带有@ syntax的python decorator参数

Question

I'm trying to use a cached property decorator that can take arguments.

I looked at this implementation: http://www.daniweb.com/software-development/python/code/217241/a-cached-property-decorator

from functools import update_wrapper 

def cachedProperty (func ,name =None ):
  if name is None :
    name =func .__name__ 
  def _get (self ):
    try :
      return self .__dict__ [name ]
    except KeyError :
      value =func (self )
      self .__dict__ [name ]=value 
      return value 
  update_wrapper (_get ,func )
  def _del (self ):
    self .__dict__ .pop (name ,None )
  return property (_get ,None ,_del )

But the problem I have is that I cannot call the decorator with the @ syntax if I want to use the parameter:

@cachedProperty(name='test') # This does NOT work
def my_func(self):
    return 'ok'

# Only this way works
cachedProperty(my_func, name='test')

How to use the @ syntax with decorators arguments?

Thanks

Answer 1

You need a decorator factory , another wrapper that produces the decorator:

from functools import wraps 

def cachedProperty(name=None):
    def decorator(func):
        if decorator.name is None:
            decorator.name = func.__name__ 
        @wraps(func)
        def _get(self):
            try:
                return self.__dict__[decorator.name]
            except KeyError:
                value = func(self)
            self.__dict__[decorator.name] = value 
            return value 
        def _del(self):
            self.__dict__.pop(decorator.name, None)
        return property(_get, None, _del)
    decorator.name = name
    return decorator

Use this as:

@cachedProperty(name='test')
def my_func(self):
    return 'ok'

A decorator is really just syntactic sugar for:

def my_func(self):
    return 'ok'
my_func = cachedProperty(name='test')(my_func)

so as long as the expression after @ returns your decorator ^[*] it doesn't matter what the expression itself actually does.

In the above example, the @cachedProperty(name='test') part first executes cachedProperty(name='test') , and the return value of that call is used as the decorator. In the above example, decorator is returned, so the my_func function is decorated by calling decorator(my_func) , and the return value of that call is property object, so that is what'll replace my_func .

---

^[*] The @ expression syntax is deliberately limited in how much it is allowed to do. You can do attribute lookups and calls, that's it, the decorator grammar rule only allows an optional call with arguments at the end of a dotted name (where dots are optional):

 decorator ::= "@" dotted_name ["(" [argument_list [","]] ")"] NEWLINE) 

This is a deliberate limitation of the syntax.