I don't think I changed any boost values, at least not on purpose. I think
the reason for the changed document id is that, to my knowledge, an update
is a delete and an add.
The code for my solrj update:
public void updateDocument(SolrDocument document) {
SolrServer server = new CommonsHttpSolrServer(SOLR_URL);
SolrInputDocument input = new SolrInputDocument();
Collection<String> fields = document.getFieldNames();
for (String field : fields) {
input.setField(field, document.getFieldValue(field));
}
input.removeField("id"); //is regenerated from the url value
input.removeField("score");
server.add(input);
}
--
Ole-Martin Mørk
On Mon, Oct 5, 2009 at 11:15 AM, Simon Willnauer <
simon.willna...@googlemail.com> wrote:
> Did you change any boost values for URL field or document while reindexing
> the document by any chance? Or do you look at different documents - one is
> internal id 0 and other is internal id 22 - this could be the updated one
> just curious if that might be the cause?!
>
> simon
>
> On Mon, Oct 5, 2009 at 10:21 AM, Ole-Martin Mørk <olemar...@gmail.com
> >wrote:
>
> > Hi. I am trying to understand Lucene's scoring algorithm. We're
> > getting some strange results. First we search for a given page by it's
> > url. We get this result:
> >
> > 0.0014793393 = fieldWeight(url:"our super secret url" in 22), product of:
> > 1.0 = tf(phraseFreq=1.0)
> > 32.31666 = idf(url: www=7327 host=321 com=7327 article=2456
> > something=2 something=44 704290075=1)
> > 4.5776367E-5 = fieldNorm(field=url, doc=22)
> >
> > When this is done, we use solrJ to read and write the document. The
> > only change is the title of the document (appends the number 2)
> >
> > We search again and the fieldNorm is changed significantly:
> >
> > 9.874598 = fieldWeight(url:"our super secret url" in 0), product of:
> > 1.0 = tf(phraseFreq=1.0)
> > 31.598713 = idf(url: www=7328 host=322 com=7328 article=2457
> > something=3 somthing=45 704290075=2)
> > 0.3125 = fieldNorm(field=url, doc=0)
> >
> > Why does the value of fieldNorm change so much?
> >
> > Looking forward to your answers.
> >
> > --
> > Ole-Martin Mørk
> > http://twitter.com/olemartin
> > http://flickr.com/olemartin
> >
>
