Hi Eric -
This appears to be a new bug, unknown before. We think it is *not*
fixed in the 2.1 release. We'll work on getting it fixed in the next
release.
Thanks for reporting it. -Marshall
Eric VACHON wrote:
Hi
As Olivier already described it in the alphaworks forum, we have a simple CPE
composed of a collection reader, an analysis engine (named ERROR) remotely
deployed as a Vinci service and an integrated
consumer. Our error handling parameters are set to "continue"
The descriptor looks like:
<?xml version="1.0" encoding="UTF-8"?>
<cpeDescription>
<collectionReader>
<collectionIterator>
<descriptor>
<include href="SimpleCollectionReader.xml"/>
</descriptor>
<configurationParameterSettings />
</collectionIterator>
</collectionReader>
<casProcessors casPoolSize="2" processingUnitThreadCount="1">
<casProcessor deployment="remote" name="ERROR">
<descriptor>
<include href="ERROR.xml"/>
</descriptor>
<filter/>
<errorHandling>
<errorRateThreshold action="continue" value="2/5"/>
<maxConsecutiveRestarts action="continue" value="2"/>
<timeout max="300000"/>
</errorHandling>
<checkpoint batch="1"/>
<deploymentParameters>
<parameter name="vnsHost" type="string" value="localhost"/>
<parameter name="vnsPort" type="string" value="9000"/>
<parameter name="service-access" type="string"
value="exclusive"/>
</deploymentParameters>
</casProcessor>
<casProcessor deployment="integrated" name="SimpleCasConsumer">
<descriptor>
<include href="SimpleCasConsumer.xml"/>
</descriptor>
<deploymentParameters/>
<filter/>
<errorHandling>
<errorRateThreshold action="continue" value="2/5"/>
<maxConsecutiveRestarts action="continue" value="2"/>
<timeout max="300000"/>
</errorHandling>
<checkpoint batch="1"/>
</casProcessor>
</casProcessors>
<cpeConfig>
<numToProcess>-1</numToProcess>
<deployAs>immediate</deployAs>
<checkpoint file="" time="300000"/>
<timerImpl/>
</cpeConfig>
</cpeDescription>
The ERROR cas processor generates AnnotatorProcessException when the processing
of some particular documents ( CAS ) went wrong.
We have a StatusCallbackListener with a entityProcessComplete method
implemented like that:
public void entityProcessComplete(CAS aCas, EntityProcessStatus aStatus) {
if (aStatus.isEntitySkipped()) {
System.err.println("The CAS has been skipped by at
least one component");
}
if (aStatus.isException()) {
List exceptions = aStatus.getExceptions();
for (int i = 0; i < exceptions.size(); i++) {
((Throwable)
exceptions.get(i)).printStackTrace(System.err);
}
return;
}
entityCount++;
size += aCas.getTCAS().getDocumentText().length();
}
Our ERROR cas processor generates exceptions that are nicely detected by the
aStatus.getExceptions() call. But after the number of 'retries' is reached for
the processor (this number is set to 2 as
you can see in the descriptor), the entityProcessComplete is called with a
aStatus.isException() set to false which is the expected behaviour but the
aStatus.isEntitySkipped() is also set to false
which is not the expected value because the CAS has actually been skipped by
the ERROR processor.
We have this bug in the UIMA 1.4.3 version and we have to stick with it for
some more time. But we will upgrade to the apache version soon and we would
like to know if this issue has been fixed in 2.1.
Can you confirm that it is a known bug and if it is already fixed ?
Thanks a lot
