

That looks very nice and seems to hit most use cases as far as I can see in my just-woken state. If it had been easier to search for, then I might now have written mine.
I’d have to try it to see how it deals with my common case of rather a lot of numpy arrays in hierarchies if ducts or pydantic models and the floating point precision issues. Maybe it’ll do that brilliantly.




There is only one of each of the other creatures as well, so?