[ 
https://issues.apache.org/jira/browse/SPARK-55038?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18054297#comment-18054297
 ] 

Anirudh Bhatnagar commented on SPARK-55038:
-------------------------------------------

I have started working on it.

> AQE + sortBased aggregation produces wrong results for array_agg(DISTINCT) in 
> correlated subqueries
> ---------------------------------------------------------------------------------------------------
>
>                 Key: SPARK-55038
>                 URL: https://issues.apache.org/jira/browse/SPARK-55038
>             Project: Spark
>          Issue Type: Bug
>          Components: SQL
>    Affects Versions: 3.5.0, 4.0.0
>         Environment:   - Spark 4.0.1 (also reproducible on 3.5.x)
>   - Reproduced in both local[4] and distributed mode
>  
>            Reporter: Feng Zhang
>            Priority: Critical
>
> h4. Summary
> When `spark.sql.adaptive.enabled=true` 
> and`spark.sql.objectHashAggregate.sortBased.fallbackThreshold=1`,  queries 
> using `array_agg(DISTINCT)` in correlated subqueries produce incorrect 
> results.
>  
> h4. Reproducer
> {code:java}
> spark.conf.set("spark.sql.adaptive.enabled", "true")
> spark.conf.set("spark.sql.objectHashAggregate.sortBased.fallbackThreshold", 
> "1")  spark.sql("""
>     WITH t AS (SELECT explode(array(15, 16, 17)) AS v)
>     SELECT (SELECT COUNT(*) FROM t WHERE array_contains(arr, v)) AS cnt, arr
>     FROM (SELECT collect_set(v) AS arr FROM t)
> """).show() 
> -- Expected: cnt = 3
> -- Actual: cnt = 0
>   {code}
>  
> h4. Root Cause Analysis
>  
> The correlated subquery is decorrelated into a hash join using the array as 
> the join key.
> When array_agg(DISTINCT) (which uses collect_list(distinct)) is computed 
> multiple times with sort-based aggregation + AQE, the computations produce 
> arrays with the same elements but different orderings. Since array equality 
> is element-by-element, the hash join fails to match.
>  
> h4. Workarounds
>  
>   - Set spark.sql.adaptive.enabled=false, OR
>   - Set spark.sql.objectHashAggregate.sortBased.fallbackThreshold to default 
> (128)
>  
> h4. Test Results
>  
>   - AQE=true, sortBased.fallbackThreshold=1: FAIL (cnt=0, expected 3)
>   - AQE=false, sortBased.fallbackThreshold=1: PASS (cnt=3)
>   - AQE=true, sortBased.fallbackThreshold=128 (default): PASS (cnt=3)
>  
> h4. Conclusion
>  
> The bug only occurs when both AQE is enabled and 
> sortBased.fallbackThreshold=1.
>  



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to