Mainline pygls in vscode.dev!

[alcarney]

Screenshot shows running our example semantic_tokens.py server running in the browser

With the recent changes to the pygls main branch (i.e. the upcoming v2 release) I'm pleased to report that it is now possible to run your language server fully in the browser in https://vscode.dev! 😄

Rather than using Pyodide as the WASM runtime, this is using VSCode's WASI host, as such your server will have to align with the current limitations of the WASI runtime

• No async-await support
• No threads

See this repository for details on how to make it work.

[tombh]

Wow, this is amazing.

That link you have in the screenshot is this right? https://vscode.dev/github/openlawlibrary/pygls I assume it's expected that merely going to that URL isn't enough to set everything up? I get this error when visiting it:

So are you planning to write up some steps at https://github.com/alcarney/vscode-wasi-pygls?

[alcarney]

I get this error when visiting it:

Are you using firefox? I saw the same, had to switch to chromium to get VSCode to load properly :/

So are you planning to write up some steps at https://github.com/alcarney/vscode-wasi-pygls?

Yes, I suppose that would help! 😅

The short version is though

• Open https://vscode.dev/github/openlawlibrary/pygls, if vscode.dev is working properly, it should ask you to sign into GitHub before loading the repo.
• Install the swyddfa.pygls-wasi-playground extension
• Tweak .vscode/settings.json to match what you see in the screenshot.
• You'll likely have encountered an error similar to the one you see here Better error handling alcarney/vscode-wasi-pygls#5 you'll have to run the Developer: Reload Window command to work around it for now
• Wait for the server to boot up (will take 2-3min), it might be worth having the network tab in the devtools open to get an idea for how progress is going

[tombh]

Nice, I made some progress, but seems to get stuck on 404s for .pyc files?

[alcarney]

I saw those also, but they didn't seem to cause an issue for me - I assume that's just Python checking for pre-compiled *.pyc files as it would normally.

Since the code in your screenshot is not highlighted, I assume the server didn't start? Is there any output in the pygls-client or pygls-server output channels?

[alcarney]

Just published a new version of the extension with some improvements, it should now be possible to use the settings.json file (almost) as-is. The only tweak that is still needed is to remove "scheme": "file" from the pygls.client.documentSelector option

 {
     "[plaintext]": {
         // Uncomment to enable `textDocument/onTypeFormatting` requests
         // "editor.formatOnType": true
     },
     // Uncomment to override Python interpreter used.
     // "pygls.server.pythonPath": "/path/to/python",
     "pygls.server.debug": false,
     // "pygls.server.debugHost": "localhost",
     // "pygls.server.debugPort": 5678,
     "pygls.server.launchScript": "code_actions.py", // This is relative to `pygls.server.cwd`
     "pygls.server.cwd": "${workspaceFolder}/examples/servers",
     "pygls.trace.server": "off",
     "pygls.client.documentSelector": [
         {
-            "scheme": "file",
             "language": "plaintext"
         }
     ],
     "python.testing.pytestArgs": [
         "."
     ],
     "python.testing.unittestEnabled": false,
     "python.testing.pytestEnabled": true,
     // "pygls.jsonServer.exampleConfiguration": "some value here",
 }

[tombh]

Nice! I got it to work this time! Evaluated 1 + 1 = .... 2 😏

All in the browser, wow.