Github user rdblue commented on a diff in the pull request:
https://github.com/apache/spark/pull/20521#discussion_r166407366
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/DataSource.scala
---
@@ -493,9 +510,23 @@ case class DataSource(
dataSource.createRelation(
sparkSession.sqlContext, mode, caseInsensitiveOptions,
Dataset.ofRows(sparkSession, data))
case format: FileFormat =>
-
sparkSession.sessionState.executePlan(planForWritingFileFormat(format, mode,
data)).toRdd
+ val cmd = planForWritingFileFormat(format, mode, data)
+ val resolvedPartCols = cmd.partitionColumns.map { col =>
+ // The partition columns created in `planForWritingFileFormat`
should always be
+ // `UnresolvedAttribute` with a single name part.
+ assert(col.isInstanceOf[UnresolvedAttribute])
+ val unresolved = col.asInstanceOf[UnresolvedAttribute]
+ assert(unresolved.nameParts.length == 1)
+ val name = unresolved.nameParts.head
+ outputColumns.find(a => equality(a.name, name)).getOrElse {
+ throw new AnalysisException(
+ s"Unable to resolve $name given
[${data.output.map(_.name).mkString(", ")}]")
+ }
+ }
+ val resolved = cmd.copy(partitionColumns = resolvedPartCols,
outputColumns = outputColumns)
--- End diff --
I get the point of passing the physical plan, and I think that's a good
idea. What I don't understand is why the command doesn't match the physical
plan that is passed in. Is that physical plan based on a different logical
plan? I would expect that the physical plan is created once and passed into
run, but that it was created from the logical plan that is also passed in.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
