有没有办法“解决”装饰器的保留状态？

Question

发生了什么事？...

所以，这就是我的情况。我正在详细说明网络抓取，因此在此脚本的某个时刻，我决定使用装饰器来处理某些 URL。这个装饰器有一个参数（URL），它应该在 for 循环语句的帮助下动态变化，如示例脚本所示：

from functools import wraps
import logging

logging.basicConfig(level=logging.INFO)

def cycle(url):
    def outer_wrapper(func):
        state = 0
        @wraps(func)
        def inner_wrapper(**kwargs):
            nonlocal state
            state += 1
            kwargs['url'] = url
            if state == 1:
                logging.info('Returning result at first execution on {} with: '
                             'state => {}, kwargs => {}'.format(func, state, kwargs))
                return func(**kwargs)
            else:
                logging.info('Returning result at upcoming executions on {} with: '
                             'state => {}, kwargs => {}'.format(func, state, kwargs))
                return func(**kwargs)

        return inner_wrapper
    return outer_wrapper


def print_url(url):
    print('Returned from print_url function:', url)

links = ['an-url', 'another-url']

for link in links:
    # Decorator
    print_url = cycle(link)(print_url)
    print_url()

但是，输出显示出意想不到的结果，至少对我来说是这样。

INFO:root:Returning result at first execution on <function print_url at 0x000002202FD68D08> with: state => 1, kwargs => {'url': 'an-url'}
Returned from print_url function: an-url
INFO:root:Returning result at first execution on <function print_url at 0x000002202FE196A8> with: state => 1, kwargs => {'url': 'another-url'}
INFO:root:Returning result at upcoming executions on <function print_url at 0x000002202FD68D08> with: state => 2, kwargs => {'url': 'an-url'}
Returned from print_url function: an-url

装饰器保留第一次调用print_url() 函数的地址和参数。 我已经阅读了很多关于垃圾收集、弱引用和标准库的functools 的文章，但我不知道如何“重新启动”这个装饰器，以便在迭代过程中接收一个新参数。

如果可能的话，有人知道如何解决这个问题吗？

Answer 1

发生这种情况是因为您正在重新装饰已装饰的函数。

第一次迭代后：

for link in links:
    # Decorator
    print_url = cycle(link)(print_url)
    print_url()

那么print_url 指的是inner_wrapper。你再次装饰inner_wrapper。这与垃圾收集无关，这只是您编写此代码的目的。

如果你删除wraps，这会变得更清楚：

def cycle(url):
    def outer_wrapper(func):
        state = 0
        def inner_wrapper(**kwargs):
            nonlocal state
            state += 1
            kwargs['url'] = url
            if state == 1:
                print('Returning result at first execution on {} with: '
                             'state => {}, kwargs => {}'.format(func, state, kwargs))
                return func(**kwargs)
            else:
                print('Returning result at upcoming executions on {} with: '
                             'state => {}, kwargs => {}'.format(func, state, kwargs))
                return func(**kwargs)

        return inner_wrapper
    return outer_wrapper


def print_url(url):
    print('Returned from print_url function:', url)

links = ['an-url', 'another-url']

for i, link in enumerate(links):
    print("Iteration :", i)
    print_url = cycle(link)(print_url)
    print_url()

以及终端输出：

Iteration : 0
Returning result at first execution on <function print_url at 0x1060892f0> with: state => 1, kwargs => {'url': 'an-url'}
Returned from print_url function: an-url
Iteration : 1
Returning result at first execution on <function cycle.<locals>.outer_wrapper.<locals>.inner_wrapper at 0x106089378> with: state => 1, kwargs => {'url': 'another-url'}
Returning result at upcoming executions on <function print_url at 0x1060892f0> with: state => 2, kwargs => {'url': 'an-url'}
Returned from print_url function: an-url

如果你想访问原始函数，我建议只在循环之外保留对它的引用，并且不要将装饰器的结果分配给该变量，例如：

original_function = print_url
for link in links:
    # Decorator
    print_url = cycle(link)(original_function)
    print_url()
    print_url = original_function

只是为了好玩，你可以在每次迭代中恢复它，做一些事情：

for link in links:
    # Decorator
    print_url = cycle(link)(print_url)
    print_url()
    closure = print_url.__closure__
    idx_func = print_url.__code__.co_freevars.index('func')
    print_url = closure[idx_func].cell_contents

但是......这只是一个可怕的混乱，暴露了一堆内部细节，最好留在引擎盖下。

从根本上说，我不知道为什么这里需要一个装饰器，也就是说，使用这个装饰器有什么好处？