The problems arising on #15149 arise from the fact that if objects occur as
keys in dictionaries, then their hashes might be required prior to
`setstate` being called on them during unpickling. If we trace where the
failing `hash` comes from, we find it's
CategoryObject.__hash__(self):
if self._hash_value == -1:
self._hash_value = hash(repr(self))
return self._hash_value
so in a way, it's CategoryObject that breaks the unpickling: If
CategoryObject introduces a custom hash, it should also ensure that (at
least by default), pickling ensures that it gets unpacked appropriately and
is usable prior to __setstate__, a requirement that may seem a little
surprising at first. Of course, CategoryObject has no idea what's required
to compute repr(self), but if it would do something like:
def __reduce__(self):
return _CategoryObjectReconstructor,
(type(self),self.hash()),self.__dict__
def _CategoryObjectReconstructor(T,hashval):
v = CategoryObject.__new__(T)
v._hash_value = hashval
return v
we'd at least recover "hash" (and I think it's safe to assume that
repr(loads(dumps(obj))) == repr(obj), so we'd get the same result).
--
You received this message because you are subscribed to the Google Groups
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/groups/opt_out.
