to solve your problem, you have to carefully understand how lucene
Lucene can handle the value of the document fields in different ways;
the different aspects to care of are:
- indexing: whether the value is indexed or not. It may look strange
to store data on a Lucene index without indexing it, but your records
IDs probably feet this scenario, as usually you never search record by
- stored: whether you want Lucene to store the full value of the
field. In your scenario, you probably need to store only the record ID
on the Lucene index, as all the data are probably better kept on the
- tokenized: in order to provide useful text search, you need to
tokenize the "text" field(s), but you will better avoid to tokenize
date fields. For "foreign keys" fields, if you need to store only a
single value, you probably better avoid tokenize it. If instead you
need to store a "list" of foreign key values, your probably set the
value of the field with a (spare separated) values and let Lucene
Lucene manages also other way to work with the field values, but you
should not immediately need to understand them for your current need.
Now, what you really need to do is realize that Lucene basically
handles only strings, but you have quite a few "knobs" to tune. After
understanding the option that Lucene offers you, go back to your
problem and try to find out how to let Lucene encode your information
in order to achieve your expected result.
If the content of the DB is updated, and you can not manage to reindex
the whole DB every time you need to synchronize its content, you will
need also to define an update strategy for the Lucene index.
Giulio Cesare Solaroli
On 7/4/06, Alexander Mashtakov <amashtakov@xxxxxxxxx> wrote:
Thank you *Giulio,*
I read some information about this approach. Is it possible to store in
several FKs ? For example, a document has serveral mappings to category
The other question is about dates - is it possible to query Lucene for date
interval (from YYYY-MM-DD to YYYY-MM-DD) ?
And the last (but no least ;) - in database, there are mappings (N:M) which
contains additional attributes, that have to be searchable - for
example one document may reference other documents (M:N) with attribute
typeOf (isPartOf, IsVersionOf). Is it possible
to include typeOf together with N:M map ?
Sorry, may be questions are stupid but I've just started to investigate
On 7/4/06, Giulio Cesare Solaroli <giulio.cesare@xxxxxxxxx> wrote:
> Hi Alexander,
> if the set of foreign keys you are using to filter the lucene results
> with some SQL query, I suggest you to include the foreign keys values
> as additional lucene fields, and let lucene return you the exact set
> of document IDs.
> I had excellent results with this technique.
> Hope this helps.
> Best regards,
> Giulio Cesare Solaroli
> On 7/4/06, Alexander Mashtakov <amashtakov@xxxxxxxxx> wrote:
> > Hi folks,
> > I'm looking for a solution/best practices concerning Lucene and SQL
> > integration.
> > The database (MySQL) is already developed and contains data. I've tried
> > MySQL full-text
> > search, but it's quite slow and doesn't have the possibility to
> > custom analyzers.
> > Phrase search is perfomed only in boolean mode and doesn't return
> > factor :(
> > The idea is to manage full-text indexes (titles, keywords, summaries)
> > perform search
> > using Lucene. The resultset will include ID's that will be appended to
> > query in order
> > to apply additional filters based on foreign keys (categories mappings,
> > etc).
> > But, the database is going to be big enough, and the list of IDs
> returned by
> > Lucene too. This
> > may cause high memory usage and slow sql query speed (for instance 1000
> > in "IN (id1, id2 ...)"
> > sql filter)
> > Any ideas, suggestions ?
> > --
> > Thank you,
> > /Alexander
> To unsubscribe, e-mail: java-user-unsubscribe@xxxxxxxxxxxxxxxxx
> For additional commands, e-mail: java-user-help@xxxxxxxxxxxxxxxxx
To unsubscribe, e-mail: java-user-unsubscribe@xxxxxxxxxxxxxxxxx
For additional commands, e-mail: java-user-help@xxxxxxxxxxxxxxxxx