Apache Druid: pagination and SQL Rank() equivalent

Question

As of Druid 0.17, there's no query pagination support (it was formerly available through select queries).

I'm trying to emulate in some way pagination and I thought that a possible approach may be creating a virtual dimension that numbers the resulting rows of a query, so that I can filter on that dimension.

This is something that can be easily done in SQL using the RANK function. I was wondering if there is anything similar in Druid.

Anyway, is there any consalidated pattern for obtaining paginated queries?

Answer 1

Apache Druid 0.17.0-1 and 0.18.0 didn't have any equivalent for the SQL RANK() function.

My idea was to perform ordering on a query, numbering the rows and get only the first one.

In my case the query is a group by and ordering has to be performed on Druid's __time field, so there is a solution. Indeed, depending on the needs, you can use the LATEST(expr)/EARLIEST(expr) aggregation functions. They work both with integers and strings (with a slightly different signature).

Reference: http://druid.apache.org/docs/latest/querying/sql.html#aggregation-functions

For pagination, it is currently unavailable