We need to support AggregateState in Parquet/Iceberg. Since there is no such a datatype, it should be stored as a binary, but interpreted as AggregateState in ClickHouse.
We can test it as follows.
- Create AggregatingMergeTree with uniq() function.
- Insert some data
- Create a counterpart in Iceberg (ideally with CREATE TABLE, but external is also ok)
- Export PARTITION to Iceberg table
- Query uniq() from the source and Iceberg table, it should give the same results.
Same should work for SimpleAggregateFunction as well.
We should look how it is done in DataBricks, since it is for sure supports aggregate states
We need to support AggregateState in Parquet/Iceberg. Since there is no such a datatype, it should be stored as a binary, but interpreted as AggregateState in ClickHouse.
We can test it as follows.
Same should work for SimpleAggregateFunction as well.
We should look how it is done in DataBricks, since it is for sure supports aggregate states