Hi,
More I'm looking, more uncharted seems this area.
1. filter: where com.xxx.uima.type.DocumentInfo:cached='true', my
assumption, it will look at the feature structure
com.xxx.uima.type.DocumentInfo, get the value of cached, check if it's
true and pass or not to pass to AE.
What I see is:
- right part of the expression, literal 'true', hmm, I thought
this suppose to be SQL like, so anything in apostrophes are strings, ok
can live with that.
- when it does feature structure lookup, it does it by:
com.xxx.uima.type.DocumentInfo_colon_cached, hmm, I would expect it will
do lookup by the part before colon, 'com.xxx.uima.type.DocumentInfo',
type name and do value by the right part after the colon, 'cached' the
feature name.
Question really is, is there anybody working in this area, I have to
finish concept ASAP, so I'll do all changes in my local copy, but would
be great if somebody could review and prorogate my changes to the main
source tree.
Thank you.
P.S.
I just got the trunk, problems still there.
-----Original Message-----
From: Thilo Goetz [mailto:[email protected]]
Sent: Wednesday, April 29, 2009 2:34 AM
To: [email protected]
Subject: Re: ProcessingContainer_Impl.java, filter
[email protected] wrote:
> 1009: if ( (filterExpression.getOperand() == null
>
> && filterExpression.getOperand().getOperand() == null
> && !exists)
>
> ....
>
>
>
> Is this some kind of null pointer exceptions generator, hard to tell
> what should be there, but if I use filter to test existing feature,
does
> not work.
Thanks for making me laugh :-)
I would think that when filterExpression.getOperand() != null, then
filterExpression.getOperand().getOperand() is also not null. So it
seems to me that the second check above, the NPE generator, is redundant
and should be removed. Alternatively, one could parens around the null
checks and change the && to an || I guess.
Can anybody confirm or provide alternative interpretation?
--Thilo
>
>
>
> Thanks.
>
>
>
>
>
>
