"functools" --- Higher-order functions and operations on callable objects
*************************************************************************

**Code source :** Lib/functools.py

======================================================================

Le module "functools" concerne les fonctions d'ordre supérieur : des
fonctions qui agissent sur, ou renvoient, d'autres fonctions. En
général, tout objet appelable peut être considéré comme une fonction
dans la description de ce module.

Le module "functools" définit les fonctions suivantes :

@functools.cache(user_function)

   Fonction de cache très simple et sans limite de taille. Cette
   technique est parfois appelée « mémoïsation ».

   Returns the same as "lru_cache(maxsize=None)", creating a thin
   wrapper around a dictionary lookup for the function arguments.
   Because it never needs to evict old values, this is smaller and
   faster than "lru_cache()" with a size limit.

   Par exemple :

      @cache
      def factorial(n):
          return n * factorial(n-1) if n else 1

      >>> factorial(10)   # no previously cached result, makes 11 recursive calls
      3628800
      >>> factorial(5)    # no new calls, just returns the cached result
      120
      >>> factorial(12)   # two new recursive calls, factorial(10) is cached
      479001600

   The cache is threadsafe so that the wrapped function can be used in
   multiple threads.  This means that the underlying data structure
   will remain coherent during concurrent updates.

   It is possible for the wrapped function to be called more than once
   if another thread makes an additional call before the initial call
   has been completed and cached.

   Ajouté dans la version 3.9.

@functools.cached_property(func)

   Transform a method of a class into a property whose value is
   computed once and then cached as a normal attribute for the life of
   the instance. Similar to "property()", with the addition of
   caching. Useful for expensive computed properties of instances that
   are otherwise effectively immutable.

   Exemple :

      class DataSet:

          def __init__(self, sequence_of_numbers):
              self._data = tuple(sequence_of_numbers)

          @cached_property
          def stdev(self):
              return statistics.stdev(self._data)

   The mechanics of "cached_property()" are somewhat different from
   "property()".  A regular property blocks attribute writes unless a
   setter is defined. In contrast, a *cached_property* allows writes.

   The *cached_property* decorator only runs on lookups and only when
   an attribute of the same name doesn't exist.  When it does run, the
   *cached_property* writes to the attribute with the same name.
   Subsequent attribute reads and writes take precedence over the
   *cached_property* method and it works like a normal attribute.

   The cached value can be cleared by deleting the attribute.  This
   allows the *cached_property* method to run again.

   The *cached_property* does not prevent a possible race condition in
   multi-threaded usage. The getter function could run more than once
   on the same instance, with the latest run setting the cached value.
   If the cached property is idempotent or otherwise not harmful to
   run more than once on an instance, this is fine. If synchronization
   is needed, implement the necessary locking inside the decorated
   getter function or around the cached property access.

   Note, this decorator interferes with the operation of **PEP 412**
   key-sharing dictionaries.  This means that instance dictionaries
   can take more space than usual.

   Also, this decorator requires that the "__dict__" attribute on each
   instance be a mutable mapping. This means it will not work with
   some types, such as metaclasses (since the "__dict__" attributes on
   type instances are read-only proxies for the class namespace), and
   those that specify "__slots__" without including "__dict__" as one
   of the defined slots (as such classes don't provide a "__dict__"
   attribute at all).

   If a mutable mapping is not available or if space-efficient key
   sharing is desired, an effect similar to "cached_property()" can
   also be achieved by stacking "property()" on top of "lru_cache()".
   See Comment mettre en cache le résultat d'une méthode ? for more
   details on how this differs from "cached_property()".

   Ajouté dans la version 3.8.

   Modifié dans la version 3.12: Prior to Python 3.12,
   "cached_property" included an undocumented lock to ensure that in
   multi-threaded usage the getter function was guaranteed to run only
   once per instance. However, the lock was per-property, not per-
   instance, which could result in unacceptably high lock contention.
   In Python 3.12+ this locking is removed.

functools.cmp_to_key(func)

   Transforme une fonction de comparaison à l'ancienne en une
   *fonction clé*.  Utilisé avec des outils qui acceptent des
   fonctions clef (comme "sorted()", "min()", "max()",
   "heapq.nlargest()", "heapq.nsmallest()", "itertools.groupby()").
   Cette fonction est destinée au portage de fonctions python 2
   utilisant des fonctions de comparaison vers Python 3.

   A comparison function is any callable that accepts two arguments,
   compares them, and returns a negative number for less-than, zero
   for equality, or a positive number for greater-than.  A key
   function is a callable that accepts one argument and returns
   another value to be used as the sort key.

   Exemple :

      sorted(iterable, key=cmp_to_key(locale.strcoll))  # locale-aware sort order

   Pour des exemples de tris et un bref tutoriel, consultez Sorting
   Techniques.

   Ajouté dans la version 3.2.

@functools.lru_cache(user_function)
@functools.lru_cache(maxsize=128, typed=False)

   Décorateur qui englobe une fonction avec un appelable mémoïsant qui
   enregistre jusqu'à *maxsize* appels récents. Cela peut gagner du
   temps quand une fonction coûteuse en ressources est souvent appelée
   avec les mêmes arguments.

   The cache is threadsafe so that the wrapped function can be used in
   multiple threads.  This means that the underlying data structure
   will remain coherent during concurrent updates.

   It is possible for the wrapped function to be called more than once
   if another thread makes an additional call before the initial call
   has been completed and cached.

   Since a dictionary is used to cache results, the positional and
   keyword arguments to the function must be *hashable*.

   Distinct argument patterns may be considered to be distinct calls
   with separate cache entries.  For example, "f(a=1, b=2)" and
   "f(b=2, a=1)" differ in their keyword argument order and may have
   two separate cache entries.

   Si *user_function* est défini, ce doit être un appelable. Ceci
   permet à *lru_cache* d'être appliqué directement sur une fonction
   de l'utilisateur, sans préciser *maxsize* (qui est alors défini à
   sa valeur par défaut, 128) :

      @lru_cache
      def count_vowels(sentence):
          return sum(sentence.count(vowel) for vowel in 'AEIOUaeiou')

   Si *maxsize* est à "None", la fonctionnalité LRU est désactivée et
   le cache peut grossir sans limite.

   If *typed* is set to true, function arguments of different types
   will be cached separately.  If *typed* is false, the implementation
   will usually regard them as equivalent calls and only cache a
   single result. (Some types such as *str* and *int* may be cached
   separately even when *typed* is false.)

   Note, type specificity applies only to the function's immediate
   arguments rather than their contents.  The scalar arguments,
   "Decimal(42)" and "Fraction(42)" are be treated as distinct calls
   with distinct results. In contrast, the tuple arguments "('answer',
   Decimal(42))" and "('answer', Fraction(42))" are treated as
   equivalent.

   The wrapped function is instrumented with a "cache_parameters()"
   function that returns a new "dict" showing the values for *maxsize*
   and *typed*.  This is for information purposes only.  Mutating the
   values has no effect.

   To help measure the effectiveness of the cache and tune the
   *maxsize* parameter, the wrapped function is instrumented with a
   "cache_info()" function that returns a *named tuple* showing
   *hits*, *misses*, *maxsize* and *currsize*.

   The decorator also provides a "cache_clear()" function for clearing
   or invalidating the cache.

   La fonction sous-jacente originale est accessible à travers
   l'attribut "__wrapped__".  Ceci est utile pour l'introspection,
   pour outrepasser le cache, ou pour ré-englober la fonction avec un
   cache différent.

   The cache keeps references to the arguments and return values until
   they age out of the cache or until the cache is cleared.

   If a method is cached, the "self" instance argument is included in
   the cache.  See Comment mettre en cache le résultat d'une méthode ?

   An LRU (least recently used) cache works best when the most recent
   calls are the best predictors of upcoming calls (for example, the
   most popular articles on a news server tend to change each day).
   The cache's size limit assures that the cache does not grow without
   bound on long-running processes such as web servers.

   In general, the LRU cache should only be used when you want to
   reuse previously computed values.  Accordingly, it doesn't make
   sense to cache functions with side-effects, functions that need to
   create distinct mutable objects on each call (such as generators
   and async functions), or impure functions such as time() or
   random().

   Exemple d'un cache LRU pour du contenu web statique :

      @lru_cache(maxsize=32)
      def get_pep(num):
          'Retrieve text of a Python Enhancement Proposal'
          resource = f'https://peps.python.org/pep-{num:04d}'
          try:
              with urllib.request.urlopen(resource) as s:
                  return s.read()
          except urllib.error.HTTPError:
              return 'Not Found'

      >>> for n in 8, 290, 308, 320, 8, 218, 320, 279, 289, 320, 9991:
      ...     pep = get_pep(n)
      ...     print(n, len(pep))

      >>> get_pep.cache_info()
      CacheInfo(hits=3, misses=8, maxsize=32, currsize=8)

   Exemple de calcul efficace de la suite de Fibonacci en utilisant un
   cache pour implémenter la technique de programmation dynamique :

      @lru_cache(maxsize=None)
      def fib(n):
          if n < 2:
              return n
          return fib(n-1) + fib(n-2)

      >>> [fib(n) for n in range(16)]
      [0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144, 233, 377, 610]

      >>> fib.cache_info()
      CacheInfo(hits=28, misses=16, maxsize=None, currsize=16)

   Ajouté dans la version 3.2.

   Modifié dans la version 3.3: L'option *typed* a été ajoutée.

   Modifié dans la version 3.8: Ajout de l'option *user_function*.

   Modifié dans la version 3.9: Added the function
   "cache_parameters()"

@functools.total_ordering

   A partir d'une classe définissant une ou plusieurs méthodes de
   comparaison riches, ce décorateur de classe fournit le reste.  Ceci
   simplifie l'effort à fournir dans la spécification de toutes les
   opérations de comparaison riche :

   The class must define one of "__lt__()", "__le__()", "__gt__()", or
   "__ge__()". In addition, the class should supply an "__eq__()"
   method.

   Par exemple :

      @total_ordering
      class Student:
          def _is_valid_operand(self, other):
              return (hasattr(other, "lastname") and
                      hasattr(other, "firstname"))
          def __eq__(self, other):
              if not self._is_valid_operand(other):
                  return NotImplemented
              return ((self.lastname.lower(), self.firstname.lower()) ==
                      (other.lastname.lower(), other.firstname.lower()))
          def __lt__(self, other):
              if not self._is_valid_operand(other):
                  return NotImplemented
              return ((self.lastname.lower(), self.firstname.lower()) <
                      (other.lastname.lower(), other.firstname.lower()))

   Note:

     Même si ce décorateur permet de créer des types ordonnables
     facilement, cela vient avec un *coût* d'exécution et des traces
     d'exécution complexes pour les méthodes de comparaison dérivées.
     Si des tests de performances le révèlent comme un goulot
     d'étranglement, l'implémentation manuelle des six méthodes de
     comparaison riches résoudra normalement vos problèmes de
     rapidité.

   Note:

     This decorator makes no attempt to override methods that have
     been declared in the class *or its superclasses*. Meaning that if
     a superclass defines a comparison operator, *total_ordering* will
     not implement it again, even if the original method is abstract.

   Ajouté dans la version 3.2.

   Modifié dans la version 3.4: Returning "NotImplemented" from the
   underlying comparison function for unrecognised types is now
   supported.

functools.partial(func, /, *args, **keywords)

   Retourne un nouvel objet partiel qui, quand il est appelé,
   fonctionne comme *func* appelée avec les arguments positionnels
   *args* et les arguments nommés *keywords*. Si plus d'arguments sont
   fournis à l'appel, ils sont ajoutés à *args*. Si plus d'arguments
   nommés sont fournis, ils étendent et surchargent *keywords*. À peu
   près équivalent à :

      def partial(func, /, *args, **keywords):
          def newfunc(*fargs, **fkeywords):
              newkeywords = {**keywords, **fkeywords}
              return func(*args, *fargs, **newkeywords)
          newfunc.func = func
          newfunc.args = args
          newfunc.keywords = keywords
          return newfunc

   "partial()" est utilisé pour une application de fonction partielle
   qui "gèle" une portion des arguments et/ou mots-clés d'une fonction
   donnant un nouvel objet avec une signature simplifiée.  Par
   exemple, "partial()" peut être utilisé pour créer un appelable qui
   se comporte comme la fonction "int()" ou l'argument *base* est deux
   par défaut :

   >>> from functools import partial
   >>> basetwo = partial(int, base=2)
   >>> basetwo.__doc__ = 'Convert base 2 string to an int.'
   >>> basetwo('10010')
   18

class functools.partialmethod(func, /, *args, **keywords)

   Retourne un nouveau descripteur "partialmethod" qui se comporte
   comme "partial" sauf qu'il est fait pour être utilisé comme une
   définition de méthode plutôt que d'être appelé directement.

   *func* doit être un *descriptor* ou un appelable (les objets qui
   sont les deux, comme les fonction normales, sont gérés comme des
   descripteurs).

   When *func* is a descriptor (such as a normal Python function,
   "classmethod()", "staticmethod()", "abstractmethod()" or another
   instance of "partialmethod"), calls to "__get__" are delegated to
   the underlying descriptor, and an appropriate partial object
   returned as the result.

   Quand *func* est un appelable non-descripteur, une méthode liée
   appropriée est crée dynamiquement. Elle se comporte comme une
   fonction Python normale quand elle est utilisée comme méthode :
   l'argument *self* sera inséré comme premier argument positionnel,
   avant les *args* et *keywords* fournis au constructeur
   "partialmethod".

   Exemple :

      >>> class Cell:
      ...     def __init__(self):
      ...         self._alive = False
      ...     @property
      ...     def alive(self):
      ...         return self._alive
      ...     def set_state(self, state):
      ...         self._alive = bool(state)
      ...     set_alive = partialmethod(set_state, True)
      ...     set_dead = partialmethod(set_state, False)
      ...
      >>> c = Cell()
      >>> c.alive
      False
      >>> c.set_alive()
      >>> c.alive
      True

   Ajouté dans la version 3.4.

functools.reduce(function, iterable, [initial, ]/)

   Apply *function* of two arguments cumulatively to the items of
   *iterable*, from left to right, so as to reduce the iterable to a
   single value.  For example, "reduce(lambda x, y: x+y, [1, 2, 3, 4,
   5])" calculates "((((1+2)+3)+4)+5)". The left argument, *x*, is the
   accumulated value and the right argument, *y*, is the update value
   from the *iterable*.  If the optional *initial* is present, it is
   placed before the items of the iterable in the calculation, and
   serves as a default when the iterable is empty.  If *initial* is
   not given and *iterable* contains only one item, the first item is
   returned.

   À peu près équivalent à :

      initial_missing = object()

      def reduce(function, iterable, initial=initial_missing, /):
          it = iter(iterable)
          if initial is initial_missing:
              value = next(it)
          else:
              value = initial
          for element in it:
              value = function(value, element)
          return value

   Voir "itertools.accumulate()" pour un itérateur qui génère toutes
   les valeurs intermédiaires.

@functools.singledispatch

   Transforme une fonction en une *fonction générique* *single-
   dispatch*.

   To define a generic function, decorate it with the
   "@singledispatch" decorator. When defining a function using
   "@singledispatch", note that the dispatch happens on the type of
   the first argument:

      >>> from functools import singledispatch
      >>> @singledispatch
      ... def fun(arg, verbose=False):
      ...     if verbose:
      ...         print("Let me just say,", end=" ")
      ...     print(arg)

   To add overloaded implementations to the function, use the
   "register()" attribute of the generic function, which can be used
   as a decorator.  For functions annotated with types, the decorator
   will infer the type of the first argument automatically:

      >>> @fun.register
      ... def _(arg: int, verbose=False):
      ...     if verbose:
      ...         print("Strength in numbers, eh?", end=" ")
      ...     print(arg)
      ...
      >>> @fun.register
      ... def _(arg: list, verbose=False):
      ...     if verbose:
      ...         print("Enumerate this:")
      ...     for i, elem in enumerate(arg):
      ...         print(i, elem)

   "types.UnionType" and "typing.Union" can also be used:

      >>> @fun.register
      ... def _(arg: int | float, verbose=False):
      ...     if verbose:
      ...         print("Strength in numbers, eh?", end=" ")
      ...     print(arg)
      ...
      >>> from typing import Union
      >>> @fun.register
      ... def _(arg: Union[list, set], verbose=False):
      ...     if verbose:
      ...         print("Enumerate this:")
      ...     for i, elem in enumerate(arg):
      ...         print(i, elem)
      ...

   Pour le code qui n’utilise pas les indications de type, le type
   souhaité peut être passé explicitement en argument au décorateur :

      >>> @fun.register(complex)
      ... def _(arg, verbose=False):
      ...     if verbose:
      ...         print("Better than complicated.", end=" ")
      ...     print(arg.real, arg.imag)
      ...

   For code that dispatches on a collections type (e.g., "list"), but
   wants to typehint the items of the collection (e.g., "list[int]"),
   the dispatch type should be passed explicitly to the decorator
   itself with the typehint going into the function definition:

      >>> @fun.register(list)
      ... def _(arg: list[int], verbose=False):
      ...     if verbose:
      ...         print("Enumerate this:")
      ...     for i, elem in enumerate(arg):
      ...         print(i, elem)

   Note:

     At runtime the function will dispatch on an instance of a list
     regardless of the type contained within the list i.e. "[1,2,3]"
     will be dispatched the same as "["foo", "bar", "baz"]". The
     annotation provided in this example is for static type checkers
     only and has no runtime impact.

   To enable registering *lambdas* and pre-existing functions, the
   "register()" attribute can also be used in a functional form:

      >>> def nothing(arg, verbose=False):
      ...     print("Nothing.")
      ...
      >>> fun.register(type(None), nothing)

   The "register()" attribute returns the undecorated function. This
   enables decorator stacking, "pickling", and the creation of unit
   tests for each variant independently:

      >>> @fun.register(float)
      ... @fun.register(Decimal)
      ... def fun_num(arg, verbose=False):
      ...     if verbose:
      ...         print("Half of your number:", end=" ")
      ...     print(arg / 2)
      ...
      >>> fun_num is fun
      False

   Quand elle est appelée, la fonction générique distribue sur le type
   du premier argument :

      >>> fun("Hello, world.")
      Hello, world.
      >>> fun("test.", verbose=True)
      Let me just say, test.
      >>> fun(42, verbose=True)
      Strength in numbers, eh? 42
      >>> fun(['spam', 'spam', 'eggs', 'spam'], verbose=True)
      Enumerate this:
      0 spam
      1 spam
      2 eggs
      3 spam
      >>> fun(None)
      Nothing.
      >>> fun(1.23)
      0.615

   Where there is no registered implementation for a specific type,
   its method resolution order is used to find a more generic
   implementation. The original function decorated with
   "@singledispatch" is registered for the base "object" type, which
   means it is used if no better implementation is found.

   If an implementation is registered to an *abstract base class*,
   virtual subclasses of the base class will be dispatched to that
   implementation:

      >>> from collections.abc import Mapping
      >>> @fun.register
      ... def _(arg: Mapping, verbose=False):
      ...     if verbose:
      ...         print("Keys & Values")
      ...     for key, value in arg.items():
      ...         print(key, "=>", value)
      ...
      >>> fun({"a": "b"})
      a => b

   To check which implementation the generic function will choose for
   a given type, use the "dispatch()" attribute:

      >>> fun.dispatch(float)
      <function fun_num at 0x1035a2840>
      >>> fun.dispatch(dict)    # note: default implementation
      <function fun at 0x103fe0000>

   Pour accéder à toutes les implémentations enregistrées, utiliser
   l'attribut en lecture seule "registry" :

      >>> fun.registry.keys()
      dict_keys([<class 'NoneType'>, <class 'int'>, <class 'object'>,
                <class 'decimal.Decimal'>, <class 'list'>,
                <class 'float'>])
      >>> fun.registry[float]
      <function fun_num at 0x1035a2840>
      >>> fun.registry[object]
      <function fun at 0x103fe0000>

   Ajouté dans la version 3.4.

   Modifié dans la version 3.7: The "register()" attribute now
   supports using type annotations.

   Modifié dans la version 3.11: The "register()" attribute now
   supports "types.UnionType" and "typing.Union" as type annotations.

class functools.singledispatchmethod(func)

   Transforme une méthode en une *fonction générique* *single-
   dispatch*.

   To define a generic method, decorate it with the
   "@singledispatchmethod" decorator. When defining a method using
   "@singledispatchmethod", note that the dispatch happens on the type
   of the first non-*self* or non-*cls* argument:

      class Negator:
          @singledispatchmethod
          def neg(self, arg):
              raise NotImplementedError("Cannot negate a")

          @neg.register
          def _(self, arg: int):
              return -arg

          @neg.register
          def _(self, arg: bool):
              return not arg

   "@singledispatchmethod" supports nesting with other decorators such
   as "@classmethod". Note that to allow for "dispatcher.register",
   "singledispatchmethod" must be the *outer most* decorator. Here is
   the "Negator" class with the "neg" methods bound to the class,
   rather than an instance of the class:

      class Negator:
          @singledispatchmethod
          @classmethod
          def neg(cls, arg):
              raise NotImplementedError("Cannot negate a")

          @neg.register
          @classmethod
          def _(cls, arg: int):
              return -arg

          @neg.register
          @classmethod
          def _(cls, arg: bool):
              return not arg

   The same pattern can be used for other similar decorators:
   "@staticmethod", "@~abc.abstractmethod", and others.

   Ajouté dans la version 3.8.

functools.update_wrapper(wrapper, wrapped, assigned=WRAPPER_ASSIGNMENTS, updated=WRAPPER_UPDATES)

   Update a *wrapper* function to look like the *wrapped* function.
   The optional arguments are tuples to specify which attributes of
   the original function are assigned directly to the matching
   attributes on the wrapper function and which attributes of the
   wrapper function are updated with the corresponding attributes from
   the original function. The default values for these arguments are
   the module level constants "WRAPPER_ASSIGNMENTS" (which assigns to
   the wrapper function's "__module__", "__name__", "__qualname__",
   "__annotations__", "__type_params__", and "__doc__", the
   documentation string) and "WRAPPER_UPDATES" (which updates the
   wrapper function's "__dict__", i.e. the instance dictionary).

   Pour autoriser l'accès à la fonction originale pour l'introspection
   ou à d'autres fins (par ex. outrepasser l'accès à un décorateur de
   cache comme "lru_cache()"), cette fonction ajoute automatiquement
   un attribut "__wrapped__" qui référence la fonction englobée.

   La principale utilisation de cette fonction est dans les
   *décorateurs* qui renvoient une nouvelle fonction. Si la fonction
   crée n'est pas mise à jour, ses métadonnées refléteront sa
   définition dans le décorateur, au lieu de la définition originale,
   métadonnées souvent bien moins utiles.

   "update_wrapper()" peut être utilisé avec des appelables autres que
   des fonctions. Tout attribut défini dans *assigned* ou *updated*
   qui ne sont pas l'objet englobé sont ignorés (cette fonction
   n'essaiera pas de les définir dans la fonction englobante).
   "AttributeError" est toujours levée si le fonction englobante elle
   même a des attributs non existants dans *updated*.

   Modifié dans la version 3.2: The "__wrapped__" attribute is now
   automatically added. The "__annotations__" attribute is now copied
   by default. Missing attributes no longer trigger an
   "AttributeError".

   Modifié dans la version 3.4: L'attribut "__wrapped__" renvoie
   toujours la fonction englobée, même si cette fonction définit un
   attribut "__wrapped__". (voir bpo-17482)

   Modifié dans la version 3.12: The "__type_params__" attribute is
   now copied by default.

@functools.wraps(wrapped, assigned=WRAPPER_ASSIGNMENTS, updated=WRAPPER_UPDATES)

   Ceci est une fonction d'aide pour appeler "update_wrapper()"  comme
   décorateur de fonction lors de la définition d'une fonction
   englobante.  C'est équivalent à  "partial(update_wrapper,
   wrapped=wrapped, assigned=assigned, updated=updated)". Par exemple
   :

      >>> from functools import wraps
      >>> def my_decorator(f):
      ...     @wraps(f)
      ...     def wrapper(*args, **kwds):
      ...         print('Calling decorated function')
      ...         return f(*args, **kwds)
      ...     return wrapper
      ...
      >>> @my_decorator
      ... def example():
      ...     """Docstring"""
      ...     print('Called example function')
      ...
      >>> example()
      Calling decorated function
      Called example function
      >>> example.__name__
      'example'
      >>> example.__doc__
      'Docstring'

   Without the use of this decorator factory, the name of the example
   function would have been "'wrapper'", and the docstring of the
   original "example()" would have been lost.


Objets "partial"
================

Les objets "partial" sont des objets appelables créés par "partial()".
Ils ont trois attributs en lecture seule :

partial.func

   Un objet ou une fonction appelable.  Les appels à l'objet "partial"
   seront transmis à "func" avec les nouveaux arguments et mots-clés.

partial.args

   Les arguments positionnels qui seront ajoutés avant les arguments
   fournis lors de l'appel d'un objet "partial".

partial.keywords

   Les arguments nommés qui seront fournis quand l'objet "partial" est
   appelé.

"partial" objects are like function objects in that they are callable,
weak referenceable, and can have attributes. There are some important
differences.  For instance, the "__name__" and "function.__doc__"
attributes are not created automatically.  Also, "partial" objects
defined in classes behave like static methods and do not transform
into bound methods during instance attribute look-up.
