Benjamin Garrett created NIFI-4647:
--------------------------------------
Summary: ConvertAvroToORC fails if the avro schema uses a union of
a null, string, and long
Key: NIFI-4647
URL: https://issues.apache.org/jira/browse/NIFI-4647
Project: Apache NiFi
Issue Type: Bug
Components: Core Framework
Affects Versions: 1.4.0
Reporter: Benjamin Garrett
If an avro schema has a field defined like this:
{"name":"myStringOrLong","type":["null","string","long"],"default":null}
then when we run ConvertAvroToORC we get the following error:
Object Type for class org.apache.avro.util.Utf8 not in Union declaration
This is because Avro uses its internal Utf8 class (instead of 'String') as a
performance optimization as discussed elsewhere on the internet, such as here:
http://apache-avro.679487.n3.nabble.com/why-Utf8-vs-String-td3247788.html
but the NiFiOrcUtils class is not expecting the Utf8 class to be used for
'string' types.
A simple workaround is to modify NiFiOrcUtils lines 72-73 to be something like
this:
Class clazzToCompareTo = o.getClass();
if (o instanceof org.apache.avro.util.Utf8 ) {
clazzToCompareTo = String.class;
}
TypeInfo objectTypeInfo =
TypeInfoUtils.getTypeInfoFromObjectInspector(
ObjectInspectorFactory.getReflectionObjectInspector(
clazzToCompareTo,
ObjectInspectorFactory.ObjectInspectorOptions.JAVA));
This solution works for me.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
