes usually isn't. ("grammar semver")

No, it shouldn't.

Previous queries targeting

(attribute_name) @attribute

For example should continue working

Also alongside the more specific queries, eg:

(attribute_name
    (attribute_identifier) @attribute
    (attribute_modifier) @function
)

[AlbertMarashi]

@amaanq could we get a look at this? We're currently using this commit reference upstream in zed, so would be good to get it merged!