Integration of generative testing

[halgari]

I'm loving tUnit so far, and would love to start integrating generative testing into some of my projects. An example project for this sort of thing is FsCheck. In order to do this we'd need some way to run tests and refine the inputs of a test based on what tests fail.

For those who haven't seen generative testing before, it mostly involves probing a test, if a test takes int it would be run first with 0 then 1, then 100, then -100, etc. until it fails. If the test fails on 50 and not 1 the testing suite then tries 25 then performs a binary search of the input space until a single failing result is found. The report for the failure is then "Test failed with input 42"

I'm not sure how best to integrate this sort of test process. But I'd love to kick off the converstion.

[thomhurst]

I did experiment with a method on the context like "ReregisterTestWithArguments(Args)" and it'd clone the test but with the new arguments and kick it off.

The logic for data generation could be custom and defined in a Data source generator attribute.

Feel free to have a play. It's marked with experimental.

I don't have any experience with FsCheck, and when I tried to explore, it went over my head tbh.

[thomhurst]

See here: https://github.dev/thomhurst/TUnit/blob/4f338e9bc6750afd720d4a4065e97925636fc841/TUnit.TestProject/DynamicallyRegisteredTests.cs

If you had time to play around with it, I'd appreciate feedback on developing it further.