New methods for weakref.Weak*Dictionary types
I'd like to commit this for Python 2.5: http://www.python.org/sf/1479988 The WeakKeyDictionary and WeakValueDictionary don't provide any API to get just the weakrefs out, instead of the usual mapping API. This can be desirable when you want to get a list of everything without creating new references to the underlying objects at that moment. This patch adds methods to make the references themselves accessible using the API, avoiding requiring client code to have to depend on the implementation. The WeakKeyDictionary gains the .iterkeyrefs() and .keyrefs() methods, and the WeakValueDictionary gains the .itervaluerefs() and .valuerefs() methods. The patch includes tests and docs. -Fred -- Fred L. Drake, Jr. <fdrake at acm.org>
[Fred L. Drake, Jr.]
I'd like to commit this for Python 2.5:
http://www.python.org/sf/1479988
The WeakKeyDictionary and WeakValueDictionary don't provide any API to get just the weakrefs out, instead of the usual mapping API. This can be desirable when you want to get a list of everything without creating new references to the underlying objects at that moment.
This patch adds methods to make the references themselves accessible using the API, avoiding requiring client code to have to depend on the implementation. The WeakKeyDictionary gains the .iterkeyrefs() and .keyrefs() methods, and the WeakValueDictionary gains the .itervaluerefs() and .valuerefs() methods.
The patch includes tests and docs.
+1. A real need for this is explained in ZODB's ZODB/util.py's WeakSet class, which contains a WeakValueDictionary: """ # Return a list of weakrefs to all the objects in the collection. # Because a weak dict is used internally, iteration is dicey (the # underlying dict may change size during iteration, due to gc or # activity from other threads). as_weakref_list() is safe. # # Something like this should really be a method of Python's weak dicts. # If we invoke self.data.values() instead, we get back a list of live # objects instead of weakrefs. If gc occurs while this list is alive, # all the objects move to an older generation (because they're strongly # referenced by the list!). They can't get collected then, until a # less frequent collection of the older generation. Before then, if we # invoke self.data.values() again, they're still alive, and if gc occurs # while that list is alive they're all moved to yet an older generation. # And so on. Stress tests showed that it was easy to get into a state # where a WeakSet grows without bounds, despite that almost all its # elements are actually trash. By returning a list of weakrefs instead, # we avoid that, although the decision to use weakrefs is now very # visible to our clients. def as_weakref_list(self): # We're cheating by breaking into the internals of Python's # WeakValueDictionary here (accessing its .data attribute). return self.data.data.values() """ As that implementation suggests, though, I'm not sure there's real payback for the extra time taken in the patch's `valuerefs` implementation to weed out weakrefs whose referents are already gone: the caller has to make this check anyway when it iterates over the returned list of weakrefs. Iterating inside the implementation, to build the list via itervalues(), also creates that much more vulnerability to "dict changed size during iteration" multi-threading surprises. For that last reason, if the patch went in as-is, I expect ZODB would still need to "cheat"; obtaining the list of weakrefs directly via plain .data.values() is atomic, and so immune to these multi-threading surprises.
On Monday 01 May 2006 16:57, Tim Peters wrote:
+1. A real need for this is explained in ZODB's ZODB/util.py's WeakSet class, which contains a WeakValueDictionary: ... As that implementation suggests, though, I'm not sure there's real payback for the extra time taken in the patch's `valuerefs` implementation to weed out weakrefs whose referents are already gone: the caller has to make this check anyway when it iterates over the
Good point; I've updated the patch accordingly. -Fred -- Fred L. Drake, Jr. <fdrake at acm.org>
Hi Tim, On Mon, May 01, 2006 at 04:57:06PM -0400, Tim Peters wrote:
""" # Return a list of weakrefs to all the objects in the collection. # Because a weak dict is used internally, iteration is dicey (the # underlying dict may change size during iteration, due to gc or # activity from other threads).
But then, isn't the real problem the fact that applications cannot safely iterate over weak dicts? This fact could be viewed as a bug, and fixed without API changes. For example, I can imagine returning to the client an iterator that "locks" the dictionary. Upon exhaustion, or via the __del__ of the iterator, or even in the 'finally:' part of the generator if that's how iteration is implemented, the dict is unlocked. Here "locking" means that weakrefs going away during this time are not eagerly removed from the dict; they will be removed only when the dict is unlocked. A bientot, Armin.
[Tim Peters]
""" # Return a list of weakrefs to all the objects in the collection. # Because a weak dict is used internally, iteration is dicey (the # underlying dict may change size during iteration, due to gc or # activity from other threads).
[Armin Rigo]
But then, isn't the real problem the fact that applications cannot safely iterate over weak dicts?
Well, in the presence of threads, iterating over any dict (weak or not) may blow up: while some thread is iterating over the dict, other threads can change the size of the dict, and then the dict iterator will blow up the next time it's invoked. In the context from which that comment was extracted, that's a potential problem (which is why the comment mentions "activity from other threads" :-)).
This fact could be viewed as a bug, and fixed without API changes. For example, I can imagine returning to the client an iterator that "locks" the dictionary. Upon exhaustion, or via the __del__ of the iterator, or even in the 'finally:' part of the generator if that's how iteration is implemented, the dict is unlocked.
Here "locking" means that weakrefs going away during this time are not eagerly removed from the dict; they will be removed only when the dict is unlocked.
That could remove one source of potential iteration surprises unique to weak dicts, due to "magical" removal of dict entries (note that it would probably need a thread-safe count of outstanding iterators, and not "unlock the dict" until the count fell to 0). Other threads could still change the dict's size _seemingly_ "by magic" (from the dict iterator's POV). I don't know whether fixing "magical" weak-dict removal without fixing "seemingly magical" weak-dict removal or addition via other threads would be worth the bother. Anyone burned by either now has learned to avoid the iter{keys,values,items}() methods. Without more support in the dict implementation (and support that would probably be difficult to add), the only thoroughly safe strategy is to atomically materialize a hidden collection of the keys, values, or items to be iterated over, and have the iterator march over those. In effect, apps do that themselves now by iterating over .keys()/values()/items() instead of their .iterXYZ() versions. Many apps can get away without that, though, so there's value in keeping the current "obvious" dict iterators.
participants (3)
-
Armin Rigo -
Fred L. Drake, Jr. -
Tim Peters