Investigate Haskell rewrite of testsuite driver
Currently, the testsuite driver is written in Python. This is a bit of a liability:
- Having to program in "not Haskell" might scare away potential contributors here
- Even developers who are proficient in Python probably prefer working with Haskell
- Managing Python as a dependency poses an additional burden and complicates deployment, setting up dev environments, CI, etc.; especially on Windows.
- We miss out on the benefits of type checks and typed programming for a substantial bit of infrastructure
- We miss out on a nice dog-fooding opportunity
So in order to judge the scope of this task, and define it better, it would be good to investigate a bit:
- What would it take to get high-level feature parity with the current solution?
- Does the Haskell ecosystem cover the concerns that we would like to address with existing libraries? (Process management, for example)
- Can we come up with a smart and frictionless way of migrating all the existing test cases in a reliable automated fashion?
- What are risks and unknowns?
#15363 (closed) is the recent case that sparked this - rather than the proposed patch there, which ports existing Haskell code to Python, we would prefer going in the other direction.