On Thu, Aug 25, 2005 at 09:23:09PM +0300, Stelios Xanthakis wrote: > The explanation is this: hash > and comparison of objects depends on the state of the memory > allocator. A sample case is this: > > class A: pass > dummy0=47 # comment this to get a different result for min > a=A() > b=A() > print min (a, b) > > the result of 'min' is not only non-deterministic but also depends > on whether other things have been allocated before. The same > thing can happen for 'dictionary.keys()' if the keys are objects > and 'iterate-over-set' when the set contains objects. > > In the sudoku solver, there is a min (number, object) which is > probably what's affected by the extistance of the dummy variable. > Now, in sudoku puzzles some times the algorithm has to suppose
Doh, I feel silly. Without an 'import random' in the program I assumed it was deterministic. I would have also expected the comparison to be a TypeError, and it sometimes is. >>> mynum = 7 >>> myob = object() >>> cmp(mynum, myob) -1 >>> mynum.__cmp__(myob) Traceback (most recent call last): File "<stdin>", line 1, in ? TypeError: int.__cmp__(x,y) requires y to be a 'int', not a 'object' Thanks for the explanation, -jackdied -- http://mail.python.org/mailman/listinfo/python-list
