Towards a Reversible Functional Language

We identify concepts of reversibility for a functional language by means of a set of semantic rules with specific properties. These properties include injectivity along with local backward determinism, an important operational property for an efficient reversible language. We define a concise reversible first-order functional language in which access to the backward semantics is provided to the programmer by inverse function calls. Reversibility guarantees that in this language a backward run (inverse interpretation) is as fast as the corresponding forward run itself. By adopting a

symmetric first-match

policy for case expressions, we can write overlapping patterns in case branches, as is customary in ordinary functional languages, and also in leaf expressions, unlike existing inverse interpreter methods, which enables concise programs. In patterns, the use of a duplication/equality operator also simplifies inverse computation and program inversion. We discuss the advantages of a reversible functional language using example programs, including run-length encoding. Program inversion is seen to be as lightweight as for imperative reversible languages and realized by recursive descent. Finally, we show that the proposed language is

r-Turing complete

.