Tree-sitter rolling fixes, 1.117 edition #974
Conversation
|
Is it worth reaching out to the "one of the main Tree-sitter contributors" and be so bold as to ask if they'd like to adopt your parser, or combine efforts until there's one parser that meets both of your respective sets of expectations? (I dunno if this suggestion is a little "out there", but it came across my mind.) Anyway, good to see language support expanding for modern tree-sitter. |
No, it's a fine suggestion. He knows about my parser, but he didn't see it until after he'd made his own. I told him about my test corpus and I expect him to adopt it at some point, but I also trust his solutions to some problems more than I trust my own (which are sometimes hacky) and his parser “extends” the Tree-sitter CSS parser more formally than mine does, so I think his is a better foundation to start from. For now, though, it's not ready. |
…which, surprising as it sounds, has a _huge_ performance benefit.
savetheclocktower
force-pushed
the
tree-sitter-may
branch
from
55da975 to
8e9e05d
Compare
|
Taking this out of draft so we have time to get it reviewed before 1.117. |
There was a problem hiding this comment.
Not a huge one this time, especially if looking at it commit-by-commit.
I am largely going by faith, since I don't know a ton of the details of the scopes/grammars myself. But this looks okay! FWIW that I say that!
Thanks as always for maintaining the grammars!
[language-gfm] Make each block-level HTML tag its own injection…
…which, surprising as it sounds, has a huge performance benefit.
Surprising indeed!
Sounds good to me!
Also sounds good!
| @@ -902,6 +903,20 @@ | |||
| "." @keyword.operator.accessor._LANG_ | |||
| "?." @keyword.operator.accessor.optional-chaining._LANG_ | |||
|
|
|||
| ; Optional chaining is illegal… | |||
There was a problem hiding this comment.
Unsure of the factuality of this comment, perhaps
| ; Optional chaining is illegal… | |
| ; Optional chaining should be illegal… |
/s (as in, just kidding!)
Ah, I understand now, from reading the corresponding comment fragments below! So, there are specific contexts in which optional chaining is illegal. Got it. Definitely never mind this, then...
|
|
||
| [ | ||
| "&&=" |
There was a problem hiding this comment.
| "+=" | ||
| "-=" | ||
| "*=" | ||
| "**=" |
There was a problem hiding this comment.
Getting a head start on this one.
This PR starts by adding a modern Tree-sitter grammar for SCSS. For ages, there wasn't a grammar out there that had good enough support for SCSS's features, but I forked this one a while back and very gradually fixed the things that needed to be fixed. I think it's good enough now to bundle and ship and see if there are any edge cases that need to be addressed.
A few days ago, one of the main Tree-sitter contributors made this parser, and I could see us migrating to it in the future (I don't love being a parser maintainer myself)… but it's not there yet. If it ever passes the tests that exist in my fork, then it'd be worth switching to.