In MLscript, patterns are first-class. They can be recursive, higher-order, and can be composed through Boolean operations (union, intersection, and complement).

Patterns can transform what they match using arbitrary computations; in that sense, they can be viewed as composable partial functions. The composition of patterns, when viewed as functions, is denoted via the binary operator >>.

A pattern can be applied repeatedly to its own output until that output reaches a fixed point. Importantly, each matching pass runs in linear time with respect to the size of the input, and fixed point patterns do not re-traverse unchanged term prefixes. With these abilities, lambda calculus evaluators (similar to Danvy-Nielsen refocusing), object validators, expression simplifiers, and even stack-based parsers can be written entirely as patterns—declarative, typed, and efficient. We call this style pattern-oriented programming, and we believe it could be the future of functional programming.

The demonstration will consist of a live coding session in which the authors develop two examples in real time: arithmetic-expression simplification, as a warm-up, and the call-by-value lambda calculus evaluator described above. Both examples will be available in a web-based MLscript IDE, allowing attendees to follow along and experiment with variations during the session.