LSC does not encode/decode characters which are not allowed in File URIs

[neowit]

Hello @natebosch ,

Thank you for your plugin!

It would appear that current version of LSC does not encode characters which are not allowed in File URIs.

According to FILE URI requirements:

Characters which are not allowed in URIs, but which are allowed in filenames, must also be percent-encoded. For example, any of "{}`^ " and all control characters. In the example above, the space in the filename is encoded as %20.

For instance, here is a Go To Definition message from VS Code (which does appropriate encoding):

{"jsonrpc":"2.0","id":1,"method":"textDocument/definition","params":{"textDocument":{"uri":"file:///Users/tester/projects/Sforce%20-%20SFDC%20Experiments/SForce%20%28lcs-test%29/src/classes/CompletionTester.cls"},"position":{"line":14,"character":10}}}

and here is the same message from vim LSC:

{"method":"textDocument/definition","jsonrpc":"2.0","id":2,"params":{"textDocument":{"uri":"file:///Users/tester/projects/Sforce - SFDC Experiments/SForce (lcs-test)/src/classes/CompletionTester.cls"},"position":{"character":10,"line":14}}}

In the opposite direction - if server does appropriate URI encoding and returns URI same way as in above VS Code example - vim LSC does not understand returned path.

e.g. example below is understood by VS Code but not understood by vim LSC

{"id":1,"result":[{"uri":"file:///Users/tester/projects/Sforce%20-%20SFDC%20Experiments/SForce%20(lsc-test)/src/classes/CompletionTester.cls","range":{"start":{"line":10,"character":8},"end":{"line":10,"character":18},"offset":{"line":0,"character":0}}}],"jsonrpc":"2.0"}

[natebosch]

Curiously in your last example the () are not encoded...

I think I can go with a relatively straightforward encode strategy since we expect the entire file path to be treated as the path component from the URI spec.

1. Allow unreserved characters
2. Allow / unescaped
3. Escape all other characters. Expect that all characters are ascii

My first approach for decoding will probably only handle %nn style encodings. Handling other cases like translating + to space or %% to % could be trickier since it's less general.

[natebosch]

Can you try updating to the latest and see if it works for you?

[neowit]

Hello @natebosch ,

Thank you for the update, looks good.
URI encoding for messages like textDocument/definition or textDocument/didOpen appears to be working, but there is one case (I found so far) where URI encoding does not get applied:

{"method":"initialize","jsonrpc":"2.0","id":1,"params":{"rootUri":"file:///Users/tester/projects/Sforce - SFDC Experiments/SForce (lsc-test)/src/classes","capabilities":{...}}

I think this line:

      \ 'rootUri': 'file://'.getcwd(),

should look something like like this:

      \ 'rootUri': lsc#uri#documentUri(getcwd()),

Curiously in your last example the () are not encoded...

that's my bad. Thank you for highlighting this.

[natebosch]

Good catch on the rootUri. Should be fixed now.

[neowit]

That's great, thank you.