µWeb

The uWeb Model automatically deeploads all items when comparing two records. This is a bug that was introduced when deeploading became available, and triggers because of the list-conversion that is done in eq. Triggering extra SQL queries at compare-time is very bad behaviour.

The secondary is a more rare bug. In most cases, two dictionaries will have the exact same key order, but this is not at all guaranteed. Observe the following:

>>> d1 = {'directory': u'/root/some/path', 'package': u'foo', 'ID': 1, 'description': u'Empty test thingy'}
>>> d2 = {'directory': u'/root/some/path', 'ID': 1, 'description': u'Empty test thingy', 'package': u'foo'}
>>> d1
{'directory': u'/root/some/path', 'description': u'Empty test thingy', 'ID': 1, 'package': u'foo'}
>>> d2
{'directory': u'/root/some/path', 'package': u'foo', 'ID': 1, 'description': u'Empty test thingy'}
>>> d1 == d2
True

As can be seen, the dictionaries compare equal, but their internal order is different (This exact behaviour shows on Python2.6.6 but is likely prevalent). Substituting the basic dicts for a Record here makes the comparison break. This is because the Record class performs its equality check on the lists of values from both records. This is bad for two reasons:

• Key names are not compared at all, leading to false positives;
• Key (and thus value) ordering is not guaranteed, leading to false negatives.