pacak
pacak
I did consider this scenario a while ago and decided not to do anything about it back then. Reexports make sense if you have a library and and you want...
:) Well, good to hear that. To get this merged we need to rebase, make `rustfmt` happy and briefly mention it somewhere in the documentation, maybe even add an example...
`rustfmt` still fails - running `cargo fmt` and committing the updated code should do the trick.
I merged the change but replaced `#[bpaf(..., bpaf_path = xxx)]` with `#[bpaf(..., path(xxx))]` mostly to keep things uniform with other annotations. Will try to release 0.9.10 soon-ish after cleaning up...
Will take a look in a few hours/tomorrow.
Problem is related to having "value" items (`complete_shell(ShellComp::File{mask: None})`) completing at the same time as named items (`--add`, etc). Past me decided that "value" items take priority here but didn't...
Hmm... Probably still related to positional items. Making slow progress on the next version so hopefully soon :)
Rubberducking/note to self. Let's consider a minimal possible parser that picks between adding a `FILE` (a positional parser with attached shell completion) and a flag `-a`. The way completion works...
All the shells provide more or less the same info and I don't think any additional info from the shell will help, it's all about the internal representation. Anyway, after...
Got something working. Can you check if branch `complete` works for you?