mbutrovich opened a new issue, #4788:
URL: https://github.com/apache/datafusion-comet/issues/4788
### Describe the bug
`flatten` on an `array<array<T>>` column returns wrong results and silently
drops nulls when a row's outer array contains a null sub-array. Spark returns
`null` for any such row; Comet returns a non-null, misaligned array. Silent
data corruption, not a crash.
### Steps to reproduce
Add to a suite extending `CometTestBase`:
```scala
test("flatten with null sub-array") {
val data = Seq(
Tuple1(Seq(Seq(1, 2, 3), Seq(4, 5))),
Tuple1(Seq[Seq[Int]](Seq(1), null)), // Spark: flatten -> null
Tuple1(Seq[Seq[Int]](null, null))) // Spark: flatten -> null
withParquetTable(data, "t") {
checkSparkAnswerAndOperator("SELECT flatten(_1) FROM t")
}
}
```
```
== Results ==
!== Spark Answer - 3 == == Comet Answer - 3 ==
struct<flatten(_1):array<int>> struct<flatten(_1):array<int>>
![List(1, 2, 3, 4, 5)] [List()]
![null] [List(1)]
![null] [List(1, 2, 3, 4, 5)]
```
### Expected behavior
Match Spark: a null sub-array makes `flatten` return `null` for that row.
### Additional context
Found while enabling `CometLocalTableScanExec` by default (#4393), but
reproduces over a plain Parquet scan. Upstream test: `DataFrameFunctionsSuite`
"flatten function".
--
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]
