diff options
| -rw-r--r-- | notes.txt | 20 | ||||
| -rw-r--r-- | readme.txt | 7 |
2 files changed, 23 insertions, 4 deletions
@@ -26,10 +26,28 @@ it's probably a programming error on my part future direction ---------------- +better commented code, because I'm going to have to maintain this later + util to list paper ids that fit specific criteria to doublecheck potential errors -replace the use of floats for transfer values with fixed points for increased accuracy +replace the use of floats for transfer values with rational fixed points for increased accuracy add proper tiebreaker handling +more counters/parsers/options to handle state elections, general stv data + +more detailed logging for preference distribution? +tracking of exhausted/fractional-loss ballots? + +multithreaded operation to speed counting up, possibly using unsafePerformIO + +factor TransferMap datatype out into its own module + +bulk exclusions to hopefully increase speed a bit + +combining preference criteria somehow to reduce necessary counts + (requires implementing rational fixed points for transfer values) + (the lack of this, transferring criteria separately, is likely a source of inaccuracy) + (possibly formulate criteria as [candidate #1, then any number of #x, #y, #z, then candidate #2, etc]) + @@ -37,10 +37,11 @@ Legislative Council data and more general STV data is a goal of future versions. -Note that the number of count logs this program outputs differs slightly from -the number that the AEC provides. This is because the AEC's program merges a +Note that the number of count logs this program outputs differs from the +number that the AEC provides. This is because the AEC's program merges a few more logs compared to this one, eg the initial count, and counts that -elect candidates. +elect candidates. On the other hand, the AEC program is also a lot more +verbose on the distribution of preferences. |