You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@datafusion.apache.org by gi...@apache.org on 2026/08/19 02:22:03 UTC
(datafusion) 01/01: fix: Streaming aggregation produces wrong result with grouping sets (#24422)
This is an automated email from the ASF dual-hosted git repository.
github-merge-queue[bot] pushed a commit to branch gh-readonly-queue/main/pr-24422-ebea06912c18563efb976c2c2698aea30ab5b6b3
in repository https://gitbox.apache.org/repos/asf/datafusion.git
commit 5c846af2caa0b59710c73a0450bf11cd35b8d88f
Author: Yongting You <20...@gmail.com>
AuthorDate: Wed Aug 19 02:21:39 2026 +0000
fix: Streaming aggregation produces wrong result with grouping sets (#24422)
## Which issue does this PR close?
<!--
We generally require a GitHub issue to be filed for all bug fixes and
enhancements and this helps us generate change logs for our releases.
You can link an issue to this PR using the GitHub syntax. For example
`Closes #123` indicates that this PR will close issue #123.
-->
- Closes #24421
## Rationale for this change
<!--
Why are you proposing this change? If this is already explained clearly
in the issue then this section is not needed.
Explaining clearly why changes are proposed helps reviewers understand
your changes and offer better suggestions for fixes.
Please explain the problem you are trying to solve in terms of the
user-visible
behavior, rather than the implementation.
For example, "The code in `foo.rs` doesn't handle nulls" is a symptom of
the
implementation. "COUNT(DISTINCT) returns wrong results when the column
contains
nulls" is the user-visible problem.
-->
The aggregation planning will enable ordering optimization when the
grouping sets is available, however they're not compatible now.
I'm also not sure if we want to enable ordering optimization for
grouping sets in the future, since this seems requires extra
implementation complexity.
This PR disables ordering/streaming aggregation optimization when
grouping set is present.
Reference for ordering optimization in aggregation:
https://github.com/apache/datafusion/blob/710e74e6764275675b117823d26d4944115f6ba2/datafusion/physical-plan/src/aggregates/ordered_partial_stream.rs#L62
## What changes are included in this PR?
<!--
There is no need to duplicate the description in the issue here, but it
is sometimes worth providing a summary of the individual changes in this
PR.
-->
- 80cb8eee1375097262a2c8e0c6b92109563128fe: reproducer that fails
without the fix
- 5673f3e: fix. It disables ordering optimization when grouping set is
present
## Are these changes tested?
<!--
We typically require tests for all PRs in order to:
1. Prevent the code from being accidentally broken by subsequent changes
2. Serve as another way to document the expected behavior of the code
If tests are not included in your PR, please explain why (for example,
are they covered by existing tests)?
-->
slt
## Are there any user-facing changes?
<!--
If there are user-facing changes then we may require documentation to be
updated before approving the PR.
If there are any breaking changes to public APIs, please add the `api
change` label.
-->
no
---
.../tests/physical_optimizer/filter_pushdown.rs | 4 +-
datafusion/physical-plan/src/aggregates/mod.rs | 48 +++++++++---
datafusion/sqllogictest/test_files/grouping.slt | 91 ++++++++++++++++++++++
3 files changed, 130 insertions(+), 13 deletions(-)
diff --git a/datafusion/core/tests/physical_optimizer/filter_pushdown.rs b/datafusion/core/tests/physical_optimizer/filter_pushdown.rs
index a98c1b7bcf..57d7dba295 100644
--- a/datafusion/core/tests/physical_optimizer/filter_pushdown.rs
+++ b/datafusion/core/tests/physical_optimizer/filter_pushdown.rs
@@ -2235,7 +2235,7 @@ fn test_pushdown_grouping_sets_filter_on_common_column() {
- DataSourceExec: file_groups={1 group: [[test.parquet]]}, projection=[a, b, c], file_type=test, pushdown_supported=true
output:
Ok:
- - AggregateExec: mode=Final, gby=[(a@0 as a, b@1 as b), (NULL as a, b@1 as b)], aggr=[cnt], ordering_mode=PartiallySorted([1])
+ - AggregateExec: mode=Final, gby=[(a@0 as a, b@1 as b), (NULL as a, b@1 as b)], aggr=[cnt]
- DataSourceExec: file_groups={1 group: [[test.parquet]]}, projection=[a, b, c], file_type=test, pushdown_supported=true, predicate=b@1 = bar
"
);
@@ -2521,7 +2521,7 @@ fn test_pushdown_through_aggregate_grouping_sets_with_reordered_input() {
- DataSourceExec: file_groups={1 group: [[test.parquet]]}, projection=[a, b, c], file_type=test, pushdown_supported=true
output:
Ok:
- - AggregateExec: mode=Final, gby=[(a@1 as a, b@2 as b), (NULL as a, b@2 as b)], aggr=[cnt], ordering_mode=PartiallySorted([1])
+ - AggregateExec: mode=Final, gby=[(a@1 as a, b@2 as b), (NULL as a, b@2 as b)], aggr=[cnt]
- ProjectionExec: expr=[c@2 as c, a@0 as a, b@1 as b]
- DataSourceExec: file_groups={1 group: [[test.parquet]]}, projection=[a, b, c], file_type=test, pushdown_supported=true, predicate=b@1 = bar
"
diff --git a/datafusion/physical-plan/src/aggregates/mod.rs b/datafusion/physical-plan/src/aggregates/mod.rs
index 1a7193a7d8..97226159da 100644
--- a/datafusion/physical-plan/src/aggregates/mod.rs
+++ b/datafusion/physical-plan/src/aggregates/mod.rs
@@ -170,7 +170,7 @@ use crate::statistics::{ChildStats, StatisticsArgs};
use crate::{ChildrenPropertiesMode, ReplaceChildrenOptions, validate_child_count};
use crate::{
DisplayFormatType, Distribution, ExecutionPlan, InputDistributionRequirements,
- InputOrderMode, SendableRecordBatchStream, Statistics,
+ InputOrderMode, Partitioning, SendableRecordBatchStream, Statistics,
};
use datafusion_common::config::ConfigOptions;
use parking_lot::Mutex;
@@ -1015,7 +1015,7 @@ impl AggregateExec {
.filter(|idx| group_by.groups.iter().all(|group| !group[*idx]))
.collect();
- let input_order_mode = if indices.len() == groupby_exprs.len()
+ let mut input_order_mode = if indices.len() == groupby_exprs.len()
&& !indices.is_empty()
&& group_by.groups.len() == 1
{
@@ -1026,19 +1026,29 @@ impl AggregateExec {
InputOrderMode::Linear
};
+ // To keep the ordering optimization simple, grouping sets are not supported.
+ // See [`OrderedPartialAggregateStream`] for more details about this optimization.
+ if group_by.has_grouping_set() {
+ input_order_mode = InputOrderMode::Linear;
+ }
+
// construct a map from the input expression to the output expression of the Aggregation group by
let group_expr_mapping =
ProjectionMapping::try_new(group_by.expr.clone(), &input.schema())?;
- let cache = Self::compute_properties(
- &input,
- Arc::clone(&schema),
- &group_expr_mapping,
- group_by.is_true_no_grouping(),
- &mode,
- &input_order_mode,
- aggr_expr.as_ref(),
- )?;
+ let cache = if group_by.has_grouping_set() {
+ Self::compute_grouping_set_properties(&input, Arc::clone(&schema))
+ } else {
+ Self::compute_properties(
+ &input,
+ Arc::clone(&schema),
+ &group_expr_mapping,
+ group_by.is_true_no_grouping(),
+ &mode,
+ &input_order_mode,
+ aggr_expr.as_ref(),
+ )?
+ };
let mut exec = AggregateExec {
mode,
@@ -1438,6 +1448,22 @@ impl AggregateExec {
))
}
+ fn compute_grouping_set_properties(
+ input: &Arc<dyn ExecutionPlan>,
+ schema: SchemaRef,
+ ) -> PlanProperties {
+ // Grouping-set expansion can replace group keys with nulls and adds a
+ // grouping ID, so input properties do not project through unchanged.
+ PlanProperties::new(
+ EquivalenceProperties::new(schema),
+ Partitioning::UnknownPartitioning(
+ input.output_partitioning().partition_count(),
+ ),
+ EmissionType::Final,
+ input.boundedness(),
+ )
+ }
+
pub fn input_order_mode(&self) -> &InputOrderMode {
&self.input_order_mode
}
diff --git a/datafusion/sqllogictest/test_files/grouping.slt b/datafusion/sqllogictest/test_files/grouping.slt
index 7d893f94be..ed7b5c26f1 100644
--- a/datafusion/sqllogictest/test_files/grouping.slt
+++ b/datafusion/sqllogictest/test_files/grouping.slt
@@ -48,6 +48,97 @@ NULL A 1 0 2 1
NULL B 1 0 2 1
NULL NULL 1 1 3 3
+# Grouping sets must not use the ordered aggregation optimization even when the
+# input is ordered by a grouping expression that is present in every set.
+statement ok
+SET datafusion.execution.target_partitions = 1;
+
+statement ok
+SET datafusion.execution.enable_migration_aggregate = false;
+
+statement ok
+CREATE EXTERNAL TABLE grouping_set_ordered (
+ a0 INTEGER,
+ a INTEGER,
+ b INTEGER,
+ c INTEGER,
+ d INTEGER
+)
+STORED AS CSV
+WITH ORDER (c)
+LOCATION '../core/tests/data/window_2.csv'
+OPTIONS ('format.has_header' 'true');
+
+# An unordered AggregateExec omits `ordering_mode`; its absence below ensures
+# that grouping sets do not select ordered/streaming aggregation.
+query TT
+EXPLAIN
+SELECT b, c, grouping(b, c) AS gid
+FROM grouping_set_ordered
+GROUP BY GROUPING SETS ((b, c), (c));
+----
+logical_plan
+01)Projection: grouping_set_ordered.b, grouping_set_ordered.c, CAST(__grouping_id & UInt8(3) AS Int32) AS gid
+02)--Aggregate: groupBy=[[GROUPING SETS ((grouping_set_ordered.b, grouping_set_ordered.c), (grouping_set_ordered.c))]], aggr=[[]]
+03)----TableScan: grouping_set_ordered projection=[b, c]
+physical_plan
+01)ProjectionExec: expr=[b@0 as b, c@1 as c, CAST(__grouping_id@2 & 3 AS Int32) as gid]
+02)--AggregateExec: mode=Final, gby=[b@0 as b, c@1 as c, __grouping_id@2 as __grouping_id], aggr=[]
+03)----AggregateExec: mode=Partial, gby=[(b@0 as b, c@1 as c), (NULL as b, c@1 as c)], aggr=[]
+04)------DataSourceExec: file_groups={1 group: [[WORKSPACE_ROOT/datafusion/core/tests/data/window_2.csv]]}, projection=[b, c], output_ordering=[c@1 ASC NULLS LAST], file_type=csv, has_header=true
+
+# The grouping sets produce two rows for each value of c.
+query I
+SELECT c
+FROM (
+ SELECT b, c, grouping(b, c) AS gid
+ FROM grouping_set_ordered
+ GROUP BY GROUPING SETS ((b, c), (c))
+)
+ORDER BY c
+LIMIT 10;
+----
+0
+0
+1
+1
+2
+2
+3
+3
+4
+4
+
+statement ok
+SET datafusion.execution.enable_migration_aggregate = true;
+
+query I
+SELECT c
+FROM (
+ SELECT b, c, grouping(b, c) AS gid
+ FROM grouping_set_ordered
+ GROUP BY GROUPING SETS ((b, c), (c))
+)
+ORDER BY c
+LIMIT 10;
+----
+0
+0
+1
+1
+2
+2
+3
+3
+4
+4
+
+statement ok
+RESET datafusion.execution.enable_migration_aggregate;
+
+statement ok
+SET datafusion.execution.target_partitions = 4;
+
# grouping_with_cube
query TTIIII
select
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@datafusion.apache.org
For additional commands, e-mail: commits-help@datafusion.apache.org