Using metaclasses to play with decorators.

[Colin J. Williams]

David MacQuigg wrote:

> On Wed, 23 Jun 2004 07:30:34 -0400, "Colin J. Williams"
> <cjw at sympatico.ca> wrote:
> 
> 
>>
>>David MacQuigg wrote:
>>
>>>On Sun, 20 Jun 2004 14:48:23 -0400, "Colin J. Williams"
>>><cjw at sympatico.ca> wrote:
>>>
>>>
>>>
>>>>I have yet to wrap my mind around decorators.
>>>
>>>
>>>Decorators are very simple.  They are just a way to provide different
>>>forms of methods, without introducing a new keyword, or some other
>>>more awkward syntax.
>>>
>>>Say you wanted to define a method that didn't have 'self' as its first
>>>argument.  You could add a new keyword to the language:
>>>
>>>noself methodA(x,y):
>>>	return x + y
>>>
>>>Or you could add a "decorator" to the existing syntax:
>>>
>>>def methodA(x,y) [noself]:
>>>	return x + y
>>>
>>>Change 'noself' to 'staticmethod' and you have one of the current
>>>proposals in PEP318.
>>>
>>>Don't get distracted by 'staticmethod' and other mumbo-jumbo
>>>terminology, and you should have no problem with decorators.
>>
>>OK, I'll ignore 'staticmethod', but could you tell me how
>>
>>  def methodA(x, y) [noself]:
>>    return x + y
>>
>>differs in substance from
>>
>>  def methodA(self, y):
>>    return self + y
>>
>>or
>>  def methodA(x, y):
>>    return x + y
>>
>>What has been gained by the added syntactic clutter?
> 
> 
> The example above was to explain decorators, not justify them.  If
> they were to be used *only* to clean up the current syntax for
> staticmethod, then I would say a better alternative would be to get
> rid of the need for a separate staticmethod form entirely. ( See the
> thread "Unification of Methods and Functions" for a thorough
> discussion of this topic.)
> 
> Are decorators on functions necessary?  I can't think of a simpler or
> more consistent way to handle all the variations proposed in PEP 318.
> 
> Assuming that we *will* have a decorator syntax with many options, I
> think that making [staticmethod] one of those options is appropriate.
> I would still prefer a word more meaningful to new users, however. The
> rational I have heard for "staticmethod" is so contorted, it is not
> worth repeating.
> 
> -- Dave
> 
PEP 318 seems to focus on the "how" to implement decorators, rather than 
the "why".  Is there some accessible explanation of the purpose of them?

Colin W.