[question] Subcommand completion support for proper parsers

[nicholasjng]

Hey, thanks for creating this!

I wanted to start using it in a CLI project of my own, but before adoption, I am facing a usability question, which I was not able to find an immediate answer to.

Does shtab support the "one ArgumentParser per command class" use case?

Briefly explained, in this model, each CLI command inherits from a Command base class, which defines a parser for that specific argument only. Resolving the subcommand can then happen in an argument parser for the bare command. An example of a project using this is pip, with the base command implemented here.

Whereas for dvc, each command gets a subparser added to the single main dvc command parser, and the completion is added to the base parser, if I understood correctly.

The completion I am looking for would then do, say for two available subcommands bb and bbcc,

prog b<TAB>
# bb bbcc

Is this possible to do in current shtab if both of my bb and bbcc subcommands have their own argument parsers, as well as the base prog parser?

Thanks!

[casperdcl]

pip ultimately relies on the deprecated optparse^pip-optparse rather than the stdlib argparse^argparse, so wouldn't work with shtab.

However in general if you use argparse or argparse.ArgumentParser-like parsers, things should work fine.

[nicholasjng]

Sweet, thank you for the sources! For clarification, in order to achieve my subcommand completion example, would I need to add a subcommand argument to the main parser, with all available subcommands as choices, and then use shtab.CHOICE on that?