Is it possible to measure how much work does LogicT do ?
Is it as simple as adding a mutable reference in the underlying monad state and updating it whenever the failure continuation (see runLogicT @ Control.Monad.Logic) returns? Is there a more principled approach? I’m not yet clear on how do state and backtracking play together.
Thanks for all pointers!
The failure continuation of runLogicT will be called just once, so that won’t help you count backtracking.
One approach is to define a monad on top of LogicT, such as a newtype around LogicT (State Int), which makes empty :: m a count every call.
State as the base monad provides a “global” state that cannot be rolled back by monad transformers you apply on top of it.
So for this approach the newtype would need also a custom MonadLogic instance, correct?
Indeed your approach corresponds with that of @econlon : Backtracking state with LogicT - Eric Conlon . Thanks!