可以做些什么来加速这个 memoization 装饰器？

Question

我想要的是一个 memoization 装饰器：

• 可以使用参数和关键字参数来记忆实例方法
• 有一个可以通过一次调用（全局）清除的缓存（与使用每个函数缓存的这个相比：python resettable instance method memoization decorator）
• 相当有效

我已经调整了我看到的一个示例并提出了以下内容：

import functools

class Memoized(object):
  """Decorator that caches a function's return value each time it is called.
  If called later with the same arguments, the cached value is returned, and
  not re-evaluated.
  """

  __cache = {}

  def __init__(self, func):
    self.func = func
    self.key = (func.__module__, func.__name__)

  def __call__(self, *args):
    try:
      return Memoized.__cache[self.key][args]
    except KeyError:
      value = self.func(*args)
      Memoized.__cache[self.key] = {args : value}
      return value
    except TypeError:
      # uncachable -- for instance, passing a list as an argument.
      # Better to not cache than to blow up entirely.
      return self.func(*args)

  def __get__(self, obj, objtype):
    """Support instance methods."""
    return functools.partial(self.__call__, obj)

  @staticmethod
  def reset():
    Memoized.__cache = {}

我的问题是缓存部分似乎涉及很多开销（例如，对于递归函数）。以下面的函数为例，我可以在非记忆版本比记忆版本更短的时间内调用 fib(30) 十次。

def fib(n):

   if n in (0, 1):
      return n
   return fib(n-1) + fib(n-2)

谁能推荐一个更好的方法来编写这个装饰器？ （或者给我一个更好（即更快）的装饰器来做我想要的）。 我对保留方法签名或帮助自省工具“了解”装饰函数的任何内容不感兴趣。

谢谢。

附：使用python 2.7

Answer 1

这是一个明显更快的版本。不幸的是，reset 不能再完全清除缓存，因为所有实例都在存储它们自己的对每个函数字典的引用的本地副本。虽然你可以让它工作：

import functools

class Memoized(object):
  """Decorator that caches a function's return value each time it is called.
  If called later with the same arguments, the cached value is returned, and
  not re-evaluated.
  """

  __cache = {}

  def __init__(self, func):
    self.func = func
    key = (func.__module__, func.__name__)
    if key not in self.__cache:
      self.__cache[key] = {}
    self.mycache = self.__cache[key]

  def __call__(self, *args):
    try:
      return self.mycache[args]
    except KeyError:
      value = self.func(*args)
      self.mycache[args] = value
      return value
    except TypeError:
      # uncachable -- for instance, passing a list as an argument.
      # Better to not cache than to blow up entirely.
      return self.func(*args)

  def __get__(self, obj, objtype):
    """Support instance methods."""
    return functools.partial(self.__call__, obj)

  @classmethod
  def reset(cls):
    for v in cls.__cache.itervalues():
      v.clear()

Answer 2

您实际上并没有缓存任何数据，因为每次设置新的缓存值时都会覆盖以前的值：

Memoized.__cache[self.key] = {args : value}

例如。

import functools

class Memoized(object):
    """Decorator that caches a function's return value each time it is called.
    If called later with the same arguments, the cached value is returned, and
    not re-evaluated.
    """

    cache = {}

    def __init__(self, func):
        self.func = func
        self.key = (func.__module__, func.__name__)
        self.cache[self.key] = {}

    def __call__(self, *args):
      try:
          return Memoized.cache[self.key][args]
      except KeyError:
          value = self.func(*args)
          Memoized.cache[self.key][args] = value
          return value
      except TypeError:
          # uncachable -- for instance, passing a list as an argument.
          # Better to not cache than to blow up entirely.
          return self.func(*args)

    def __get__(self, obj, objtype):
        """Support instance methods."""
        return functools.partial(self.__call__, obj)

    @staticmethod
    def reset():
        Memoized.cache = {}

• fib(30) 没有缓存：2.86742401123
• 带有缓存的fib(30)：0.000198125839233

其他一些注意事项：

• 不要使用__prefix；没有理由在这里，它只会丑化代码。
• 不要使用单一的、整体的、类属性字典，而是为Memoized 的每个实例提供自己的字典，并保留Memoized 对象的注册表。这改进了封装性，并消除了依赖模块和函数名称的奇怪之处。

.

import functools
import weakref

class Memoized(object):
    """Decorator that caches a function's return value each time it is called.
    If called later with the same arguments, the cached value is returned, and
    not re-evaluated.

    >>> counter = 0
    >>> @Memoized
    ... def count():
    ...     global counter
    ...     counter += 1
    ...     return counter

    >>> counter = 0
    >>> Memoized.reset()
    >>> count()
    1
    >>> count()
    1
    >>> Memoized.reset()
    >>> count()
    2

    >>> class test(object):
    ...     @Memoized
    ...     def func(self):
    ...         global counter
    ...         counter += 1
    ...         return counter
    >>> testobject = test()
    >>> counter = 0
    >>> testobject.func()
    1
    >>> testobject.func()
    1
    >>> Memoized.reset()
    >>> testobject.func()
    2
    """

    caches = weakref.WeakSet()

    def __init__(self, func):
        self.func = func
        self.cache = {}
        Memoized.caches.add(self)

    def __call__(self, *args):
      try:
          return self.cache[args]
      except KeyError:
          value = self.func(*args)
          self.cache[args] = value
          return value
      except TypeError:
          # uncachable -- for instance, passing a list as an argument.
          # Better to not cache than to blow up entirely.
          return self.func(*args)

    def __get__(self, obj, objtype):
        """Support instance methods."""
        return functools.partial(self.__call__, obj)

    @staticmethod
    def reset():
        for memo in Memoized.caches:
            memo.cache = {}

if __name__ == '__main__':
    import doctest
    doctest.testmod()

编辑：添加测试，并使用weakref.WeakSet。请注意，WeakSet 仅在 2.7（OP 正在使用）中可用；对于适用于 2.6 的版本，请参阅编辑历史记录。