hyperUnique vs. cardinality aggregator


I am unclear on the difference between the hyperUnique and cardinality aggregators. Are both available for use at indexing time? Do both store the indexed data as HLL?


hyperUnique is intended for use at indexing time and stores the data
in an HLL with 2^11 buckets, which should provide ~2.2% error at 1
std. dev, and 4.4% at 2 std. dev.

The cardinality aggregator was created with the intent of it being
used at query time to determine the cardinalities of things that are
already in the data. One potential use could be estimating how many
results will be in your result set to see if it might be a really
large query.

The cardinality aggregator *might* also work at indexing time, but I'm
not sure if it's been wired up for that.


Awesome, thanks Eric!