Mustafa İman created HIVE-24532:
-----------------------------------
Summary: Reduce sink vectorization mixes column types
Key: HIVE-24532
URL: https://issues.apache.org/jira/browse/HIVE-24532
Project: Hive
Issue Type: Bug
Reporter: Mustafa İman
Attachments: castexception.txt, explainplan.txt
I do insert overwrite select on a partitioned table. Partition column is
specified dynamically from select query. "ceil" function is applied on a string
column to specify partition for each row. Reduce sink gets confused about the
type of partition column. It leads to following cast exception in runtime:
{code:java}
Caused by: java.lang.ClassCastException:
org.apache.hadoop.hive.ql.exec.vector.LongColumnVector cannot be cast to
org.apache.hadoop.hive.ql.exec.vector.BytesColumnVector
at
org.apache.hadoop.hive.ql.exec.vector.VectorSerializeRow.serializePrimitiveWrite(VectorSerializeRow.java:452)
at
org.apache.hadoop.hive.ql.exec.vector.VectorSerializeRow.serializeWrite(VectorSerializeRow.java:279)
at
org.apache.hadoop.hive.ql.exec.vector.VectorSerializeRow.serializeWrite(VectorSerializeRow.java:258)
at
org.apache.hadoop.hive.ql.exec.vector.reducesink.VectorReduceSinkObjectHashOperator.processKey(VectorReduceSinkObjectHashOperator.java:305)
... 28 more
{code}
The problem is reproducible by running mvn test
-Dtest=TestMiniLlapLocalCliDriver -Dqfile=insert0.q with "set
hive.stats.autogather=false". The additional config option causes insert
statements to be vectorized so the vectorization bug appears.
insert0.q:
[https://github.com/apache/hive/blob/fb046c77257d648d0ee232356bdf665772b28bdd/ql/src/test/queries/clientpositive/insert0.q]
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
