How to deal with umlauts (and probably other non-ASCII characters in branch/bookmark names?

[turion]

My workflow looks like this:

1. I have a Github issue with an umlaut ("ü" in this case) in the title
2. I create a branch in Github, this also has the umlaut in the branch name
3. jj git fetch: This correctly shows the new branch with the umlaut
4. jj b t: Now I'd like it to autocomplete after I've typed the issue number, this seems not to work. But I can still copy & paste the branch name and create a tracked bookmark Update: Can't reproduce, this works at least sometimes.
5. jj new 1234-my-issüe: This doesn't work at all. Autocompleting the bookmark name doesn't work, pasting the bookmark name doesn't work, I tried to escape it with " and ' to no avail. It's always Error: Failed to parse revset: Syntax error.

Workarounds:

• Give the Github branch a name that doesn't contain umlauts. Tedious. As a native German speaker I don't really notice umlauts until it's too late.
• Address the revision by id, not bookmark name. Again a bit tedious.
• Maybe escape the umlaut in some way? But if this is possible, autocompletion should really have shown me how.

I think jujutsu should just be able to deal with umlauts.

[turion]

By trial and error I've found that double escaping like this works:

jj new '"1234-my-issüe"'

I'm not sure why this is. Probably the shell does something weird here. But the error message was a jj error message, not a shell error message. And besides, autocompletion should really guide me here, I believe.

[turion]

The story continues. jj bookmark move doesn't accept the doubly escaped name. Probably because it expects a bookmark name and not a revset?

[yuja]

Which jj version and platform? Recent versions of jj should accept unicode identifier characters as symbols. If you're on macOS, you might suffer from weird NFC/NFD handling on filesystem (or maybe API boundary.)

[turion]

That sounds great, I'm on jj 0.30.0 on nixos. I'll try with 0.31 soon.

[yuja]

If you're using 0.30, there should be no change. The given example works at my end (zsh on Linux), so something would go wrong at filesystem/shell layer?

% jj debug revset --no-resolve '1234-my-issüe'
-- Parsed:
CommitRef(
    Symbol(
        "1234-my-issüe",
    ),
)

-- Optimized:
CommitRef(
    Symbol(
        "1234-my-issüe",
    ),
)

I don't have a good English reference about unicode normalization mess on macOS, but this page contains a list of references.
https://qiita.com/ko1nksm/items/3a66197efd1c096a801f

[turion]

❯ jj debug revset --no-resolve '1234-my-issüe'
error: unexpected argument '--no-resolve' found

  tip: to pass '--no-resolve' as a value, use '-- --no-resolve'

Usage: jj debug revset [OPTIONS] <REVISION>

For more information, try '--help'.

❯ jj debug revset  '1234-my-issüe'
-- Parsed:
CommitRef(
    Symbol(
        "1234-my-issüe",
    ),
)

Error: Revision `1234-my-issüe` doesn't exist

Strange. So it is able to parse this.

[martinvonz]

I suppose you tried jj new '1234-my-issüe' (just single quotes)?

Is it specific to revsets? Do things like jj new -m 1234-my-issüe work?

[turion]

As of lately, I cannot reproduce. It's very strange. Sorry. I'll comment again when I have a solid reproducer.