Details
-
Sub-task
-
Status: Open
-
Minor
-
Resolution: Unresolved
-
3.1.0
-
None
-
None
Description
To allow Spark to write Hive 3 compatible bucketed table, we need to follow the same hash function with Hive/Presto. Hive murmur3hash is quite some different with Spark murmur3hash (with different default seed, different logic for NULL, array, map, struct, detail in https://github.com/apache/hive/blob/ece58fff1b53ea451bfc524c4c15f63ee12eca00/serde/src/java/org/apache/hadoop/hive/serde2/objectinspector/ObjectInspectorUtils.java#L813). So here introduce a Hive murmur3hash expression.