SchulzePR bug with ties

[itamaro]

It seems SchulzePR is misbehaving, at least with the ballot input I have "engineered":

Python 2.7.4 (default, Apr 19 2013, 18:28:01) 
[GCC 4.7.3] on linux2
>>> from pyvotecore.schulze_pr import SchulzePR
>>> b=[{'count': 1, 'ballot': [[1, 2, 3, 4], [0]]},
       {'count': 1, 'ballot': [[3], [0, 1, 2, 4]]},
       {'count': 1, 'ballot': [[0], [1, 2, 3, 4]]}]
>>> SchulzePR(b, ballot_notation="grouping").as_dict()
{'tie_breaker': [4, 0, 3, 1, 2],
 'candidates': set([0, 1, 2, 3, 4]),
 'order': [3, 0, 4, 1, 2],
 'rounds': [{'winner': 3},
            {'winner': 0},
            {'winner': 4},
            {'winner': 1, 'tied_winners': set([1, 2])},
            {'winner': 2}]}

The anomaly, as I see it, lies in the fact that in the ballot the candidates 1, 2 and 4 are always in the same group, so essentially they must be also grouped together in the results. But in the output you can see that for some reason candidate 4 is above candidates 1 and 2 that are tied.

Am I missing something?

[bradbeattie]

My intuition certainly agrees with yours. We might be able to construct a simpler scenario to test this as well.

>>> SchulzePR([{"ballot": [[1,2,3],[4]]}, {"ballot": [[1],[2,3,4]]}], ballot_notation="grouping").rounds
[{'winner': 1}, {'winner': 3, 'tied_winners': set([2, 3])}, {'winner': 2}, {'winner': 4}]

>>> SchulzePR([{"ballot": [[1,2,3],[4]]}, {"ballot": [[1],[2,3,4]]}, {"ballot":[[4],[1,2,3]]}], ballot_notation="grouping").rounds
[{'winner': 1}, {'winner': 4}, {'winner': 3, 'tied_winners': set([2, 3])}, {'winner': 2}]

>>> SchulzePR([{"ballot": [[1,2,3],[4,5]]}, {"ballot": [[1],[2,3,4,5]]}, {"ballot":[[4,5],[1,2,3]]}], ballot_notation="grouping").rounds
[{'winner': 1}, {'winner': 5, 'tied_winners': set([4, 5])}, {'winner': 3, 'tied_winners': set([2, 3])}, {'winner': 4}, {'winner': 2}]

Okay, I'm veering away from simpler here. I'll take a deeper look into it next week when I have a chance. Thanks for reporting this.