This puts all the SQL stuff in one place. It also makes life easier by
pointing out that configs be made in either common.runtime.properties
or the broker runtime.properties.
* SQL: Ditch CalciteConnection layer and add DruidMeta, extension aggregators.
Switched from CalciteConnection to Planner, bringing benefits:
- CalciteConnection's JDBC interface no longer sits between the SQL server
(HTTP/Avatica) and Druid's query layer. Instead, the SQL servers can use
Druid Sequence objects directly, reducing overhead in the query return path.
- Implemented our own Planner-based Avatica Meta, letting us control
connection timeouts and connection / statement limits. The previous
CalciteConnection-based implementation didn't have any limits or timeouts.
- The Planner interface lets us override the operator table, opening up
SQL language extensions. This patch includes two: APPROX_COUNT_DISTINCT
in core, and a QUANTILE aggregator in the druid-histogram extension.
Also:
- Added INFORMATION_SCHEMA metadata schema.
- Added tests for Unicode literals and escapes.
* Verify statement is actually open before closing it.
* More detailed INFORMATION_SCHEMA docs.
* SQL support for nested groupBys.
Allows, for example, doing exact count distinct by writing:
SELECT COUNT(*) FROM (SELECT DISTINCT col FROM druid.foo)
Contrast with approximate count distinct, which is:
SELECT COUNT(DISTINCT col) FROM druid.foo
* Add deeply-nested groupBy docs, tests, and maxQueryCount config.
* Extract magic constants into statics.
* Rework rules to put preconditions in the "matches" method.
This patch introduces a GroupByStrategy concept and two strategies: "v1"
is the current groupBy strategy and "v2" is a new one. It also introduces
a merge buffers concept in DruidProcessingModule, to try to better
manage memory used for merging.
Both of these are described in more detail in #2987.
There are two goals of this patch:
1. Make it possible for historical/realtime nodes to return larger groupBy
result sets, faster, with better memory management.
2. Make it possible for brokers to merge streams when there are no order-by
columns, avoiding materialization.
This patch does not do anything to help with memory management on the broker
when there are order-by columns or when there are nested queries. That could
potentially be done in a future patch.
* make isSingleThreaded groupBy query processing overridable at query time
* refactor code in GroupByMergedQueryRunner to make processing of single threaded and parallel merging of runners consistent
* Add back FilteredServerView removed in a32906c7fd to reduce memory usage using watched tiers.
* Add functionality to specify "druid.broker.segment.watchedDataSources"
Gian Merlino
Jonathan Wei
Himanshu
Nishant
jisookim
Xavier Léauté
Charles Allen
Xavier Léauté
jon-wei