You are viewing a plain text version of this content. The canonical link for it is here.
Posted to github@datafusion.apache.org by "hareshkh (via GitHub)" <gi...@apache.org> on 2026/03/24 22:59:33 UTC
[PR] [branch-52] Fix push_down_filter for children with non-empty fetch fields (#21057) [datafusion]
hareshkh opened a new pull request, #21141:
URL: https://github.com/apache/datafusion/pull/21141
## 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.
-->
- Related to https://github.com/apache/datafusion/issues/21063
- Related to https://github.com/apache/datafusion/issues/21078
## 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.
-->
Currently if we see a filter with a limit underneath, we don't push the filter past the limit. However, sort nodes and table scan nodes can have fetch fields which do essentially the same thing, and we don't stop filters being pushed past them. This is a correctness bug that can lead to undefined behaviour.
I added checks for exactly this condition so we don't push the filter down. I think the prior expectation was that there would be a limit node between any of these nodes, but this is also not true. In `push_down_limit.rs`, there's code that does this optimisation when a limit has a sort under it:
```rust
LogicalPlan::Sort(mut sort) => {
let new_fetch = {
let sort_fetch = skip + fetch;
Some(sort.fetch.map(|f| f.min(sort_fetch)).unwrap_or(sort_fetch))
};
if new_fetch == sort.fetch {
if skip > 0 {
original_limit(skip, fetch, LogicalPlan::Sort(sort))
} else {
Ok(Transformed::yes(LogicalPlan::Sort(sort)))
}
} else {
sort.fetch = new_fetch;
limit.input = Arc::new(LogicalPlan::Sort(sort));
Ok(Transformed::yes(LogicalPlan::Limit(limit)))
}
}
```
The first time this runs, it sets the internal fetch of the sort to new_fetch, and on the second optimisation pass it hits the branch where we just get rid of the limit node altogether, leaving the sort node exposed to potential filters which can now push down into it.
There is also a related fix in `gather_filters_for_pushdown` in `SortExec`, which does the same thing for physical plan nodes. If we see that a given execution plan has non-empty fetch, it should not allow any parent filters to be pushed down.
## 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.
-->
Added checks in the optimisation rule to avoid pushing filters past children with built-in limits.
## 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)?
-->
Yes:
- Unit tests in `push_down_filter.rs`
- Fixed an existing test in `window.slt`
- Unit tests for the physical plan change in `sort.rs`
- New slt test in `push_down_filter_sort_fetch.slt` for this exact behaviour
## 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
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] [branch-52] Fix push_down_filter for children with non-empty fetch fields (#21057) [datafusion]
Posted by "hareshkh (via GitHub)" <gi...@apache.org>.
hareshkh commented on code in PR #21141:
URL: https://github.com/apache/datafusion/pull/21141#discussion_r2984925611
##########
datafusion/physical-plan/src/filter_pushdown.rs:
##########
@@ -359,6 +358,17 @@ impl ChildFilterDescription {
})
}
+ /// Mark all parent filters as unsupported for this child.
Review Comment:
This is present on branch-53/main but not here - it is referenced in the logic, so added this too.
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] [branch-52] Fix push_down_filter for children with non-empty fetch fields (#21057) [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #21141:
URL: https://github.com/apache/datafusion/pull/21141#issuecomment-4144412829
There is no rush - I am thinking perhaps I'll make a new 54 release candidate next week
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] [branch-52] Fix push_down_filter for children with non-empty fetch fields (#21057) [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #21141:
URL: https://github.com/apache/datafusion/pull/21141#issuecomment-4173214414
Thanks again @hareshkh and @AndreaBozzo
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] [branch-52] Fix push_down_filter for children with non-empty fetch fields (#21057) [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #21141:
URL: https://github.com/apache/datafusion/pull/21141#issuecomment-4164567185
Any updates on testing @AndreaBozzo ?
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] [branch-52] Fix push_down_filter for children with non-empty fetch fields (#21057) [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb merged PR #21141:
URL: https://github.com/apache/datafusion/pull/21141
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] [branch-52] Fix push_down_filter for children with non-empty fetch fields (#21057) [datafusion]
Posted by "hareshkh (via GitHub)" <gi...@apache.org>.
hareshkh commented on PR #21141:
URL: https://github.com/apache/datafusion/pull/21141#issuecomment-4130722039
@alamb : Can you please take a look at this one?
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] [branch-52] Fix push_down_filter for children with non-empty fetch fields (#21057) [datafusion]
Posted by "hareshkh (via GitHub)" <gi...@apache.org>.
hareshkh commented on code in PR #21141:
URL: https://github.com/apache/datafusion/pull/21141#discussion_r2984925611
##########
datafusion/physical-plan/src/filter_pushdown.rs:
##########
@@ -359,6 +358,17 @@ impl ChildFilterDescription {
})
}
+ /// Mark all parent filters as unsupported for this child.
Review Comment:
This utility is present on branch-53/main but not here - it is referenced in the logic, so added this too.
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] [branch-52] Fix push_down_filter for children with non-empty fetch fields (#21057) [datafusion]
Posted by "AndreaBozzo (via GitHub)" <gi...@apache.org>.
AndreaBozzo commented on PR #21141:
URL: https://github.com/apache/datafusion/pull/21141#issuecomment-4164896115
> Any updates on testing @AndreaBozzo ?
LGTM @alamb @hareshkh , i just wanted to run few additional slts (which pass), also leaving them in the dropdown menu
<details>
<summary>Additional stress tests (click to expand)</summary>
```sql
# Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements. See the NOTICE file
# distributed with this work for additional information
# regarding copyright ownership. The ASF licenses this file
# to you under the Apache License, Version 2.0 (the
# "License"); you may not use this file except in compliance
# with the License. You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing,
# software distributed under the License is distributed on an
# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
# KIND, either express or implied. See the License for the
# specific language governing permissions and limitations
# under the License.
statement ok
CREATE TABLE data(id INT, value INT, category TEXT) AS VALUES
(1, 10, 'A'),
(2, 20, 'B'),
(3, 30, 'A'),
(4, 40, 'B'),
(5, 50, 'A'),
(6, 60, 'B'),
(7, 70, 'A'),
(8, 80, 'B'),
(9, 90, 'A'),
(10, 100, 'B');
###############################################
# 1. OFFSET (skip) — filter must NOT be pushed past OFFSET
###############################################
# Skip first 3 rows (by value ASC: 10,20,30), then filter value < 50.
# Correct: after skipping, rows are 40,50,60,70,80,90,100 → filter keeps only 40.
# Bug: if filter pushed below OFFSET, it would filter first then skip.
query II
SELECT id, value FROM (SELECT * FROM data ORDER BY value LIMIT 7 OFFSET 3) sub WHERE sub.value < 50;
----
4 40
# Verify the plan shape: filter stays above sort+fetch
query TT
EXPLAIN SELECT id, value FROM (SELECT * FROM data ORDER BY value LIMIT 7 OFFSET 3) sub WHERE sub.value < 50;
----
logical_plan
01)SubqueryAlias: sub
02)--Filter: data.value < Int32(50)
03)----Limit: skip=3, fetch=7
04)------Sort: data.value ASC NULLS LAST, fetch=10
05)--------TableScan: data projection=[id, value]
physical_plan
01)FilterExec: value@1 < 50
02)--GlobalLimitExec: skip=3, fetch=7
03)----SortExec: TopK(fetch=10), expr=[value@1 ASC NULLS LAST], preserve_partitioning=[false]
04)------DataSourceExec: partitions=1, partition_sizes=[1]
###############################################
# 2. OFFSET only (no LIMIT) — filter must NOT be pushed past OFFSET
###############################################
# Skip first 7 rows (values 10..70), then filter value > 90.
# Correct result: after skipping, rows are 80,90,100 → filter keeps 100.
query II
SELECT id, value FROM (SELECT * FROM data ORDER BY value OFFSET 7) sub WHERE sub.value > 90;
----
10 100
###############################################
# 3. Sort with fetch + filter on a different column
###############################################
# Top 5 by value ASC (ids 1-5, values 10-50), then filter by category = 'A'.
# Correct: ids 1,3,5 (values 10,30,50).
# Bug: if filter pushed below sort, it would filter category='A' first (ids 1,3,5,7,9)
# then take top 5 → ids 1,3,5,7,9 — WRONG!
query IIT
SELECT * FROM (SELECT * FROM data ORDER BY value LIMIT 5) sub WHERE sub.category = 'A';
----
1 10 A
3 30 A
5 50 A
###############################################
# 4. CTE with LIMIT — filter must not be pushed into CTE
###############################################
query II
WITH top3 AS (SELECT id, value FROM data ORDER BY value LIMIT 3)
SELECT * FROM top3 WHERE top3.value > 15;
----
2 20
3 30
# Verify plan shape
query TT
EXPLAIN WITH top3 AS (SELECT id, value FROM data ORDER BY value LIMIT 3)
SELECT * FROM top3 WHERE top3.value > 15;
----
logical_plan
01)SubqueryAlias: top3
02)--Filter: data.value > Int32(15)
03)----Sort: data.value ASC NULLS LAST, fetch=3
04)------TableScan: data projection=[id, value]
physical_plan
01)FilterExec: value@1 > 15
02)--SortExec: TopK(fetch=3), expr=[value@1 ASC NULLS LAST], preserve_partitioning=[false]
03)----DataSourceExec: partitions=1, partition_sizes=[1]
###############################################
# 5. Nested subqueries with multiple limits
###############################################
# Inner: top 5 by value ASC (10,20,30,40,50)
# Middle: filter value > 20 → (30,'A',40,'B',50,'A')
# Outer: take top 2 of that → (30,'A'),(40,'B')
query IIT
SELECT * FROM (
SELECT * FROM (
SELECT * FROM data ORDER BY value LIMIT 5
) inner_sub WHERE inner_sub.value > 20
ORDER BY value LIMIT 2
) outer_sub;
----
3 30 A
4 40 B
###############################################
# 6. Filter after LIMIT on TableScan (no Sort)
###############################################
# Without ORDER BY, LIMIT on a TableScan is non-deterministic,
# but the filter must still not be pushed past it.
# We just verify the plan shape here.
query TT
EXPLAIN SELECT * FROM (SELECT * FROM data LIMIT 4) sub WHERE sub.value > 30;
----
logical_plan
01)SubqueryAlias: sub
02)--Filter: data.value > Int32(30)
03)----Limit: skip=0, fetch=4
04)------TableScan: data projection=[id, value, category], fetch=4
physical_plan
01)FilterExec: value@1 > 30
02)--DataSourceExec: partitions=1, partition_sizes=[1], fetch=4
###############################################
# 7. Multiple filters — none should be pushed past sort+fetch
###############################################
query IIT
SELECT * FROM (SELECT * FROM data ORDER BY value LIMIT 6) sub
WHERE sub.value > 15 AND sub.category = 'A';
----
3 30 A
5 50 A
###############################################
# 8. Filter after LIMIT in a JOIN
###############################################
# Create a small dimension table
statement ok
CREATE TABLE dim(category TEXT, label TEXT) AS VALUES
('A', 'Alpha'),
('B', 'Beta');
# Join the top-3 by value with the dimension table, then filter.
# Top 3: (1,10,'A'), (2,20,'B'), (3,30,'A')
# After join: (1,10,'A','Alpha'), (2,20,'B','Beta'), (3,30,'A','Alpha')
# Filter value > 15: (2,20,'B','Beta'), (3,30,'A','Alpha')
query IITT
SELECT sub.id, sub.value, sub.category, dim.label
FROM (SELECT * FROM data ORDER BY value LIMIT 3) sub
JOIN dim ON sub.category = dim.category
WHERE sub.value > 15
ORDER BY sub.value;
----
2 20 B Beta
3 30 A Alpha
###############################################
# 9. UNION ALL with LIMIT — filter must respect limits
###############################################
# Each branch takes top 2; filter must apply AFTER the limit in each branch.
# Category A top 2 by value: (1,10), (3,30). Category B top 2: (2,20), (4,40).
# Union: (1,10),(3,30),(2,20),(4,40). Filter value > 15: (3,30),(2,20),(4,40).
query II
SELECT * FROM (
SELECT * FROM (SELECT id, value FROM data WHERE category = 'A' ORDER BY value LIMIT 2) a
UNION ALL
SELECT * FROM (SELECT id, value FROM data WHERE category = 'B' ORDER BY value LIMIT 2) b
) sub WHERE sub.value > 15
ORDER BY value;
----
2 20
3 30
4 40
###############################################
# 10. Correctness: verify exact row counts
###############################################
# Top 4 by value → {10,20,30,40}, filter >25 → {30,40} → count=2
query I
SELECT COUNT(*) FROM (SELECT * FROM data ORDER BY value LIMIT 4) sub WHERE sub.value > 25;
----
2
# If filter were incorrectly pushed down: filter first gives {30,40,50,60,70,80,90,100},
# then limit 4 → {30,40,50,60} → count=4. So count=2 proves correctness.
statement ok
DROP TABLE dim;
statement ok
DROP TABLE data;
```
</details>
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org