lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Shai Erera <ser...@gmail.com>
Subject Re: Bug in DocInvertedPerField?
Date Mon, 06 Jul 2009 11:12:24 GMT
Yes they have the same field name.
Can we use the default posIncr? If I want to create an IndexWriter w/o an
Analyzer, why should I be forced to do new IndexWriter(new SimpleAnalyzer()
/* for example */ ...), when the analyzer will never be used?

It is an edge case though which I can easily reproduce, but not sure how
important it is to fix it (the fix should be simple though).

Shai

On Mon, Jul 6, 2009 at 2:05 PM, Michael McCandless <
lucene@mikemccandless.com> wrote:

> Were the two fields that you added to the doc the same field name?
>
> In which case, the pos incr gap is in fact needed, even if the fields
> are pre-analyzed (have TokenStream values)?
>
> Mike
>
> On Thu, Jul 2, 2009 at 10:25 AM, Shai Erera<serera@gmail.com> wrote:
> > I hit NPE in DocInvertedPerField in the following scenario:
> > 1) Create IndexWriter w/ a null Analyzer.
> > 2) Add two fields to a Document, w/ a TokenStream (or one as not analyzed
> > and one as TokenStream, or two not analyzed).
> > 3) Add the document to the IndexWriter.
> > The NPE comes from processFields (line 79), when it tries to read the
> > position increment gap.
> >
> > I think I'm using it right, but want to verify. If I always add fields as
> > TokenStream, or untokenized, I don't really need the Analyzer on
> > IndexWriter. Also, in case of a TokenStream field, DocInverterPerField
> just
> > uses the TokenStream and not the Analyzer. So maybe it should read the
> > posIncrGap only if analyzer is not null?
> >
> > Is this a bug or a misuse of IndexWriter by me?
> >
> > Shai
> >
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: java-dev-unsubscribe@lucene.apache.org
> For additional commands, e-mail: java-dev-help@lucene.apache.org
>
>

Mime
View raw message