On 10/18/13 6:17 AM, Andrew McIntyre wrote:
Hi derby-dev!
I had a couple of free afternoons and so I took a swing at a simple
lucene integration demo, since full-text search is the most requested
Derby feature. Attached is my very rough but working results. I don't
know how much time I'll have to continue working on it, but I thought
maybe if I could produce at least a working demo then maybe that could
get the ball rolling. I could certainly at least clean this up to the
Derby code standards and put together a wiki page primer that could be
included in the demo package at the very least. :-)
As it stands, I think it will be good for discussion as to whether
this would be a useful feature to continue working on. It seems like
it would be pretty easy to wire the index creation into the sql
grammar and storing the lucene indexes in the database would be pretty
straightforward by replacing those FSDirectory.open calls.
I realized as I was typing this that the test data I was using might
be considered copyrighted. So I put the ij script in my personal
directory here:
http://people.apache.org/~fuzzylogic/lucene_titles.sql
<http://people.apache.org/%7Efuzzylogic/lucene_titles.sql>
until I can find some public domain data. That includes some queries
that give you an idea of what is possible. I attached one query with
results as lucene_query.txt so you could see some results without
having to apply the patch and run a build. Speaking of which, to
build, you'll need to drop the lucene-core, lucene-analyzers-common,
and lucene-queryparsers jars from lucene-4.5.0 into tools/java after
applying the patch to build (although it shouldn't break your build
after applying the patch if you don't).
Anyway, enjoy, and let me know what you think.
cheers,
andrew
Hey Andrew,
Good to hear from you. I haven't studied the code carefully, but yes,
indeed, this optional tool looks like a very useful contribution to
Derby. I agree with Kathey that
https://issues.apache.org/jira/browse/DERBY-590 sounds like a good place
to attach this work and continue the discussion. I am interested in your
thoughts about deeper integration via the parser.
Thanks!
-Rick
