Question

I’m trying to develop a command line interface for a tool I’m writing. I’m tossing up between an interface similar to Fabric’s fab tool and one like svn’s command line tool.

fab‘s usage message:

Usage: fab [options] <command>[:arg1,arg2=val2,host=foo,hosts='h1;h2',...] ...

svn‘s usage message:

usage: svn <subcommand> [options] [args]

Note that svn also supports two options --version and --quiet (at least the one I’m using).

My questions are:

1. Is fab’s cli (with multiple commands, each of which can have arguments) design common among command-line tools? To me, the svn tool’s cli seems more commonly used (not sure though).
2. Would it be straightforward implementing fab’s cli using plac(or argparse)?
3. If I do decide to go with svn’s cli, can be this be implemented using plac? I like plac, though I haven’t been able to figure out how to add options to the tool directly, instead of the subcommands. It can definitely be implemented using argparse.

P.S.: The only reason fab’s cli appeals to me is that one can accomplish multiple tasks in one step, however ease of implementation is my main concern

Answer 1

1. subversion’s way is more standard.
2. Probably not.
3. You can cut butter with a hand axe.

If you want to simplify documentation and usage, go with SVN’s way. I can see why fab’s way seems appealing, it’s logical for a programmer. But it’s actually complicated and messed up for users.

Plac looks cool, but you need to install it. Your users too.

I like to follow a simple rule when it comes to software engineering:
Keep it simple and standard, use bleeding edge technologies when you really need it.
Retro-compatibility should be achieved when possible.

However, you’re just fooling around and don’t plan on distributing your work, then go on, flab’s way and using flac ;).