Short answer is that Datalog is originally a logic/proof language, while SQL is an arithmetic / statistical / tabular aggregation language. One can make one so

Short answer is that Datalog is originally a logic/proof language, while SQL is an arithmetic / statistical / tabular aggregation language. One can make one solve problems of the other but they're really different mental models, which impacts the representation of the data needed to enable their use. For Datalog one needs relationships- a graph- not what SQL calls relations, which is just data. Datalog requires a richer set of opinions- at least conceptually- about the data. Languages like Datalog will not become mainstream until graph modeling is mainstream.