random elections and spatial generation #130
Conversation
There was a problem hiding this comment.
Hi Kevin, this is great. My big comment is that I think we can refactor some stuff and add a little more flexibility to these things. Let's Zoom to strategize.
@peterrrock2 , can you take a look at where I tagged you? I'll save you from a bigger review until after we refactor.
| # Choose via proportional to squares | ||
| candidate_votes = {c: 0 for c in remaining} | ||
| for ballot in self.state.profile.get_ballots(): | ||
| top_choice = list(ballot.ranking[0])[0] |
There was a problem hiding this comment.
same as above, does list randomize the order of the set?
There was a problem hiding this comment.
random.choice(random_ballot.ranking[0])[0]
| self.state.profile.ballots, weights=weights, k=1 | ||
| )[0] | ||
| # randomly choose a winner according to first place rankings | ||
| winning_candidate = list(random_ballot.ranking[0])[0] |
There was a problem hiding this comment.
same comment as above, does list randomize the order of the set? @peterrrock2
There was a problem hiding this comment.
random.choice(random_ballot.ranking[0])[0]
There was a problem hiding this comment.
After some googling I believe list does not randomize the order of the set. Unfortunately random.choice() doesn't work on sets either. It could be slow but a good option might be: random.choice(list(random_ballot.ranking[0])). What do you think?
There was a problem hiding this comment.
So the order of the set is determined by the hash of the underlying type. There is an environment variable, PYTHONHASHSEED, that controls this for security reasons. If you want this to be deterministic, you will need to use something other than a set. You can make use of the OrderedSet class in the collections module, or just use a dictionary. Convertiing to a list is also fine here. It doesn't really take up a lot of overhead in the grand scheme of things.
| cands_elected = [len(s) for s in self.state.winners()] | ||
| return sum(cands_elected) < self.seats | ||
|
|
||
| def run_step(self): |
There was a problem hiding this comment.
Okay: let's refactor this. Each round should end when one candidate is eliminated.
Then remove them from the profile. Start a new round.
You'll need to ensure that the random order is preserved. Should each ballot have weight 1? Yes: de-coupling the ballots with weight bigger than 1. In the init method, validate that each weight is an integer.
There was a problem hiding this comment.
To check that a Fraction type is an int, just do int(Fraction(4)) == Fraction(4).
There was a problem hiding this comment.
Do the decoupling in the init method.
There was a problem hiding this comment.
Mention in the doc string that you decondense.
Initial work on random election mechanisms and spatial generation of preference profiles. Includes new election mechanisms RandomDictator and BoostedRandomDictator as well as Uniform/Clustered spatial generation classes.