Labelled Lambda-calculi with Explicit Copy and Erase

M. Florido and I. Mackie (Eds.): First International Workshop on Linearity (LINEARITY 2009) EPTCS 22, 2010, pp. 49–64, doi:10.4204/EPTCS.22.5 Labelled λ-calculi with Explicit Copy and Erase Maribel Fern´andez, Nikolaos Siafakas King’s College London, Dept. Computer Science, Strand, London WC2R 2LS, UK [maribel.fernandez, nikolaos.siafakas]@kcl.ac.uk We present two rewriting systems that define labelled explicit substitution λ-calculi. Our work is motivated by the close correspondence between L´evy’s labelled λ-calculus and paths in proof-nets, which played an important role in the understanding of the Geometry of Interaction. The structure of the labels in L´evy’s labelled λ-calculus relates to the multiplicative information of paths; the novelty of our work is that we design labelled explicit substitution calculi that also keep track of exponential information present in call-by-value and call-by-name translations of the λ-calculus into linear logic proof-nets. 1 Introduction Labelled λ-calculi have been used for a variety of applications, for instance, as a technology to keep track of residuals of redexes [6], and in the context of optimal reduction, using L´evy’s labels [14]. In L´evy’s work, labels give information about the history of redex creation, which allows the identification and classification of copied β-redexes. A different point of view is proposed in [4], where it is shown that a label is actually encoding a path in the syntax tree of a λ-term. This established a tight correspondence between the labelled λ-calculus and the Geometry of Interaction interpretation of cut elimination in linear logic proof-nets [13]. Inspired by L´evy’s labelled λ-calculus, we define labelled λ-calculi where the labels attached to terms capture reduction traces. However, in contrast with L´evy’s work, our aim is to use the dynamics of substitution to include information in the labels about the use of resources, which corresponds to the exponentials in proof-nets. Exponential structure in proof-nets involves box-structures and connectives that deal with their management, for instance, copying and erasing of boxes. Different translations of the λ-calculus into proof-nets place boxes at different positions; such choices are also reflected in the paths of the nets. In L´evy’s calculus substitution is a meta-operation: substitutions are propagated exhaustively and in an uncontrolled way. We would like to exploit the fact that substitutions copy labelled terms and hence paths, but it is difficult to tell with a definition such as (MN)α[P/x] = (M[P/x]N[P/x])α, whether the labels in P are actually copied or not: P may substitute one or several occurrences of a variable, or it may simply get discarded. In order to track substitutions we use calculi of explicit substitutions, where substitution is defined at the same level as β-reduction. Over the last years a whole range of explicit substitution calculi have been proposed, starting with the work of de Bruijn [9] and the λσ-calculus [1]. Since we need to track copy and erasing of substitutions, we will use a calculus where not only substitutions are explicit, but also copy and erase operations are part of the syntax. Specifically, in this paper we use explicit substitution calculi that implement closed reduction strategies [10, 11]. This may be thought of as a more powerful form of combinatory reduction [8] in the sense that β-redexes may be contracted when the argument part or the function part of the redex is closed. This essentially allows more reductions to take place under abstractions. The different possibilities of placing restrictions on the β-rule give rise to different closed reduction strategies, corresponding to different translations of the λ-calculus into proof-nets (a survey of 50 Labelled λ-calculi with explicit copy and erase available translations can be found in [16]). Closed reduction strategies date back to the late 1980’s, in fact, such a strategy was used in the proof of soundness of the Geometry of Interaction [13]. Labelled λ-calculi are a useful tool to understand the structure of paths in the Geometry of Interaction: L´evy’s labels were used to devise optimisations in GoI abstract machines, defining new strategies of evaluation and techniques for the analysis of λ-calculus programs [5, 3]. The labels in our calculi of explicit substitutions contain, in addition to the multiplicative information contained in L´evy’s labels, also information about the exponential part of paths in proof-nets. In other words, our labels relate a static concept—a path—with a dynamic one: copying and erasing of substitutions. Thus, the labels can be used not only to identify caller-callee pairs, but also copy and erasing operations. This is demonstrated in this paper using two different labelled calculi. In the first system, the β-rule applies only if the function part of the redex is closed. We relate this labelled system with proof-nets using the so-called call-by-value transla

…(Full text truncated)…

This content is AI-processed based on ArXiv data.