Github user icexelloss commented on a diff in the pull request:
https://github.com/apache/spark/pull/22104#discussion_r210414941
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/execution/python/EvalPythonExec.scala
---
@@ -117,15 +117,18 @@ abstract class EvalPythonExec(udfs: Seq[PythonUDF],
output: Seq[Attribute], chil
}
}.toArray
}.toArray
- val projection = newMutableProjection(allInputs, child.output)
+
+ // Project input rows to unsafe row so we can put it in the row queue
+ val unsafeProjection = UnsafeProjection.create(child.output,
child.output)
--- End diff --
This requires some discussion.
This is probably another bug I found in testing this - If the input node to
EvalPythonExec doesn't produce UnsafeRow, and cast here will fail. I don't know
if we require data sources to produce unsafe rows, if not, then this is a
problem.
I also don't know if this will introduce additional copy if input is
already UnsafeRow - it seems like UnsafeProject should be smart to skip the
copy but I am not sure if it's actually the case
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
