VeloDB Cloud
SQL Manual
Data Types
HLL (HyperLogLog)

HLL (HyperLogLog)



HLL cannot be used as a key column. The aggregation type when creating a table is HLL_UNION.

The user does not need to specify the length and default value. The system will determine the length according to the data aggregation degree.

The HLL column can only be queried or used by its supporting functions such as hll_union_agg, hll_raw_agg, hll_cardinality, and hll_hash.

HLL uses fuzzy deduplication. When handling large data volumes, its performance is better than Count Distinct.

The error rate of HLL is usually around 1% and sometimes 2%.


select hour, HLL_UNION_AGG(pv) over(order by hour) uv from(
   select hour, HLL_RAW_AGG(device_id) as pv
   from metric_table -- Query the accumulated UV per hour
   where datekey=20200922
group by hour order by 1
) final;