I'm having a bit of a hard time grasping how to make meaningful tests based on the randomly generated data vs making my test in a transaction where I feed it initial data and check for an expected outcome.
For example's sake if I had an SP called DivideTwoCols that would divide colA by colB in table0...
If I were making tests with known data, I'd probably make a test using two positives, one negative and one positive, two negatives, colA = 0, colB = 0, and no rows in table0.
If I were just testing randomly generated data how would I know I had my cases covered would the random test just pick a random row to try it with using select top 1 colA, colB from table0 order by newid()
I can appreciate throwing in some randomness in testing to check for conditions you may have forgotten to check for, but it seems fairly hit or miss and not a reliable indicator that I'm checking all things properly.
Am I missing some core idea or plan for using random testing Are there any tutorials, best practices, sites available that go over ideas on random testing with the generator