Update syntax: alternative block comment and nvim-lsp based highlights for bold and italic

[yangwenbo99]

Hi,

I just have some time to fix two problems I have encountered in recent months.

The first one is inserting a "/*" in a large document stresses my CPU and destroys the whole file's syntax. I use Vim's C syntax highlight file's strategy to replace the block comment.

The second issue is about italic and bold fonts. I added an option "typst_syntax_use_lsp" to use LSP for highlighting. See the attachment for the effects.

Since there are so many marginal cases for Typst syntax, I am thinking that syntax highlighting should probably be done using LSP in the future. However, the current LSP is rather limited in highlighting, as it does not provide any tree-based highlighting.

Before:

After:

[kaarmu]

Thanks for the PR!

I've seperated the comment fix to branch fix-106-comment. Can you just double-check that I didn't miss anything/that it fixes your issue with performance.

Regarding bold/italics, I really appreciate relying on LSP. Though, I want to hold on to that a little bit. Moreover, even though LSPs can perhaps better/more easily capture the right highlight group, I am also aiming to have standalone support as much as possible.

[yangwenbo99]

I think fix-106-comment is good.

For bold/italics, on the one hand, I do think it is very challenging to do it using vim's syntax file. On the other hand, there are a lot of restrictions for LSP-based highlighting, even though it is faster and more robust:

• the built-in LSP only exists in Neovim. I am not even sure whether other LSP clients for vim provide the same interface
• current Typst LSP server only provides token-based highlighting. It does not give support tree-based information. This means, we cannot use it to support collapse.

In my proposed (temporary) modification, I added an option for users to choose whether to use LSP or not.

In the long run, I think it would be better to offer two options, (a) pure vim syntax file highlighting, with limited capability and (b) (mostly) LSP/treesitter-based highlighting.

There is already a treesitter implementation for typst, and I am looking at it.