Subtyping appears in a variety of programming languages, in the form of the ‘automatic coercion’ of integers to reals, Pascal subranges, and subtypes arising from class hierarchies in languages with inheritance. A general framework based on untyped lambda calculus provides a simple semantic model of subtyping and is used to demonstrate that an extension of Curry's type inference rules are semantically complete. An algorithm G for computing the most general typing associated with any given expression, and a restricted, optimized algorithm GA using only atomic subtyping hypotheses are developed. Both algorithms may be extended to insert type conversion functions at compile time or allow polymorphic function declarations as in ML.

Programming languages often come with type systems. Some of these are simple, others are sophisticated. As a stylistic representation of types in programming languages several versions of typed lambda calculus are studied. During the last 20 years many of these systems have appeared, so there is some need of classification. Working towards a taxonomy, Barendregt (1991) gives a fine-structure of the theory of constructions (Coquand and Huet 1988) in the form of a canonical cube of eight type systems ordered by inclusion. Berardi (1988) and Terlouw (1988) have independently generalized the method of constructing systems in the λ-cube. Moreover, Berardi (1988, 1990) showed that the generalized type systems are flexible enough to describe many logical systems. In that way the well-known propositions-as-types interpretation obtains a nice canonical form.

The λσ-calculus is a refinement of the λ-calculus where substitutions are manipulated explicitly. The λσ-calculus provides a setting for studying the theory of substitutions, with pleasant mathematical properties. It is also a useful bridge between the classical λ-calculus and concrete implementations.

A parallel program may be indeterminate so that it can adapt its behavior to the number of processors available.

Indeterminate programs are hard to write, understand, modify or verify. They are impossible to debug, since they may not behave the same from one run to the next.

We propose a new construct, a polymorphic abstract data type called an improving value, with operations that have indeterminate behavior but simple determinate semantics. These operations allow the type of indeterminate behavior required by many parallel algorithms.

We define improving values in the context of a functional programming language, but the technique can be used in procedural programs as well.

Quest is a programming language based on impredicative type quantifiers and subtyping within a three-level structure of kinds, types and type operators, and values.

The semantics of Quest is rather challenging. In particular, difficulties arise when we try to model simultaneously features such as contravariant function spaces, record types, subtyping, recursive types and fixpoints.

In this paper we describe in detail the type inference rules for Quest, and give them meaning using a partial equivalence relation model of types. Subtyping is interpreted as in previous work by Bruce and Longo (1989), but the interpretation of some aspects – namely subsumption, power kinds, and record subtyping – is novel. The latter is based on a new encoding of record types.

We concentrate on modelling quantifiers and subtyping; recursion is the subject of current work.

We present a modular proof of strong normalization for the Calculus of Constructions of Coquand and Huet (1985, 1988). This result was first proved by Coquand (1986), but our proof is more perspicious. The method consists of a little juggling with some systems in the cube of Barendregt (1989), which provides a fine structure of the calculus of constructions. It is proved that the strong normalization of the calculus of constructions is equivalent with the strong normalization of Fω.

In order to give the proof, we first establish some properties of various type systems. Therefore, we present a general framework of typed lambda calculi, including many well-known ones.

Traditionally the view has been that direct expression of control and store mechanisms and clear mathematical semantics are incompatible requirements. This paper shows that adding objects with memory to the call-by-value lambda calculus results in a language with a rich equational theory, satisfying many of the usual laws. Combined with other recent work, this provides evidence that expressive, mathematically clean programming languages are indeed possible.

The evaluation transformer model of reduction generalizes lazy evaluation in two ways: it can start the evaluation of expressions before their first use, and it can evaluate expressions further than weak head normal form. Moreover, the amount of evaluation required of an argument to a function may depend on the amount of evaluation required of the function application. It is a suitable candidate model for implementing lazy functional languages on parallel machines.

In this paper we explore the implementation of lazy functional languages on parallel machines, both shared and distributed memory architectures, using the evaluation transformer model of reduction. We will see that the same code can be produced for both styles of architecture, and the definition of the instruction set is virtually the same for each style. The essential difference is that a distributed memory architecture has one extra node type for non-local pointers, and instructions which involve the value of such nodes need their definitions extended to cover this new type of node.

To make our presentation accessible, we base our description on a variant of the well-known G-machine, an abstract machine for executing lazy functional programs.

Traditional functional languages do not have an explicit distinction between binding times. It arises implicitly, however, as one typically instantiates a higher-order function with the arguments that are known, whereas the unknown arguments remain to be taken as parameters. The distinction between ‘known’ and ‘unknown’ is closely related to the distinction between binding times like ‘compile-time’ and ‘run-time’. This leads to the use of a combination of polymorphic type inference and binding time analysis for obtaining the required information about which arguments are known.

Following the current trend in the implementation of functional languages we then transform the run-time level of the program (not the compile-time level) into categorical combinators. At this stage we have a natural distinction between two kinds of program transformations: partial evaluation, which involves the compile-time level of our notation, and algebraic transformations (i.e., the application of algebraic laws), which involves the run-time level of our notation.

By reinterpreting the combinators in suitable ways we obtain specifications of abstract interpretations (or data flow analyses). In particular, the use of combinators makes it possible to use a general framework for specifying both forward and backward analyses. The results of these analyses will be used to enable program transformations that are not applicable under all circumstances.

Finally, the combinators can also be reinterpreted to specify code generation for various (abstract) machines. To improve the efficiency of the code generated, one may apply abstract interpretations to collect extra information for guiding the code generation. Peephole optimizations may be used to improve the code at the lowest level.

This article describes theoretical and practical aspects of an implemented self-applicable partial evaluator for the untyped lambda-calculus with constants and a fixed point operator. To the best of our knowledge, it is the first partial evaluator that is simultaneously higher-order, non-trivial, and self-applicable.

Partial evaluation produces a residual program from a source program and some of its input data. When given the remaining input data the residual program yields the same result that the source program would when given all its input data. Our partial evaluator produces a residual lambda-expression given a source lambda-expression and the values of some of its free variables. By self-application, the partial evaluator can be used to compile and to generate stand-alone compilers from a denotational or interpretive specification of a programming language.

An essential component in our self-applicable partial evaluator is the use of explicitbinding time information. We use this to annotate the source program, marking asresidual the parts for which residual code is to be generated and marking aseliminable the parts that can be evaluated using only the data that is known during partial evaluation. We give a simple criterion,well-annotatedness, that can be used to check that the partial evaluator can handle the annotated higher-order programs without committing errors.

Our partial evaluator is simple, is implemented in a side-effect free subset of Scheme, and has been used to compile and to generate compilers and a compiler generator. In this article we examine two machine-generated compilers and find that their structures are surprisingly natural.

Polymorphic types are labels classifying both (a) defined components in a library and (b) contexts of free variables in partially written programs. It is proposed to help programmers make better use of software libraries by providing a system that, given (b), identifies candidates from (a) with matching types. Assuming at first that matching means unifying (i.e. having a common instance), efficient ways of implementing such a retrieval system are discussed and its likely effectiveness based on a quantitative study of currently available libraries is indicated. The applicative instance relation between types, which captures some intuitions about generalization/specialization is then introduced, and its use as the basis of a more flexible system is discussed.

This paper presents an application of functional programming: searching a domain for elements which satisfy certain constraints. We give a very general formulation of the problem and describe ‘generate and test’, ‘backtracking’ and ‘forward checking’ algorithms. We then introduce the concept of domain generating functions to capture a common optimization during the search process: using partial solutions to reduce the size of the search space. We compare the efficiency of the original algorithms and those using domain generating functions first with the ‘classical’ n-queens example, and then with a problem having larger domains to search which was inspired by an application in macromolecular structure determination. Using algorithms coded in Miranda, Haskell and Common Lisp, we show that a high order (lazy) functional language is a useful and efficient tool for prototyping search methods in large complex domains.

For numeral systems in untyped λ-calculus the definability of a successor, a predecessor and a test for zero implies the definability of all recursive functions on that system. Towards a disproof of the converse statement, H. P. Barendregt and the author constructed a numeral system consisting of unsolvable λ-terms, being adequate for unary functions. Then, independently, B. Intrigila found an analogous system for all computable functions.

A method is proposed to search for an identifier in a functional program library by using its Hindley–Milner type as a key. This can be seen as an approximation of using the specification as a key.

Functions that only differ in their argument order or currying are essentially the same, which is expressed by a congruence relation on types. During a library search, congruent types are identified. If a programmer is not satisfied with the type of a found value, he can use a conversion function (like curry), which must exist between congruent types, to convert the value into the type of his choice.

Types are congruent if they are isomorphic in all cartesian closed categories. To put it more simply, types are congruent if they are equal under an arithmetical interpretation, with cartesian product as multiplication and function space as exponentiation. This congruence relation is characterized by seven equational axioms. There is a simple term-rewriting algorithm to decide congruence, using which a search system has been implemented for the functional language Lazy ML, with good performance.

The congruence relation can also be used as a basis for other search strategies, e.g. searching for identifiers of a more general type, modulo congruence or allowing free type variables in queries.

Programming languages which are capable of interpreting themselves have been fascinating computer scientists. Indeed, if this is possible then a ‘strange loop’ (in the sense of Hofstadter, 1979) is involved. Nevertheless, the phenomenon is a direct consequence of the existence of universal languages. Indeed, if all computable functions can be captured by a language, then so can the particular job of interpreting the code of a program of that language. Self-interpretation will be shown here to be possible in lambda calculus.

The set of λ-terms, notation Λ, is defined by the following abstract syntax

where

is the set {v, v′, v″, v′″,…} of variables. Arbitrary variables are usually denoted by x, y,z,… and λ-terms by M,N,L,…. A redex is a λ-term of the form

that is, the result of substituting N for (the free occurrences of) x in M. Stylistically, it can be said that λ-terms represent functional programs including their input. A reduction machine executes such terms by trying to reduce them to normal form; that is, redexes are continuously replaced by their contracta until hopefully no more redexes are present. If such a normal form can be reached, then this is the output of the functional program; otherwise, the program diverges.

Abstract interpretation is the collective name for a family of semantics-based techniques for compile-time analysis of programs. One of the most costly operations in automating such analyses is the computation of fixed points. The frontiers algorithm is an elegant method, invented by Chris Clack and Simon Peyton Jones, which addresses this issue.

In this article we present a new approach to the frontiers algorithm based on the insight that frontiers represent upper and lower subsets of a function's argument domain. This insight leads to a new formulation of the frontiers algorithm for higher-order functions, which is considerably more concise than previous versions.

We go on to argue that for many functions, especially in the higher-order case, finding fixed points is an intractable problem unless the sizes of the abstract domains are reduced. We show how the semantic machinery of abstract interpretation allows us to place upper and lower bounds on the values of fixed points in large lattices by working within smaller ones.