Github user cloud-fan commented on a diff in the pull request:
https://github.com/apache/spark/pull/23208#discussion_r239682984
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/DataFrameWriter.scala ---
@@ -241,32 +241,28 @@ final class DataFrameWriter[T] private[sql](ds:
Dataset[T]) {
assertNotBucketed("save")
- val cls = DataSource.lookupDataSource(source,
df.sparkSession.sessionState.conf)
- if (classOf[DataSourceV2].isAssignableFrom(cls)) {
- val source =
cls.getConstructor().newInstance().asInstanceOf[DataSourceV2]
- source match {
- case provider: BatchWriteSupportProvider =>
- val sessionOptions = DataSourceV2Utils.extractSessionConfigs(
- source,
- df.sparkSession.sessionState.conf)
- val options = sessionOptions ++ extraOptions
-
+ val session = df.sparkSession
+ val cls = DataSource.lookupDataSource(source,
session.sessionState.conf)
+ if (classOf[TableProvider].isAssignableFrom(cls)) {
+ val provider =
cls.getConstructor().newInstance().asInstanceOf[TableProvider]
+ val sessionOptions = DataSourceV2Utils.extractSessionConfigs(
+ provider, session.sessionState.conf)
+ val options = sessionOptions ++ extraOptions
+ val dsOptions = new DataSourceOptions(options.asJava)
+ provider.getTable(dsOptions) match {
+ case table: SupportsBatchWrite =>
+ val relation = DataSourceV2Relation.create(table, dsOptions)
+ // TODO: revisit it. We should not create the `AppendData`
operator for `SaveMode.Append`.
+ // We should create new end-users APIs for the `AppendData`
operator.
--- End diff --
yea, that's why I only left a comment and just ask for revisiting later. I
think we can see a clearer picture after we migrating the file source.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
