Issue1534

This is a bit strange, here is the sample:

class Attr(object):
    def __init__(self, name):
        self.name = name

    def __set__(self, instance, value):
        instance.__dict__[self.name] = value


class Test(object):
    attr = Attr('attr')

    def test(self):
        self.attr = 5

t = Test()
print t.attr
t.test()
print t.attr

In python, if you have a newstyle object called x, and in some point define the attribute __dict__["x"] as 5 then getting the value of x. So in python that program would return:

<__main__.Attr object at 0x7f1b820b7f90>
5

On jython it returns:

<__main__.Attr object at 0x1>
<__main__.Attr object at 0x1>

A workaround would be to add a __get__ method something like this:

    def __get__(self, instance, owner):
        return instance.__dict__[self.name]

But for compatibility it would be nice to change this.

Thanks!

Thanks for logging this, I just went over it with Chris McDonough yesterday. pypy also suffers from this

Jython/Pypy seem more correct than CPython according to the rules on http://docs.python.org/reference/datamodel.html

"For instance bindings, the precedence of descriptor invocation depends on the which descriptor methods are defined. Normally, data descriptors define both __get__() and __set__(), while non-data descriptors have just the __get__() method. Data descriptors always override a redefinition in an instance dictionary. In contrast, non-data descriptors can be overridden by instances."

"A descriptor can define any combination of __get__(), __set__() and __delete__(). If it does not define __get__(), then accessing the attribute even on an instance will return the descriptor object itself. If the descriptor defines __set__() and/or __delete__(), it is a data descriptor; if it defines neither, it is a non-data descriptor."

But I don't think we can fault CPython as it's had this behavior forever

fixed in r7007

I also brought this up with the PyPy guys as their behavior matched jython's https://codespeak.net/issue/pypy-dev/issue491

There was a discussion on python-dev about it. The conclusion came to was that we need to match CPython and that those data descriptor docs needed to be fixed too