[HIVE-6414] ParquetInputFormat provides data values that do not match the object inspectors

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.13.0
• Fix Version/s: 0.13.0
• Component/s: Serializers/Deserializers
• Labels: Parquet
• Assignee: Justin Coffey
• Reporter: Remus Rusanu
• Created: 12/Feb/14 16:50
• Updated: 10/Mar/14 18:22
• Resolved: 08/Mar/14 00:57

Description

While working on HIVE-5998 I noticed that the ParquetRecordReader returns IntWritable for all 'int like' types, in disaccord with the row object inspectors. I though fine, and I worked my way around it. But I see now that the issue trigger failuers in other places, eg. in aggregates:

Caused by: org.apache.hadoop.hive.ql.metadata.HiveException: Hive Runtime Error while processing row {"cint":528534767,"ctinyint":31,"csmallint":4963,"cfloat":31.0,"cdouble":4963.0,"cstring1":"cvLH6Eat2yFsyy7p"}
        at org.apache.hadoop.hive.ql.exec.MapOperator.process(MapOperator.java:534)
        at org.apache.hadoop.hive.ql.exec.mr.ExecMapper.map(ExecMapper.java:177)
        ... 8 more
Caused by: org.apache.hadoop.hive.ql.metadata.HiveException: java.lang.ClassCastException: org.apache.hadoop.io.IntWritable cannot be cast to java.lang.Short
        at org.apache.hadoop.hive.ql.exec.GroupByOperator.processOp(GroupByOperator.java:808)
        at org.apache.hadoop.hive.ql.exec.Operator.forward(Operator.java:790)
        at org.apache.hadoop.hive.ql.exec.SelectOperator.processOp(SelectOperator.java:87)
        at org.apache.hadoop.hive.ql.exec.Operator.forward(Operator.java:790)
        at org.apache.hadoop.hive.ql.exec.TableScanOperator.processOp(TableScanOperator.java:92)
        at org.apache.hadoop.hive.ql.exec.Operator.forward(Operator.java:790)
        at org.apache.hadoop.hive.ql.exec.MapOperator.process(MapOperator.java:524)
        ... 9 more
Caused by: java.lang.ClassCastException: org.apache.hadoop.io.IntWritable cannot be cast to java.lang.Short
        at org.apache.hadoop.hive.serde2.objectinspector.primitive.JavaShortObjectInspector.get(JavaShortObjectInspector.java:41)
        at org.apache.hadoop.hive.serde2.objectinspector.ObjectInspectorUtils.compare(ObjectInspectorUtils.java:671)
        at org.apache.hadoop.hive.serde2.objectinspector.ObjectInspectorUtils.compare(ObjectInspectorUtils.java:631)
        at org.apache.hadoop.hive.ql.udf.generic.GenericUDAFMin$GenericUDAFMinEvaluator.merge(GenericUDAFMin.java:109)
        at org.apache.hadoop.hive.ql.udf.generic.GenericUDAFMin$GenericUDAFMinEvaluator.iterate(GenericUDAFMin.java:96)
        at org.apache.hadoop.hive.ql.udf.generic.GenericUDAFEvaluator.aggregate(GenericUDAFEvaluator.java:183)
        at org.apache.hadoop.hive.ql.exec.GroupByOperator.updateAggregations(GroupByOperator.java:641)
        at org.apache.hadoop.hive.ql.exec.GroupByOperator.processHashAggr(GroupByOperator.java:838)
        at org.apache.hadoop.hive.ql.exec.GroupByOperator.processKey(GroupByOperator.java:735)
        at org.apache.hadoop.hive.ql.exec.GroupByOperator.processOp(GroupByOperator.java:803)
        ... 15 more

My test is (I'm writing a test .q from HIVE-5998, but the repro does not involve vectorization):

create table if not exists alltypes_parquet (
  cint int,
  ctinyint tinyint,
  csmallint smallint,
  cfloat float,
  cdouble double,
  cstring1 string) stored as parquet;

insert overwrite table alltypes_parquet
  select cint,
    ctinyint,
    csmallint,
    cfloat,
    cdouble,
    cstring1
  from alltypesorc;

explain select * from alltypes_parquet limit 10; select * from alltypes_parquet limit 10;

explain select ctinyint,
  max(cint),
  min(csmallint),
  count(cstring1),
  avg(cfloat),
  stddev_pop(cdouble)
  from alltypes_parquet
  group by ctinyint;
select ctinyint,
  max(cint),
  min(csmallint),
  count(cstring1),
  avg(cfloat),
  stddev_pop(cdouble)
  from alltypes_parquet
  group by ctinyint;

Issue Links

• is duplicated by: HIVE-6477 Aggregation functions for tiny/smallint broken with parquet