"Datalog is declarative, optimizable, and relatively expressive. However, writing Datalog often involves repeating yourself; the restrictions that make Datalog practical also prevent ““higher-order”” idioms ─ for example, you cannot define a generic transitive closure operator. Datafun is a higher-order, typed functional language with Datalog-inspired semantics. In this talk, I’ll:

1. Explain what Datafun is and how to write Datalog-style programs in Datafun.

2. Give examples of Datafun’s increased expressivity.

3. Show how Datafun’s support for semilattice aggregations is a natural fit for program analysis.

4. Discuss why I think Datafun, although still in the design stage, has the potential to be implemented as efficiently as Datalog. In particular, we have recent work on generalizing seminaïve evaluation to Datafun."