Mailing-List: contact java-dev-help@lucene.apache.org; run by ezmlm
Precedence: bulk
Reply-To: java-dev@lucene.apache.org
Received-SPF: pass (hermes.apache.org: local policy)
From: Paul Elschot <paul.elschot@xs4all.nl>
To: java-dev@lucene.apache.org
Subject: Re: search on stored, unindexed fields?
Date: Thu, 21 Apr 2005 21:40:02 +0200
User-Agent: KMail/1.5.4
References: <LMENLAOACIBLMOIILNNNKEEJEOAA.rengels@ix.netcom.com>
In-Reply-To: <LMENLAOACIBLMOIILNNNKEEJEOAA.rengels@ix.netcom.com>
MIME-Version: 1.0
Content-Type: text/plain;
  charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
Message-Id: <200504212140.02091.paul.elschot@xs4all.nl>

On Thursday 21 April 2005 20:47, Robert Engels wrote:
> Has there ever been any work done to allow the searching on stored, but
> unindexed fields?

Only indexed fields can be searched, but any field can be both indexed and
stored.

> It does not seems to be trivial, since a user could use and/or constructs
> that mixed indexed and non-indexed fields.
> 
> On a related note, does lucene optimize the search based on term frequency?
> 
> So that if I use +field1:value1 and +field2:value0 and it knows that the
> second term is much more unique, it will do that search first, and use
> skipTo on the first one? If so, does the same happen for range searches?

The implementation of conjunctions (both terms required) is actually
symmetrical on both terms, but the effect is what you expect.
Range queries are disjunctions and they are implemented
differently.
Please have a look at BooleanScorer and ConjunctionScorer in
package org.apache.lucene.search for the details.

Regards,
Paul Elschot.


---------------------------------------------------------------------
To unsubscribe, e-mail: java-dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: java-dev-help@lucene.apache.org