Skip to main content

A cute helper for argparse in Python 3

Project description

ArgCat - A cute helper for argparse in Python 3

License: MIT linting: pylint Static Badge Static Badge

ArgCat is a tiny tool designed to make it more joyful to use argparse module in Python 3.

As a bridge between developer and argparse module, it wraps argparse and tries to take on all the "dirty works" for you, improving both the ''building" and "handling" parts of an argparse process, allowing you to focus more on business logic without worring about creating and configuring parsers and arguments of argparse in command-line interfaces.

Installation

pip install argcat

Once installation done, you should get a v0.4.x version in your python package library , as the latest stable version is v0.4.x.

Features

Build parsers and arguments of argparse in a more straightforward and clearer way

A typical code snippet to create a program supports sub-commands using argparse:

(Snippet from "Sub-commands" section in argparse)

# Create the top-level parser
parser = argparse.ArgumentParser(prog='PROG')
parser.add_argument('--foo', action='store_true', help='foo help')
subparsers = parser.add_subparsers(help='sub-command help')

# Create the parser for the "a" command
parser_a = subparsers.add_parser('a', help='a help')
parser_a.add_argument('bar', type=int, help='bar help')

# Create the parser for the "b" command
parser_b = subparsers.add_parser('b', help='b help')
parser_b.add_argument('--baz', choices='XYZ', help='baz help')

# Parse some argument lists
parser.parse_args(['a', '12'])
parser.parse_args(['--foo', 'b', '--baz', 'Z'])

But, using ArgCat

argcat = ArgCat()

# Build parsers and arguments
with argcat.build() as builder:
    # Set descriptive information of the program
    builder.set_prog_info(prog='PROG')
    builder.set_subparsers_info(help='sub-command help')
            
    # Add an argument to the main parser
    builder.main_parser().add_argument('--foo', action='store_true', help='foo help')
            
    # Create the parser for the "a" command
    builder.add_subparser('a', help='a help')
    builder.subparser('a').add_argument('bar', type=int, help='bar help')
            
    # Create the parser for the "b" command
    builder.add_subparser('b', help='b help')
    builder.subparser('b').add_argument('--baz', choices='XYZ', help='baz help')

# Parse some argument lists
argcat.parse_args(['a', '12'])
argcat.parse_args(['--foo', 'b', '--baz', 'Z'])

Handle parsed result from argparse directly and easily

A typical code snippet to create a program supports sub-commands and handles the parsed result using argparse:

(Snippet from "Sub-commands" section in argparse)

# Sub-command functions
def foo(args):
    print(args.x * args.y)

def bar(args):
    print('((%s))' % args.z)

# Create the top-level parser
parser = argparse.ArgumentParser()
subparsers = parser.add_subparsers()

# Create the parser for the "foo" command
parser_foo = subparsers.add_parser('foo')
parser_foo.add_argument('-x', type=int, default=1)
parser_foo.add_argument('y', type=float)
parser_foo.set_defaults(func=foo)

# Create the parser for the "bar" command
parser_bar = subparsers.add_parser('bar')
parser_bar.add_argument('z')
parser_bar.set_defaults(func=bar)

# Parse the args and call whatever function was selected
args = parser.parse_args('foo 1 -x 2'.split())
args.func(args)

# Parse the args and call whatever function was selected
args = parser.parse_args('bar XYZYX'.split())
args.func(args)

Instead, using ArgCat

# Sub-command handler functions.
# Note that the parameters of the functions should be the dests of the arguments for 
# the parsers. ArgCat helps you to route the argument's inputs from the parsed args into
# the parsers' handler functions respectively.
def foo(x, y):
    print(x * y)

def bar(z):
    print('((%s))' % z)

argcat = ArgCat()
with argcat.build() as builder:
    # Create the parser for the "foo" command.
    builder.add_subparser('foo')
    builder.subparser('foo').add_argument('-x', type=int, default=1)
    builder.subparser('foo').add_argument('y', type=float)

    # Create the parser for the "bar" command.
    builder.add_subparser('bar')
    builder.subparser('bar').add_argument('z')

# Set handler functions for receiving and processing parsed result.
argcat.set_parser_handler('foo', foo)
argcat.set_parser_handler('bar', bar)

# Parse the args and the handler function `foo()` will be called with required 
# parameters automatically.
argcat.parse_args('foo 1 -x 2'.split())
# Parse the args and the handler function `bar()` will be called with required 
# parameters automatically.
argcat.parse_args('bar XYZYX'.split())

And, if you would like to set all handler functions at once, using ArgCat

# Sub-command handler functions, which are decorated for the respective parsers.
@ArgCat.handler(parser_name='foo')
def decorated_foo(x, y):
    print(x * y)
        
@ArgCat.handler(parser_name='bar')
def decorated_bar(z):
    print('((%s))' % z)    

argcat = ArgCat()
with argcat.build() as builder:
    # Create the parser for the "foo" command.
    builder.add_subparser('foo')
    builder.subparser('foo').add_argument('-x', type=int, default=1)
    builder.subparser('foo').add_argument('y', type=float)

    # Create the parser for the "bar" command.
    builder.add_subparser('bar')
    builder.subparser('bar').add_argument('z')

# Set handler functions all at once from a handler provider
# Supposed this code snippet is in a __main__ module, taking the module as the provider.
# ArgCat will try to find all functions decorated by @ArgCat.handler() and set them as 
# the the handlers for the parsers according to the `parser_name` set in the decorator
# respectively.
argcat.add_main_module_as_handler_provider()
# The above function call is a convenient way for below:
# argcat.add_handler_provider(sys.modules['__main__'])

# Parse the args and the handler function `foo()` will be called with required 
# parameters automatically.
argcat.parse_args('foo 1 -x 2'.split())
# Parse the args and the handler function `bar()` will be called with required 
# parameters automatically.
argcat.parse_args('bar XYZYX'.split())

A Few Further Explanations: (You may be curious about)

  1. Handler Provider

    • A handler provider can be any object owns any functions decorated by @ArgCat.handler() with a parameter as a parser name.

    • When setting a handler provider, ArgCat tries to find all handler functions of the provider at first. Then it collects all the parsers of the specified names from the decorated functions. In the end, it tries to link the functions to each parser picked respectively.

  2. Handler Functions

    • ArgCat supports all kinds of functions to be handler functions, such as instance methods, @staticmethod, @classmethod and the other callables.

    • When decorating a callable to be a handler function, please make sure to place @ArgCat.handler() in the nearest place to the function definition, like below:

    @classmethod
    @ArgCat.handler(parser_name='foo')
    def foo(cls):
        ...
    
    • A handler function's signature must match arguments' dests of its parsers. For example, if a parser has two arguments ['--x'] and ['-y'], its handler function should have a signature contains the exact two parameters: (x, y) . By contrast, the name of the function is arbitary. So, either foo(x, y) or go(x, y) can be the parser's handler function.

      Note: There is a scenario which you may need to be aware of. Supposed you have one argument ['verbose'] for the main parser and also have a ['--file'] argument for the subparser 'load'. Then the handler function for the subparser should be in a form as foo(verbose, file) instead of foo(file) . Becase the subparser's handler function will also take the arguments from the main parser by default, unless the argument added to the main handler is through add_exlusive_argument() . Nevertheless, don't panic. Both set_parser_handler() and add_handler_provider() will check the signature for you and let you know what is the correct one.

License

MIT License

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

argcat-0.4.3.tar.gz (10.0 kB view details)

Uploaded Source

Built Distribution

argcat-0.4.3-py3-none-any.whl (4.6 kB view details)

Uploaded Python 3

File details

Details for the file argcat-0.4.3.tar.gz.

File metadata

  • Download URL: argcat-0.4.3.tar.gz
  • Upload date:
  • Size: 10.0 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/4.0.2 CPython/3.11.5

File hashes

Hashes for argcat-0.4.3.tar.gz
Algorithm Hash digest
SHA256 65893c8f674091532221f37d9b7a55d4e9ab4f84623da2a4917ad0e86b3bbd2d
MD5 b9327dbb7c1bacee827b0f8fbf90256b
BLAKE2b-256 c8f22700327b6fbc18b26d9d4f95638d6367e46a518321ddbd20764cb77985ed

See more details on using hashes here.

File details

Details for the file argcat-0.4.3-py3-none-any.whl.

File metadata

  • Download URL: argcat-0.4.3-py3-none-any.whl
  • Upload date:
  • Size: 4.6 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/4.0.2 CPython/3.11.5

File hashes

Hashes for argcat-0.4.3-py3-none-any.whl
Algorithm Hash digest
SHA256 c2c3b4745724fa405dd22e40c8e3a9c2f2f82fffcca89a50a39ed42d7f69bb78
MD5 d4b20e2a60943323c23bd561251a5522
BLAKE2b-256 b47f8ef92aa563e5ea5d1777c2f07edc01a7ceccce08636838afa47b86834b38

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page