SaveMode
SaveMode
SaveMode
SaveMode
This is like SDLSaveMode.Overwrite but processed partitions are manually deleted instead of using dynamic partitioning mode.
This is like SDLSaveMode.Overwrite but processed partitions are manually deleted instead of using dynamic partitioning mode. Then it uses Sparks append mode to add the new partitions. This helps if there are performance problems when using dynamic partitioning mode with hive tables and many partitions.
Implementation: This save mode will delete processed partition directories manually. If no partition values are present when writing to a partitioned data object, all partitions are deleted. This is different to Sparks dynamic partitioning, which only deletes partitions where data is present in the DataFrame to be written (enabled by default in SDL). To stop if no partition values are present, configure executionMode.type = FailIfNoPartitionValuesMode on the Action.
This is like SDLSaveMode.Overwrite but doesnt delete the directory of the DataObject and its partition, but only the files inside.
This is like SDLSaveMode.Overwrite but doesnt delete the directory of the DataObject and its partition, but only the files inside. Then it uses Sparks append mode to add the new files. Like that ACLs set on the base directory are preserved.
Implementation: This save mode will delete all files inside the base directory, but not the directory itself. If no partition values are present when writing to a partitioned data object, all files in all partitions are deleted, but not the partition directories itself. This is different to Sparks dynamic partitioning, which only deletes partitions where data is present in the DataFrame to be written (enabled by default in SDL). To stop if no partition values are present, configure executionMode.type = FailIfNoPartitionValuesMode on the Action.
SDL supports more SaveModes than Spark, that's why there is an own definition of SDLSaveMode.