Preprocessor-based folding

[llchan]

This is just an idea, but could YCM communicate preprocessor conditions to vim as folding hints? e.g. if I have a #ifndef FOO section, and it's defined in the flags in the YCM config file, could vim automatically add a collapsed fold for that? I know this isn't the primary goal of YCM, but since it already has all the heavy machinery to parse the file asynchronously with the right flags, it seems like a good place to add that functionality.

[RobertBuhren]

I'd love that! In addition to folding a specific highlight group could also be set. So the code that is not defined is marked as e.g. a comment.

[bstaletic]

This, along with other semantic highlightng, is aleady implemented in dyevim, which uses a fork of YCM. So you do get the benefit of single compile for all of the features upstream YCM has plus semantic highlighting.

[RobertBuhren]

Ah, nice. Thx for pointing that out.

[mellery451]

dyevim has fallen out of sync with ycm (not surprising). Can we accomplish something like this with clangd or another c++ lang server ?

[bstaletic]

Here's a little bit of history. What I didn't mention above, is that the creator of dyevim made a PR to upstream the changes. #1873 ycm-core/ycmd#291 Now this was long before LSP was a thing. Today, if this feature would be implemented, thinks would be much different:

• Clangd would do the heavy lifting of parsing and providing semantic tokens, which would significantly reduce the complexity on the ycmd's side.
• Ycmd would only have to implement that part of the LSP protocol.
• YCM would still need to figure out what to do with those tokens, which would also be easier than before, because Vim now has something called "text properties".

But before that, we need the protocol to support this.

• There's a feature request for it: Support semantic highlighting microsoft/language-server-protocol#18

• There's a PR for it: Proposal for (Semantic) Coloring (see #18) microsoft/language-server-protocol#124

• The above PR was closed and superseded by: Proposal of the semantic highlighting protocol extension microsoft/vscode-languageserver-node#367

• The above PR is on hold until vscode ("the de facto client" ugh) is able to implement the protocol extension: Tokenization overhaul microsoft/vscode#77140

• Once the protocol supports semantic highlighting, we would need at least one of the officially supported LSP servers to implement the protocol.

  • That is: clangd, jdt.ls, gopls and rls.

• As soon as one of them implements the feature, YCM could start working on it as it would have something to test against.

  • Assuming the protocol won't require server-specific knowledge this feature would work even with the unofficially supported servers. https://github.com/ycm-core/lsp-examples

Note that ccls is capable of semantic highlighting as a protocol extension, but I don't think we should migrate to a different server over an unofficial protocol extension.

[mellery451]

makes sense - thanks for the update/background info.

[puremourning]

Neither folding nor semantic completion are currently considered to be in scope for YCM.