CPM setAnalysisEngine throws IndexOutOfBounds
---------------------------------------------
Key: UIMA-1732
URL: https://issues.apache.org/jira/browse/UIMA-1732
Project: UIMA
Issue Type: Bug
Components: Collection Processing
Affects Versions: 2.3
Reporter: Marshall Schor
Assignee: Jerry Cwiklik
Priority: Minor
A user tried to do several calls to aCPM.setAnalysisEngine(xxx). This method
is set up to remove a previous AE if it exists, replacing it with the new one.
See http://markmail.org/message/3s2w7hxb24l3czky
There are several issues:
The method CPMEngine removeCasProcessor(int aCasProcessorIndex) has a bad
bounds test, need to change
{noformat}
if (aCasProcessorIndex < 0 || aCasProcessorIndex > annotatorList.size()) to
if (aCasProcessorIndex < 0 || aCasProcessorIndex >= annotatorList.size())
{noformat}
The CPMEngine class uses two different collections to represent lists of
analysis engines:
* casprocessorList (an array - includes AEs and Cas Consumers)
* annotatorList (a linked list, not sure if it has both AEs and Cas Consumers)
The CPMImpl setAnalysisEngine seems to have set something which shows up in the
array from getCasProcessors call, while the "annotatorList" is not updated and
remains empty.
This method:
{noformat}
public void removeCasProcessor(CasProcessor aCasProcessor) {
cpEngine.removeCasProcessor(0);
}
{noformat}
ignores its argument, and removes the 0'th CasProcessor. This seems wrong,
given the Javadoc.
The Javadoc for setAnalysisEngine's javadoc should emphasize that this is a
convenience method for the common case of supporting a single AE. If it is
called multiple times, it will replace the first AE (if it exists) . The
Javadocs should be expanded to say if you want to add multiple AEs to a pipe
line, to not use this method, but use addCasProcessor instead.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
