On 2/6/06, Tim Dudgeon <[EMAIL PROTECTED]> wrote: > thanks for that info. I understand your explaination, but doesn't this > significantly limit the use of DDLUtils? > If DDLUtils can make uncalled for changes to any database that has been > created, or modified, by anything other than DDLUtils, then that's quite > a high risk to accept?
Its not so much limiting but ruther a matter of focus. In its current version DdlUtils primarily intends to allow Java applications to create and initialize databases without having to care about the actual database and its native types, SQL structure, etc. One particular point is that DdlUtils must ensure that the native type that a JDBC type is mapped to, is suited (large enough, the driver accepts the corresponding Java type etc.). The problem now is that when reading back a model, DdlUtils has to rely on the JDBC driver's implementation of the database metadata. And if the driver decides that multiple types map to the same JDBC type (think Oracle which maps numeric types back to JDBC type DECIMAL), then we're in trouble because DdlUtils has no way to distinguish between them (perhaps only via the size as DdlUtils does in the case of Oracle) and hence problems as in your case are bound to arise. That being said, it would surely be possible to invest more work and make this more robust. But this means a lot of work, and thus cannot be a target for the 1.0 release which should come out real-soon-now(tm). Tom
