Type Inference for the Mixture of Matching and Implicit Subtyping

We present a type inference system for a language with object-oriented features such as polymorphic record access and subtyping. We separate the notion of matching and subtyping in order to avoid difficulties caused by recursive type constraints. We will use two kinds of type constraints --- record (variant) constraints and subtyping constraints. Then, we discuss simplification of such mixed type constraints and show that simplification becomes much easier by the separation. 1. Introduction In this paper, we propose an ML-style type inference system which uses constrained types and allows both polymorphic record access and implicit subtyping. We will then discuss simplification of type constraints. In the design of type systems for object-oriented languages, much attention is paid in order to avoid difficulties caused by binary methods [1] --- methods which take another object of the same class as an argument. The problem is that if a class has binary methods, its subclasses do not ..