[FLINK-17334] Flink does not support HIVE UDFs with primitive return types

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.10.0
• Fix Version/s: 1.10.1
• Component/s: Connectors / Hive
• Labels: pull-request-available
• Flags: Patch
• Assignee: roy.ru
• Reporter: roy.ru
• Created: 23/Apr/20 05:05
• Updated: 29/Apr/20 08:16
• Resolved: 29/Apr/20 08:16

Description

We are currently migrating Hive UDF to Flink. While testing compatibility, we found that Flink cannot support primitive types like boolean, int, etc.

Hive UDF:

public class UDFTest extends UDF {
public boolean evaluate(String content) {
if (StringUtils.isEmpty(content))

{ return false; }

else

{ return true; }

}

}

We found that the following error will be reported:
Caused by: org.apache.flink.table.functions.hive.FlinkHiveUDFException: Class boolean is not supported yet
at org.apache.flink.table.functions.hive.conversion.HiveInspectors.getObjectInspector(HiveInspectors.java:372)
at org.apache.flink.table.functions.hive.HiveSimpleUDF.getHiveResultType(HiveSimpleUDF.java:133)

I found that if I add the type comparison in HiveInspectors.getObjectInspector to the primitive type, I can get the correct result.

as follows：
public static ObjectInspector getObjectInspector(HiveShim hiveShim, Class clazz){       

   ..........

          else if (clazz.equals(boolean.class) || clazz.equals(Boolean.class) ||                  clazz.equals(BooleanWritable.class))

{                                       typeInfo = TypeInfoFactory.booleanTypeInfo;                                 }

         ..........

}

 

 

Issue Links

• links to: GitHub Pull Request #11876
• links to: GitHub Pull Request #11939