loose methods: Smalltalk asPython

[Steve Holden]

Carl Banks wrote:
> Jan Theodore Galkowski wrote:
>>> We've not had
>>> an excellent dynamic OO language since Smalltalk, IMO.
> 
> I would say that "excellence" in object oriented programming is not a
> strong design goal of Python.  Python tries to support OOP well, but
> not to enhance OOP to the detriment of other, more important goals.
> 
> [snip]
>>> Guido was opposed to modifying builtin types before Java existed.  It's
>>> similar to his opposition to dynamic syntax.
>> Opposition or not, the language definition is there.  Surely Smalltalk's
>> OO style can be respected.  Smalltalkers have been doing OO in a dynamic
>> context longer than many.  There are things to learn there.
> 
> Well, that Smalltalk and Ruby do this at least suggests that it won't
> be the end of the world as we know it if it were to happen in Python.
> Personally, I'm with Guido here.  Allowing this might make Python's OOP
> more "excellent", but it would be to the detriment of other, more
> important goals, like minimizing surprise.
> 
>> There are two things here that need to be kept straight, I think.
>>
>> One is the primacy of builtin types.
>>
>> The other is placing those builtins at the top of their own object
>> hierarchy.  Is "int", properly speaking, a descendent of "object"?  No,
>> it's the root of its own tree.
> 
> Are you saying that it shouldn't be a descendant of "object", or that
> it isn't?  "int" is certainly a subclass of "object" in Python (as are
> all objects).
> 
>> Smalltalk has "int" as well, as a
>> *subclass* of Number.  It's primitive.
> 
> Number is primitive?  I thought all types in Smalltalk had a common
> base type.
> 
> Anyways, Python 3000 is exploring the idea of an abstract hierarchy
> that has classes (interfaces?) like Number and Sequence.  In fact, I
> even saw some talk about somehow being able to retrofit classes to
> support one of these interfaces/ABCs, though it was pretty conjectural.
> 
>> In any case, as respectful of Mr/Dr van Rossum as I am, simply because
>> so-and-so says something is bad and therefore it must be true is a
>> fallacious response.
> 
> The same could be said of Smalltalk, i.e., because Smalltalk does it
> therefore it is good is just as fallacious.
> 
> Regardless, whatever Guido pronounces is what happens.  If he's
> adamantly against something, it's pretty much an academic discussion.
> 
Just to be perverse, there was an example recently where a developer 
asked (on python-dev) about the best way to trap and note the creation 
of all objects.

With a proper hierarchy, in which everything calls its parent's 
constructor all the way up to object, in theory (and ignoring old-style 
classes) it would be possible to simply modify the behaviour of object. 
Unfortunately this is a no-no because object's behaviour is hard-wired 
into the CPython interpreter.

The general feeling on python-dev seems to be that this is a step too 
far, and that the confusion engendered by this kind of patching would 
not be worth the benefits gained. Certainly it appears to this reader 
that such a feature woudl be more abused than used.

The principle of least surprise thus reigns for now.

regards
  Steve
-- 
Steve Holden       +44 150 684 7255  +1 800 494 3119
Holden Web LLC/Ltd          http://www.holdenweb.com
Skype: holdenweb     http://del.icio.us/steve.holden
Blog of Note:          http://holdenweb.blogspot.com