On Jul 7, 2008, at 5:27 AM, Rick McGuire wrote:
I've finally gotten back around to looking at this. I notice that the validation for the fields that CmpJpaConversion adds as implicit CMP fields, there's no validation performed on the setters (which the generator will generate unconditionally). Shouldn't there be a check for a matching setting method with a compatible name and parameters?
In OpenEJB we perform as much validation as possible in the Validation phase of deployment. This vastly simplifies the remaining deployment code because it can assume the inputs are correct. Before David made this optimization, you would typically have several lines of validation code for each step of deployment which really made deployment confusing and hard to follow. Additionally, the deployment would fail at the very first error, so when a developing an app you would get in a loop of fix one error, redeploy, fix the next, redeploy and so on. The validation phase works more like a compiler in that it collects all deployment problems and returns all errors to the happier user.
Also, the selection of the "is" methods is not checking that the return type is boolean. Shouldn't that be included as one of the selection criteria?
It should. -dain
