[Public API Reform] Do a really independant Pairwise object about condorcet HOT 5 CLOSED

@LiamM32

from condorcet.

Sure. This is to allow only a specified portion of all votes to be processed for pairwise results, right?

Earlier I was thinking that some of the Election functions related to processing information from votes can be moved into a trait in a new file, and be included in both the Election class and a new class.

But you would know better than me how it's best to achieve the desired functionality. It would be nice to have a good way to determine pairwise results for a specified set of voters identified by a tag. Even for the conventional Schulze method or Ranked Pairs, it might be desired to get some statistics for certain demographics of voters, even if this isn't used to determine the end result.

from condorcet.

The Schulze STV method that I am working on has a function to determine how many votes prefer a candidate to all other candidates in a set. Perhaps this belongs in pairwise. Though I didn't put it there because I don't really understand how the pairwise object works.

from condorcet.

No, it's about iterating over votes.

        foreach ($this->getElection()->getVotesValidUnderConstraintGenerator() as $oneVote) {
            $weight = $oneVote->getWeight($election);
            $oneRanking = $oneVote->getContextualRankingWithoutSort($election);

           if (  count($oneRanking) >= 1 &&
                 count($oneRanking[1]) === 1 &&
                $yourExpectedWinnerKey === $this->getElection()->getCandidateKey($oneRanking[1])
           ) {
              // $yourExpectedWinner is the first
           }
        }

from condorcet.

To be honest, I don't really understand why Pairwise needs to be it's own object class. It's massively complicated for what it does. Why not just have a function in the vote object itself to determine which of two candidates is preferred, & methods can iterate over all votes in a foreach loop? If you just want cache for pairwise results, isn't there a simpler implementation that isn't any slower than the current one?

from condorcet.