methods2test VS algebra-driven-design

Datasets: * https://github.com/microsoft/methods2test -- dataset of Java code and unit tests -- so one may used supervised learning to generate tests. Similar dataset may be available at your company for ABAP, and you may try training an encoder-decoder / transformer model for test case generation.

You might want to read on Algebra-Driven Design and Data Oriented Design. If you are greedy, both books were pirated.

I would like to do this more. At the moment I don't do formal verification. However, I often borrow methodology from the excellent book Algebra-Driven Design (text available on GitHub but support the author if you find it useful!) when designing systems. This means I define algebraic data types to describe the program and the laws that relate the different types, and use that to guide implementation in a language that doesn't support ADTs.

Their approach might be to express their business logic as a carefully selected collection of types and laws, as described in Algebra-Driven Design (the full text is available on GitHub, please support the author if you find it useful though). I recommend this book to everyone because, even if you don't use this approach to design your programs, it's an excellent way to think about problems and better understand the problem space.

If you have an interest in computer science, category theory is very useful there. Here's a good book on the topic (Algebra-Driven Design, by Sandy Maguire).

This reminds me of QuickSpec (different from QuickCheck) in Haskell. It takes Haskell code and finds the Mathematical laws that the code supports. It does this using a sort of smart random search. I learned about this as part of the Algebra Driven Design book.

Sorry probably worth pointing out that I'm not referring to an Algebra in the pure math sense. I'm more referring to specifying algebraic laws that an API would have to satisfy as talked about in this book https://leanpub.com/algebra-driven-design.