Tell us more about your usage

[dubzzz]

We want to share with the community about how our users are using fast-check, what they think of it, what they struggle with, what would be their dreamed feature... or anything that come to your mind.

We would like to share these journeys on our blog, with possibly one article for each feedback. If you're interested into it, we are looking for your article.

[gruhn]

Currently, I'd love to use something like model-based testing against a local-first, real-time text editor similar to Google docs. There are "commands" like user-edits-text, user-makes-something-italic, etc. Every action is send to the server as a message. At any time the connection can be lost but the application should still be usable.

The problem with testing it, is that I can only control the client. So I either have to:

1. Ignore the problem and talk to the real backend during tests: Makes command sequences non-deterministic which breaks shrinking.
2. Mock the backend: Very complex. Hard to mock the entire backend accurately. Even if done well, there is a risk of drift over time.
3. Create a programmatic API to control backend: A lot of refactoring effort. Also a lot of non-determinism comes from the network.
4. After every client action, add a ton of awaits to make sure the backend "settles": Works ok but won't be able to reproduce all the hard/interesting race conditions where a ton of things happen in quick succession.

None of these options are satisfying but maybe we can improve on (1). I recently discovered git bayesect: A version of git bisect that can handle a flaky oracle. Very interesting approach. I was wondering if one could do something similar for shrinking. I.e. a probabilistic shrinking algorithm that is resilient to false negatives (also false positives?).

If that works well, maybe that would make PBT in general more easily adoptable because there is less overhead to make your system PBT friendly.

[dubzzz]

If one day you want to share your experience of fast-check in the blog of the library please tell me. Clearly welcomed 👍