Inconsistent behavior of descriptors

[Denis S. Otkidach]

On Wed, 1 Oct 2003, John Roth wrote:

JR> The behavior of descriptors is documented in the article on
JR> the python web site. You might have to hunt a bit to find
JR> it.

http://users.rcn.com/python/download/Descriptor.htm
"The implementation works through a precedence chain that gives
data descriptors priority over instance variables, instance
variables priority over non-data descriptors, and assigns lowest
priority to __getattr__ if provided."

I can't see any reason why "data descriptors" have priority over
instance variables anyway.

JR> The reason for the inconsistency is simply that descriptors
JR> with only a __get__ method are used for methods, functions
JR> and similar entities, while descriptors with both __get__
JR> and
JR> __put__ are used for properties and similar entities.

There are many examples when data (in common sense) descriptors
are used with the only __get__ method defined, i.e. non-data
descriptors in your terms.  Here are two I use most frequently:

class CachedAttribute(object):

    def __init__(self, method, name=None):
        self.method = method
        self.name = name or method.__name__

    def __get__(self, inst, cls):
        # Some tricks here to walk around bug in Python 2.2 are
        # skiped
        result = self.method(inst)
        setattr(inst, self.name, result)
        return result


class ReadAliasAttribute(object):

    def __init__(self, name):
        self.name = name

    def __get__(self, inst, cls):
        return getattr(inst, self.name)


JR> It's useful to be able to put a unique version of a function
JR> or method in an instance, it would completely subvert the
JR> meaning of a property if you could do so.

Why?  You can't do it directly anyway if __set__ method is
defined.

-- 
Denis S. Otkidach
http://www.python.ru/      [ru]