Investigate moving cpp grammar to https://github.com/jeff-hykin/cpp-textmate-grammar

[aeschli]

https://github.com/jeff-hykin/cpp-textmate-grammar is a c++ grammar derived from https://github.com/atom/language-c fixing several of the issues..

Given that https://github.com/atom/language-c is no longer actively maintained (atom switched to tree-sitter), we could consider to use https://github.com/jeff-hykin/cpp-textmate-grammar as the built-in text mate grammar.
@jeff-hykin Would that be ok for you?

[Astrantia]

I have another suggestion. We could consider moving to tree sitter too @aeschli

[sean-mcmanus]

Yeah, the C/C++ extension team would like VS Code to switch to that better TextMate grammar, i.e. so that all the C/C++ users can benefit "out-of-the-box" without needing to install another extension.

[jeff-hykin]

I'd be fine with this! I'm a university student so my availability may fluctuate with exams, but I'd be happy adding collaborators and I certainly plan to continue rapidly adding improvements until the code is easily maintainable.

There's a lot of behind the scenes work that needs to be done to get it to to that point, but the code should always be a strict improvement from what it is now.

[jeff-hykin]

@Astrantia Tree sitters are really cool, I've been looking into them. It'd be great to move towards that or something even better. It's going to take some time though, and some of the issues in the TextMate grammar were solved by literally just backspacing non-existent keywords; thats how easy is to improve the existing code.

Textmate grammars have fundamental problems, we need to switch at some point if we want things like custom class tagging and other advanced highlighting. But there's incredible room for improvement within Textmate and lots of the improvements are easy to implement.

The json/cson file itself is unmaintainable, but we can use an actual language like Ruby to generate the TextMate grammar. It'll make it maintainable and we can get 50% of the tree sitter benefits right now, without changing VS Code at all.

[alexr00]

I've moved this to March so that we can have it for longer in insiders before it releases to stable. That means I'll try pulling in the new grammar around March 4th. Thanks @jeff-hykin!

[jeff-hykin]

That sounds good, glad I could help! This weekend I'll do an update to make the repo more collaborator friendly and update the project board with a roadmap.

[sean-mcmanus]

@jeff-hykin FYI, the bug atom/language-c#127 got a couple more upvotes (#67161) if you have time to fix that.

[jeff-hykin]

Alright I went ahead and created an issue for it there. jeff-hykin/better-cpp-syntax#4

I made some progress on it and pushed an update with the improvements; there's only one edge case its failing on now. The last edge case is going to be a little harder to fix because its conflicting with the single-character (ex: '1') syntax.

I went ahead an finished adding support, all of them should work now.

The C99 hexadecimal floating constant syntax (ex: 0x0.5p10) was also missing so I added that as well. There were a lot of other changes in the v1.4 today too. Next week I'll try to copy over all the still-unfixed atom issues into https://github.com/jeff-hykin/cpp-textmate-grammar.

[sean-mcmanus]

@jeff-hykin Cool, that is really awesome. Thanks.

[alexr00]

With a1481e3 I've made the switch. Thanks @jeff-hykin!

[jeff-hykin]

All relevant atom/language-c issues have been moved to
https://github.com/jeff-hykin/cpp-textmate-grammar/issues

All the following from atom/language-c have been fixed:
atom/language-c#320
atom/language-c#318
atom/language-c#312
atom/language-c#309
atom/language-c#289
atom/language-c#270
atom/language-c#260
atom/language-c#264
atom/language-c#246
atom/language-c#240
atom/language-c#233
atom/language-c#215
atom/language-c#177
atom/language-c#153
atom/language-c#121
atom/language-c#101

A framework for unit testing patterns has been implemented.

And lambda support was added today 🎉

[ a, b, c ] (int thing1) mutable -> Ret { }
[ a, b, c ] (Args... args) { }
[=]() mutable { return; }
[=] -> int { }
[ a, b, c ] { }

[rbx]

How can I revert to the old grammar in 1.33 ?!

[jeff-hykin]

@rbx I can probably make an extension this weekend that would revert it. Is there anything in particular that you do not like though?

The syntax should only be adding information, all color differences can be changed without changing the syntax (but not vice-versa).

[rbx]

@jeff-hykin in particular I encountered several inconsistent highlighting issues which were mostly already reported (I also opened an issue today here: jeff-hykin/better-cpp-syntax#87).

Since I didn't have any problems with the old grammar, I would like to be able to revert until new one is fixed.

Anyway, thanks for your effort, I hope this will end up being better. I like the improved namespace highlighting.

[alexr00]

The grammar will be reverted in 1.33.1 since that is the lowest risk fix for C++ syntax highlighting. In Insiders 1.34, the new grammar is still in place. I've pulled in @jeff-hykin's recent changes, and I've started updating the default themes to better handle some of the changes in the new grammar. Things that should be fixed in the next Insiders: using, new, delete, typedef.

No one is maintaining the old grammar, so once we've stablaized the grammar + theme combos it should be a significant improvement!