TOWARDS PROGRAMMING WITH FIRST-CLASS PATTERNS

Abstract

Pattern matching is a powerful programming paradigm which first appeared in functional programming languages to make data structure analysis and decomposition more declarative. Promoting patterns to first-class status does not increase the computational power of a programming language, but it does increase its expressiveness allowing for brand new ways of solving problems. First-class patterns were studied in the context of the lambda calculus. Today, almost all modern programming languages incorporate some form of pattern matching. However, with only a few exceptions, all programming languages we are aware of that support pattern matching stop short of treating patterns as first-class citizens. Consequently, many interesting use cases of pattern matching lie beyond the reach of those languages. We have implemented first-class patterns in Asteroid, a dynamically typed, multi-paradigm programming language, in order to assess and experiment with first-class patterns. Here we report some of our initial findings. The idea of first-class patterns is not new but we feel that the insights provided here are novel and highlight the impact that first-class patterns can have on programming languages and the discipline of programming itself .