Same iterator object yields different result in for loop?

Same iterator object yields different result in for loop?

I came across a very strange behaviour in Python. Using a class derived
from UserDict, the iterator a.items() behaves differently in a for loop
than a.data.items(), even though the two are identical:
Python 3.3.1 (default, Apr 17 2013, 22:32:14)
[GCC 4.7.3] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> from datastruct import QueueDict
>>> a=QueueDict(maxsize=1700)
>>> for i in range(1000):
... a[str(i)]=1/(i+1)
...
>>> a.items()
ItemsView(OrderedDict([('991', 0.0010080645161290322), ('992',
0.0010070493454179255), ('993', 0.001006036217303823), ('994',
0.0010050251256281408), ('995', 0.001004016064257028), ('996',
0.0010030090270812437), ('997', 0.001002004008016032), ('998',
0.001001001001001001), ('999', 0.001)]))
>>> a.data.items()
ItemsView(OrderedDict([('991', 0.0010080645161290322), ('992',
0.0010070493454179255), ('993', 0.001006036217303823), ('994',
0.0010050251256281408), ('995', 0.001004016064257028), ('996',
0.0010030090270812437), ('997', 0.001002004008016032), ('998',
0.001001001001001001), ('999', 0.001)]))
>>> a.items()==a.data.items()
True
>>> # nevertheless:
...
>>> for item in a.items(): print(item)
...
('992', 0.0010070493454179255)
>>> for item in a.data.items(): print(item)
...
('993', 0.001006036217303823)
('994', 0.0010050251256281408)
('995', 0.001004016064257028)
('996', 0.0010030090270812437)
('997', 0.001002004008016032)
('998', 0.001001001001001001)
('999', 0.001)
('991', 0.0010080645161290322)
('992', 0.0010070493454179255)
>>>
The class definition is as follows:
import collections, sys
class QueueDict(collections.UserDict):
def __init__(self, maxsize=1*((2**10)**2), *args, **kwargs ):
self._maxsize=maxsize
super().__init__(*args, **kwargs)
self.data=collections.OrderedDict(self.data)
def __getitem__(self, key):
self.data.move_to_end(key)
return super().__getitem__(key)
def __setitem__(self, key, value):
super().__setitem__(key, value)
self._purge()
def _purge(self):
while sys.getsizeof(self.data) > self._maxsize:
self.data.popitem(last=False)
This is quite disturbing. Any ideas how the same object [by "visual"
inspection, and also by (a.items()==a.data.items()) == True] can, and why
it does, behave differently in the for loop?
Thanks for your help and ideas!