Welcome to jaraco.functools documentation!#

For Enterprise

Professional support for jaraco.functools is available as part of the Tidelift Subscription. Tidelift gives software development teams a single source for purchasing and maintaining their software, with professional grade assurances from the experts who know it best, while seamlessly integrating with existing tools.

Learn more Request a Demo

class jaraco.functools.Throttler(func, max_rate=float('Inf'))#

Bases: object

Rate-limit a function (or other callable).


Decorate a function with a transform function that is invoked on results returned from the decorated function.

>>> @apply(reversed)
... def get_numbers(start):
...     "doc for get_numbers"
...     return range(start, start+3)
>>> list(get_numbers(4))
[6, 5, 4]
>>> get_numbers.__doc__
'doc for get_numbers'
jaraco.functools.assign_params(func, namespace)#

Assign parameters from namespace where func solicits.

>>> def func(x, y=3):
...     print(x, y)
>>> assigned = assign_params(func, dict(x=2, z=4))
>>> assigned()
2 3

The usual errors are raised if a function doesn’t receive its required parameters:

>>> assigned = assign_params(func, dict(y=3, z=4))
>>> assigned()
Traceback (most recent call last):
TypeError: func() ...argument...

It even works on methods:

>>> class Handler:
...     def meth(self, arg):
...         print(arg)
>>> assign_params(Handler().meth, dict(arg='crystal', foo='clear'))()

Decorate a function to return its parameter unless check.

>>> enabled = [object()]  # True
>>> @bypass_unless(enabled)
... def double(x):
...     return x * 2
>>> double(2)
>>> del enabled[:]  # False
>>> double(2)
jaraco.functools.bypass_when(check, *, _op=identity)#

Decorate a function to return its parameter when check.

>>> bypassed = []  # False
>>> @bypass_when(bypassed)
... def double(x):
...     return x * 2
>>> double(2)
>>> bypassed[:] = [object()]  # True
>>> double(2)

Compose any number of unary functions into a single unary function.

>>> import textwrap
>>> expected = str.strip(textwrap.dedent(compose.__doc__))
>>> strip_and_dedent = compose(str.strip, textwrap.dedent)
>>> strip_and_dedent(compose.__doc__) == expected

Compose also allows the innermost function to take arbitrary arguments.

>>> round_three = lambda x: round(x, ndigits=3)
>>> f = compose(round_three, int.__truediv__)
>>> [f(3*x, x+1) for x in range(1,10)]
[1.5, 2.0, 2.25, 2.4, 2.5, 2.571, 2.625, 2.667, 2.7]
jaraco.functools.except_(*exceptions, replace=None, use=None)#

Replace the indicated exceptions, if raised, with the indicated literal replacement or evaluated expression (if present).

>>> safe_int = except_(ValueError)(int)
>>> safe_int('five')
>>> safe_int('5')

Specify a literal replacement with replace.

>>> safe_int_r = except_(ValueError, replace=0)(int)
>>> safe_int_r('five')

Provide an expression to use to pass through particular parameters.

>>> safe_int_pt = except_(ValueError, use='args[0]')(int)
>>> safe_int_pt('five')
jaraco.functools.first_invoke(func1, func2)#

Return a function that when invoked will invoke func1 without any parameters (for its side effect) and then invoke func2 with whatever parameters were passed, returning its result.


Return the argument.

>>> o = object()
>>> identity(o) is o
jaraco.functools.invoke(f, /, *args, **kwargs)#

Call a function for its side effect after initialization.

The benefit of using the decorator instead of simply invoking a function after defining it is that it makes explicit the author’s intent for the function to be called immediately. Whereas if one simply calls the function immediately, it’s less obvious if that was intentional or incidental. It also avoids repeating the name - the two actions, defining the function and calling it immediately are modeled separately, but linked by the decorator construct.

The benefit of having a function construct (opposed to just invoking some behavior inline) is to serve as a scope in which the behavior occurs. It avoids polluting the global namespace with local variables, provides an anchor on which to attach documentation (docstring), keeps the behavior logically separated (instead of conceptually separated or not separated at all), and provides potential to re-use the behavior for testing or other purposes.

This function is named as a pithy way to communicate, “call this function primarily for its side effect”, or “while defining this function, also take it aside and call it”. It exists because there’s no Python construct for “define and call” (nor should there be, as decorators serve this need just fine). The behavior happens immediately and synchronously.

>>> @invoke
... def func(): print("called")
>>> func()

Use functools.partial to pass parameters to the initial call

>>> @functools.partial(invoke, name='bingo')
... def func(name): print('called with', name)
called with bingo
jaraco.functools.method_cache(method, cache_wrapper=functools.lru_cache())#

Wrap lru_cache to support storing the cache data in the object instances.

Abstracts the common paradigm where the method explicitly saves an underscore-prefixed protected property on first call and returns that subsequently.

>>> class MyClass:
...     calls = 0
...     @method_cache
...     def method(self, value):
...         self.calls += 1
...         return value
>>> a = MyClass()
>>> a.method(3)
>>> for x in range(75):
...     res = a.method(x)
>>> a.calls

Note that the apparent behavior will be exactly like that of lru_cache except that the cache is stored on each instance, so values in one instance will not flush values from another, and when an instance is deleted, so are the cached values for that instance.

>>> b = MyClass()
>>> for x in range(35):
...     res = b.method(x)
>>> b.calls
>>> a.method(0)
>>> a.calls

Note that if method had been decorated with functools.lru_cache(), a.calls would have been 76 (due to the cached value of 0 having been flushed by the ‘b’ instance).

Clear the cache with .cache_clear()

>>> a.method.cache_clear()

Same for a method that hasn’t yet been called.

>>> c = MyClass()
>>> c.method.cache_clear()

Another cache wrapper may be supplied:

>>> cache = functools.lru_cache(maxsize=2)
>>> MyClass.method2 = method_cache(lambda self: 3, cache_wrapper=cache)
>>> a = MyClass()
>>> a.method2()

Caution - do not subsequently wrap the method with another decorator, such as @property, which changes the semantics of the function.

See also http://code.activestate.com/recipes/577452-a-memoize-decorator-for-instance-methods/ for another implementation and additional justification.

jaraco.functools.method_caller(*args, **kwargs)#

Decorate func so it’s only ever called the first time.

This decorator can ensure that an expensive or non-idempotent function will not be expensive on subsequent calls and is idempotent.

>>> add_three = once(lambda a: a+3)
>>> add_three(3)
>>> add_three(9)
>>> add_three('12')

To reset the stored value, simply clear the property saved_result.

>>> del add_three.saved_result
>>> add_three(9)
>>> add_three(8)

Or invoke ‘reset()’ on it.

>>> add_three.reset()
>>> add_three(-3)
>>> add_three(0)

Wrap func so it’s not called if its first param is None.

>>> print_text = pass_none(print)
>>> print_text('text')
>>> print_text(None)

Convert a generator into a function that prints all yielded elements.

>>> @print_yielded
... def x():
...     yield 3; yield None
>>> x()

Decorate a function with an action function that is invoked on the results returned from the decorated function (for its side effect), then return the original result.

>>> @result_invoke(print)
... def add_two(a, b):
...     return a + b
>>> x = add_two(2, 3)
>>> x
jaraco.functools.retry(*r_args, **r_kwargs)#

Decorator wrapper for retry_call. Accepts arguments to retry_call except func and then returns a decorator for the decorated function.


>>> @retry(retries=3)
... def my_func(a, b):
...     "this is my funk"
...     print(a, b)
>>> my_func.__doc__
'this is my funk'
jaraco.functools.retry_call(func, cleanup=lambda : ..., retries=0, trap=())#

Given a callable func, trap the indicated exceptions for up to ‘retries’ times, invoking cleanup on the exception. On the final attempt, allow any exceptions to propagate.


Wrap a method such that when it is called, the args and kwargs are saved on the method.

>>> class MyClass:
...     @save_method_args
...     def method(self, a, b):
...         print(a, b)
>>> my_ob = MyClass()
>>> my_ob.method(1, 2)
1 2
>>> my_ob._saved_method.args
(1, 2)
>>> my_ob._saved_method.kwargs
>>> my_ob.method(a=3, b='foo')
3 foo
>>> my_ob._saved_method.args
>>> my_ob._saved_method.kwargs == dict(a=3, b='foo')

The arguments are stored on the instance, allowing for different instance to save different args.

>>> your_ob = MyClass()
>>> your_ob.method({str('x'): 3}, b=[4])
{'x': 3} [4]
>>> your_ob._saved_method.args
({'x': 3},)
>>> my_ob._saved_method.args

Wrap func to expect its parameters to be passed positionally in a tuple.

Has a similar effect to that of itertools.starmap over simple map.

>>> pairs = [(-1, 1), (0, 2)]
>>> more_itertools.consume(itertools.starmap(print, pairs))
-1 1
0 2
>>> more_itertools.consume(map(splat(print), pairs))
-1 1
0 2

The approach generalizes to other iterators that don’t have a “star” equivalent, such as a “starfilter”.

>>> list(filter(splat(operator.add), pairs))
[(0, 2)]

Splat also accepts a mapping argument.

>>> def is_nice(msg, code):
...     return "smile" in msg or code == 0
>>> msgs = [
...     dict(msg='smile!', code=20),
...     dict(msg='error :(', code=1),
...     dict(msg='unknown', code=0),
... ]
>>> for msg in filter(splat(is_nice), msgs):
...     print(msg)
{'msg': 'smile!', 'code': 20}
{'msg': 'unknown', 'code': 0}

Indices and tables#