Abstract

Compilers for programming languages such as Standard ML are able to find many programming errors at compile time, however the diagnostic messages from the type inference algorithm do not always clearly identify the source of type errors. We argue that an extended type definition, which assigns types to open expressions as well as closed expressions, can serve as the basis for a programming environment that helps programmers debug type errors. We present such a type definition, which is closely related to the Damas/Milner definition, but which in addition provides principal typings for open expressions. We present an algorithm that performs type inference with respect to our type system and give a simple direct proof of its correctness. Finally we describe a prototype implementation. Keywords: Programming environments, Type Inference 1 Introduction The Standard ML type system permits ML [MTH90] compilers to identify certain kinds of programming errors at compile time by inferring a "m...

Citations