问题
tldnr: given a function, is there a way to automatically create an ArgumentParser from its signature?
I've got a bunch of functions that I'd like to expose to the command line. So basically, a module:
def copy(foo, bar, baz):
...
def move(from, to):
...
def unlink(parrot, nomore=True):
...
if __name__ == '__main__':
argparse stuff
which can be called from the command line like this:
python commands.py move spam ham
python commands.py unlink --parrot Polly
Although this is pretty straightforward to implement, there's a lot of wiring involved:
parser = argparse.ArgumentParser(...)
subparsers = parser.add_subparsers()
...
c = subparsers.add_parser('unlink', description='Unlink a parrot')
c.add_argument('--parrot', help='parrots name', required=True)
c.add_argument('--nomore', help='this parrot is no more', action='store_true')
...
c = subparsers.add_parser('move', description='Move stuff')
...
and so on, for each function. The worst thing, should function arguments change (and they do), the argparse stuff needs to be synchronized manually.
It would be much nicer if the functions could provide argparse stuff for themselves, so that the main code would be like:
parser = argparse.ArgumentParser(...)
subparsers = parser.add_subparsers()
copy.register(subparsers)
move.register(subparsers)
unlink.register(subparsers)
...
I thought of something along these lines:
@args(
description='Unlink a parrot',
parrot={'required':True, 'help':'parrots name'},
nomore={'action': 'store_true', 'help': 'this parrot is no more'}
)
def unlink(parrot, nomore=True):
...
My questions:
- is there a library that does something like this?
- if no, is it possible to write such a decorator, and how?
- is there a other/better way to implement what I want?
Upd:
plac appears to be the solution. Here's how to do what I want with plac:
commands module: cmds.py:
import plac
@plac.annotations(
foo=('the foo thing'),
bar=('the bar thing'),
fast=('do a fast copy', 'flag')
)
def copy(foo, bar, fast=False):
"""Copy some foo to bar."""
pass
@plac.annotations(
parrots=('parrots names'),
nomore=('these parrots are no more', 'flag'),
repeat=('repeat n times', 'option', 'r', int)
)
def unlink(nomore=False, repeat=1, *parrots):
"""Unlink some parrots."""
pass
#more commands...
# export commands so that plac knows about them
commands = 'copy', 'unlink'
and here's the main module:
import plac
import cmds
plac.call(cmds)
Quite neat if you ask me.
回答1:
Have you tried plac?
An example in docs:
# dbcli.py
import plac
from sqlalchemy.ext.sqlsoup import SqlSoup
@plac.annotations(
db=plac.Annotation("Connection string", type=SqlSoup),
header=plac.Annotation("Header", 'flag', 'H'),
sqlcmd=plac.Annotation("SQL command", 'option', 'c', str, metavar="SQL"),
delimiter=plac.Annotation("Column separator", 'option', 'd'),
scripts=plac.Annotation("SQL scripts"),
)
def main(db, header, sqlcmd, delimiter="|", *scripts):
"A script to run queries and SQL scripts on a database"
yield 'Working on %s' % db.bind.url
if sqlcmd:
result = db.bind.execute(sqlcmd)
if header: # print the header
yield delimiter.join(result.keys())
for row in result: # print the rows
yield delimiter.join(map(str, row))
for script in scripts:
db.bind.execute(open(script).read())
yield 'executed %s' % script
if __name__ == '__main__':
for output in plac.call(main):
print(output)
Output:
usage: dbcli.py [-h] [-H] [-c SQL] [-d |] db [scripts [scripts ...]]
A script to run queries and SQL scripts on a database
positional arguments:
db Connection string
scripts SQL scripts
optional arguments:
-h, --help show this help message and exit
-H, --header Header
-c SQL, --sqlcmd SQL SQL command
-d |, --delimiter | Column separator
回答2:
Functionality akin to plac is provided by argh, which particularly features simple creation of subparsers (like those found in git or django-admin.py).
An example from its docs:
from argh import *
def dump(args):
return db.find()
@command
def load(path, format='json'):
print loaders[format].load(path)
p = ArghParser()
p.add_commands([load, dump])
if __name__ == '__main__':
p.dispatch()
Produces the following --help response:
usage: prog.py [-h] {load,dump} ...
positional arguments:
{load,dump}
load
dump
optional arguments:
-h, --help show this help message and exit
and the following with load --help:
usage: prog.py load [-h] [-f FORMAT] path
positional arguments:
path
optional arguments:
-h, --help show this help message and exit
-f FORMAT, --format FORMAT
Arguments may be annotated:
@arg('path')
@arg('--format', choices=['yaml','json'], default='json')
@arg('--dry-run', default=False)
@arg('-v', '--verbosity', choices=range(0,3), default=1)
def load(args, LOADERS={'json': json.load, 'yaml': yaml.load}):
loader = loaders[args.format]
data = loader(open(args.path))
...
And with @plain_signature, the args argument to load is expanded into keyword arguments:
@arg('path')
@arg('--format', choices=['yaml','json'], default='json')
@arg('--dry-run', default=False)
@arg('-v', '--verbosity', choices=range(0,3), default=1)
@plain_signature
def load(path, format, dry_run, verbosity):
...
回答3:
You can use the inspect module to look at your own function definitions. This way you can at least write a rudimentary argparse skeleton. However, you will probably need more information than just the argument names and possibly default values.
For instance, you will also need to give a description. This info you can largely provide by creating a docstring in an appropriate format. There are parsers for docstrings (e.g. Sphynx) Using this extra information, I think you will be able to automatically generate argparse invocations for your functions.
I do not think a decorator is required, as probably all the information can be stored in your docstring.
Let me know how you fare, I am interested in the results of your project.
回答4:
Another insteresting alternative is commando python module as declarative interface to argparse with additional utilities.
Example
Without commando:
def main():
parser = argparse.ArgumentParser(description='hyde - a python static website generator',
epilog='Use %(prog)s {command} -h to get help on individual commands')
parser.add_argument('-v', '--version', action='version', version='%(prog)s ' + __version__)
parser.add_argument('-s', '--sitepath', action='store', default='.', help="Location of the hyde site")
subcommands = parser.add_subparsers(title="Hyde commands",
description="Entry points for hyde")
init_command = subcommands.add_parser('init', help='Create a new hyde site')
init_command.set_defaults(run=init)
init_command.add_argument('-t', '--template', action='store', default='basic', dest='template',
help='Overwrite the current site if it exists')
init_command.add_argument('-f', '--force', action='store_true', default=False, dest='force',
help='Overwrite the current site if it exists')
args = parser.parse_args()
args.run(args)
def init(self, params):
print params.sitepath
print params.template
print params.overwrite
With commando:
class Engine(Application):
@command(description='hyde - a python static website generator',
epilog='Use %(prog)s {command} -h to get help on individual commands')
@param('-v', '--version', action='version', version='%(prog)s ' + __version__)
@param('-s', '--sitepath', action='store', default='.', help="Location of the hyde site")
def main(self, params): pass
@subcommand('init', help='Create a new hyde site')
@param('-t', '--template', action='store', default='basic', dest='template',
help='Overwrite the current site if it exists')
@param('-f', '--force', action='store_true', default=False, dest='overwrite',
help='Overwrite the current site if it exists')
def init(self, params):
print params.sitepath
print params.template
print params.overwrite
回答5:
The "least boilerplate" library I have found is fire (pip install fire).
Creating a command line parser for your example is as easy as:
import fire
def copy(foo, bar, baz):
...
def unlink(parrot, nomore=True):
...
if __name__ == '__main__':
fire.Fire()
and this turns your module into "Fire" CLI:
python your_module.py copy sim sala bim
来源：https://stackoverflow.com/questions/13248487/semi-automatic-generation-of-argparsers-for-functions