We describe an extension of ML with records where inheritance is given by ML generic polymorphism. All common operations on records but concatenation are supported, in particular the free extension of records. Other operations such as renaming of fields are added. The solution relies on an extension of ML, where the language of types is sorted and considered modulo equations, and on a record extension of types. The solution is simple and modular and the type inference algorithm is efficient in practice.

representing the official policies, either expressed or implied, of the U.S. Government.

Type systems for operations on extensible records form a foundation for statically typed languages addressing some aspects of object oriented programming and database applications. A number of primitive operations have been proposed: extending a record with a new field, overwriting an existing field, removing a field, and various kinds of concatenation. We show here that a record calculus based on a symmetric concatenation operator, where two records may be concatenated only if they have no overlapping fields, also captures the types of many other useful primitive record operations. "Mergeability constraints" are expressed directly using explicit annotations on type variables and constrained second-order type quantification instead of a rule of subsumption; we argue that the resulting system is more straightforward than subsumption-based alternatives. This research was supported in part by the Office of Naval Research and in part by the Defense Advanced Research Projects Agency (DOD), ...

Type systems based on intersection types have been studied extensively in recent years, both as tools for the analysis of the pure -calculus and, more recently, as the basis for practical programming languages. The dual notion, union types, also appears to have practical interest. For example, by refining types ordinarily considered as atomic, union types allow a restricted form of abstract interpretation to be performed during typechecking. The addition of second-order polymorphic types further increases the power of the type system, allowing interesting variants of many common datatypes to be encoded in the "pure" fragment with no type or term constants. This report summarizes a preliminary investigation of the expressiveness of a programming language combining intersection types, union types, and polymorphism.

We show that any functional language with record extension possesses record concatenation for free. We exhibit a translation from the latter into the former. We obtain a type system for a language with record concatenation by composing the translation with typechecking in a language with record extension. We apply this method to a version of ML with record extension and obtain an extension of ML with either asymmetric or symmetric concatenation. The latter extension is simple, flexible and has a very efficient type inference algorithm in practice. Concatenation together with removal of fields needs one more construct than extension of records. It can be added to the version of ML with record extension. However, many typed languages with record cannot type such a construct. The method still applies to them, producing type systems for record concatenation without removal of fields. Object systems also benefit of the encoding which shows that multiple inheritance does not actually require...

with constraints. The basic idea is to factor out the common core of previous extensions of the Hindley/Milner system. I present a Hindley/Milner system where the constraint part is a parameter. Speci c applications can be obtained by providing speci c constraint systems which capture the application in mind. For instance, the Hindley/Milner system can be recovered by instantiating the constraint part to the standard Herbrand constraint system. Type system instances of the general framework are sound if the underlying constraint system is sound. Furthermore, I give a generic type inference algorithm for the general framework, under sucient conditions on the speci c constraint system type inference yields principal types.

We extend the Abadi-Cardelli calculus of primitive objects with object extension. We enrich object types with a more precise, uniform, and flexible type structure. This enables to type object extension under both width and depth subtyping. Objects may also have extendonly or virtual contra-variant methods and read-only co-variant methods. The resulting subtyping relation is richer, and types of objects can be weaken progressively from a class level to a more traditional object level along the subtype relationship.

This paper explains how the high-level treatment of datatypes in functional languages—using features like constructor functions and pattern matching—can be made to coexist with bitdata.Weuse this term to describe the bit-level representations of data that are required in the construction of many different applications, including operating systems, device drivers, and assemblers. We explain our approach as a combination of two language extensions, each of which could potentially be adapted to any modern functional language. The first adds simple and elegant constructs for manipulating raw bitfield values, while the second provides a view-like mechanism for defining distinct new bitdata types with fine-control over the underlying representation. Our design leverages polymorphic type inference, as well as techniques for improvement of qualified types, to track both the type and the width of bitdata structures. We have implemented our extensions in a small functional language interpreter, and used it to show that our approach can handle a wide range of practical bitdata types.

: Current programming languages that are grounded in a formal logic --- such as pure Lisp (based on the lambda calculus) and Prolog (based on Horn clause logic) --- do not support the use of complex, pointer-based data structures. The lack of this important feature in logically grounded languages contrasts sharply with its strong support in the imperative programming languages that have enjoyed wide application, of which C is a prime example. Unfortunately, the formal methods for reasoning about imperative languages have not proved broadly useful for reasoning about programs that manipulate complex, pointer-based data structures. Between these two camps resides an open question: How can we verify programs involving complex, pointer-based data structures? This work gives an answer to this question. It describes a programming language in which a programmer can define logical predicates on data structures and pointers, and use these predicates to specify programs that manipulate complex, ...