The sequential execution behavior is implied by Calcite as a SQL execution engine, not much of Kylin's work. The entry point is QueryService.execute() where the control is passed to Calcite.
On Thu, Aug 10, 2017 at 3:14 PM, Alexander Sterligov <[email protected]> wrote: > As I understand it is also single threader for joins (not with lookup > tables, but with different OLAP), am I right? > > Could you please guide me where it happens in the code? I would like to > contribute. Maybe there are already some tickets about it? > > On Sun, Aug 6, 2017 at 9:12 AM, Li Yang <[email protected]> wrote: > >> That is right. Sub-queries are executed sequentially as of Kylin 2.0. >> >> On Fri, Jul 28, 2017 at 2:16 AM, Alexander Sterligov <[email protected] >> > wrote: >> >>> Hello! >>> >>> My fact table has 12 boolean fields and user id. I need to count >>> distinct users with has certain combinations of these flags. So I do >>> several sub-queries and union all of them. >>> This query may take up to one minute and it doesn't depend on number of >>> regionservers in hbase. >>> >>> It looks like sub-queries are executed sequentially. Or maybe segment >>> pruning by dictionary is done sequentially. >>> >>> Am I right? >>> >>> Best regards, >>> Alexander >>> >> >> >
