Question

I'm writing a relatively simple Python script which supports a couple of different commands. The different commands support different options and I want to be able to pass the options parsed by argparse to the correct method for the specified command.

The usage string looks like so:

usage: script.py [-h]

            {a, b, c}
            ...
script.py: error: too few arguments

I can easily call the appropriate method:

def a():
    ...

def b():
    ...

def c():
    ...

if __name__ == "__main__":
    parser = argparse.ArgumentParser()
    parser.set_defaults(method = a)
    ...

    arguments = parser.parse_args()
    arguments.method()

However, I have to pass arguments to these methods and they all accept different sets of arguments.

Currently, I just pass the Namespace object returned by argparse, like so:

 def a(arguments):
     arg1 = getattr(arguments, 'arg1', None)
     ...

This seems a little awkward, and makes the methods a little harder to reuse as I have to pass arguments as a dict or namespace rather than as usual parameters.

I would like someway of defining the methods with parameters (as you would a normal function) and still be able to call them dynamically while passing appropriate parameters. Like so:

def a(arg1, arg2):
    ...

if __name__ == "__main__":
    parser = argparse.ArgumentParser()
    parser.set_defaults(method = a)
    ...

    arguments = parser.parse_args()
    arguments.method() # <<<< Arguments passed here somehow

Any ideas?

Was it helpful?

Solution

I found quite a nice solution:

import argparse

def a(arg1, arg2, **kwargs):
    print arg1
    print arg2

if __name__ == "__main__":
        parser = argparse.ArgumentParser()
        parser.set_defaults(method = a)
        parser.add_argument('arg1', type = str)
        parser.add_argument('arg2', type = str)

        arguments = parser.parse_args()
        arguments.method(**vars(arguments))

Of course there's a minor problem if the arguments of the method clash with the names of the arguments argparse uses, though I think this is preferable to passing the Namespace object around and using getattr.

OTHER TIPS

You're probably trying to achieve the functionality that sub-commands provide: http://docs.python.org/dev/library/argparse.html#sub-commands

Not sure how practical this is, but by using inspect you can leave out the extraneous **kwargs parameter on your functions, like so:

import argparse
import inspect


def sleep(seconds=0):
    print "sleeping", seconds, "seconds"

def foo(a, b=2, **kwargs):
    print "a=",a
    print "b=",b
    print "kwargs=",kwargs

parser = argparse.ArgumentParser()

subparsers = parser.add_subparsers(title="subcommand")

parser_sleep = subparsers.add_parser('sleep')
parser_sleep.add_argument("seconds", type=int, default=0)
parser_sleep.set_defaults(func=sleep)

parser_foo = subparsers.add_parser('foo')
parser_foo.add_argument("-a", type=int, default=101)
parser_foo.add_argument("-b", type=int, default=201)
parser_foo.add_argument("--wacky", default=argparse.SUPPRESS)
parser_foo.set_defaults(func=foo)

args = parser.parse_args()

arg_spec = inspect.getargspec(args.func)
if arg_spec.keywords:
    ## convert args to a dictionary
    args_for_func = vars(args)
else:
    ## get a subset of the dictionary containing just the arguments of func
    args_for_func = {k:getattr(args, k) for k in arg_spec.args}

args.func(**args_for_func)

Examples:

$ python test.py sleep 23
sleeping 23 seconds

$ python test.py foo -a 333 -b 444
a= 333
b= 444
kwargs= {'func': <function foo at 0x10993dd70>}

$ python test.py foo -a 333 -b 444 --wacky "this is wacky"
a= 333
b= 444
kwargs= {'func': <function foo at 0x10a321d70>, 'wacky': 'this is wacky'}

Have fun!

Licensed under: CC-BY-SA with attribution
Not affiliated with StackOverflow
scroll top