Use django-subcommander to write Django management commands that have
subcommands, each optionally having its own distinct options, help, and other
typically per-command behavior. Subcommands are just normal Django
BaseCommand subclasses, so there's very little new to learn. Here's an
example top-level command; you'd put this in your app's
from django_subcommander import SubcommandDispatcher class Command(SubcommandDispatcher): """The top-level "dessert" command which has several subcommands""" help = 'Eat, top, and do various things with desserts.' args = '<subcommand> [more arguments and options]' def _subcommand(self, name): """ Return a management command that implements the subcommand of the given name. """ if name == 'eat': return EatCommand() elif name == 'top': ,,, class EatCommand(BaseCommand): # You could put this in another module or wherever you want. help = 'Eat a dessert.' args = '[number of bites]' # Add options here with make_option(), in the usual way. def handle(self, *args, **options): ...
To invoke the subcommand for eating a dessert in 5 bites...
./manage.py dessert eat 5
To see the help for the
./manage.py dessert eat --help
To see the help for the top-level command...
./manage.py dessert --help
The help for the top-level command will list its subcommands if you implement
class Command(SubcommandDispatcher): ... def _subcommand(self, name): ... def _subcommand_names(self): """Return a list of the names of all the subcommands.""" return ['eat', 'top']
./manage.py dessert --help will result in something like this:
Usage: ... Options: ... Subcommands: eat [number of bites] top <topping> [more toppings]
- Notice that
_subcommand()returns a command instance, not a class or a module path. Not only does this give you the freedom to put your subcommand code wherever you wish, but it also means you can generate or parametrize subcommands dynamically, at runtime.
- There's no reason you shouldn't be able to have one
SubcommandDispatcherreturn another, thereby implementing multi-level subcommands.
Django's management command framework is built on
optparse, not the more
argparse, which supports subcommands natively. It would be quite a
bit of gluing to get
argparse working with Django's management command
infrastructure, so I took the simple road. This lets authors reuse everything
they already know about writing Django management commands. For example, I had
several groups of pre-existing commands I wanted to organize under a handful of
subcommands. This let me avoid having hundreds of individual files under
management/commands; it gave me the freedom to locate all that command code
elsewhere and organize it in a more natural way. Turning the commands into
subcommands required no changes to them at all.
django-subcommander is definitely a "worse is better" solution. It's an eminently practical solution to the profusion of files in
management/commands. If it had turned out long and complicated, I probably would have rigged Django to support argparse-based commands instead...hmm, what about making a BaseCommand alternative that's argparse-based?
- Tests. I'm using it all day now, but I've been testing it only "in situ".
- More flexibility in how to display the list of subcommands when getting
--helpon the top-level command
- Support for Django's other command superclasses like
LabelCommand(if, in fact, they don't work already and if there's demand)
- I went a short way down the path of giving
_subcommand()access to the whole
argvand having it return any args that weren't used to do dispatch. This ended up complicating things significantly for unclear benefit. If you'd find this useful, please file a bug.
- First release.