null fsIndexRepository

[Aaron Kaplan <Aa...@xrce.xerox.com>]

My annotator's process method begins like this:

	public void process(JCas jcas) throws AnalysisEngineProcessException {
		FSIndexRepository repo = jcas.getFSIndexRepository();

I'm getting a NullPointerException at the line after that, because repo 
is null.  How can this happen?

Possibly relevant: the CAS I'm processing was loaded by the CVD from an 
XMI file.  In the CVD I see an index called 
blah.blah.blah.MyType_GeneratedIndex, whereas I expected to see an index 
  called MyTypeIndex that we defined.

-Aaron

Re: null fsIndexRepository

[Marshall Schor <ms...@schor.com>]

Aaron Kaplan wrote:
> My annotator's process method begins like this:
>
>     public void process(JCas jcas) throws 
> AnalysisEngineProcessException {
>         FSIndexRepository repo = jcas.getFSIndexRepository();
>
> I'm getting a NullPointerException at the line after that, because 
> repo is null.  How can this happen?
>
> Possibly relevant: the CAS I'm processing was loaded by the CVD from 
> an XMI file.  In the CVD I see an index called 
> blah.blah.blah.MyType_GeneratedIndex, whereas I expected to see an 
> index  called MyTypeIndex that we defined.
>
I'm a bit out of my area of expertise here, but I saw this in the 
documentation for CVD:

Important: XMI CAS is a serialization format that serializes a CAS 
without type system and index information. It is therefore impossible to 
read in a stand-alone XMI CAS file. XMI CAS files can only be 
interpreted in the context of an existing type system. Consequently, you 
need to load the AE that was used to create the XMI file to be able to 
load it.

Did you first load the Analysis Engine (that has the type system 
specified that matches the type system of the saved XMI CAS), before 
loading the XMI Cas?

-Marshall

Re: null fsIndexRepository

[Aaron Kaplan <Aa...@xrce.xerox.com>]

Eddie Epstein wrote:

> By chance does the annotator descriptor declare Sofa input or output
> capabilities? If so, the CAS passed to process is a base CAS, which has no
> index repository.

Indeed, that was the problem.

Actually, what I described in my previous mail was actually two separate 
problems.  I introduced the bug you pointed out when constructing a 
simplified test case for a different problem.  The original problem was 
that an index I was expecting to see wasn't there.  That turned out to 
be because I had failed to declare the index in the annotator's 
descriptor, and as a result it wasn't getting created when the CAS was 
deserialized.

Thanks once again for your quick and effective help.
-Aaron

Re: null fsIndexRepository

[Eddie Epstein <ea...@gmail.com>]

Hi Aaron,

By chance does the annotator descriptor declare Sofa input or output
capabilities? If so, the CAS passed to process is a base CAS, which has no
index repository.

Here's some code from CasImpl.java

  public FSIndexRepository getIndexRepository() {
    if (this == this.svd.baseCAS) {
      // BaseCas has no indexes for users
      return null;
    }

Eddie

On Tue, Jun 24, 2008 at 9:53 AM, Aaron Kaplan <Aa...@xrce.xerox.com>
wrote:

> My annotator's process method begins like this:
>
>        public void process(JCas jcas) throws AnalysisEngineProcessException
> {
>                FSIndexRepository repo = jcas.getFSIndexRepository();
>
> I'm getting a NullPointerException at the line after that, because repo is
> null.  How can this happen?
>
> Possibly relevant: the CAS I'm processing was loaded by the CVD from an XMI
> file.  In the CVD I see an index called
> blah.blah.blah.MyType_GeneratedIndex, whereas I expected to see an index
>  called MyTypeIndex that we defined.
>
> -Aaron
>