You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@druid.apache.org by "LakshSingla (via GitHub)" <gi...@apache.org> on 2023/06/20 06:16:20 UTC
[GitHub] [druid] LakshSingla opened a new pull request, #14450: Fix planning bug while using sort merge frame processor
LakshSingla opened a new pull request, #14450:
URL: https://github.com/apache/druid/pull/14450
### Description
Sort merge join plans the join queries in a different way where both the sides of the joins are subquery. This introduces a regression in planning when the subquery itself becomes a boolean flag as to whether or not the row is to be included. This PR aims to fix the working of the Sort merge join frame processor based on that understanding.
<Test case to be added>
<hr>
##### Key changed/added classes in this PR
* `MyFoo`
* `OurBar`
* `TheirBaz`
<hr>
<!-- Check the items by putting "x" in the brackets for the done things. Not all of these items apply to every PR. Remove the items which are not done or not relevant to the PR. None of the items from the checklist below are strictly necessary, but it would be very helpful if you at least self-review the PR. -->
This PR has:
- [ ] been self-reviewed.
- [ ] using the [concurrency checklist](https://github.com/apache/druid/blob/master/dev/code-review/concurrency.md) (Remove this item if the PR doesn't have any relation to concurrency.)
- [ ] added documentation for new or modified features or behaviors.
- [ ] a release note entry in the PR description.
- [ ] added Javadocs for most classes and all non-trivial methods. Linked related entities via Javadoc links.
- [ ] added or updated version, license, or notice information in [licenses.yaml](https://github.com/apache/druid/blob/master/dev/license.md)
- [ ] added comments explaining the "why" and the intent of the code wherever would not be obvious for an unfamiliar reader.
- [ ] added unit tests or modified existing tests to cover new code paths, ensuring the threshold for [code coverage](https://github.com/apache/druid/blob/master/dev/code-review/code-coverage.md) is met.
- [ ] added integration tests.
- [ ] been tested in a test Druid cluster.
--
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: commits-unsubscribe@druid.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@druid.apache.org
For additional commands, e-mail: commits-help@druid.apache.org
Re: [PR] Fix planning bug while using sort merge frame processor (druid)
Posted by "LakshSingla (via GitHub)" <gi...@apache.org>.
LakshSingla merged PR #14450:
URL: https://github.com/apache/druid/pull/14450
--
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: commits-unsubscribe@druid.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@druid.apache.org
For additional commands, e-mail: commits-help@druid.apache.org
Re: [PR] Fix planning bug while using sort merge frame processor (druid)
Posted by "LakshSingla (via GitHub)" <gi...@apache.org>.
LakshSingla commented on code in PR #14450:
URL: https://github.com/apache/druid/pull/14450#discussion_r1252602160
##########
sql/src/test/java/org/apache/druid/sql/calcite/CalciteJoinQueryTest.java:
##########
@@ -5652,4 +5652,128 @@ public void testJoinsWithThreeConditions()
)
);
}
+
+ @Test
+ public void testJoinWithInputRefCondition()
+ {
+ cannotVectorize();
+ Map<String, Object> context = new HashMap<>(QUERY_CONTEXT_DEFAULT);
+
+ Query expectedQuery;
+
+ if (!NullHandling.sqlCompatible()) {
+ expectedQuery = Druids.newTimeseriesQueryBuilder()
+ .dataSource(
+ join(
+ new TableDataSource(CalciteTests.DATASOURCE1),
+ new QueryDataSource(
+ GroupByQuery.builder()
+ .setInterval(querySegmentSpec(Filtration.eternity()))
+ .setGranularity(Granularities.ALL)
+ .setDataSource(new TableDataSource(CalciteTests.DATASOURCE1))
+ .setVirtualColumns(expressionVirtualColumn(
+ "v0",
+ "1",
+ ColumnType.LONG
+ ))
+ .setDimensions(
+ new DefaultDimensionSpec("m1", "d0", ColumnType.FLOAT),
+ new DefaultDimensionSpec("v0", "d1", ColumnType.LONG)
+ )
+ .build()
+ ),
+ "j0.",
+ "(floor(100) == \"j0.d0\")",
+ JoinType.LEFT
+ )
+ )
+ .granularity(Granularities.ALL)
+ .aggregators(aggregators(
+ new FilteredAggregatorFactory(
+ new CountAggregatorFactory("a0"),
+ new SelectorDimFilter("j0.d1", null, null)
+ )
+ ))
+ .context(getTimeseriesContextWithFloorTime(TIMESERIES_CONTEXT_BY_GRAN, "d0"))
+ .intervals(querySegmentSpec(Filtration.eternity()))
+ .context(context)
+ .build();
+
+ } else {
+ expectedQuery = Druids.newTimeseriesQueryBuilder()
+ .dataSource(
+ join(
+ join(
+ new TableDataSource("foo"),
+ new QueryDataSource(
+ Druids.newTimeseriesQueryBuilder()
+ .dataSource("foo")
+ .aggregators(
+ new CountAggregatorFactory("a0"),
+ new FilteredAggregatorFactory(
+ new CountAggregatorFactory("a1"),
+ not(selector("m1", null, null)),
+ "a1"
+ )
+ )
+ .intervals(querySegmentSpec(Filtration.eternity()))
+ .context(context)
+ .build()
+ ),
+ "j0.",
+ "1",
+ JoinType.INNER
+ ),
+ new QueryDataSource(
+ GroupByQuery.builder()
+ .setInterval(querySegmentSpec(Filtration.eternity()))
+ .setGranularity(Granularities.ALL)
+ .setDataSource(new TableDataSource(CalciteTests.DATASOURCE1))
+ .setVirtualColumns(expressionVirtualColumn(
+ "v0",
+ "1",
+ ColumnType.LONG
+ ))
+ .setDimensions(
+ new DefaultDimensionSpec("m1", "d0", ColumnType.FLOAT),
+ new DefaultDimensionSpec("v0", "d1", ColumnType.LONG)
+ )
+ .build()
+ ),
+ "_j0.",
+ "(floor(100) == \"_j0.d0\")",
+ JoinType.LEFT
+ )
+ )
+ .granularity(Granularities.ALL)
+ .aggregators(aggregators(
+ new FilteredAggregatorFactory(
+ new CountAggregatorFactory("a0"),
+ or(
+ new SelectorDimFilter("j0.a0", "0", null),
+ and(
+ selector("_j0.d1", null, null),
+ expressionFilter("(\"j0.a1\" >= \"j0.a0\")")
+ )
+
+ )
+ )
+ ))
+ .context(getTimeseriesContextWithFloorTime(TIMESERIES_CONTEXT_BY_GRAN, "d0"))
+ .intervals(querySegmentSpec(Filtration.eternity()))
+ .context(context)
+ .build();
+
+ }
+
+ testQuery(
+ "SELECT COUNT(*) FILTER (WHERE FLOOR(100) NOT IN (SELECT m1 FROM foo)) "
+ + "FROM foo",
+ context,
Review Comment:
This is a test in that sense since this class is subclassed to execute with both the joins, however, there isn't any assertion made (since it stems from the parent test class).
I'll add a similar test in the MSQ engine and see if I can add some assertions on the algorithm used.
--
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: commits-unsubscribe@druid.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@druid.apache.org
For additional commands, e-mail: commits-help@druid.apache.org
Re: [PR] Fix planning bug while using sort merge frame processor (druid)
Posted by "LakshSingla (via GitHub)" <gi...@apache.org>.
LakshSingla commented on PR #14450:
URL: https://github.com/apache/druid/pull/14450#issuecomment-1630529206
Thanks @cryptoe @adarshsanjeev for the review
--
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: commits-unsubscribe@druid.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@druid.apache.org
For additional commands, e-mail: commits-help@druid.apache.org
Re: [PR] Fix planning bug while using sort merge frame processor (druid)
Posted by "LakshSingla (via GitHub)" <gi...@apache.org>.
LakshSingla commented on PR #14450:
URL: https://github.com/apache/druid/pull/14450#issuecomment-1630110323
Thanks, fixed the test case that was throwing.
--
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: commits-unsubscribe@druid.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@druid.apache.org
For additional commands, e-mail: commits-help@druid.apache.org
Re: [PR] Fix planning bug while using sort merge frame processor (druid)
Posted by "adarshsanjeev (via GitHub)" <gi...@apache.org>.
adarshsanjeev commented on code in PR #14450:
URL: https://github.com/apache/druid/pull/14450#discussion_r1251905160
##########
extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/querykit/DataSourcePlan.java:
##########
@@ -198,6 +203,28 @@ public Optional<QueryDefinitionBuilder> getSubQueryDefBuilder()
return Optional.ofNullable(subQueryDefBuilder);
}
+ private static JoinAlgorithm deduceJoinAlgorithm(JoinAlgorithm preferredJoinAlgorithm, JoinDataSource joinDataSource)
+ {
+
+ if (JoinAlgorithm.BROADCAST.equals(preferredJoinAlgorithm)) {
+ return JoinAlgorithm.BROADCAST;
+ }
+
+ // preferredJoinAlgorithm would only be sortMerge now
+
+ if (isConditionEqualityOnLeftAndRightColumns(joinDataSource.getConditionAnalysis())) {
Review Comment:
We might want to update the MSQ docs for join algorithm, that the join algorithm is more of a hint now, and might be a different one than expected.
##########
extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/querykit/DataSourcePlan.java:
##########
@@ -198,6 +203,28 @@ public Optional<QueryDefinitionBuilder> getSubQueryDefBuilder()
return Optional.ofNullable(subQueryDefBuilder);
}
+ private static JoinAlgorithm deduceJoinAlgorithm(JoinAlgorithm preferredJoinAlgorithm, JoinDataSource joinDataSource)
Review Comment:
We might want to log the return here. If the query fails before we get the full plan, we might not know the join algorithm.
##########
sql/src/test/java/org/apache/druid/sql/calcite/CalciteJoinQueryTest.java:
##########
@@ -5652,4 +5652,128 @@ public void testJoinsWithThreeConditions()
)
);
}
+
+ @Test
+ public void testJoinWithInputRefCondition()
+ {
+ cannotVectorize();
+ Map<String, Object> context = new HashMap<>(QUERY_CONTEXT_DEFAULT);
+
+ Query expectedQuery;
+
+ if (!NullHandling.sqlCompatible()) {
+ expectedQuery = Druids.newTimeseriesQueryBuilder()
+ .dataSource(
+ join(
+ new TableDataSource(CalciteTests.DATASOURCE1),
+ new QueryDataSource(
+ GroupByQuery.builder()
+ .setInterval(querySegmentSpec(Filtration.eternity()))
+ .setGranularity(Granularities.ALL)
+ .setDataSource(new TableDataSource(CalciteTests.DATASOURCE1))
+ .setVirtualColumns(expressionVirtualColumn(
+ "v0",
+ "1",
+ ColumnType.LONG
+ ))
+ .setDimensions(
+ new DefaultDimensionSpec("m1", "d0", ColumnType.FLOAT),
+ new DefaultDimensionSpec("v0", "d1", ColumnType.LONG)
+ )
+ .build()
+ ),
+ "j0.",
+ "(floor(100) == \"j0.d0\")",
+ JoinType.LEFT
+ )
+ )
+ .granularity(Granularities.ALL)
+ .aggregators(aggregators(
+ new FilteredAggregatorFactory(
+ new CountAggregatorFactory("a0"),
+ new SelectorDimFilter("j0.d1", null, null)
+ )
+ ))
+ .context(getTimeseriesContextWithFloorTime(TIMESERIES_CONTEXT_BY_GRAN, "d0"))
+ .intervals(querySegmentSpec(Filtration.eternity()))
+ .context(context)
+ .build();
+
+ } else {
+ expectedQuery = Druids.newTimeseriesQueryBuilder()
+ .dataSource(
+ join(
+ join(
+ new TableDataSource("foo"),
+ new QueryDataSource(
+ Druids.newTimeseriesQueryBuilder()
+ .dataSource("foo")
+ .aggregators(
+ new CountAggregatorFactory("a0"),
+ new FilteredAggregatorFactory(
+ new CountAggregatorFactory("a1"),
+ not(selector("m1", null, null)),
+ "a1"
+ )
+ )
+ .intervals(querySegmentSpec(Filtration.eternity()))
+ .context(context)
+ .build()
+ ),
+ "j0.",
+ "1",
+ JoinType.INNER
+ ),
+ new QueryDataSource(
+ GroupByQuery.builder()
+ .setInterval(querySegmentSpec(Filtration.eternity()))
+ .setGranularity(Granularities.ALL)
+ .setDataSource(new TableDataSource(CalciteTests.DATASOURCE1))
+ .setVirtualColumns(expressionVirtualColumn(
+ "v0",
+ "1",
+ ColumnType.LONG
+ ))
+ .setDimensions(
+ new DefaultDimensionSpec("m1", "d0", ColumnType.FLOAT),
+ new DefaultDimensionSpec("v0", "d1", ColumnType.LONG)
+ )
+ .build()
+ ),
+ "_j0.",
+ "(floor(100) == \"_j0.d0\")",
+ JoinType.LEFT
+ )
+ )
+ .granularity(Granularities.ALL)
+ .aggregators(aggregators(
+ new FilteredAggregatorFactory(
+ new CountAggregatorFactory("a0"),
+ or(
+ new SelectorDimFilter("j0.a0", "0", null),
+ and(
+ selector("_j0.d1", null, null),
+ expressionFilter("(\"j0.a1\" >= \"j0.a0\")")
+ )
+
+ )
+ )
+ ))
+ .context(getTimeseriesContextWithFloorTime(TIMESERIES_CONTEXT_BY_GRAN, "d0"))
+ .intervals(querySegmentSpec(Filtration.eternity()))
+ .context(context)
+ .build();
+
+ }
+
+ testQuery(
+ "SELECT COUNT(*) FILTER (WHERE FLOOR(100) NOT IN (SELECT m1 FROM foo)) "
+ + "FROM foo",
+ context,
Review Comment:
Should we add a test with join in the query context as "Sort merge", but because of the query structure, msq instead uses broadcast joins?
--
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: commits-unsubscribe@druid.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@druid.apache.org
For additional commands, e-mail: commits-help@druid.apache.org
Re: [PR] Fix planning bug while using sort merge frame processor (druid)
Posted by "LakshSingla (via GitHub)" <gi...@apache.org>.
LakshSingla commented on code in PR #14450:
URL: https://github.com/apache/druid/pull/14450#discussion_r1252602415
##########
extensions-core/multi-stage-query/src/main/java/org/apache/druid/msq/querykit/DataSourcePlan.java:
##########
@@ -198,6 +203,28 @@ public Optional<QueryDefinitionBuilder> getSubQueryDefBuilder()
return Optional.ofNullable(subQueryDefBuilder);
}
+ private static JoinAlgorithm deduceJoinAlgorithm(JoinAlgorithm preferredJoinAlgorithm, JoinDataSource joinDataSource)
+ {
+
+ if (JoinAlgorithm.BROADCAST.equals(preferredJoinAlgorithm)) {
+ return JoinAlgorithm.BROADCAST;
+ }
+
+ // preferredJoinAlgorithm would only be sortMerge now
+
+ if (isConditionEqualityOnLeftAndRightColumns(joinDataSource.getConditionAnalysis())) {
Review Comment:
Thanks for pointing this out, updated it along with the release notes.
--
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: commits-unsubscribe@druid.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@druid.apache.org
For additional commands, e-mail: commits-help@druid.apache.org