Thursday, March 17, 2016

Datafun: A Functional Datalog

Together with Michael Arntzenius, I have a new draft paper, Datafun: a Functional Datalog
Datalog may be considered either an unusually powerful query language or a carefully limited logic programming language. It has been applied successfully in a wide variety of problem domains thanks to its "sweet spot" combination of expressivity, optimizability, and declarativeness. However, most use-cases require extending Datalog in an application-specific manner. In this paper we define Datafun, an analogue of Datalog supporting higher-order functional programming. The key idea is to track monotonicity via types.

I've always liked domain specific languages, but have never perpetrated one before. Thanks to Michael, now I have! Even better, it's a higher-order version of Datalog, which is the language behind some of my favorite applications, such as John Whaley and Monica Lam's BDDBDDB tool for writing source code analyses.

You can find the Github repo here, as well. Michael decided to implement it in Racket, which I had not looked at closely in several years. It's quite nice how little code it took to implement everything!

No comments:

Post a Comment