Documentation

A Term-indexed associative data structure supporting approximate (conservative) lookup. Rather than using a Trie keyed by Key directly, a DiscrimTree is instead represented more like a case tree.

This allows us to exploit the fact that instance selection often focuses on a small part of the term: Only that critical chain is represented in the tree. As an example, level parameters are unlikely to contribute to narrowing a search problem, so it would be wasteful to have an indirection in the tree for every FlexK standing for a level parameter.

Merge a pair of discrimination trees. This function tries to build the minimal discrimination tree that yields the union of the inputs' results, though it does so slightly naïvely, without considerable optimisations (e.g. it does not turn single-alternative CaseDTs into DoneDTs).

Construct the case tree corresponding to only performing proper matches on the given key. In this context, a "proper match" is any Key that is not FlexK.