the opinion about tight or loose coupling is also influenced by the frequency of changes in the RDBMS schema.
If my application is - through Torque - tightly coupled with the RDBMS, it will almost certainly fail with an exception upon a moderately significant change in the RDBMS. Which is good, because it will precisely pinpoint the change, and makes 'sure' (well, fairly sure that is :-) that my appliction only runs against the RDBMS that is was designed for and none other.
But I agree, having the possibility of making Torque more loosely coupled from the RDBMS would be a nice feature. It could be implemented by allowing specifying aliases for db objects in the XML schema definition which does seem to be fairly simple to implement, but maybe a more sophisticated abstraction layer isn't that hard to make either.
gr. Michel
Manske, Michael wrote:
hi,
i knew that such a discussion would come up and it depends on the point of view of each indivual user. :)
I don't know, I think I would Torque rather see more tightly coupled with the RDBMS and dump the XML schema entirely.if you have control over database structure and changes of the database
structure, then you will
perhaps prefer a strict coupling. But if not (like me), you will always
prefer loose coupling to be more independent of changes made by another dev
team.
My RDBMS already has a schema, which would be the metadatabase in the systems tables. So why create another definition in XML of the same database and tables?
If you have to support different RDBMS the metadescription in some "system tables" will get useless.
Torque's capability of abstraction of the RDBMS-specific isssues comes in quite handy here. The process could be automated by having Torque generate the XML definition from a JDBC conncection, and then generate the om from that XML, but I haven't tried that yet.
Thats what i'm talking about, we are working with torque this way because we have to deal with a couple of already existing databases. And yes, torques abstraction is somewhat of handy - thats why we use it. :-)
Loose coupling means among other things to hide the physical database structure completely from the objects, which have to access the database. A layer (like torque) will then act as mediator between objects and database. So if you would have problematic identifiers like "short", you would be easily able to map them to another name, which could then be used in java objects, e.g. map "short" to "short_descr". There is already some kind of support for this but at the moment it isn't suitable at all.
I guess torque is so popular because of his abilities to generate more or less useable code and the usage of a xml schema at runtime (respectively at application startup) would possibly be contradictory to the generator BUT it would also provide more independency from used database structure.
I'm not sure wheter this is a mainly intention of torque but i would be glad if the devs would expand support for loose coupling (at least for mapping of table/column names to java names) in future versions...
regards, Michael
PS: pros and cons of loose coupling will always be a matter of opinion
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
