I wasn't able to reproduce the Tests using AHashMap, but it still should be slightly faster, especially considering that our Vecs are about 5 times as small as the ones in the tests.
It wouldn't help much in terms of speed, but when moving from FxHashSet to something else, we could straight up us Vecs and thus remove some imports from a few files.
The main ones:
initial_pause_indicators
exclude_rows
I base this request off of those benchmarks: https://gist.github.com/daboross/976978d8200caf86e02acb6805961195
AHashMap
, but it still should be slightly faster, especially considering that ourVec
s are about 5 times as small as the ones in the tests.FxHashSet
to something else, we could straight up usVec
s and thus remove some imports from a few files.