The tabling implementation could use some attending to. Currently the tabling implementation generates a lot of redundant information instead of sharing information across tabled calls. For example currently after some calls to a tabled goal - the goal table looks like this:

t ={arg: #{ans0, ans1, ans2, ..., ansN} ans0: #{ans1, ans2, ..., ansN} ans1: #{ans2, ..., ansN} ... } 

Yuck! :)

It maybe the case that tabling interacts oddly with constraints. In Nada Amin's TAPL repo, the tabled reducer shows the correct results but also shows some spurious contraints. We should investigate.