lucene-java-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Erik Hatcher <>
Subject Re: Configurable indexing of an RDBMS, has it been done before?
Date Wed, 09 Feb 2005 15:39:26 GMT
On Feb 9, 2005, at 4:51 AM, mark harwood wrote:
> A GUI plugin for Squirrel SQL (
> would make a
> great way of configuring the mapping.

That would be slick!

> 1) Should we build this mapper into Luke instead? We
> would have to lift a LOT of the DB handling "smarts"
> from Squirrel. Luke however is doing a lot with
> Analyzer configuration which would certainly be useful
> code in any mapping tool (can we lift those and use in
> Squirrel?).

The dilemma with Luke is that its not ASL'd (because of the thinlet 
integration).  Anyone up for a Swing conversion project?  :)

It would be quite cool if Lucene had a built-in UI tool (like or 
actually Luke).  Luke itself is ASL'd and I believe Andrzej has said 
he'd gladly donate it to Lucene's codebase, but the Thinlet LGPL is an 

> 2) What should the XML for the batch-driven
> configuration look like? Is it ANT tasks or a custom
> framework?

Don't concern yourselves with Ant at the moment.  Anything that is 
easily callable from Java can be made into an Ant task.  In fact, the 
minimum requirements for an Ant task is a "public void execute()" 
method.  Whatever Java infrastructure you come up with, I'll gladly 
create the Ant task wrapper for it when its ready.

> 3) If our mapping understands the make-up of the rdbms
> and the Lucene index should we introduce a
> higher-level software layer for searching which sits
> over the rdbms and Lucene and abstracts them to some
> extent? This layer would know where to go to retrieve
> field values or construct filters ie understands
> whether to retrieve "title" field for display from
> database column or a Lucene "stored" field and whether
> the "price< $100" search criteria is resolved by a
> lucene query or an RDBMS-query to produce a Lucene
> filter. It seems like currently, every DB+Lucene
> integration project struggles with designing a
> solution to manage this divide and handcodes the
> solution.

Wow... that is getting pretty clever.  I like it!

I don't personally have a need for relational database indexing, but I 
support this effort to make a generalized mapping facility.


To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message