Sorry for my belated reply to this hot topic. I've been too busy
writing code around Lucene to tune into e-mail :)
There are techniques you can use to keep a Lucene index in sync with
a DB. If you're using Hibernate, there are hooks you can use to
manage the index on updates, deletes, adds.
There is an inherent mismatch between relational database querying
and full-text searching, and thus it is a bit of duplication, but it
is a necessary one to have strong full-text searching capabilities.
There are full-text searching capabilities within SQL Server, Oracle,
MySQL, and probably others, but they are vender specific and many
have claimed these are inferior to Lucene's capabilities.
As for Lucene being used within a database - that can, and has, been
done. A friend of mine has written a custom Lucene Directory
subclass that does this very thing, and I continually urge him to
contribute it (but he hasn't yet). There have been other efforts
that you can find in the archives of the Lucene user e-mail list, but
nothing I'd consider that viable. There is a Berkeley DB Directory
implementation that provides transactional support, etc - but that
might not be the best fit for your situation.
Good luck finding something better than Lucene! :) However, I think
you'll find that figuring out a way to work Lucene into your
architecture will be the best solution, as there is nothing else that
I know of that compares. Then again, I'm biased! - http://
www.lucenebook.com
Erik
On Oct 5, 2005, at 12:18 PM, Patrick Casey wrote:
Sorry for the OTness here, but I find this mailing list
has more
java knowledge in it than anywhere else I'm likely to look so .
I'm considering adding freeform text search to a web
application
I'm working on. I've done the same thing before with other
applications
using Lucene, but that approach was, to be blunt, frail because I
had to
keep (and manage) two parallel data stores. There was Lucene's file
based
storage listing all the text and object keys, and there was my
database that
held the objects themselves. Keeping the two in synch through errors,
transaction failures, backups, and recoveries was like pulling
teeth and
never *quite* worked right, resulting in the occasional "oh, just
rebuild
the whole lucene index" approach to fault recovery.
What I'd really like is some form of vendor-neutral
freeform
text search that uses an RDBMS (not the file system) as its
repository. Does
anybody know of a java based solution that offers that kind of
functionality?
--- Pat
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
