[Python-Dev] Re: decorators and 2.4

[Jeff Bone]

On Jun 28, 2004, at 1:16 PM, Chermside, Michael wrote:

>
>   1. Registering with a framework:
>     [FooFramework]
>     def some_func():
>         body()
>
>   2. Take actions before or after a call:
>     [traceable]
>     def some_func():
>         body()
>
>
>   3. Enforce pre- and post- conditions:
>     [precondition(lambda x,y: x<y),
>      postcondition(lambda x,y,result: x<result<y)]
>     def some_func(x, y):
>         body()
>
>
>  4. Storing Private Data:
>     class Some_Class:
>         [private]
>         def some_method(self, private, x):
>             body(self, private, x)
>

All of these have the effect of either "wrapping" the function (i.e., 
can be thought of as higher-order functions that take functions and 
return functions with the same type as the input function) or 
"bracketing" the function call (in the case of your pre- and post.  
Subtly different from the former.)  The former's trivial to deal with, 
semantically;  the latter slightly less so but tractable.

I would say that all such uses --- when the "outer" function doesn't 
impact state or type of the wrapped function, except perhaps to extend 
it (w/, e.g., exceptions, such as potentially in pre- and post-) --- 
can more of less be thought of as "declarative."

(Side note:  were Python to be extended with true anonymous functions, 
this would all be rendered moot;  the result would IMHO be both 
syntactically and semantically clearer.)

> You follow with a few questions:
>
>> (1)  Does a function have access to its own decorators?
>
> I wouldn't think so. They are NOT "metadata", that's just
> one possible use. The function no more has access to its
> decorators than it has access to its source code! If we
> just needed a place to stick metadata we'd use the
> function's __dict__.

+1.  I waffle on this.  In general I'd feel better if functions etc. 
don't have access to their own decorators at runtime, though you can 
make the general argument that introspection is a good thing.

>> (2)  Do decorators define literal constant values at
>> design-time or do
>> they define a scope of mutable bindings?
>
> The decorators are simply functions that get executed.

IMHO, that's one way to think about it.  From what I can tell from the 
discussion, the PEP, and reading between the lines, there are three 
legit uses and a few wacky ones.  The legit and common / consensus ones 
seem to be:

	(1)  Wrap a function w/o impacting its internal runtime scope
	(2)  Provide pre- and post- about a function
	(3)  Provide metadata (despite the claim above, the below is IMHO a 
metadata application.)

>> 	x = "Guido van Rossum"
>> 	def mymethod(f) [attrs(versionadded="2.2", author=x)]:
>
> That's fine. author is set to Guido.

This is bad because it might inhibit typing applications.  Consider 
this from the PEP:

	def func(arg1, arg2) [accepts(int, (int,float)),
                       returns((int,float))]:
    	 return arg1 * arg2

Should I be able to DYNAMICALLY define the type signature of the 
function at runtime?  Consider, say:

	if todayIsThursday():
		mysig = (int, (int, float))
	else:
		mysig = (str, (int, float))
	[accepts(sig)]
	def func(arg1, arg2):
		if todayIsThursday():
			return foo() # foo is (int, (int, float))
		else:
			return bar() # bar is (str, (int, float))

That's worse than useless.  :-/

>> 	x = "Guido van Rossum"
>> 	def mymethod(f) [attrs(versionadded="2.2", author=x)]:
>>                     oldAuthor = someHandleToMyDecorators.atts.author
>> 		someHandleToMyDecorators.atts.author = "Eric Idle"
>
> That's meaningless unless I know what "someHandleToMyDecorators"
> means.

Right, but that's my point.  Dangerous to give the handle in the first 
place,  even more so if you can set things up dynamically based on 
runtime state, much more so if you can mutate those bindings once 
they're set, *particularly* if you can do some from within the scope of 
the thing being decorated.

BTW, call "someHandleToMyDecorators" "self" or "__dict__" and think 
about it...

>> if foo = 3:
>> 	[someDecoratorThatModifiesBarsRuntimeBehavior]
>> else:
>> 	[someOtherDecorator]
>> def bar(...):
>> 	...
>
> Hmm... I hadn't considered that one. I'll reserve judgment.
> I wouldn't mind if it worked, but I'd certainly try never to
> use it. I *WOULD* however make use of things like this:
>
>     if foo = 3:
>         standard_decoration = someDecoratorWithALongName
>     else:
>         standard_decoration = someOtherDecorator
>     # ...
>     [standard_decoration]
>     def bar(...):
>         ...

Ugh.  Slippery slope to #ifdef / macro hell ala Pike et. al.  But even 
worse:  at least #ifdef can be evaluated statically!  :-/

> You write "The following, IMHO, is truly scary and inhibits many
> of the intended uses of decorators". I'm not sure that intended
> uses it is inhibiting. The only one I can imagine is using
> decorators to create a type-aware compiler that generates type-
> specific optimized code. I don't think this IS a possible use
> for decorators. So... what use *would* be inhibited if
> decorators were used in this manner?

Exactly that.  As for that being an anticipated use, it's at least 
implied in the PEP (see above example from the PEP) and indeed 
something I've been fiddling with a bit.

Point in all of this being:  the semantics of decorators are largely 
not defined at this point;  any such description appears to be an 
anecdotal side-effect of the implementations people are playing with 
--- there's certainly no normative description of their intended 
semantics, either in the PEP of elsewhere at this point to my 
knowledge.  IMHO, decorators are a GOOD idea;  but the syntax of 
decorators is in my opinion FAR less of a concern than should be their 
semantics.

jb