Switch to lsp4j #31
Switch to lsp4j #31
Conversation
| ) | ||
| } | ||
| } | ||
|
|
||
| // Unimplemented features | ||
| override def codeAction(params: CodeActionParams) = ??? |
There was a problem hiding this comment.
Nice to get these listed out 👍
| @@ -6,19 +6,6 @@ inThisBuild( | |||
| ) | |||
| ) | |||
|
|
|||
| lazy val languageserver = project | |||
|
Does this change mean that you have no plans for building the server for Scala.js? |
|
@laughedelic There are no plans to build the server for Scala.js, even with the language-server we have in current master. We are pretty married to the JVM for the server since we invoke the compiler and classload for example custom scalafix rules. |
|
I see. Thanks for clarifying this 👍 |
| @@ -26,7 +26,11 @@ object Main extends LazyLogging { | |||
| System.setErr(err) | |||
| logger.info(s"Starting server in $workspace") | |||
| logger.info(s"Classpath: ${Properties.javaClassPath}") | |||
| server.start() | |||
| val server = new ScalametaLanguageServer(cwd, out) | |||
| val launcher = LSPLauncher.createServerLauncher(server, System.in, out) | |||
There was a problem hiding this comment.
A question not related to lsp4j itself, but more in the context of rewriting this:
Would it be hard to get rid of this mechanism of passing cwd through a var? I think that in LSP supposes that you can launch a server independently and then tell it where your project is through the initialization parameters (rootUri).
There was a problem hiding this comment.
Yes, I think that's the way to do it. I wanted to take on that, as soon as I get it up and running
This is to work around a bug with Scala 2.12 and Java runtime reflection, that affects lsp4j
|
I got the server starting with this commit here (after a looong debug session!) The presentation compiler does not start and it seems file watch events don't trigger, but it's at least a start :) |
|
I got the PC and scalafix starting by making the client lazy (it was null causing crashes). However, textDocument/onOpen and didChange don't trigger when I'm running the client. My WIP branch is here https://github.com/scalameta/language-server/compare/master...olafurpg:lsp4j-get?expand=1 it's based off the first commit in this PR before implementing the server in java. I suspect the problem may be related to trait encodings in 2.11, even if it compiles it could be that the runtime reflection is not invoking our methods. I'm honestly starting to think we might be better off to continue with dragos/language-server, I believe would have taken me less time to write out the full protocol in our current master that I've spent trying to debug this java reflection madness 😅 |
This PR replaces dragos-vscode-scala LSP bindings with lsp4j.
The reason for this change is that lsp4j is a complete implementation of the LSP, as per the specification. This saves us adding new bindings on demand.
The downside is some "javosity" creeping into
ScalametaLanguageServercode, but nothing too awful.Currently this compiles but it throws a runtime exception about some JSON-RPC error that I'm trying to debug.