更改生成器的__name__

Question

鉴于以下设置:

def mapper(f):
    def wrapper(items):
        for x in items:
            yield f(x)
    wrapper.__name__ = f.__name__ # This has no effect!
    return wrapper

@mapper
def double(x):
    return x * 2

装饰器按预期工作:

>>> [x for x in double([1,2,3])]
[2, 4, 6]

然而,它__name__不是double所希望的:

>>> double([1,2]).__name__
"wrapper"

是否可以强制生成器的名称？或者,是否可以在生成器对象中挖掘并检索名称double？

Answer 1

您只能将名称复制到该功能.生成的生成器对象仍然使用在编译时设置的旧名称.

每次生成新的生成器对象时都必须设置该名称; 遗憾的是,该属性对于生成器是只读的:

>>> def gen():
...     yield None
... 
>>> gen().__name__
'gen'
>>> gen().__name__ = 'foo'
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: attribute '__name__' of 'generator' objects is not writable

该名称被烘焙到代码对象中:

>>> gen.__code__.co_name
'gen'

所以你可以改变它的唯一方法是重新构建代码对象(以及扩展,函数):

def rename_code_object(func, new_name):
    code_object = func.__code__
    function, code = type(func), type(code_object)
    return function(
        code(
            code_object.co_argcount, code_object.co_nlocals,
            code_object.co_stacksize, code_object.co_flags,
            code_object.co_code, code_object.co_consts,
            code_object.co_names, code_object.co_varnames,
            code_object.co_filename, new_name,
            code_object.co_firstlineno, code_object.co_lnotab,
            code_object.co_freevars, code_object.co_cellvars),
        func.__globals__, new_name, func.__defaults__, func.__closure__)

这将创建一个新函数和代码对象,new_name用于替换旧名称:

>>> new_name = rename_code_object(gen, 'new_name')
>>> new_name.__name__
'new_name'
>>> new_name().__name__
'new_name'
>>> new_name()
<generator object new_name at 0x10075f280>

在装饰器中使用它:

def mapper(f):
    def wrapper(items):
        for x in items:
            yield f(x)
    return rename_code_object(wrapper, f.__name__)

演示:

>>> def mapper(f):
...     def wrapper(items):
...         for x in items:
...             yield f(x)
...     return rename_code_object(wrapper, f.__name__)
... 
>>> @mapper
... def double(x):
...     return x * 2
... 
>>> double([1, 2])
<generator object double at 0x10075f370>
>>> list(double([1, 2]))
[2, 4]

从Python 3.5开始,生成器对象__name__从函数对象而不是从它们的代码对象co_name属性中获取它们,参见问题#21205 ; 该属性在此过程中变得可写.