This is an automated email from the ASF dual-hosted git repository.
yiguolei pushed a commit to branch branch-4.1
in repository https://gitbox.apache.org/repos/asf/doris.git
The following commit(s) were added to refs/heads/branch-4.1 by this push:
new 052eb8ded8c branch-4.1: [test](regression) Stabilize search score
cache test #65894 (#68252)
052eb8ded8c is described below
commit 052eb8ded8c1f023b04acb6943af865c595bd9a4
Author: Jack <[email protected]>
AuthorDate: Sun Sep 20 19:49:48 2026 +0800
branch-4.1: [test](regression) Stabilize search score cache test #65894
(#68252)
### What problem does this PR solve?
Issue Number: N/A
Related PR: #65894
Problem Summary:
Backport of apache/doris#65894 to `branch-4.1`. Test-only change.
`search/test_search_score_cache` fails intermittently when the
regression cluster runs with fuzzy session variables.
`SessionVariable.initFuzzyModeVariables()` randomizes the VARIANT
defaults used at CREATE TABLE time: doc mode is enabled with 50%
probability, and `default_variant_doc_materialization_min_rows` is then
set to `nextInt(20)` with 50% probability. The suite inserts only 8
rows, so whenever the threshold lands in 9..19 the `host` path is kept
only in the doc-value column, is not materialized as a subcolumn and
gets no inverted index. `CAST(v["host"] AS STRING) MATCH_PHRASE ...`
then falls back to the non-index match path, no BM25 scorer runs,
`score()` is filled with 0 and the `score() > 0` assertion fails at
`test_search_score_cache.groovy:24 / :36 / :189`.
The backport pins the three VARIANT session variables before CREATE
TABLE so `v.host` is always materialized as an indexed subcolumn. It is
unrelated to the inverted index query cache.
Source PR: https://github.com/apache/doris/pull/65894
Source commit: `1197ae1fa9800b7026317e3bd2fd2f26b527053b` (squash merge,
single parent), picked with `git cherry-pick -x`, no conflict, original
author preserved.
No revert or follow-up fix of #65894 exists on master. #66858 later
touched the same suite, but it is the "Variant V2 by default" feature,
not a fix of this change, and is not part of this backport.
### Backport hunk audit
| # | Source file and hunk | Result | Target evidence |
| --- | --- | --- | --- |
| 1 | `regression-test/suites/search/test_search_score_cache.groovy @@
-43,6 +43,11 @@ suite("test_search_score_cache", "p0") {` | Ported
(verbatim) | Target context lines 43-46 are identical to the source; the
resulting hunk is byte-for-byte equal to the source diff. All three
session variables exist in `SessionVariable.java` on `branch-4.1`. |
### Verification
Format, compile and unit tests do not apply: the only changed file is a
Groovy regression suite. Because the source PR changes a regression
suite, the suite was actually run on the target branch.
Cluster: single FE + single BE, shared-nothing, fuzzy session variables
off, binaries `doris-4.1.4-rc04` (built from branch-4.1 on 2026-09-16;
the change is test-only, so no rebuild is needed). Command:
`./run-regression-test.sh --run --conf <conf> -d search -s
test_search_score_cache`.
| Step | Setup | Result |
| --- | --- | --- |
| RED | Unpatched suite from `origin/branch-4.1`; `set global
default_variant_enable_doc_mode=true`,
`default_variant_max_subcolumns_count=0`,
`default_variant_doc_materialization_min_rows=10` | `Test 1 suites,
failed 1 suites`; stack `test_search_score_cache.groovy:24` -> `:36` ->
`:189`. The table is created with `variant_enable_doc_mode=true`,
`variant_doc_materialization_min_rows=10`; the score query returns ids 4
and 2 with `score() = 0`; BE logs `collection_statistics.cpp:63] Index
statistics collection failed: [E-6003] ... No index with id ...`. |
| GREEN | This PR's suite, same forced globals | `Test 1 suites, failed
0 suites` |
| GREEN | This PR's suite, default globals | `Test 1 suites, failed 0
suites` |
### Release note
None
### Check List (For Author)
- Test
- [x] Regression test
- [ ] Unit Test
- [ ] Manual test (add detailed scripts or steps below)
- [ ] No need to test or manual test. Explain why:
- [ ] This is a refactor/code format and no logic has been changed.
- [ ] Previous test can cover this change.
- [ ] No code files have been changed.
- [ ] Other reason
- Behavior changed:
- [x] No.
- [ ] Yes.
- Does this need documentation?
- [x] No.
- [ ] Yes.
### Check List (For Reviewer who merge this PR)
- [ ] Confirm the release note
- [ ] Confirm test cases
- [ ] Confirm document
- [ ] Add branch pick label
Co-authored-by: liangj777
<[email protected]>
---
regression-test/suites/search/test_search_score_cache.groovy | 5 +++++
1 file changed, 5 insertions(+)
diff --git a/regression-test/suites/search/test_search_score_cache.groovy
b/regression-test/suites/search/test_search_score_cache.groovy
index caa3259195d..5c393980a5e 100644
--- a/regression-test/suites/search/test_search_score_cache.groovy
+++ b/regression-test/suites/search/test_search_score_cache.groovy
@@ -43,6 +43,11 @@ suite("test_search_score_cache", "p0") {
""")
}
+ // Keep v.host materialized as an indexed scalar subcolumn so score() can
collect BM25.
+ sql """ set default_variant_enable_doc_mode = false """
+ sql """ set default_variant_max_subcolumns_count = 0 """
+ sql """ set default_variant_doc_materialization_min_rows = 0 """
+
sql "DROP TABLE IF EXISTS ${tableName}"
sql """
CREATE TABLE ${tableName} (
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
