ryukobayashi opened a new pull request, #6486:
URL: https://github.com/apache/hive/pull/6486
<!--
Thanks for sending a pull request! Here are some tips for you:
1. If this is your first time, please read our contributor guidelines:
https://cwiki.apache.org/confluence/display/Hive/HowToContribute
2. Ensure that you have created an issue on the Hive project JIRA:
https://issues.apache.org/jira/projects/HIVE/summary
3. Ensure you have added or run the appropriate tests for your PR:
4. If the PR is unfinished, add '[WIP]' in your PR title, e.g.,
'[WIP]HIVE-XXXXX: Your PR title ...'.
5. Be sure to keep the PR description updated to reflect all changes.
6. Please write your PR title to summarize what this PR proposes.
7. If possible, provide a concise example to reproduce the issue for a
faster review.
-->
### What changes were proposed in this pull request?
<!--
Please clarify what changes you are proposing. The purpose of this section
is to outline the changes and how this PR fixes the issue.
If possible, please consider writing useful notes for better and faster
reviews in your PR. See the examples below.
1. If you refactor some codes with changing classes, showing the class
hierarchy will help reviewers.
2. If you fix some SQL features, you can provide some references of other
DBMSes.
3. If there is design documentation, please add the link.
4. If there is a discussion in the mailing list, please add the link.
-->
In vectorized outer join, `generateOuterNulls()` and
`generateOuterNullsRepeatedAll()` set `isNull[i] = true` on scratch columns but
leave `vector[i]` untouched. When `hive.vectorized.reuse.scratch.columns=true`
(the default), a scratch column slot
freed after an expression evaluation (e.g. `CastStringToLong`) can be
reused for the outer join's null-marking column. After `reset()` clears
`isNull[]`, the expression overwrites `vector[i]` with a fresh value (e.g.
2025). Later, `generateOuterNulls()`
sets `isNull[i] = true` without clearing `vector[i]`, leaving a stale
non-zero value.
Downstream operators such as `ColOrCol` read `vector[i]` directly to
distinguish "false" (`== 0`) from "null" (`!= 0`). The stale value causes null
rows to be misinterpreted as "true", producing wrong OR/AND/CASE WHEN results.
The fix adds `clearVectorValue()`, called whenever `isNull[i]` is set to
`true` in the outer join null-marking paths, zeroing `vector[i]` for all
supported column vector types (`LongColumnVector`, `DoubleColumnVector`,
`BytesColumnVector`,
`TimestampColumnVector`, `IntervalDayTimeColumnVector`).
### Why are the changes needed?
<!--
Please clarify why the changes are needed. For instance,
1. If you propose a new API, clarify the use case for a new API.
2. If you fix a bug, you can clarify why it is a bug.
-->
Without the fix, vectorized outer joins silently return wrong results when
scratch column reuse is enabled (the default). The bug is non-obvious because
it only triggers when a specific combination of conditions is met: a
type-casting expression allocates a scratch column that is later reused for the
outer join's null-marking column, and the join result is consumed by a boolean
operator that reads the raw vector value for null discrimination. Users have no
indication that results are wrong; workarounds require disabling vectorization
entirely (`hive.vectorization.enabled=false`) or disabling scratch column reuse
(`hive.vectorized.reuse.scratch.columns=false`), both of which carry a
significant performance cost.
### Does this PR introduce _any_ user-facing change?
<!--
Note that it means *any* user-facing change including all aspects such as
the documentation fix.
If yes, please clarify the previous behavior and the change this PR proposes
- provide the console output, description, screenshot and/or a reproducable
example to show the behavior difference if possible.
If possible, please also clarify if this is a user-facing change compared to
the released Hive versions or within the unreleased branches such as master.
If no, write 'No'.
-->
No
### How was this patch tested?
<!--
If tests were added, say they were added here. Please make sure to add some
test cases that check the changes thoroughly including negative and positive
cases if possible.
If it was tested in a way different from regular unit tests, please clarify
how you tested step by step, ideally copy and paste-able, so that other
reviewers can test and check, and descendants can verify in the future.
If tests were not added, please describe why they were not added and/or why
it was difficult to add.
-->
The existing TestMapJoinOperator suite (17 tests) passes without
regression. The bug can also be verified manually with the minimal SQL
reproducer below; with the fix applied, the result matches the expected output
(C 2026 new, D 2026 new) that was previously only obtainable by disabling
vectorization.
```
CREATE TABLE src (k STRING, v STRING);
INSERT INTO src VALUES
('p','1'),('p','2'),('p','3'),
('q','2'),('q','3'),
('r','3'),
('s','3');
WITH base AS (
SELECT k, v FROM src GROUP BY k, v
),
classified AS (
SELECT t1.k, t1.v,
CASE WHEN COALESCE(t2.k,'') = '' THEN 'new'
WHEN COALESCE(t3.k,'') = '' THEN 'two_step'
ELSE 'three_step' END AS status
FROM base t1
LEFT JOIN base t2
ON t1.k = t2.k
AND CAST(t1.v AS INT) - 1 = CAST(t2.v AS INT)
LEFT JOIN base t3
ON t1.k = t3.k
AND CAST(t1.v AS INT) - 2 = CAST(t3.v AS INT)
WHERE CAST(t1.v AS INT) >= 3
GROUP BY t1.k, t1.v,
CASE WHEN COALESCE(t2.k,'') = '' THEN 'new'
WHEN COALESCE(t3.k,'') = '' THEN 'two_step'
ELSE 'three_step' END
)
SELECT * FROM classified WHERE status = 'new';
```
--
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]
