Problems related to LSP/Monaco integration

[MarkZ3]

I have looked a bit at the LSP/Monaco integration. Here are my observations so far. I tested with the Java and C/C++ servers.

• Semantic Syntax Coloring: This was mentioned before but AFAIK this is not in the protocol. Is this really meant to be added at this moment? Perhaps via a protocol extension connecting to a different server? If so, how can I test?
• Code Completion: I see duplicated entries for Java. I think this could be a server problem? Seems to be fine for C/C++ (Clangd).
• Find References: Looks good!
• Go To Declaration: Doesn't seem to select the full range correctly, the cursor is placed at the beginning of the range. Is this a Monaco bug/by-design-behavior?
• Formatting: Looks good!
• Mark Occurrences: Looks good!
• Hover Information: Looks good!
• Quick Fixes: Does not seem to work? For Java and C/C++ (Clangd) at least. I tried to do "Quick Fix" from the command palette when my cursor at the location where there should be a code action. I also tried the "CMD+." shortcut displayed in the command palette.
• Signature Information: I'm not sure how to trigger it. Is there a way right now?
• Code lens: Seems to display "references" twice for Java?

[svenefftinge]

re: semantic coloring There is no server nor client speaking such an extension, so there is no point in implementing it atm. But Anton put everything in place in order to enable arbitrary extensions.

re: signature information It usually pops up when you type the opening bracket in a method call.

[akosyakov]

https://code.visualstudio.com/docs/extensionAPI/language-support#_help-with-function-and-method-signatures - Method Signatures in VS code

[MarkZ3]

https://code.visualstudio.com/docs/extensionAPI/language-support#_help-with-function-and-method-signatures - Method Signatures in VS code

Nice! I didn't know about that page, it's quite useful! Thanks!

[MarkZ3]

It usually pops up when you type the opening bracket in a method call.

It doesn't seem to work with the Java server, I don't see the request from the client either (perhaps the server doesn't have the capability??). Does it work for you? For example:

    public static void main(String[] args) {
        main   // <--- type the opening parenthesis here
    }

[akosyakov]

It is not implemented: https://github.com/eclipse/eclipse.jdt.ls/blob/f58ad76f7f258926a3579dd6c6ea4eb008d44d88/org.eclipse.jdt.ls.core/src/org/eclipse/jdt/ls/core/internal/handlers/JDTLanguageServer.java#L220

The list of features supported by jdt.ls: https://github.com/eclipse/eclipse.jdt.ls#features

[svenefftinge]

But the python LS has it and it works in Theia.

[MarkZ3]

But the python LS has it and it works in Theia.

Perfect, I'll try that!

[MarkZ3]

Yup, it's working with "pyls", thanks!

[akosyakov]

• [x] duplicate providers: code lens, content assist, because java language is registered twice
• [x] a selection is not revealed or revealed bogusly for go to declaration/reference
• [x] quick fixes does not work

[akosyakov]

Go To Declaration: Doesn't seem to select the full range correctly, the cursor is placed at the beginning of the range. Is this a Monaco bug/by-design-behavior?

It's by design behavior.

[akosyakov]

I've filed an issue for bogus hover ranges: https://github.com/Microsoft/vscode/issues/26747

[svenefftinge]

The outstanding bug has been fixed by the vs code team.

[MarkZ3]

@svenefftinge Do you know if the "quick fixes" work now? I haven't had the chance to test them.

[svenefftinge]

Yes, I added that while working on yangster. https://github.com/theia-ide/theia/pull/327

[MarkZ3]

Thanks!