Post #2479129

@zwarich@hachyderm.io @MartinEscardo@mathstodon.xyz OK, this is a quite elegant idea. I feel like in practice I would want to write the tree automata as an automata over parametrized states (e.g. with integer precedence) but that's a minor detail. A tool that took in a grammar and a tree automata and produced random samples from their intersection would be quite useful for testing parsers.