Hi Alan, ----- Original Message ---- From: Alan Burlison <[EMAIL PROTECTED]> To: solr-user@lucene.apache.org Sent: Saturday, December 23, 2006 8:19:21 AM Subject: Re: Handling disparate data sources in Solr
Otis Gospodnetic wrote: > omitNorms let's you not use field norms for certain field when > calculating document matching score. This can save you some RAM. > See http://issues.apache.org/jira/browse/LUCENE-448. Thanks. What eddect does this have on the quality of the returned matches? Are there any guidelines as to when you would disable field norms, and on which type of fields? OG: fields that are really used as booleans and not as full-text searchable fields. For example, a date field, or a numeric field are good candidates. A username is a good field. Untokenized fields whose length should not impact the score. > For position > increment gap, have a look at > http://lucene.apache.org/java/docs/api/org/apache/lucene/analysis/Token.html#setPositionIncrement(int) > , it is described pretty well there. Right - so whereas in most cases it is set to 100, in fact a number of 5 or so would be plenty, is that correct? In fact, isn't it more-ore-less a boolean switch? OG: depends on what you want to do, as described in the javadoc... Otis
