User reference

Clize deduces what kind of parameter to pick for the CLI depending on what kind of parameter is found on the Python function as well as its annotations.


For how parameters are converted between Python and CLI:

Using annotations

You can use annotations defined throughout this document to refine parameter conversion (or change it completely). Here’s a primer on how to use these annotations.


Python 2 compatibility

In this document we will be showing examples that use Python 3 syntax such as annotations and keyword-only parameters for conciseness. To translate those into Python 2, you can use sigtools.modifiers.kwoargs and sigtools.modifiers.annotate.

For instance, given this Python 3 function:

def func(ab:int, *, cd:'c'=2):

You would write in Python 2:

from sigtools import modifiers

@modifiers.annotate(ab=int, cd='c')
def func(ab, cd=2):

You can pass annotations as a sequence:

def func(*, param:('p', int)): pass
def func(*, param:['p', int]): pass

When only one annotation is needed, you can omit the sequence:

def func(*, param:'p'): pass
def func(*, param:('p',)): pass
def func(*, param:['p']): pass

Annotations for parameters that handle a value

The positional and options parameters detailed later both handle the following features:

Specifying a value converter

A function or callable decorated with parser.value_converter passed as annotation will be used during parsing to convert the value from the string found in sys.argv into a value suitable for the annotated function.

from clize import run, parser

def wrap_xy(arg):
    return 'x' + arg + 'y'

def func(a, b:wrap_xy):
    print(repr(a), repr(b))

$ python abc def
'abc' 'xdefy'

def was transformed into xdefy because of the value converter.

Besides callables decorated with parser.value_converter, the built-in functions int, float and bool are also recognized as value converters.

Included value converters


Parses a date into a datetime value

Requires dateutil to be installed.

clize.converters.file(arg=<unset>, *, stdio='-', keep_stdio_open=False, **kwargs)[source]

Takes a file argument and provides a Python object that opens a file

def main(in_: file(), out: file(mode='w')):
    with in_ as infile, out as outfile:
  • stdio – If this value is passed as argument, it will be interpreted as stdin or stdout depending on the mode parameter supplied.
  • keep_stdio_open – If true, does not close the file if it is stdin or stdout.

Other arguments will be relayed to

This converter also opens the file or stream designated by the default value:

def main(inf: file()='-'):
    with inf as f:
$ python3 ./
<_io.TextIOWrapper name='<stdin>' mode='r' encoding='UTF-8'>

Specifying a default value

The parameter’s default value is used without conversion. If no value converter is specified, its type is used instead. When a default value is specified, the parameter becomes optional.

>>> def func(par=3):
...     print(repr(par))
>>> run(func, exit=False, args=['func', '46'])
>>> run(func, exit=False, args=['func'])

Therefore, be careful not to use values of types for which the constructor does not handle strings, unless you specify a converter:

>>> from datetime import datetime
>>> now = datetime.utcnow()
>>> def fail(par=now):
...     print(repr(par))
>>> run(fail, exit=False, args=['func', '...'])
Traceback (most recent call last):
TypeError: an integer is required (got type str)
>>> from dateutil.parser import parse
>>> from datetime import datetime
>>> now = datetime.utcnow()
>>> def func(par:parse=now):
...     print(par)
>>> run(func, exit=False, args=['func', '1/1/2016'])
2016-01-01 00:00:00

Ignoring the source parameter’s default value


Annotate a parameter with this to force it to be required, even if there is a default value in the source.

>>> from clize import run, Parameter
>>> def func(par:Parameter.REQUIRED=3):
...     pass
>>> run(func, exit=False, args=['func'])
func: Missing required arguments: par
Usage: func par

Positional parameters

Normal parameters in Python are turned into positional parameters on the CLI. Plain arguments on the command line (those that don’t start with a -) are processed by those and assigned in the order they appear:

from clize import run

def func(posparam1, posparam2, posparam3):
    print('posparam1', posparam1)
    print('posparam2', posparam2)
    print('posparam3', posparam3)

$ python one two three
posparam1 one
posparam2 two
posparam3 three

It also shares the features detailed in Annotations for parameters that handle a value.

Parameter that collects remaining positional arguments

An *args-like parameter in Python becomes a repeatable positional parameter on the CLI:

from clize import run

def func(param, *args):
    print('param', param)
    print('args', args)

$ python one two three
param one
args ('two', 'three')

When used on an *args parameter, requires at least one value to be provided.

You can use clize.parameters.multi for more options.

Named parameters

Clize treats keyword-only parameters as named parameters, which, instead of their position, get designated by they name preceeded by --, or by - if the name is only one character long.

There are a couple kinds of named parameters detailed along with examples in the sections below.

They all understand annotations to specify alternate names for them: Simply pass them as strings in the parameter’s annotation:

from clize import run

def func(*, par:('p', 'param')):
    print('par', par)

$ python --par value
par value
$ python -p value
par value
$ python --param value
par value

Python 2 support for named parameters

Python 2 has no keyword-only parameters. To fill that gap, you can use the decorators from sigtools.modifiers to emulate them.

from sigtools.modifiers import DECORATOR

def func(ab, cd, de=None, fg=None, hi=None):
@DECORATOR Parameters that become keyword-only
@kwoargs('cd', 'fg') cd and fg
@kwoargs(start='fg') fg and all following parameters
@autokwoargs All parameters with defaut values
@autokwoargs(exceptions=['fg']) Same, except for fg

Name conversion

So you can best follow Python’s naming conventions, Clize alters parameter names to CLI conventions so that they always match --kebap-case. While Python recommends snake_case for identifiers such as parameter names, Clize also handles words denoted by capitalization:

Python parameter name CLI parameter name
param --param
p -p
snake_case --snake-case
list_ --list
capitalizedOnce --capitalized-once
CapitalizedTwice --capitalized-twice

Clize replaces underscores (_) with dashes (-), and uppercase characters with a dash followed by the equivalent lowercase character. Dashes are deduplicated and removed from the extremities.

Short option names (those that are only one letter) are prepended with one dash (-). Long option names are prepended with two dashes (--). No dash is prepended for subcommand names.


You do not need to consider this when documenting parameters. Simply match the way the parameter is written in your Python sources.

Named parameters that take an argument

Keyword-only parameters in Python become named parameters on the CLI: They get designated by their name rather than by their position:

from clize import run

def func(arg, *, o, par):
    print('arg', arg)
    print('o', o)
    print('par', par)

$ python -o abc def --par ghi
arg def
o abc
par ghi
$ python -oabc --par=def ghi
arg ghi
o abc
par def

The parameter is designated by prefixing its name with two dashes (eg. --par) or just one dash if the name is only one character long (eg. -o). The value is given either as a second argument (first example) or glued to the parameter name for the short form, glued using an equals sign (=) for the long form.

It shares the features of Annotations for parameters that handle a value and Named parameters.

Named parameters that take an integer argument

A variant of Named parameters that take an argument used when the value type from Annotations for parameters that handle a value is int. The only difference is that when designating the parameter using the short glued form, you can chain other short named parameters afterwards:

from clize import run

def func(*, i: int, o):
    print('i', i)
    print('o', o)

$ python -i42o abcdef
i 42
o abcdef

Flag parameters

Flag parameters are named parameters that unlike options do not take an argument. Instead, they set their corresponding parameter in Python to True if mentionned.

You can create them by having a keyword-only parameter take False as default value:

from clize import run

def func(*, flag=False):
    print('flag', flag)

$ python
flag False
$ python --flag
flag True
$ python --flag=0
flag False

Additionally, you can chain their short form on the command line with other short parameters:

from clize import run

def func(*, flag:'f'=False, other_flag:'g'=False, opt:'o'):
    print('flag', flag)
    print('other_flag', other_flag)
    print('opt', opt)

$ python -fgo arg
flag True
other_flag True
opt arg
$ python -fo arg
flag True
other_flag False
opt arg

Mapped parameters

clize.parameters.mapped(values, *, list_name='list', case_sensitive=None)[source]

Creates an annotation for parameters that maps input values to Python objects.

  • values (sequence) – A sequence of pyobj, names, description tuples. For each item, the user can specify a name from names and the parameter will receive the corresponding pyobj value. description is used when listing the possible values.
  • list_name (str) – The value the user can use to show a list of possible values and their description.
  • case_sensitive (bool) – Force case-sensitiveness for the input values. The default is to guess based on the contents of values.
greeting = parameters.mapped([
    ('Hello', ['hello', 'hi'], 'A welcoming message'),
    ('Goodbye', ['goodbye', 'bye'], 'A parting message'),

def main(name='world', *, kind:('k', greeting)='Hello'):
    :param name: Who is the message for?
    :param kind: What kind of message should be given to name?
    return '{0} {1}!'.format(kind, name)
$ python -m examples.mapped -k list
python -m examples.mapped: Possible values for --kind:

  hello, hi      A welcoming message
  goodbye, bye   A parting message
$ python -m examples.mapped -k hello
Hello world!
$ python -m examples.mapped -k hi
Hello world!
$ python -m examples.mapped -k bye
Goodbye world!
$ python -m examples.mapped
Hello world!
clize.parameters.one_of(*values, case_sensitive=None, list_name='list')[source]

Creates an annotation for a parameter that only accepts the given values.

  • valuesvalue, description tuples, or just the accepted values
  • list_name (str) – The value the user can use to show a list of possible values and their description.
  • case_sensitive (bool) – Force case-sensitiveness for the input values. The default is to guess based on the contents of values.

Repeatable parameters

clize.parameters.multi(min=0, max=None)[source]

For option parameters, allows the parameter to be repeated on the command-line with an optional minimum or maximum. For *args-like parameters, just adds the optional bounds.

def main(*, listen:('l', parameters.multi(min=1, max=3))):
    """Listens on the given addresses

    :param listen: An address to listen on.
    for address in listen:
        print('Listening on {0}'.format(address))
$ python -m examples.multi -l bacon
Listening on bacon
$ python -m examples.multi -l bacon -l ham -l eggs
Listening on bacon
Listening on ham
Listening on eggs
$ python -m examples.multi -l bacon -l ham -l eggs -l spam
python -m examples.multi: Received too many values for --listen
Usage: python -m examples.multi [OPTIONS]
$ python -m examples.multi
python -m examples.multi: Missing required arguments: --listen
Usage: python -m examples.multi [OPTIONS]

Decorated arguments


Decorates a function to create an annotation for adding parameters to qualify another.

def capitalize(arg, *, capitalize:('c', 'upper')=False, reverse:'r'=False):
    Options to qualify {param}:

    :param capitalize: Make {param} uppercased
    :param reverse: Reverse {param}
    if capitalize:
        arg = arg.upper()
    if reverse:
        arg = arg[::-1]
    return arg

def main(*args:capitalize):
    :param args: Words to print
    return ' '.join(args)
$ python -m examples.argdeco --help
Usage: python -m examples.argdeco [OPTIONS] [[-c] [-r] args...]

  args...         stuff

Options to qualify args:
  -c, --capitalize, --upper
                  Make args uppercased
  -r, --reverse   Reverse args

Other actions:
  -h, --help      Show the help
$ python -m examples.argdeco abc -c def ghi
abc DEF ghi

Annotations that work on any parameter

The following objects can be used as annotation on any parameter:

Parameter converters

Callables decorated with clize.parser.parameter_converter are used instead of the default converter to construct a CLI parameter for the annotated Python parameter.

The callable can return a Parameter instance or Parameter.IGNORE to instruct clize to drop the parameter.

>>> from clize import run, parser
>>> @parser.parameter_converter
... def print_and_ignore(param, annotations):
...     print(repr(param), annotations)
...     return parser.Parameter.IGNORE
>>> def func(par:(print_and_ignore,int)):
...     pass
>>> run(func, exit=False, args=['func', '--help'])
<Parameter at 0x7fc6b0c3dae8 'par'> (<class 'int'>,)
Usage: func

Other actions:
  -h, --help   Show the help

Unless you are creating new kinds of parameters this should not be useful to you directly. Note that custom parameter converters are likely to use different conventions than those described in this reference.

Parameter instances

A Parameter instance seen in an annotation will be used to represent that parameter on the CLI, without any further processing. Using a parameter converter is recommended over this.

Skipping parameters

Parameter.IGNORE = clize.Parameter.IGNORE

Annotate a parameter with this and it will be dropped from the resulting CLI signature.

Note that it is dangerous to use this on anything except:

  • On *args and **kwargs-like parameters,
  • On keyword parameters with defaults.

For instance, clize’s default converter does not handle **kwargs:

>>> from clize import run, Parameter
>>> def fail(**kwargs):
...     pass
>>> run(fail, exit=False)
Traceback (most recent call last):
ValueError: This converter cannot convert parameter 'kwargs'.

However, if we decorate that parameter with Parameter.IGNORE, clize ignores it:

>>> def func(**kwargs:Parameter.IGNORE):
...     pass
>>> run(func, exit=False)
>>> run(func, exit=False, args=['func', '--help'])
Usage: func

Other actions:
  -h, --help   Show the help

Hiding parameters from the help

Parameter.UNDOCUMENTED = clize.Parameter.UNDOCUMENTED

Parameters annotated with this will be omitted from the documentation (--help).

>>> from clize import run, Parameter
>>> def func(*, o1, o2:Parameter.UNDOCUMENTED):
...     pass
>>> run(func, exit=False, args=['func', '--help'])
Usage: func [OPTIONS]


Other actions:
  -h, --help   Show the help

Forcing arguments to be treated as positional

Parameter.LAST_OPTION = clize.Parameter.LAST_OPTION

Annotate a parameter with this and all following arguments will be processed as positional.

>>> from clize import run, Parameter
>>> def func(a, b, c, *, d:Parameter.LAST_OPTION, e=''):
...     print("a:", a)
...     print("b:", b)
...     print("c:", c)
...     print("d:", d)
...     print("e:", e)
Usage:  [OPTIONS] a b c
>>> run(func, exit=False, args=['func', 'one', '-d', 'alpha', '-e', 'beta'])
a: one
b: -e
c: beta
d: alpha

Here, -e beta was received by the b and c parameters rather than e, because it was processed after -d alpha, which triggered the parameter d which had the annotation.

Retrieving the executable name

clize.parameters.pass_name(param, annotations)[source]

Parameters decorated with this will receive the executable name as argument.

This can be either the path to a Python file, or python -m some.module. It is also appended with sub-command names.

from clize import run, parameters

def main(name:parameters.pass_name, arg):
    print('name:', name)
    print('arg:', arg)

def alt(arg, *, name:parameters.pass_name):
    print('arg:', arg)
    print('name:', name)

run(main, alt=alt)
$ python ham
arg: ham
$ python -m pn ham
name: python -m pn
arg: ham
$ python --alt spam
arg: spam
name: --alt
$ python -m pn --alt spam
arg: spam
name: python -m pn --alt

Inserting arbitrary values


Create an annotation that hides a parameter from the command-line and always gives it the result of a function.

Parameters:value_factory (function) – Called to determine the value to provide for the parameter. The current parser.CliBoundArguments instance is passed as argument, ie. value_factory(ba).
from clize import run, parameters

def insert_ultimate_answer(ba):
    return 42

def main(arg, ans:insert_ultimate_answer):
    print('arg:', arg)
    print('ans:', ans)

$ python eggs
arg: eggs
ans: 42