Questions about shell autocompletion

[hverlin]

Hi,
Thank you for this library!

Some questions about auto-completion.

Send additional context to complete() function

Currently, the complete has no context apart from the option/argument that it is completing. complete() does not take any arguments.
Would it be feasible in your opinion to send the full command (maybe as is) to the complete() function?

Here is an example of what I would like to achieve

my_cli run <program> <some_args>

I can auto-complete <program> dynamically (e.g., by reading the content of some configuration listing all possible programs).
However, then I am not able to auto-complete <some_args> as I would need to know the name of the program.

For now, the workaround seems to be that one has to register all possible combinations dynamically in advance. Due to the fact that some the command name gets converted to shell functions, it's not possible to use : or other special characters (whereas it's possible when using complete)

I think one way to implement this could be to update
https://github.com/c4spar/deno-cliffy/blob/main/command/completions/_zsh_completions_generator.ts#L50-58
with

lines="$(fry completions complete "${action}" "${@}" -- $words $state)"

Auto-completion for variadic arguments

Another question is how do I get auto-completion working for variadic arguments?

If I use the following, it appears that this will only auto-complete once instead of several times?

    new Command()
        .arguments('[args...:string:args]')
        .complete('args', () => {
          return ['arg1', 'arg2']
        })

For zsh, one way to achieve I believe is to replace

  _arguments -w -s -S -C \
    '(- *)'{-h,--help}'[Show this help.]' \
    '1:command:_commands'

with

  _arguments -w -s -S -C \
    '(- *)'{-h,--help}'[Show this help.]' \
    '*:command:_commands'

here: https://github.com/c4spar/deno-cliffy/blob/main/command/completions/_zsh_completions_generator.ts#L173

https://zsh.sourceforge.io/Doc/Release/Completion-System.html#:~:text=%3A*pattern%3Amessage%3Aaction

Description for arguments

Currently, the complete() function returns an array of possible arguments. Do you think it would be possible to return them as additional arguments?

zsh at least seems to support this:
https://zsh.sourceforge.io/Doc/Release/Completion-Widgets.html#Completion-Widgets:~:text=This%20adds%20per%2Dcompletion%20display%20strings

This could then automatically create a new ${action}_description function which would return the corresponding list of descriptions.
https://github.com/c4spar/deno-cliffy/blob/main/command/completions/_zsh_completions_generator.ts#L55

lines="$(my_cli completions complete "${action}" "${@}")"
values=("${(ps:\n:)lines}")

local -a _descriptions _descriptions_values
_descriptions="$(my_cli completions complete "${action}_description" "${@}")"
_descriptions_values=("${(ps:\n:)_descriptions}")

compadd -d _descriptions_values -a values

zsh also supports using -l to put the autocompletion in lines instead of columns, as well as -X 'top-level description' to print a description at the top

compadd -X 'my explanation' -l  -d _descriptions_values -a values

I am happy to contribute if this is of interest to you.

[c4spar]

Hi @hverlin , thx for your detailed feature request!
All your feature request are very wellcome and i would be happy to receive some PR's!

Some thoughts:

Send additional context to complete() function

Instead of passing $words and $state as additional arguments to the complete command:

lines="$(fry completions complete "${action}" "${@}" -- $words $state)"

I think i would use options instead of arguments.

lines="$(fry completions complete --words $words --state $state "${action}" "${@}")"

Description for arguments

I don't think I quite understand what you mean by "return them as additional arguments". Can you show me an example of what that would look like?