We propose a notion of syntax with metavariables that generalises Miller’s decidable pattern fragment of second-order unification for simply typed $𝜆$ -calculus. Using categorical semantics, we show that, under some conditions, a generalisation of Miller’s unification algorithm applies. To illustrate our semantic analysis, we implemented our generic unification algorithm in Agda. The syntax with metavariables given as input of the algorithm is specified by a notion of signature generalising binding signatures, covering a wide range of examples, including ordered $𝜆$ -calculus and (intrinsic) polymorphic syntax such as System F. Although we do not explicitly handle equations, we also tackle simply typed $𝜆$ -calculus modulo $𝛽$ - and $𝜂$ -equations (Miller’s original setting) by working on the syntax of normal forms.

Higher-order unification has been shown to be undecidable. Miller discovered the pattern fragment and subsequently showed that higher-order pattern unification is decidable and has most general unifiers. We extend the algorithm to higher-order rational terms (a.k.a. regular Böhm trees, a form of cyclic λ-terms) and show that pattern unification on higher-order rational terms is decidable and has most general unifiers. We prove the soundness and completeness of the algorithm.

First-order unification algorithms (Robinson, 1965) are traditionally implemented via general recursion, with separate proofs for partial correctness and termination. The latter tends to involve counting the number of unsolved variables and showing that this total decreases each time a substitution enlarges the terms. There are many such proofs in the literature (Manna & Waldinger, 1981; Paulson, 1985; Coen, 1992; Rouyer, 1992; Jaume, 1997; Bove, 1999). This paper shows how a dependent type can relate terms to the set of variables over which they are constructed. As a consequence, first-order unification becomes a structurally recursive program, and a termination proof is no longer required. Both the program and its correctness proof have been checked using the proof assistant LEGO (Luo & Pollack, 1992; McBride, 1999).