gianm opened a new pull request, #20593:
URL: https://github.com/apache/druid/pull/20593
This patch adds the ability for SQL UNION to run over relations that involve
projections and filters of base tables, such as:
```
SELECT dim1, SUM(m1), COUNT(*)
FROM (SELECT dim1, m1 FROM foo2 WHERE dim1 = '1'
UNION ALL
SELECT dim1, m1 FROM foo)
GROUP BY 1;
```
Such queries are planned into stacks of the existing "filter" datasource and
the new "project" datasource.
Other changes, outside the SQL planner:
1) DataSource method "createSegmentPruner" has DimFilter and
VirtualColumns added as parameters. This enables DataSource
implementations, such as FilteredDataSource and ProjectDataSource,
to rewrite the filter and contribute additional virtual columns to
the pruning logic. ExecutionVertex#isBaseColumn is removed and more
scenarios are now pruneable.
2) DataSource gets "getReservedColumns" with a set of column names to
avoid when generating higher-level virtual columns. This is used
by the SQL planner to avoid name collisions.
3) DataSource gets "mayFilterRows", which allows time-boundary queries
to properly recognize that filtering DataSources should go down the
"query has filter" path.
4) Fix an issue where various CursorHolder implementations return
lists from "getOrdering" that erroneously include base columns
which were shadowed by virtual columns.
5) Updates to MSQ to call the new ExecutionVertex#splitUnion and
splitUnionSegmentMapFunctions methods to provide each arm with its
own SegmentPruner and SegmentMapFunction.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
