Repository navigation
test: use WITH ORDER tables for date_bin/date_trunc ordering plan tests - #26332
Merged
Merged
Conversation
Read the sorted input of the ordering propagation EXPLAIN tests in timestamps.slt from CSV tables declared with `WITH ORDER`, instead of an `ORDER BY ... LIMIT n` subquery. The subquery left a TopK in every plan, so a plan where the outer sort was removed still showed a SortExec. Positive cases now have no SortExec at all, and the named-timezone case keeps exactly the one required sort.
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #26332 +/- ##
========================================
Coverage 82.79% 82.80%
========================================
Files 1149 1149
Lines 454146 454480 +334
Branches 454146 454480 +334
========================================
+ Hits 376032 376330 +298
- Misses 55262 55280 +18
- Partials 22852 22870 +18 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
alamb
approved these changes
Oct 15, 2026
| 562129259 562129200 | ||
| # The ordering propagation tests below read from tables declared sorted with | ||
| # `WITH ORDER`, so a plan has no sort at all when the ordering propagates, and | ||
| # exactly the final sort when it must not. Each file is written in its |
| 01)ProjectionExec: expr=[date_trunc(hour, from_unixtime(column1@0)) as truncated] | ||
| 02)--SortExec: TopK(fetch=2), expr=[column1@0 ASC NULLS LAST], preserve_partitioning=[false] | ||
| 03)----DataSourceExec: partitions=1, partition_sizes=[1] | ||
| 01)SortPreservingMergeExec: [truncated@0 ASC NULLS LAST] |
Member
Author
|
Thank you @alamb ! |
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
Which issue does this PR close?
Rationale for this change
In #25815, @alamb asked why a plan still had a
SortExecwhen ordering propagation should remove it. That sort was theTopK(fetch=n)from theORDER BY ... LIMIT nsubquery used to produce ordered input. The outer sort, the one the test was about, had already been removed. Reading the input from a table declared sorted makes the plans show this directly: no sort at all when the ordering propagates, and only the final sort when it does not.What changes are included in this PR?
Test-only. In
datetime/timestamps.slt, the seven ordering-propagationEXPLAINtests throughdate_bin,date_truncandfrom_unixtime(plus the result queries next to them) now read from CSV files written withCOPY ... TOand registered withCREATE EXTERNAL TABLE ... WITH ORDER. This follows the pattern inorder.slt. EachDataSourceExecreports the declaredoutput_ordering.from_unixtime, timezone-freedate_bin, coarse-precision values outside the nanosecond range, positive month stride, nesteddate_trunc) now have noSortExec.America/Goose_Baycase keeps exactly oneSortExec, the required final sort.Timestamp(Second)columns come fromTIMESTAMP(0), the coarse-precision case still uses ±10,000,000,000 seconds, and every query result is unchanged. The Goose Bay result query moved next to itsEXPLAIN.I did not find any other
date_bin/date_truncordering plan tests that use theORDER BY ... LIMITsubquery pattern. The related tests ingroup_by.sltandsort_pushdown.sltalready useWITH ORDER.What is the testing strategy for this PR?
This PR only changes tests.
datetime/timestamps.sltpasses.Are there any user-facing changes?
No.