Github user viirya commented on a diff in the pull request:
https://github.com/apache/spark/pull/19803#discussion_r152882655
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/execution/DataSourceScanExec.scala
---
@@ -355,19 +355,12 @@ case class FileSourceScanExec(
// PhysicalRDD always just has one input
val input = ctx.freshName("input")
ctx.addMutableState("scala.collection.Iterator", input, s"$input =
inputs[0];")
- val exprRows = output.zipWithIndex.map{ case (a, i) =>
- BoundReference(i, a.dataType, a.nullable)
- }
val row = ctx.freshName("row")
- ctx.INPUT_ROW = row
- ctx.currentVars = null
- val columnsRowInput = exprRows.map(_.genCode(ctx))
- val inputRow = if (needsUnsafeRowConversion) null else row
--- End diff --
When `needsUnsafeRowConversion` is true, previously we pass in a null and
`consume` will generate code for a unsafe row projection. Now we always pass in
the `row` which can possibly not a unsafe row. This looks changing behavior?
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
