Add ability to seed vanilla state sampler for testing
e-cal opened this issue · comments
There isn't a way to reliably test that the effects of an extracted model are correct in a vanilla state sampler -> Observer integration test as the traces generated will be different everytime. Some actions may be used or not used across different generations, and in a different order, causing the extracted pre-conditions to vary.
In order to reliably test the accuracy of an extracted model, the same state trace data needs to be generated each time.
This do the trick?
https://www.w3schools.com/python/ref_random_seed.asp
Even just running the end-to-end and asserting it completes would be useful (so we know the README code still works without crashing ;)).
Just completion is what I have now, figured that's good enough for the readme test, but I'm assuming we'll want proper integration tests at some point.