lucene-java-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "d rj" <drjat...@gmail.com>
Subject Re: Sharing Documents between Lucene and DotLucene
Date Mon, 28 Aug 2006 19:51:10 GMT
Anything could be in the client's sub-index (for example, say if searches
are being done against a large mail archive one may want to keep an local
index of the just the user's to/from as that may be the most often searched
data).

Building the index on the server than transferring it across to the client
is a feasible solution (in fact it is the method that I currently employ).
However, building a temporary index in a temporary directory on the server,
then transferring all files (minimally 3) to the client, then merging this
temporary index into the clients already existing index and cleaning up the
temporary index on both server and client it is not as clean as simply
transferring a Document or list of Documents to a requesting client.
Although it does work.

I think that the best method to transfer Document objects across the wire
from Lucene to Lucene.Net is to write the appropriate xml schema using xsdl,
then write the necessary translation code for both Java and C# that would
marshall Lucene Document objects to schema based Document objects and vise
versa.  The schema based Documents being what is now transferable over the
wire.

This is my current approach.  I wanted to post to the group to know if
schema descriptors for Lucene Documents already exists or if anyone has
employed some other successful technique for transferring Documents from
Lucene to Lucene.Net.

Thanks.

-drj

On 8/28/06, Erik Hatcher <erik@ehatchersolutions.com> wrote:
>
>
> On Aug 28, 2006, at 9:52 AM, d rj wrote:
> > The primary reason for building a smaller client side sub-index is
> > quality
> > of service.  Building a client side index will avoid both network
> > latency
> > and network outage issues.  Rather than have the client application
> > attempt
> > to query the server's Lucene index for search results I would like to
> > transfer a minimal index to the client that can be queried in a
> > fast and
> > 100% reliable manner regardless of network latency, network
> > connectivity and
> > server load.
>
> What would define the documents in the sub-index?   What about
> building client-specific indexes and transfer the actual index to the
> client rather than pulling out individual documents?
>
> > So, I was wondering if anyone had any approaches to transferring
> > Documents
> > between Lucene and Lucene.Net.  Or if there is something similar to
> > an xsdl
> > schema for the Document class that would allow one to transfer
> > Document via
> > a webservice between Lucene and Lucene.Net.
>
> There really isn't anything special about a Document object - pulling
> its fields into Map would do the trick for transferring it.  But,
> again, what about unstored fields?  What about analysis?  You'd have
> to re-index the a document on the client, which may be an issue to
> keep Java and .NET analyzers in sync exactly with one another.
>
>         Erik
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
> For additional commands, e-mail: java-user-help@lucene.apache.org
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message