Does the tool figure it out by some heuristics? Or we tell it the "normal" data ranges -
somehow affecting the generation logic? Heuristics would be great -
- but somehow never thought of/missed would get covered...
As far as GenRocket is concerned, realistic test data simply means data that could be conceived as real. To explain this i'll compare it to patterned data. Patterned data, which GenRocket can generate, might look like this:
Realistic data might look like this
The difference is simply a Generator choice. For the realistic data you might choose a NameGen to generate first and last names. Whereas for patterned data you might choose a ConcatGen to join the ID and some other value together.
The important thing to remember is that your test data should be reliable and predictable. You'll need to be able to assert that what you get is what you expect. If you leave too much of that data up to chance and randomness, you will not have very reliable tests.