Unable to autocomplete Unicode characters with Julia language server

[ChrHorn]

The Julia language server is able to autocomplete certain Unicode characters. These always start with a backslash, for example \alpha should complete to α.

Issue seems to be that helix gets the wrong prefix because it doesn't include the backslash (not a word).

helix/helix-term/src/commands.rs

Lines 3493 to 3495 in 05161aa

	let offset = iter.take_while(|ch| chars::char_is_word(*ch)).count();
	let start_offset = cursor.saturating_sub(offset);
	let prefix = text.slice(start_offset..cursor).to_string();

The correct results from the server, that start with a backslash, are then filtered out.

helix/helix-term/src/commands.rs

Lines 3519 to 3527 in 05161aa

	items = items
	.into_iter()
	.filter(|item| {
	item.filter_text
	.as_ref()
	.unwrap_or(&item.label)
	.starts_with(&prefix)
	})
	.collect();

Reproduction steps

File with \alpha, see logs.

Environment

• Platform: Windows 10, WSL2
• Terminal emulator: Windows Terminal
• Helix version: 0.6.0

~/.cache/helix/helix.log

2022-03-13T20:51:22.283 helix_lsp::transport [INFO] -> {"jsonrpc":"2.0","method":"textDocument/completion","params":{"position":{"character":6,"line":0},"textDocument":{"uri":"..."}},"id":5}
2022-03-13T20:51:22.285 helix_lsp::transport [INFO] <- {"id":5,"jsonrpc":"2.0","result":{"isIncomplete":true,"items":[{"label":"\\biAlpha","kind":11,"detail":"𝜜","documentation":"𝜜","textEdit":{"range":{"start":{"line":0,"character":0},"end":{"line":0,"character":6}},"newText":"𝜜"}},{"label":"\\bfalpha","kind":11,"detail":"𝛂","documentation":"𝛂","textEdit":{"range":{"start":{"line":0,"character":0},"end":{"line":0,"character":6}},"newText":"𝛂"}},{"label":"\\bisansAlpha","kind":11,"detail":"𝞐","documentation":"𝞐","textEdit":{"range":{"start":{"line":0,"character":0},"end":{"line":0,"character":6}},"newText":"𝞐"}},{"label":"\\alpha","kind":11,"detail":"α","documentation":"α","textEdit":{"range":{"start":{"line":0,"character":0},"end":{"line":0,"character":6}},"newText":"α"}},{"label":"\\bfAlpha","kind":11,"detail":"𝚨","documentation":"𝚨","textEdit":{"range":{"start":{"line":0,"character":0},"end":{"line":0,"character":6}},"newText":"𝚨"}},{"label":"\\Alpha","kind":11,"detail":"Α","documentation":"Α","textEdit":{"range":{"start":{"line":0,"character":0},"end":{"line":0,"character":6}},"newText":"Α"}},{"label":"\\itAlpha","kind":11,"detail":"𝛢","documentation":"𝛢","textEdit":{"range":{"start":{"line":0,"character":0},"end":{"line":0,"character":6}},"newText":"𝛢"}},{"label":"\\bisansalpha","kind":11,"detail":"𝞪","documentation":"𝞪","textEdit":{"range":{"start":{"line":0,"character":0},"end":{"line":0,"character":6}},"newText":"𝞪"}},{"label":"\\bialpha","kind":11,"detail":"𝜶","documentation":"𝜶","textEdit":{"range":{"start":{"line":0,"character":0},"end":{"line":0,"character":6}},"newText":"𝜶"}},{"label":"\\^alpha","kind":11,"detail":"ᵅ","documentation":"ᵅ","textEdit":{"range":{"start":{"line":0,"character":0},"end":{"line":0,"character":6}},"newText":"ᵅ"}},{"label":"\\bsansAlpha","kind":11,"detail":"𝝖","documentation":"𝝖","textEdit":{"range":{"start":{"line":0,"character":0},"end":{"line":0,"character":6}},"newText":"𝝖"}},{"label":"\\italpha","kind":11,"detail":"𝛼","documentation":"𝛼","textEdit":{"range":{"start":{"line":0,"character":0},"end":{"line":0,"character":6}},"newText":"𝛼"}},{"label":"\\bsansalpha","kind":11,"detail":"𝝰","documentation":"𝝰","textEdit":{"range":{"start":{"line":0,"character":0},"end":{"line":0,"character":6}},"newText":"𝝰"}}]}}
2022-03-13T20:51:22.286 helix_lsp::transport [INFO] <- {"isIncomplete":true,"items":[{"detail":"𝜜","documentation":"𝜜","kind":11,"label":"\\biAlpha","textEdit":{"newText":"𝜜","range":{"end":{"character":6,"line":0},"start":{"character":0,"line":0}}}},{"detail":"𝛂","documentation":"𝛂","kind":11,"label":"\\bfalpha","textEdit":{"newText":"𝛂","range":{"end":{"character":6,"line":0},"start":{"character":0,"line":0}}}},{"detail":"𝞐","documentation":"𝞐","kind":11,"label":"\\bisansAlpha","textEdit":{"newText":"𝞐","range":{"end":{"character":6,"line":0},"start":{"character":0,"line":0}}}},{"detail":"α","documentation":"α","kind":11,"label":"\\alpha","textEdit":{"newText":"α","range":{"end":{"character":6,"line":0},"start":{"character":0,"line":0}}}},{"detail":"𝚨","documentation":"𝚨","kind":11,"label":"\\bfAlpha","textEdit":{"newText":"𝚨","range":{"end":{"character":6,"line":0},"start":{"character":0,"line":0}}}},{"detail":"Α","documentation":"Α","kind":11,"label":"\\Alpha","textEdit":{"newText":"Α","range":{"end":{"character":6,"line":0},"start":{"character":0,"line":0}}}},{"detail":"𝛢","documentation":"𝛢","kind":11,"label":"\\itAlpha","textEdit":{"newText":"𝛢","range":{"end":{"character":6,"line":0},"start":{"character":0,"line":0}}}},{"detail":"𝞪","documentation":"𝞪","kind":11,"label":"\\bisansalpha","textEdit":{"newText":"𝞪","range":{"end":{"character":6,"line":0},"start":{"character":0,"line":0}}}},{"detail":"𝜶","documentation":"𝜶","kind":11,"label":"\\bialpha","textEdit":{"newText":"𝜶","range":{"end":{"character":6,"line":0},"start":{"character":0,"line":0}}}},{"detail":"ᵅ","documentation":"ᵅ","kind":11,"label":"\\^alpha","textEdit":{"newText":"ᵅ","range":{"end":{"character":6,"line":0},"start":{"character":0,"line":0}}}},{"detail":"𝝖","documentation":"𝝖","kind":11,"label":"\\bsansAlpha","textEdit":{"newText":"𝝖","range":{"end":{"character":6,"line":0},"start":{"character":0,"line":0}}}},{"detail":"𝛼","documentation":"𝛼","kind":11,"label":"\\italpha","textEdit":{"newText":"𝛼","range":{"end":{"character":6,"line":0},"start":{"character":0,"line":0}}}},{"detail":"𝝰","documentation":"𝝰","kind":11,"label":"\\bsansalpha","textEdit":{"newText":"𝝰","range":{"end":{"character":6,"line":0},"start":{"character":0,"line":0}}}}]}

[MichaelByrneAU]

Just thought I would chime in to say that I have 'solved' this by modifying the char_is_word function to also accept backslashes for my local copy of Helix.

I don't know what broader implications this has on the rest of the application so I haven't opened up a pull request. It does resolve this issue though in allowing for the autocomplete of unicode characters.

[archseer]

I think the intent there is to keep scanning while char != whitespace, I'd accept a change that does that.

[ChrHorn]

Just thought I would chime in to say that I have 'solved' this by modifying the char_is_word function to also accept backslashes for my local copy of Helix.

I did something similar. The problem with this solution is, that you can't chain multiple completions. For example \alpha<Tab>\beta<Tab> does not work.

To solve this you need an exemption for \.
When the nonword char is a \: treat it as a boundary, but include the \.
For every other nonword char: don't include it.

[leoflotor]

Just thought I would chime in to say that I have 'solved' this by modifying the char_is_word function to also accept backslashes for my local copy of Helix.

I did something similar. The problem with this solution is, that you can't chain multiple completions. For example \alpha<Tab>\beta<Tab> does not work.

To solve this you need an exemption for \. When the nonword char is a \: treat it as a boundary, but include the \. For every other nonword char: don't include it.

Hi! Sorry to ask, I am not really fluent using rust. Could you explain me how can I include this exemption and how to treat \ as a boundary and include it?

[ChrHorn]

Sorry, I somehow missed this.

I did not implement this so far, just described how it should work. Not really fluent in Rust either. So far I just replaced line 3493 with this hacky workaround:

    let offset = iter.take_while(|ch| chars::char_is_word(*ch) || *ch == '\\').count();