eldenmoon opened a new pull request, #64205:
URL: https://github.com/apache/doris/pull/64205
### What problem does this PR solve?
Issue Number: None
Related PR: #64102
Problem Summary: The TeamCity regression
`variant_p0/predefine/test_variant_compaction_with_sparse_limit` could miss
rows when `default_variant_max_subcolumns_count` was 1 and sparse column
statistics reached the configured limit. A segment could record the exact
sparse path `b` while omitting child or prefix information such as `b.c`, so
the reader chose binary extraction for `v["b"]` and skipped hidden child sparse
paths. This change treats truncated sparse statistics as inconclusive for both
recorded exact sparse paths and missing child paths, forcing hierarchical reads
so exact and child sparse values are merged. The regression now fixes DUPLICATE
and UNIQUE tables to `variant_max_subcolumns_count = 1` to cover the failing
shape deterministically while keeping AGGREGATE on the default larger limit.
### Release note
Fix incorrect variant sparse path reads when sparse column statistics are
truncated.
### Check List (For Author)
- Test: Regression test
- `PATH=/mnt/disk1/claude-max/ldb_toolchain16/bin:$PATH
build-support/clang-format.sh`
- `BUILD_TYPE=ASAN USE_MEM_TRACKER=ON ./build.sh --be`
- `./run-regression-test.sh --conf tmp/regression-conf.auto.groovy --run
-d variant_p0/predefine -s test_compaction_variant_predefine_with_sparse_limit
-forceGenOut`
- `./run-regression-test.sh --conf tmp/regression-conf.auto.groovy --run
-d variant_p0/predefine -s test_compaction_variant_predefine_with_sparse_limit`
- Behavior changed: Yes. Fixes incorrect query results for variant sparse
paths when sparse statistics are truncated.
- Does this need documentation: 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: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
