I vaguely remember trying Shapiro’s code a long time ago. The papers looked good – but the code was clearly “research quality” and would have needed a lot of work to become useable. I suspect that a combination of trace/2 and wrap_predicate/4 would work better for large programs.