Bump C3 tree-sitter parser to 0.2.3

[PgBiel]

Keeping it as draft before I run some simple usage tests.

Fixes #91 (fix yet to be confirmed).

Applied suggestions from #91 (comment), except that I added a temporary fix for the AST convert function, where we parse initializer_list directly. That makes the test pass (although isn't a general fix for all cases).

All tests are passing.

[pherrymason]

Thanks! Just let me know when you feel is ready

[PgBiel]

The fix seems to work, however, for some reason, the LSP is now consuming much more CPU on completion compared to the latest version's binary (obtained through nixpkgs). Is there a way to make an optimized build that I'm missing? (I'm using make build-linux)

[pherrymason]

Mm now that you mention, go build says CGO could be optimized (cgo is the feature of compiling c code... which refers to the embedded treesitter library )

[PgBiel]

Now that I think about it, it's possible that it's just doing more work, because now it can parse the entirety of raylib.c3l properly. Before, each raylib file would just be a single error node due to <* *>.

In comparison, the CPU usage seems to be roughly the same for a very simple project.

Perhaps you can confirm this on your own projects? (I don't have many to test on)

Still, it'd be nice to somehow solve this problem, of course.

I wonder if it's somehow related to the usage of the .c3l format specifically, or whether it's constantly rebuilding C files or something, but I'm not sure.
We could consider delaying a proper solution to this problem if we can confirm that it's not that different from before.

[PgBiel]

I've found the problem :)
Opened a separate PR to fix it: #100

[pherrymason]

Is this PR ready then?

[PgBiel]

Added unit tests and tested a bit more locally as well. Should be ready for merge 😄