The precision of using HLL.INIT(...)
and HLL.MERGE(...)
is described here:
https://cloud.google.com/bigquery/docs/reference/standard-sql/hll_functions
However, I'm wondering if there is ever a cardinality size, under which point HLL is guaranteed to be accurate. For example, if I have 10 distinct values, and I'm "merging" them together in various areas, will this be 100% accurate every time, or will it be subject to the HLL error range? In other words, is there a guaranteed upper bound for each level of HLL Precision
, under which there is 100% guaranteed accuracy? If so, how can that be determined for each Precision level?