argparse option of options

Question

I am trying to add option of options in argparse. Currently I have:

group = parser.add_mutually_exclusive_group()
group.add_argument("--md", help="Create xyz file for each ionic step for"
 " visualization", action='store_true')
group.add_argument("--force", help="See which atom has maximum force",
 action='store_true')
group.add_argument("--opt", help="grep string from file",
 nargs=2, metavar=("str", "file"))
parser.add_argument("--xsf", help="Create xsf file for md(default is xyz)"
 " visualization", action='store_true')
parser.add_argument("-N", help="Showing first N line",
 metavar='integer', type=int)
parser.add_argument("-n", help="Showing last n line",
 metavar='integer', type=int)
args = parser.parse_args()

which gives:

./foo.py --h
usage: foo.py [-h]
 [--md | --force | --opt str file]
 [--xsf] [-N integer] [-n integer]

But I want --xsf as a suboption for --md, -N,-n for --opt; e.g.

./foo.py --h
 usage: foo.py [-h]
 [--md [--xsf]| --force | --opt str file [-N integer] [-n integer]]

But I dont know how to achieve that. May be I am missing something, but there is no option like that in argparse doc

Is there any other way of getting that?

• BaRud: If I add xsf to group, it can't be used with md, they will be mutually exclusive to one another.

Answer 1

The mutually_exclusive_group mechanism is quite simple, and does not work with any kind of nesting, or subgrouping.

There is a Python bug/issue requesting a more comprehensive grouping mechanism, but the proposed patch is rather complicated. The problem isn't just with testing, it's with defining the groups in a user friendly way, and with generating the usage line. It's nice that you included a desired usage, but that format is well beyond the capabilities of the current help formatter.

You might look into recasting your problem as a subparser one. subparsers are mutually exclusive (you can only give one command name), and you could specify --xsf as an argument for md, and -N as argument for --opt. But subparsers has its own help issues.

Another route is to write your own usage, and do your own testing of arguments after parsing. With a suitable choice of defaults you can usually tell whether an argument has been provided or not (the user can't specify None) or you can ignore unnecessary ones.