Raviston6296 commented on PR #59841:
URL: https://github.com/apache/spark/pull/59841#issuecomment-6570998419
While checking #59842, I applied this PR on top of current master
(`4fecd73323f`) and found a regression when SortMergeJoin functions are spilled
into a nested class.
```sql
-- spark.sql.codegen.wholeStage.splitExpressions.methodLimit=0, AQE off,
broadcast off
SELECT /*+ MERGE(r) */ l.*, r.*, CASE WHEN r.c1 = 1 THEN r.c1 * 1 ... (2000
branches) ELSE 0 END AS v
FROM l JOIN r ON l.k = r.k
-- l, r: spark.range(0, 10) / spark.range(5, 15) with `id AS k` plus 1, 40
or 150 extra columns
```
On master, all 27 combinations (inner / left outer / right outer, 1 / 40 /
150 extra columns, CASE WHEN on `l.k`, `r.k` or `r.c1`) compile and spill into
a nested class. With this PR applied, all 27 fail to compile:
```
org.codehaus.commons.compiler.InternalCompilerException: Code grows beyond
64 KB
... Compiling "processNext()"
```
--
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]
