[Fix](nereids) fix NormalizeRepeat, change the outputExpression rewrite logic #34196
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Thank you for your contribution to Apache Doris. Since 2024-03-18, the Document has been moved to doris-website. |
|
run buildall |
TPC-H: Total hot run time: 41172 ms |
|
run buildall |
TPC-H: Total hot run time: 41828 ms |
TPC-DS: Total hot run time: 186321 ms |
feiniaofeiafei
force-pushed
the
grouping_fix_rewrite
branch
from
9cef386
to
ffabae2
Compare
|
run buildall |
TPC-H: Total hot run time: 41947 ms |
TPC-DS: Total hot run time: 188715 ms |
|
run buildall |
TPC-H: Total hot run time: 40778 ms |
TPC-DS: Total hot run time: 187086 ms |
|
run external |
|
run cloud_p1 |
|
run buildall |
TPC-H: Total hot run time: 40729 ms |
TPC-DS: Total hot run time: 185318 ms |
feiniaofeiafei
changed the title
morrySnow
previously approved these changes
May 6, 2024
|
PR approved by at least one committer and no changes requested. |
github-actions
bot
added
the
approved
|
PR approved by anyone and no changes requested. |
morrySnow
reviewed
May 6, 2024
| // rewrite the arguments of grouping scalar function to slots | ||
| // rewrite grouping scalar function to virtual slots | ||
| // rewrite the arguments of agg function to slots | ||
| List<NamedExpression> normalizedAggOutput = repeat.getOutputExpressions().stream() |
There was a problem hiding this comment.
could we replace all stream api by for loop?
|
run buildall |
github-actions
bot
removed
the
approved
TPC-H: Total hot run time: 40878 ms |
TPC-DS: Total hot run time: 185654 ms |
|
run p0 |
|
run cloud_p1 |
…calar function with bottom projects
…with needToSlots in repeat outputExpressions
… the agg arguments and grouping scalar function
feiniaofeiafei
force-pushed
the
grouping_fix_rewrite
branch
from
4f81946
to
9ae14b1
Compare
|
run buildall |
morrySnow
approved these changes
May 8, 2024
github-actions
bot
added
the
approved
|
PR approved by at least one committer and no changes requested. |
924060929
approved these changes
May 8, 2024
yiguolei
pushed a commit
that referenced
this pull request
May 10, 2024
…te logic (#34196) In NormalizeRepeat, three parts of the outputExpression of LogicalRepeat need to be pushed down and outputted by bottom project: flattenGroupingSetExpr, argumentsOfGroupingScalarFunction, argumentsOfAggregateFunction. In the original code, use these three parts to rewrite the outputExpressions of LogicalRepeat to slots.This can cause problems in some cases, for example: ```sql SELECT ROUND( SUM(pk + 1) - 3) col_alias1, pk + 1 AS col_alias3 FROM table_20_undef_partitions2_keys3_properties4_distributed_by53 GROUP BY GROUPING SETS ((pk), ()) ; ``` The three parts expression needed to be pushed down are: pk, pk+1. The original code use pk+1 to rewrite the pk + 1 AS col_alias3 to slot. But the pk+1 is not in the list of grouping outputs, and then report error. This pr change the rewrite process, divide the expression needed to be pushed down into 2 parts: one is (flattenGroupingSetExpr) and the other one is (argumentsOfGroupingScalarFunction, argumentsOfAggregateFunction). and use the flattenGroupingSetExpr rewrite all LogicalRepeat outputExpressions, and use the argumentsOfGroupingScalarFunction, argumentsOfAggregateFunction to rewrite only the agg function arguments and the grouping scalar function. So, in the above sql, the pk + 1 AS col_alias3 will not be rewritten to slot, and can be computed.
ByteYue
pushed a commit
to ByteYue/doris
that referenced
this pull request
May 15, 2024
…te logic (apache#34196) In NormalizeRepeat, three parts of the outputExpression of LogicalRepeat need to be pushed down and outputted by bottom project: flattenGroupingSetExpr, argumentsOfGroupingScalarFunction, argumentsOfAggregateFunction. In the original code, use these three parts to rewrite the outputExpressions of LogicalRepeat to slots.This can cause problems in some cases, for example: ```sql SELECT ROUND( SUM(pk + 1) - 3) col_alias1, pk + 1 AS col_alias3 FROM table_20_undef_partitions2_keys3_properties4_distributed_by53 GROUP BY GROUPING SETS ((pk), ()) ; ``` The three parts expression needed to be pushed down are: pk, pk+1. The original code use pk+1 to rewrite the pk + 1 AS col_alias3 to slot. But the pk+1 is not in the list of grouping outputs, and then report error. This pr change the rewrite process, divide the expression needed to be pushed down into 2 parts: one is (flattenGroupingSetExpr) and the other one is (argumentsOfGroupingScalarFunction, argumentsOfAggregateFunction). and use the flattenGroupingSetExpr rewrite all LogicalRepeat outputExpressions, and use the argumentsOfGroupingScalarFunction, argumentsOfAggregateFunction to rewrite only the agg function arguments and the grouping scalar function. So, in the above sql, the pk + 1 AS col_alias3 will not be rewritten to slot, and can be computed.
morrySnow
pushed a commit
that referenced
this pull request
Jul 26, 2024
Induced by #34196. In NormalizeRepeat, when NormalizeToSlot is called, aggregate function parameters, grouping scalar function parameters, and all expressions in grouping sets (including columns and column aliases) are pushed down to the lower-level project output. In the previous PR #34196, the context was split into two, but the two contexts were not consistent. It is possible that the triplets in one context save (id, c1, id as c1), and the triplets in the other context save (id, id, id). This causes id as c1 to be pushed down, but there is a reference to id in the upper-level LogicalRepeat, which causes the slot to be not found. This pr has been modified. If the same slot in the projection column has different aliases, for example, select id as c1, id, id as c3, grouping(id) from table1 group by grouping sets((id, value2),(id)); then id as c1 (using the first alias) will be pushed down to the project. In both the LogicalRepeat operator and the LogicalAggregate operator, c1 is referenced as the input slot, and id and c3 will not be used as input slots. before NormalizeRepeat: LogicalResultSink[32] ( outputExprs=[c1#3, id#0, c3#4, __grouping_3#5] ) +--LogicalRepeat ( groupingSets=[[id#0, value2#2], [id#0]], outputExpressions=[id#0 AS `c1`#3, id#0, id#0 AS `c3`#4, Grouping(id#0) AS `Grouping(id)`#5] ) +--LogicalOlapScan (qualified=table1) After NormalizeRepeat: LogicalResultSink[33] (outputExprs=[c1#3, id#0, c3#4, __grouping_3#5]) +--LogicalAggregate[30] (groupByExpr=[c1#3, value2#2, GROUPING_ID#7, GROUPING_PREFIX_c1#6 originExpression=Grouping(c1#3)], outputExpr=[c1#3, c1#3 AS `id`#0, c1#3 AS `c3`#4, GROUPING_PREFIX_c1#6 originExpression=Grouping(c1#3) AS `GROUPING_PREFIX_c1`#5], hasRepeat=true ) +--LogicalRepeat (groupingSets=[[c1#3, value2#2], [c1#3]], outputExpressions=[c1#3, value2#2, GROUPING_ID#7, GROUPING_PREFIX_c1#6 originExpression=Grouping(c1#3)] ) +--LogicalProject[28] (projects=[id#0 AS `c1`#3, value2#2]) +--LogicalOlapScan (qualified=table1)
feiniaofeiafei
added a commit
to feiniaofeiafei/doris
that referenced
this pull request
Jul 29, 2024
Induced by apache#34196. In NormalizeRepeat, when NormalizeToSlot is called, aggregate function parameters, grouping scalar function parameters, and all expressions in grouping sets (including columns and column aliases) are pushed down to the lower-level project output. In the previous PR apache#34196, the context was split into two, but the two contexts were not consistent. It is possible that the triplets in one context save (id, c1, id as c1), and the triplets in the other context save (id, id, id). This causes id as c1 to be pushed down, but there is a reference to id in the upper-level LogicalRepeat, which causes the slot to be not found. This pr has been modified. If the same slot in the projection column has different aliases, for example, select id as c1, id, id as c3, grouping(id) from table1 group by grouping sets((id, value2),(id)); then id as c1 (using the first alias) will be pushed down to the project. In both the LogicalRepeat operator and the LogicalAggregate operator, c1 is referenced as the input slot, and id and c3 will not be used as input slots. before NormalizeRepeat: LogicalResultSink[32] ( outputExprs=[c1#3, id#0, c3#4, __grouping_3#5] ) +--LogicalRepeat ( groupingSets=[[id#0, value2#2], [id#0]], outputExpressions=[id#0 AS `c1`apache#3, id#0, id#0 AS `c3`apache#4, Grouping(id#0) AS `Grouping(id)`apache#5] ) +--LogicalOlapScan (qualified=table1) After NormalizeRepeat: LogicalResultSink[33] (outputExprs=[c1#3, id#0, c3#4, __grouping_3#5]) +--LogicalAggregate[30] (groupByExpr=[c1#3, value2#2, GROUPING_ID#7, GROUPING_PREFIX_c1#6 originExpression=Grouping(c1#3)], outputExpr=[c1#3, c1#3 AS `id`#0, c1#3 AS `c3`apache#4, GROUPING_PREFIX_c1#6 originExpression=Grouping(c1#3) AS `GROUPING_PREFIX_c1`apache#5], hasRepeat=true ) +--LogicalRepeat (groupingSets=[[c1#3, value2#2], [c1#3]], outputExpressions=[c1#3, value2#2, GROUPING_ID#7, GROUPING_PREFIX_c1#6 originExpression=Grouping(c1#3)] ) +--LogicalProject[28] (projects=[id#0 AS `c1`apache#3, value2#2]) +--LogicalOlapScan (qualified=table1)
feiniaofeiafei
added a commit
to feiniaofeiafei/doris
that referenced
this pull request
Jul 30, 2024
Induced by apache#34196. In NormalizeRepeat, when NormalizeToSlot is called, aggregate function parameters, grouping scalar function parameters, and all expressions in grouping sets (including columns and column aliases) are pushed down to the lower-level project output. In the previous PR apache#34196, the context was split into two, but the two contexts were not consistent. It is possible that the triplets in one context save (id, c1, id as c1), and the triplets in the other context save (id, id, id). This causes id as c1 to be pushed down, but there is a reference to id in the upper-level LogicalRepeat, which causes the slot to be not found. This pr has been modified. If the same slot in the projection column has different aliases, for example, select id as c1, id, id as c3, grouping(id) from table1 group by grouping sets((id, value2),(id)); then id as c1 (using the first alias) will be pushed down to the project. In both the LogicalRepeat operator and the LogicalAggregate operator, c1 is referenced as the input slot, and id and c3 will not be used as input slots. before NormalizeRepeat: LogicalResultSink[32] ( outputExprs=[c1#3, id#0, c3#4, __grouping_3#5] ) +--LogicalRepeat ( groupingSets=[[id#0, value2#2], [id#0]], outputExpressions=[id#0 AS `c1`apache#3, id#0, id#0 AS `c3`apache#4, Grouping(id#0) AS `Grouping(id)`apache#5] ) +--LogicalOlapScan (qualified=table1) After NormalizeRepeat: LogicalResultSink[33] (outputExprs=[c1#3, id#0, c3#4, __grouping_3#5]) +--LogicalAggregate[30] (groupByExpr=[c1#3, value2#2, GROUPING_ID#7, GROUPING_PREFIX_c1#6 originExpression=Grouping(c1#3)], outputExpr=[c1#3, c1#3 AS `id`#0, c1#3 AS `c3`apache#4, GROUPING_PREFIX_c1#6 originExpression=Grouping(c1#3) AS `GROUPING_PREFIX_c1`apache#5], hasRepeat=true ) +--LogicalRepeat (groupingSets=[[c1#3, value2#2], [c1#3]], outputExpressions=[c1#3, value2#2, GROUPING_ID#7, GROUPING_PREFIX_c1#6 originExpression=Grouping(c1#3)] ) +--LogicalProject[28] (projects=[id#0 AS `c1`apache#3, value2#2]) +--LogicalOlapScan (qualified=table1)
dataroaring
pushed a commit
that referenced
this pull request
Aug 2, 2024
Induced by #34196. In NormalizeRepeat, when NormalizeToSlot is called, aggregate function parameters, grouping scalar function parameters, and all expressions in grouping sets (including columns and column aliases) are pushed down to the lower-level project output. In the previous PR #34196, the context was split into two, but the two contexts were not consistent. It is possible that the triplets in one context save (id, c1, id as c1), and the triplets in the other context save (id, id, id). This causes id as c1 to be pushed down, but there is a reference to id in the upper-level LogicalRepeat, which causes the slot to be not found. This pr has been modified. If the same slot in the projection column has different aliases, for example, select id as c1, id, id as c3, grouping(id) from table1 group by grouping sets((id, value2),(id)); then id as c1 (using the first alias) will be pushed down to the project. In both the LogicalRepeat operator and the LogicalAggregate operator, c1 is referenced as the input slot, and id and c3 will not be used as input slots. before NormalizeRepeat: LogicalResultSink[32] ( outputExprs=[c1#3, id#0, c3#4, __grouping_3#5] ) +--LogicalRepeat ( groupingSets=[[id#0, value2#2], [id#0]], outputExpressions=[id#0 AS `c1`#3, id#0, id#0 AS `c3`#4, Grouping(id#0) AS `Grouping(id)`#5] ) +--LogicalOlapScan (qualified=table1) After NormalizeRepeat: LogicalResultSink[33] (outputExprs=[c1#3, id#0, c3#4, __grouping_3#5]) +--LogicalAggregate[30] (groupByExpr=[c1#3, value2#2, GROUPING_ID#7, GROUPING_PREFIX_c1#6 originExpression=Grouping(c1#3)], outputExpr=[c1#3, c1#3 AS `id`#0, c1#3 AS `c3`#4, GROUPING_PREFIX_c1#6 originExpression=Grouping(c1#3) AS `GROUPING_PREFIX_c1`#5], hasRepeat=true ) +--LogicalRepeat (groupingSets=[[c1#3, value2#2], [c1#3]], outputExpressions=[c1#3, value2#2, GROUPING_ID#7, GROUPING_PREFIX_c1#6 originExpression=Grouping(c1#3)] ) +--LogicalProject[28] (projects=[id#0 AS `c1`#3, value2#2]) +--LogicalOlapScan (qualified=table1)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
In NormalizeRepeat, three parts of the outputExpression of LogicalRepeat need to be pushed down and outputted by bottom project: flattenGroupingSetExpr, argumentsOfGroupingScalarFunction, argumentsOfAggregateFunction.
In the original code, use these three parts to rewrite the outputExpressions of LogicalRepeat to slots.This can cause problems in some cases, for example:
The three parts expression needed to be pushed down are: pk, pk+1. The original code use pk+1 to rewrite the pk + 1 AS col_alias3 to slot. But the pk+1 is not in the list of grouping outputs, and then report error.
This pr change the rewrite process, divide the expression needed to be pushed down into 2 parts: one is (flattenGroupingSetExpr) and the other one is (argumentsOfGroupingScalarFunction, argumentsOfAggregateFunction).
and use the flattenGroupingSetExpr rewrite all LogicalRepeat outputExpressions, and use the argumentsOfGroupingScalarFunction, argumentsOfAggregateFunction to rewrite only the agg function arguments and the grouping scalar function.
So, in the above sql, the pk + 1 AS col_alias3 will not be rewritten to slot, and can be computed.