RE: Ozone applicable for my faceted classification?

Hi,

To your specific questions:

- Will I be able to structure the above in ozone (probably yes) and what

kind of problems may I run into (if I follow the ozone-way-of-work, I 
should run into none, right?) ? >Yes.  To avoid problems, it is
important to define your storage-classes as subclasses of OzoneObject.
Also define your query methods so that as much of the work as possible
can be done on the server side, without having to send intermediate
results back to your application for next-level-query, etc.

- Will I be able to efficiently retrieve a set of 
facetelements/documents from the database that fully comply to the 
requested set of criteria or with a minimal amount of work? >Yes. I
think your application is well-suited to Ozone.

Further comments:

With Ozone, you can approach things *as if* you were keeping everything
in memory.  Just make sure that *all* classes that store your data are
defined as Ozone objects.  Roughly, each unique class might correspond
to a "table" or "record type" in conventional database terminology.

(Not all Ozone objects need to be, nor should they be, named objects.
It is better to have one or a few named objects representing the root or
key starting points in your database.)

So if you extend your classes from OzoneObject, when you first reference
an instance of one of your classes, Ozone will only then load that
object into its cache.  Ozone's cache serves as a kind of buffer between
your application memory and disk, so that everything does not really
have to be in memory all the time, although your code will look as if it
were.

So following your proposed structure with facets and intersecting
query-result-sets:

You might want to try to organize your data so that you have an index
table (say, hashtable) that has a unique key for each unique facet-type.
Then for each unique facet-type-key, have as its value a
facet-value-hashtable.  The facet-value-hashtable would have a unique
key for each unique facet-value for its facet-type.  Finally, each
unique facet-value-key would have as its value a set of references to
the next level in a hierarchy or to the terminal objects (your
documents) that match the facet-value-key.

The "terminal objects" could be Ozone objects.  Then as you described,
you just have to intersect the sets of references and return the result.
I am not sure but I think that in the newer versions of Ozone, for
"references" you can get something like a unique, persistent object
handle or ID for each Ozone object so that you can intersect sets of IDs
without having to load all the objects.  Perhaps someone else on the
list could advise us.

Best Regards,
John Dixon

-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Arjen van
der Meijden
Sent: Tuesday, February 24, 2004 10:28 AM
To: [email protected]
Subject: [Ozone-users] Ozone applicable for my faceted classification?


Hi list,

I'm afraid this email became a bit long, but if someone can help me, I'd

be very grateful.

I'm going to decide upon a storage system for a custom designed faceted 
classification system. I have a preliminary design of my classes, which 
you can find here: http://arethusa.tweakers.net/~acm/facettree.png

The storage system will have to store all those objects, or a similar 
representation, in some persistent way since keeping everything in 
memory is probably a no-go (maybe, I can keep the structure mapped in 
memory and retrieve the documents). I do see a way to construct this 
setup in a sql-database. The model itself shows that it can be done in
Java.

The problems start as soon as I'm starting to think of the document 
retrieval process. Documents can be retrieved based on certain criteria.

For instance, one could want to have the set of documents that:
- are in the main FacetClass -> Computer Hardware -> Laptops (all 
FacetObjects "below" this Class),
- are in a facet Who-facet -> Manufacturer -> Dell (all FacetObjects 
related to this facet, directly or indirectly),
- are in a facet Part-specification -> screen size -> 15 inch (""),
- has to be found in the location "forum",
- should be dated later than 1-jan-2003.

One needs a simple cross-section of the result sets of the above set of 
specifications and has a list of complying documents.

Since I have found a few ways in SQL to do single-query tree-traversal I

can get the above criteria in a few simple queries which can then be 
literally intersected using the INTERSECT-statement, but that is slow. A

good performing query will, however, probably become very easily very 
complex and as a result also be (a bit less) slow. I'm therefore not too

happy with the ways SQL allows me to do the above.

A straightforward solution in an OO-environment is to simply retrieve 
all documents which are connected (potentially using multiple levels of 
the hierarchy) to each selected FacetElement or Facet, intersecting 
those sets and afterwards filtering the location and date.
With a few thousand documents, this can be done relatively fast, but I 
may get tens or even hundreds of thousands documents in our setup.

My questions are simple:
- Will I be able to structure the above in ozone (probably yes) and what

kind of problems may I run into (if I follow the ozone-way-of-work, I 
should run into none, right?) ?
- Will I be able to efficiently retrieve a set of 
facetelements/documents from the database that fully comply to the 
requested set of criteria or with a minimal amount of work?

If the last question is a result of my sql-focusness, please point me in

the right direction of reformulating that question :)

Thanks in advance and best regards,

Arjen van der Meijden




-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
_______________________________________________
Ozone-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ozone-users


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id56&alloc_id438&op=click