lucene-solr-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Erick Erickson <erickerick...@gmail.com>
Subject Re: HI
Date Tue, 15 Dec 2009 00:51:15 GMT
I know this seems like a smart-aleck remark,
but just don't go there <G>.

It goes against all your DB instincts, but you need
to think about *de*normalizing our data. Maybe
a useful mental trick is to think about what it would
take to make a single table in a DB where all the
information you needed was in a single row of that
table, then index *that* schema in Lucene/SOLR.

The point here is that SOLR/Lucene is built to search
lots and lots of text very quickly. We call these documents
(*somewhat* similar to a database row). But SOLR/Lucene
doesn't do a very good job of dealing with relations
between rows. When you find yourself trying this latter,
back up, take a deep breath, and re-think what you're doing.

This is not a knock on SOLR/Lucene, any more than
saying that full text searching in a database isn't very
well supported. It's a statement on the appropriateness
of the tool for the application.

So if you're looking for a way to map database tables
into SOLR, I think you're looking at the problem from
the wrong perspective. Instead, clearly define what you
want it to *do*, then how SOLR/Lucene can help.

It may be that you stick with a DB. It may be that you
create a hybrid solution that searches text with SOLR,
say returns a list of DB  ids, and then does some DB
work with the result. It may be that you can put everything
in SOLR and be happy. It entirely depends on your problem
space.

Note that there are some db-like capabilities built into SOLR
(I'm thinking faceting here), but those are really a distraction
from the above. Decide what you want it to do. Outline some
use cases, and *then* decide on the tools/approaches.

Hand-waving and not very concrete I know, but expressing
my absolute conviction that the problem you're trying to solve
must be defined before diving into the mechanics...

Best
Erick

On Sun, Dec 13, 2009 at 1:01 PM, Faire Mii <faire.mii@gmail.com> wrote:

> Hi,
>
> I am a beginner and i wonder what a document, entity and a field relates to
> in a database?
>
> And i wonder if there are some good tutorials that learn you how to design
> your schema. Because all other articles i have
>
> read aren't very helpful for beginners.
>
> Regards
>
> Fayer
>

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