Bulk Operation

SPARK-5180 Conversion is applied twice on partitioned data sources

SPARK-5180 Speed up path construction in DynamicPartitionWriterContainer.outputWriterForRow

SPARK-5180 Move all internal data source related classes out of sources package

SPARK-5180 Reduce memory consumption for dynamic partition insert

SPARK-5180 Replace the hash map in DynamicPartitionWriterContainer.outputWriterForRow with java.util.HashMap

SPARK-5180 DataFrame partitionBy memory pressure scales extremely poorly

SPARK-5180 Do not use FloatType in partition column inference

SPARK-5180 FSBasedRelation interface tweaks

SPARK-5180 Add common string filters to data sources

SPARK-5180 Finalize DDL and write support APIs

SPARK-5180 Preinsert casting and renaming rule is needed in the Analyzer

SPARK-5180 In memory data cache should be invalidated after insert into/overwrite

SPARK-5180 Python API for the write support of the data source API

SPARK-5180 Write support for the data source API

SPARK-5180 Improve the performance of metadata operations

SPARK-5180 Document data source API

SPARK-5180 Partitioning support for tables created by the data source API

SPARK-5180 Persistent data source tables

SPARK-5180 Adding support for defining schema in foreign DDL commands.