jdx
commented
9 months ago yes, this is table-stakes. In fact what I need usage for currently is the ability to "mount" one CLI into another for mise run .... I haven't fully figured out how this will work but right now but in usage-cli I'm thinking the commands to build completions and stuff could accept a command string, e.g.:
$ usage complete-word -x "mycli --usage"And it would exec mycli --usage, get the dynamic definition, and be able to complete against it.
But it may make sense to configure this by putting it into the spec itself, maybe we could add a param like this:
fetch_spec "mycli --usage"I don't love that name but something like that.
heya,
this looks fab! One thing I've been wondering about is some sort of self-documenting/completing mechanism - e.g. some way for bins/scripts to act as their own completer.
having a way to get a usage spec out of a script/bin (e.g. your double-shebang mechanism) is one half, but the other is for the command itself to be recursively involved in completions. Think e.g. a field that wants an s3:// url, or an scp remotehost:/remotepath field. Not fleshed out sorry, I'll come back to this :)