Usually you'd have a set-like storage of n-tuples/structs per relation, as a row/tuple of a relation has fixed arity. Tree-sets of rows, for example. But you co

Usually you'd have a set-like storage of n-tuples/structs per relation, as a row/tuple of a relation has fixed arity. Tree-sets of rows, for example. But you could just use SQL tables as a storage engine with one table per datalog relation.