I'm starting a new project, and trying very very hard to use TDD to drive the design. I've been pushing for years, and finally got approval to spend the extra time on this project to use it while I learn how to do it properly.
This is a new module, to tie into an existing system. Currently, all data access happens through webservices, which for the most part are just a thin wrapper over database stored procedures.
One requirement is that for a given store, I return all Purchase Orders that are considered valid for this application. A PO is considered valid if it's ship date falls withing a given range from the stores opening date (this is for new stores).
Now, I can't put this logic in the application code, as I'm not going to bring back a million POs just to get the dozen that apply to could apply to this store given the constraint above.
I was thinking, I could pass the date range to a GetValidPOs proc, and have it use those values to return the valid POs. But, what if we add another requirement to what is considered a valid PO?
And how do I test this and verify it keeps working? We are not using an ORM, and it's unlikely to happen. And I can't call the DB in my test.
My other thought, is have some mocks that return valid data, others that return some bad data, and have the local repository throw an exception if bad data occurs, and test that the exception gets thrown if invalid data is returned by GetValidPOs proc (or the mock used in testing).
Does this make sense? Or is there a better way?
UPDATE: I am able to use EF it would seem. Now I just need to figure out how to use it, and make it testable, while still being able to rely on stored procedures, and the difficulty of having data scattered across several databases.