1
0

Split insert() into insert() & bulkInsert() (#69)

- Behaviour change for existing insert() users
 - Made the current insert() implementation, as something to use for bulkInsert()
 - Normal inserts now share a lot of code with upsert, which provides benefits like small file handling
 - Refactored/Cleaned up code in HoodieWriteClient for reuse
 - Added a unit test, switching few tests to call bulkInsert() and few to call insert()
This commit is contained in:
vinoth chandar
2017-01-27 10:51:00 -08:00
committed by prazanna
parent 54409b07ea
commit 8e72ed69b4
6 changed files with 309 additions and 159 deletions

View File

@@ -84,6 +84,17 @@ public abstract class HoodieTable<T extends HoodieRecordPayload> implements Seri
Iterator<HoodieRecord<T>> recordIterator,
Partitioner partitioner);
/**
* Perform the ultimate IO for a given inserted (RDD) partition
*
* @param partition
* @param recordIterator
* @param partitioner
*/
public abstract Iterator<List<WriteStatus>> handleInsertPartition(Integer partition,
Iterator<HoodieRecord<T>> recordIterator,
Partitioner partitioner);
public static HoodieTable getHoodieTable(HoodieTableType type,
String commitTime,