Title: argparse: parser aliases in subparsers stores alias in dest variable
Type: enhancement Stage:
Components: Library (Lib) Versions: Python 3.7
Status: open Resolution:
Dependencies: Superseder:
Assigned To: Nosy List: Peter McEldowney, nemeskeyd, paul.j3
Priority: normal Keywords:

Created on 2019-04-18 23:30 by Peter McEldowney, last changed 2019-11-01 01:51 by paul.j3.

File name Uploaded Description Edit Peter McEldowney, 2019-04-18 23:30 Sample code with subparser (also in comment)
Messages (5)
msg340515 - (view) Author: pcm1289 (Peter McEldowney) Date: 2019-04-18 23:30
I noticed that I have to add a lot more code to handle contexts in subparsers that I was expecting would be necessary. This is something I feel should be handled by the argparse library. What are your thoughts on this?

If you run the sample code with the commands below, you can see that although I would want them to do the same thing, I have to add more lines into my code to achieve this. This becomes cumbersome/annoying when dealing with subparser trees.

python3 subsection
python3 s

Sample code (also attached):

import argparse

def get_args(args=None):
	parser = argparse.ArgumentParser()
	subparser = parser.add_subparsers(dest='context')
	sub = subparser.add_parser('subsection', aliases=['s', 'sub', 'subsect'])
	return parser.parse_args(args)
def my_subsection_function(args):
	print('my subsection was called')
def invalid_context(args):
	print('my functon was not called <sadface>')

def main(args=get_args()):	
	return {
		'subsection': my_subsection_function
	}.get(args.context, invalid_context)(args)

if __name__ == "__main__":
msg340525 - (view) Author: paul j3 (paul.j3) * (Python triager) Date: 2019-04-19 05:27
I added a `print(args)` to clarify what you are talking about:

2148:~/mypy$ python3 subsection
my subsection was called
2148:~/mypy$ python3 s
my functon was not called <sadface>
2148:~/mypy$ python3 sub
my functon was not called <sadface>

The value of `args.context` depends on what alias was used, not the primary name of the subparser.

The help lists all aliases

2148:~/mypy$ python3 -h
usage: [-h] {subsection,s,sub,subsect} ...

The sub-parser doesn't actually have a name.  In self._name_parser_map each alias is a key with a parser object value.  Multiple keys for a single value.  The only thing that distinguishes 'subsection' is that was the first key in that dictionary.  

In effect the subparser Action object does not maintain a mapping from the aliases to the 'subsection' name.  I can imagine some ways of deducing that mapping, but it's not going to be a trivial task.

Unless someone comes up with a clever patch, I think the best choice is for you maintain your own mapping.  For example write a utility that takes a 'name' and alias list, calls

    sub = subparser.add_parser('subsection', aliases=['s', 'sub', 'subsect'])

and saves some sort of mapping from the aliases to 'subsection'.  Then use that later when you use `args.context`.
msg340595 - (view) Author: paul j3 (paul.j3) * (Python triager) Date: 2019-04-21 03:29

shows how to use `set_defaults` in a parser to set a 'func' attribute.  That method could just as well be used to set the true 'name' or any other kind of attribute that is unique to that parser.  

I think we should leave the code as is, storing the aliases in the subparsers' 'dest'.  It's been that way since aliases were added, and changing it runs the risk of creating backward compatibility issues.  

There are enough work arounds if you don't want the aliases.
msg355759 - (view) Author: Dávid Nemeskey (nemeskeyd) Date: 2019-10-31 16:08
I ran into the same problem. I know of the set_defaults() method, in fact, that is what I have always been using. But why was dest added if not to provide a better solution?

I agree that changing it would _perhaps_ break some code, so I agree that this issue cannot be "fixed" for Python 3.7 (and probably 3.8). But what about 3.9 or 4.0?

BTW much higher profile backward-incompatible changes do happen in Python; take PEP 563 (, for instance.
msg355779 - (view) Author: paul j3 (paul.j3) * (Python triager) Date: 2019-11-01 01:51
Just clarify how the code currently works.  `subparsers` is a positional Action of subclass _SubParsersAction.  It has a nargs='+...', requiring at least one string, and taking all remaining strings.  Its __call__ has the standard signature.  So everything that's special about subparsers is embodied in this Action subclass.  

    def __call__(self, parser, namespace, values, option_string=None):
        parser_name = values[0]
        arg_strings = values[1:]

        # set the parser name if requested
        if self.dest is not SUPPRESS:
            setattr(namespace, self.dest, parser_name)

        # select the parser
            parser = self._name_parser_map[parser_name]

So the `parser_name` is first string, the actual alias that user provided.  It is added to the namespace if a `dest` was provided (the default `dest` is SUPPRESS).  That's all of the relevant code - the alias is simply added to to Namespace.

As mentioned before `parser_name` is used find the actual sub parser, which is called with the remaining `arg_strings`.

Earlier in the subclasses `add_parser` method, the 'name' and 'alias' list are used to populate the '_name_parser_map' mapping, and also create the metavar that's used in the help display.  But neither is saved as an attribute.


I still think 'set_defaults' is the cleanest way of saving a unique name for a sub parser.

    parser_foo.set_defaults(func=foo, name='foo')


One further note - if you make subparsers required, you should also set a dest name:

    parser.add_subparsers(dest='cmd', required=True)

otherwise the missing-subparsers error message will raise an error.  It needs to identify the missing action in some way.  Functionally, this might be the most important reason to set the 'dest'.
Date User Action Args
2019-11-01 01:51:18paul.j3setmessages: + msg355779
2019-10-31 16:08:45nemeskeydsetnosy: + nemeskeyd
messages: + msg355759
2019-04-21 03:29:55paul.j3setmessages: + msg340595
2019-04-19 05:27:11paul.j3setmessages: + msg340525
2019-04-19 02:34:09xtreaksetnosy: + paul.j3
2019-04-18 23:30:32Peter McEldowneycreate