Complex File technology and pre Java 7 bug

Problem statement

How many times have you tried to use a user-defined file structure using Native Format and registered it with ODI Topology as a Complex File Data server - only to have it fail when testing the connection with an error saying :

ORABPEL-11100

Translation Failure.

[Offset=7] Translation from native failed. For input string: "...............".

The incoming data does not conform to the NXSD schema. Please correct the problem.

Well, it could be problem with the input data, but if you are absolutely sure that your data is conformant with the XSD you have created (btw you can use commandline to check this : see here) then there is one thing you can look for : in the error message does the reported string start with "+" and is the datatype for this element as defined in the XSD xsd:long?

If you can say yes to all of this then congratulations! You have been bitten by a Java bug tht has been fixed in Java 7 onwards. The bug is in Java's Long.valueOf where the code only checks for "-" and not "+". So when it finds "+" in there it throws a NumberFormatException. This exception however gets swallowed up and the impression that is created is that it is the fault of the data in the file.

Solution

If you are using ODI 11.1.1.5.0, tough luck! With ODI 11.1.1.6.x, you will need to set up a standalone or JEE agent. These agents must run using Java 7. For standalone agent edit /oracledi/agent/bin/odiparams.sh/bat where ODI_JAVA_HOME will point to your Java location. For JEE agent when you are configuring the domain choose a Java 7 installation. Then for 'test connection' use this agent. For reverse engineering choose 'Customized' rather than 'Standard' on the reverse engineering finger tab and also choose this agent. All ODI Sessions that use this data server must execute on this agent.