[Python] classic vs new-style Python classes

[Bob Miller]

Rob Hudson wrote:

> Anyone know more about "new-style" classes?
> 
> See this:
> http://jcalderone.livejournal.com/39678.html
> 
> Should you or shouldn't you derive from `object`?  Based on that post
> it seems buggy if you do, but then why is it the "new-style"?

Last question first.  It's "new-style" because old-style classes
were implemented first.  They were found wanting, so their
replacement was designed.  In Python 3000, all classes are new-style.

My personal strategy is to use old-style classes until I need
a new-style feature.

The advantage of my strategy is that I don't clutter up the source
code with a bunch of "derived from object" declarations.

The disadvantage is that I occasionally have to track down bugs where
the property or super() call I just wrote isn't working because I
inserted it into a member of an old-style class hierarchy.  Usually
these are immediately obvious, but not always.

I disagree with the 'blog post you linked.  While there are some
subtle differences, they're rarely differences that you rely on.  I
personally have never broken code by converting a class to new-style.
And I have a tendency to use too much magic in my code.

FWIW, note that you can implicitly create a new-style class by
setting its metaclass to <type> or a subclass of <type>.

    >>> class C: __metaclass__ = type
    ... 
    >>> C.mro()
    [<class '__main__.C'>, <type 'object'>]
    >>> 

-- 
Bob Miller                              K<bob>
                                        kbob at jogger-egg.com