VS Code editor sends incorrect range params to LSP when special characters are in the line

[petetronic]

If a line has a character code point > U+FFFF, e.g. an emoji, the range params sent to the LSP for various operations, like hover, are invalid and beyond the range of the line length (i.e. the column number is greater than the logical line length).

[kevinushey]

You might be able to configure the LSP to use UTF-8 position encodings (https://microsoft.github.io/language-server-protocol/specifications/lsp/3.17/specification/#serverCapabilities). This is the approach I took solving a similar bug in ark; see 9486c98.

[petetronic]

Interesting lead, though when I change the server capabilities for position encoding to utf-8, the vscode language server client throws an error "Unsupported position encoding (utf-8) received from server ..."
https://github.com/microsoft/vscode-languageserver-node/blob/main/client/src/common/client.ts#L1241

[petetronic]

Also see recent fixes in ARK posit-dev/ark#209

[DavisVaughan]

Yea if you have questions about this feel free to ask me. The ranges likely aren't invalid, they are just in UTF-16 and this is likely unexpected by your LSP implementation