Semantic Token - Highlighting problem on multilines

[theogiraudet]

Hello,
I don't really know if it's a bug or not, so I prefer post here first.
In my model, I have a semantic element named "Documentation" that can be on multiple lines (as many languages do).
I have defined a SemanticTokenProvider in wich I do this:

else if(isDocumentation(node)) {
    acceptor({ node, feature: "doc", type: SemanticTokenTypes.comment, modifier: SemanticTokenModifiers.documentation });
}

But when I test the result:

The highlighting doesn't work when the documentation in on multiple lines.

Is there a bug or I doing something wrong?

Here is the code:
https://github.com/theogiraudet/Minimal-Langium-Semantic-Token-Issue

Thanks!

[msujew]

Hi @theogiraudet,

this isn't necessarily a bug nor intended behavior, but a vscode limitation. Language clients provide a boolean multilineTokenSupport for the semanticTokens request. If that's false, the client will discard any response data which tries to do semantic highlighting for multiple lines. Unfortunately, that's the case for vscode:

Generally, I would recommend you to use the textMate grammar support in that case. Alternatively, you can manually identify the cst node that belongs to the doc assignment (using the findNodeForFeature method) and generate a semantic token for each line that belongs to the comment node.

[theogiraudet]

Thanks for the solution! In my case, the targeted front isn't VSCode and doesn't natively support TextMate so I prefer the solution with findNodeForFeature to stay in the LSP protocole.

[msujew]

@theogiraudet By the way, with #553 the default semantic token provider will automatically split tokens into multiple lines (if the multilineTokenSupport is false). So for 0.5, your workaround won't be necessary anymore 👍

[theogiraudet]

Oh really nice! Thanks for the information ^^