Mailing-List: contact lucene-user-help@jakarta.apache.org; run by ezmlm
Precedence: bulk
Reply-To: "Lucene Users List" <lucene-user@jakarta.apache.org>
From: Morus Walter <morus.walter@tanto.de>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: quoted-printable
Message-ID: <16610.35228.247157.936985@tanto-xipolis.de>
Date: Wed, 30 Jun 2004 11:36:28 +0200
To: "Lucene Users List" <lucene-user@jakarta.apache.org>
Subject: Re: return value of terms()
In-Reply-To: <20040629135404.5225.qmail@web12708.mail.yahoo.com>
References: <143088710.1088516807953.JavaMail.dummy@smb-tec.com>
	<20040629135404.5225.qmail@web12708.mail.yahoo.com>

Otis Gospodnetic writes:
> I see.  A search for that Term still gets Hits.  I don't think this
> should be happening.  Maybe Erik or one of the other Lucene developer=
s
> will have some ideas.
>=20
Maybe I missunderstood something, but a search shouldn't get hits, sinc=
e a=20
search removes hits from deleted documents (given that the term doen't
occur in other undeleted documents).

OTOH looking at the TermEnum one will still find terms of deleted docum=
ents.

AFAIK deleting a document doesn't mean removing a document from the ind=
ex.
Deleting a document means marking the document deleted.
The document will be removed at the next index optimization (or=20
implicit merge I guess; but I'm not sure about that)

I don't know when the frequency numbers are updated though.

So if I was Lars I'd call optimize after deleting and be on the safe si=
de...
(which might be a problem for frequent deletions on large indexes thoug=
h)

Morus

>=20
>=20
> --- Lars Martin <Lars.Martin@smb-tec.com> wrote:
> > -----Urspr=FCngliche Nachricht-----
> > Von: Otis Gospodnetic <otis_gospodnetic@yahoo.com>
> > Gesendet am: 29. Jun 2004, 13:46:41
> >=20
> > > I would try using the delete(Term) method, to ensure all document=
s
> > > with the given Term are removed:
> > >=20
> > >    IndexReader indexReader =3D IndexReader.open( indexPath );=20
> > >    indexReader.delete( new Term( "body", "YourTermHere" ) );
> > >    indexReader.close();
> > >    ...
> > >    IndexReader indexReader =3D IndexReader.open( indexPath );
> > >    TermEnum enum =3D indexReader.terms( new Term( "body", "" ) );=

> > >=20
> > > Something like that...
> >=20
> >=20
> > Thanks for your reply.
> >=20
> > I do not want to delete documents by terms. All my indexed document=
s=20
> > are referenced by id, so I have to use delete( id ). What makes me
> > insecure is the fact, that there are still terms in index from
> > documents
> > which are already deleted. This would mean that TermEnum is a
> > continously
> > growing beast. No problem when I query such a term, because no
> > document
> > is matching the query. But when I do computation based on indexed
> > terms
> > I heavily depend on the number of terms in the current TermEnum. Ev=
en
> > if such terms don't impact my computation - because the freq is
> > always
> > 0 - it would have an impact on my runtime behavior and complexity.
> >=20
> > Regards, Lars
> >=20
> >=20
> > > Otis
> > >=20
> > > --- Lars Martin <Lars.Martin@smb-tec.com> wrote:
> > > > Hi.
> > > >=20
> > > > Is it the normal behavior that IndexReader.terms( Term t ) stil=
l
> > > > returns Terms which are not any longer to be found in the index=
,
> > > > e.g. after removing the document containing these Terms?
> > > > I've removed nearly all documents from index but the terms()
> > method
> > > > is still returning all terms.
> > > >=20
> > > >   IndexReader indexReader =3D IndexReader.open( indexPath );=20=

> > > >   indexReader.delete( docId );
> > > >   indexReader.close();
> > > >   ...
> > > >   IndexReader indexReader =3D IndexReader.open( indexPath );
> > > >   TermEnum enum =3D indexReader.terms( new Term( "body", "" ) )=
;
> > > >=20
> > > > Any hints? Regards, Lars
>=20
>=20
> ---------------------------------------------------------------------=

> To unsubscribe, e-mail: lucene-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: lucene-user-help@jakarta.apache.org

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