Xtpacz opened a new issue, #12260:
URL: https://github.com/apache/gluten/issues/12260
### Backend
VL (Velox)
### Bug description
### Description
When a Project contains decimal arithmetic of the form `<bigint_aggregate> /
<integer-valued decimal literal>`, the entire Project is incorrectly falled
back to JVM by `ColumnarPartialProjectRule`, losing native Velox acceleration.
The root cause is in `CheckOverflowTransformer`: it uses
`original.child.dataType` (Spark's declared type) instead of `child.dataType`
(the Gluten transformer's actual output type) when deciding whether to insert a
cast, generating a substrait plan that fails Velox SimpleFunction type
validation.
### Reproducer
```sql
CREATE TABLE t1 (val BIGINT) USING parquet;
CREATE TABLE t2 (val BIGINT) USING parquet;
INSERT INTO t1 VALUES (200);
INSERT INTO t2 VALUES (100), (100), (100), (100), (100);
SELECT
a.val,
(a.val - COALESCE(SUM(b.val), 0) / 5.0)
/ (COALESCE(SUM(b.val), 0) / 5.0) AS growth_rate
FROM t1 a CROSS JOIN t2 b
GROUP BY a.val;
```
Spark UI shows `Project` (JVM) instead of `ProjectExecTransformer`.
### Trigger conditions (all required)
1. Aggregate over an integer column (e.g. `SUM(bigint_col)`)
2. Divided by an integer-valued decimal literal (e.g. `5.0`, `3.0`, `2.0`)
3. Literal's `BigDecimal.isValidLong` is true (triggers `rescaleLiteral`)
4. Non-constant column reference + `GROUP BY` (prevents constant folding)
### Gluten version
main branch
### Spark version
Spark-3.3.x
### Spark configurations
_No response_
### System information
_No response_
### Relevant logs
```bash
```
--
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]
