I agree with Chris, a Hibernate-based solution would probably be the best. I have never used MySQL "user defined functions" but I assume they are similar to oracle's "stored procedures". In my opinion such a solution has some disadvantages; they tend to be slow and also break the DB-independence that Hibernate provides. What's the disadvantage of just creating a CDK-DB module with an API for accessing hibernate-DAO's?
As I see it, the ideal approach would be if someone knowledgable of the core CDK-classes added XDoclet tags so that hibernate mapping files and DAO-objects could be generated with Ant/XDoclet. I know this is a big undertaking (and far from trivial), but in my opinion it would be the most versatile. Adding business-logic above this layer would then be totally separated from the database access methods. Another point: Wouldn't a quick solution be to just store the entire molecule/reaction etc as CML and duplicate all elements that you need indexed (like ID, fingerprints, hashes, names etc.) as common data types in a regular RDB? That (together with a custom annotation system) was the quick solution I had in mind for a simple Bioclipse-database plugin based on Hibernate. Very simple, and very easy/fast to implement. As long as you specify the additional elements to be persisted, and write an API that takes care of this, it would in my opinion scale quite well. Can anyone think of a solution that would scale better and still keep all information in the database? Well, just my thoughts on the matter. I'm sure there are many comments on this! Cheers, .../Ola On Thu, 2005-12-29 at 20:56, Christoph Steinbeck wrote: > Egon Willighagen wrote: > > > Java uses a general JDBC for interaction with database; it's the java > > version > > of the general ODBC architecture. Databases provide ODBC drivers, and users > > just know the JDBC/ODBC layer. > > > > The DBReader/DBWriter use JDBC, and I assume NMRShiftDB does too. > > Indirectly, yes. > We use an object-relational mapper, like anyone else :-) > Due to the "age" of NMRShiftDB, it's the Torque mapper from the apache > project. > I guess, nowadays one would use Hibernate. > > The trick is not the OR-Mapping but the code to do the chemistry searches. > Substructure searching simply cannot be done efficiently based on SQL, so one > needs to do fingerprint-based prescreening. > This again needed custom code, in MySQL called "user defined functions". > > Cheers, > > Chris ------------------------------------------------------- This SF.net email is sponsored by: Splunk Inc. Do you grep through log files for problems? Stop! Download the new AJAX search engine that makes searching your log files as easy as surfing the web. DOWNLOAD SPLUNK! http://ads.osdn.com/?ad_id=7637&alloc_id=16865&op=click _______________________________________________ Cdk-user mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/cdk-user
