Andrzej Filinski

Abstracting control

The last few years have seen a renewed interest in continuations for expressing advanced control structures in programming languages, and new models such as Abstract Continuations have been proposed to capture these dimensions. This article investigates an alternative formulation, exploiting the latent expressive power of the standard continuation-passing style (CPS) instead of introducing yet other new concepts. We build on a single foundation: abstracting control as a hierarchy of continuations, each one modeling a specific language feature as acting on nested evaluation contexts. We show how iterating the continuation-passing conversion allows us to specify a wide range of control behavior. For example, two conversions yield an abstraction of Prolog-style backtracking. A number of other constructs can likewise be expressed in this framework; each is defined independently of the others, but all are arranged in a hierarchy making any interactions between them explicit. This approach preserves all the traditional results about CPS, e.g., its evaluation order independence. Accordingly, our semantics is directly implementable in a call-by-value language such as Scheme or ML. Furthermore, because the control operators denote simple, typable lambda-terms in CPS, they themselves can be statically typed. Contrary to intuition, the iterated CPS transformation does not yield huge results: except where explicitly needed, all continuations beyond the first one disappear due to the extensionality principle (&eegr;-reduction). Besides presenting a new motivation for control operators, this paper also describes an improved conversion into applicative-order CPS. The conversion operates in one pass by performing all administrative reductions at translation time; interestingly, it can be expressed very concisely using the new control operators. The paper also presents some examples of nondeterministic programming in direct style.

Representing Control: a Study of the CPS Transformation

This paper investigates the transformation of λ ν -terms into continuation-passing style (CPS). We show that by appropriate η-expansion of Fisher and Plotkins two-pass equational specification of the CPS transform, we can obtain a static and context-free separation of the result terms into “essential” and “administrative” constructs. Interpreting the former as syntax builders and the latter as directly executable code, We obtain a simple and efficient one-pass transformation algorithm, easily extended to conditional expressions, recursive definitions, and similar constructs. This new transformation algorithm leads to a simpler proof of Plotkins simulation and indifference results. We go on to show how CPS-based control operators similar to, more general then, Schemes call/cc can be naturally accommodated by the new transformation algorithm. To demonstrate the expressive power of these operators, we use them to present an equivalent but even more concise formulation of the efficient CPS transformation algorithm. Finally, we relate the fundamental ideas underlying this derivation to similar concepts from other work on program manipulation; we derive a one-pass CPS transformation of λ n -terms; and we outline some promising areas for future research.

Representing monads

We show that any monad whose unit and extension operations are expressible as purely functional terms can be embedded in a call-by-value language with “composable continuations”. As part of the development, we extend Meyer and Wands characterization of the relationship between continuation-passing and direct style to one for continuation-passing vs. general “monadic” style. We further show that the composable-continuations construct can itself be represented using ordinary, non-composable first-class continuations and a single piece of state. Thus, in the presence of two specific computational effects - storage and escapes - any expressible monadic structure (e.g., nondeterminism as represented by the list monad) can be added as a purely definitional extension, without requiring a reinterpretation of the whole language. The paper includes an implementation of the construction (in Standard ML with some New Jersey extensions) and several examples.

Representing layered monads

There has already been considerable research on constructing modular, monad-based specifications of computational effects (state, exceptions, nondeterminism, etc.) in programming languages. We present a simple framework in this tradition, based on a Church-style effect-typing system for an ML-like language. The semantics of this language is formally defined by a series of monadic translations, each one expanding away a layer of effects. Such a layered specification is easy to reason about, but its direct implementation (whether by parameterized interpretation or by actual translation) is often prohibitively inefficient.By exploiting deeper semantic properties of monads, however, it is also possible to derive a vastly more efficient implementation: we show that each layer of effects can be uniformly simulated by continuation-passing, and further that multiple such layers can themselves be simulated by a standard semantics for call/cc and mutable state. Thus, even multi-effect programs can be executed in Scheme or SML/NJ at full native speed, generalizing an earlier single-effect result. As an example, we show how a simple resumption-based semantics of concurrency allows us to directly simulate a shared-state program across all possible dynamic interleavings of execution threads.

Declarative Continuations: an Investigation of Duality in Programming Language Semantics

This paper presents a formalism for including first-class continuations in a programming language as a declarative concept, rather than an imperative one. A symmetric extension of the typed λ-calculus is introduced, where values and continuations play dual roles, permitting mirror-image syntax for dual categorical concepts like products and coproducts. An implementable semantic description and a static type system for this calculus are presented. We also give a categorical description of the language, by presenting a correspondence with a system of combinatory logic, similar to a cartesian closed category, but with a completely symmetrical set of axioms.

Normalization by evaluation for the computational lambda-calculus

We show how a simple semantic characterization of normalization by evaluation for the λβη-calculus can be extended to a similar construction for normalization of terms in the computational λ-calculus. Specifically, we show that a suitable residualizing interpretation of base types, constants, and computational effects allows us to extract a syntactic normal form from a terms denotation. The required interpretation can itself be constructed as the meaning of a suitable functional program in an ML-like language, leading directly to a practical normalization algorithm. The results extend easily to product and sum types, and can be seen as a formal basis for call-by-value type-directed partial evaluation.

Normalization and Partial Evaluation

We give an introduction to normalization by evaluation and type-directed partial evaluation. We first present normalization by evaluation for a combinatory version of Godel System T. Then we show normalization by evaluation for typed lambda calculus with s and ? conversion. Finally, we introduce the notion of binding time, and explain the method of type-directed partial evaluation for a small PCF-style functional programming language. We give algorithms for both call-byname and call-by-value versions of this language.

Linear continuations

We present a functional interpretation of classical linear logic based on the concept of linear continuations. Unlike their non-linear counterparts, such continuations lead to a model of control that does not inherently impose any particular evaluation strategy. Instead, such additional structure is expressed by admitting closely controlled copying and discarding of continuations. We also emphasize the importance of classicality in obtaining computationally appealing categorical models of linear logic and propose a simple “coreflective subcategory” interpretation of the modality “!”.

Monads in action

In functional programming, monadic characterizations of computational effects are normally understood denotationally: they describe how an effectful program can be systematically expanded or translated into a larger, pure program, which can then be evaluated according to an effect-free semantics. Any effect-specific operations expressible in the monad are also given purely functional definitions, but these definitions are only directly executable in the context of an already translated program. This approach thus takes an inherently Church-style view of effects: the nominal meaning of every effectful term in the program depends crucially on its type. We present here a complementary, operational view of monadic effects, in which an effect definition directly induces an imperative behavior of the new operations expressible in the monad. This behavior is formalized as additional operational rules for only the new constructs; it does not require any structural changes to the evaluation judgment. Specifically, we give a small-step operational semantics of a prototypical functional language supporting programmer-definable, layered effects, and show how this semantics naturally supports reasoning by familiar syntactic techniques, such as showing soundness of a Curry-style effect-type system by the progress+preservation method.

A Denotational Account of Untyped Normalization by Evaluation

We show that the standard normalization-by-evaluation construction for the simply-typed λ βη -calculus has a natural counterpart for the untyped λ β -calculus, with the central type-indexed logical relation replaced by a recursively defined invariant relation, in the style of Pitts. In fact, the construction can be seen as generalizing a computational-adequacy argument for an untyped, call-by-name language to normalization instead of evaluation. In the untyped setting, not all terms have normal forms, so the normalization function is necessarily partial. We establish its correctness in the senses of soundness (the output term, if any, is β-equivalent to the input term); standardization (β-equivalent terms are mapped to the same result); and completeness (the function is defined for all terms that do have normal forms). We also show how the semantic construction enables a simple yet formal correctness proof for the normalization algorithm, expressed as a functional program in an ML-like call-by-value language.