Github user cloud-fan commented on a diff in the pull request:
https://github.com/apache/spark/pull/22009#discussion_r210155543
--- Diff:
sql/core/src/main/java/org/apache/spark/sql/sources/v2/BatchWriteSupportProvider.java
---
@@ -21,33 +21,39 @@
import org.apache.spark.annotation.InterfaceStability;
import org.apache.spark.sql.SaveMode;
-import org.apache.spark.sql.sources.v2.writer.DataSourceWriter;
+import org.apache.spark.sql.sources.v2.writer.BatchWriteSupport;
import org.apache.spark.sql.types.StructType;
/**
* A mix-in interface for {@link DataSourceV2}. Data sources can implement
this interface to
- * provide data writing ability and save the data to the data source.
+ * provide data writing ability for batch processing.
+ *
+ * This interface is used when end users want to use a data source
implementation directly, e.g.
+ * {@code Dataset.write.format(...).option(...).save()}.
*/
@InterfaceStability.Evolving
-public interface WriteSupport extends DataSourceV2 {
+public interface BatchWriteSupportProvider extends DataSourceV2 {
/**
- * Creates an optional {@link DataSourceWriter} to save the data to this
data source. Data
+ * Creates an optional {@link BatchWriteSupport} to save the data to
this data source. Data
* sources can return None if there is no writing needed to be done
according to the save mode.
*
* If this method fails (by throwing an exception), the action will fail
and no Spark job will be
* submitted.
*
- * @param writeUUID A unique string for the writing job. It's possible
that there are many writing
- * jobs running at the same time, and the returned
{@link DataSourceWriter} can
- * use this job id to distinguish itself from other
jobs.
+ * @param queryId A unique string for the writing query. It's possible
that there are many
+ * writing queries running at the same time, and the
returned
+ * {@link BatchWriteSupport} can use this id to
distinguish itself from others.
* @param schema the schema of the data to be written.
* @param mode the save mode which determines what to do when the data
are already in this data
* source, please refer to {@link SaveMode} for more details.
* @param options the options for the returned data source writer, which
is an immutable
* case-insensitive string-to-string map.
- * @return a writer to append data to this data source
+ * @return a write support to write data to this data source.
*/
- Optional<DataSourceWriter> createWriter(
- String writeUUID, StructType schema, SaveMode mode,
DataSourceOptions options);
+ Optional<BatchWriteSupport> createBatchWriteSupport(
+ String queryId,
+ StructType schema,
+ SaveMode mode,
--- End diff --
I totally agree that `SaveMode` is a bad API which leads to undefined
behavior. That's why we start a project to design new DDL logical plans and
write APIs. However, I believe we had an agreement before that we can't remove
existing APIs, so the `DataFrameWriter` and `SaveMode` will still be there in
Spark. If I'm a data source developer, even I've implemented the new write APIs
(assuming it's finished), I would still support `SaveMode` to attract more
users. `DataFrameWriter` is a very widely used API and the end users may need a
long time to migrate to the new write APIs. BTW, file source (without catalog)
does have a clearly defined behavior regarding `SaveMode`, we should make it
possible to migrate file source to data source v2.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
