Atze Dijkstra

A generic deriving mechanism for Haskell

Haskells deriving mechanism supports the automatic generation of instances for a number of functions. The Haskell 98 Report only specifies how to generate instances for the Eq, Ord, Enum, Bounded, Show, and Read classes. The description of how to generate instances is largely informal. The generation of instances imposes restrictions on the shape of datatypes, depending on the particular class to derive. As a consequence, the portability of instances across different compilers is not guaranteed. We propose a new approach to Haskells deriving mechanism, which allows users to specify how to derive arbitrary class instances using standard datatype-generic programming techniques. Generic functions, including the methods from six standard Haskell 98 derivable classes, can be specified entirely within Haskell 98 plus multi-parameter type classes, making them lightweight and portable. We can also express Functor, Typeable, and many other derivable classes with our technique. We implemented our deriving mechanism together with many new derivable classes in the Utrecht Haskell Compiler.

The architecture of the Utrecht Haskell compiler

In this paper we describe the architecture of the Utrecht Haskell Compiler (UHC). UHC is a new Haskell compiler, that supports most (but not all) Haskell 98 features, plus some experimental extensions. It targets multiple backends, including a bytecode interpreter backend and a whole-program analysis backend, both via C. The implementation is rigorously organized as stepwise transformations through some explicit intermediate languages. The tree walks of all transformations are expressed as an algebra, with the aid of an Attribute Grammar based preprocessor. The compiler is just one materialization of a framework that supports experimentation with language variants, thanks to an aspect-oriented internal organization.

Ruler: programming type rules

Some type systems are first described formally, to be sometimes followed by an implementation. Other type systems are first implemented as language extensions, to be sometimes retrofitted with a formal description. In neither case it is an easy task to keep both artefacts consistent. In this paper we introduce Ruler, a domain specific language for describing type rules. Type rules can be incrementally described, thus providing a means for building complex type systems on top of simpler ones. Besides checking well-formedness of Ruler programs we use them to generate (1) a visual LATEX rendering, suitable for use in the presentation of formal aspects, and (2) an attribute grammar based implementation. Combining these two aspects in Ruler contributes to bridging the gap between theory and practice: mutually consistent representations can be generated for use in both theoretical and practical settings.

Iterative type inference with attribute grammars

Type inference is the process of constructing a typing derivation while gradually discovering type information. During this process, inference algorithms typically make subtle decisions based on the derivation constructed so far. Because a typing derivation is a decorated tree we aim to use attribute grammars as the main implementation tool. Unfortunately, we can neither express iteration, nor express decisions based on intermediate derivations in such grammars. We present the language ruler-front, a conservative extension to ordered attribute grammars, that deals with the aforementioned problems. We show why this extension is suitable for the description of constraint-based inference algorithms.

The Structure of the Essential Haskell Compiler, or Coping with Compiler Complexity

In this paper we describe the structure of the Essential Haskell Compiler (EHC) and how we manage its complexity, despite its growth from essentials to a full Haskell compiler. Our approach splits both language and implementation into smaller, manageable steps, and uses specific tools to generate parts of the compiler from higher level descriptions.

Grammar fragments fly first-class

We present a Embedded Domain Specific language in the form of a Haskell library (EDSL) for expressing grammar (fragments), using typed abstract syntax with observable references. Thus grammar fragments have become fully typed, first-class Haskell values. We can analyze and transform such representations and generate parsers from them. What makes our approach special is that we can combine embedded grammar fragments on the fly, i.e. after they have been compiled and that the Haskell type system is used to check the consistency of a collection of combined fragments. We show how we can extend an initial, limited grammar embedded in a compiler with new syntactic constructs, either by introducing new non-terminals or by adding new productions for existing non-terminals. We do not impose any restrictions on the individual grammar fragments, nor on the structure as a whole.

Building JavaScript Applications with Haskell

We introduce the Utrecht Haskell Compiler JavaScript backend, which allows one to compile Haskell code to JavaScript, so it can be run in the browser. To interface with JavaScript and overcome part of the impedance mismatch between the two languages, we introduce the Foreign Expression Language; a small subset of JavaScript for use in Foreign Function Interface imports. Finally we discuss the implementation of a JavaScript application, completely written in Haskell, with which we show that it is now possible to write an entire JavaScript application completely in Haskell.

Lazy stateless incremental evaluation machinery for attribute grammars

Many computer programs work with data that changes over time. Computations done over such data usually are repeated completely after a change in the data. For complex computations such repetitive recomputation can become too inefficient. When these recomputations take place on data which has only changed slightly, it often is possible to reformulate the computation to an incremental version which reuses the result of the computation on previous data. Such a situation typically occurs in compilers and editors for structured data (like a program) where program analyses and transformations (for example error checking) are done while editing. Although rewriting to incremental versions thus offers a solution to this problem, a manual rewrite of an already complex computation to its incremental counterpart is tedious, error prone, and inhibits further development of the original computation. We therefore intend to generate such incremental counterparts (semi)automatically by focusing on computations expressed using Attribute Grammars (AGs). In this paper we do groundwork for this goal and develop machinery for incremental attribute grammar evaluation based on change propagation and pure functions. We use pretty printing with free variable annotation to explain our techniques. Furthermore, our techniques also expose rules of conduct for a programmer desiring incrementality: the automatic translation of code to an incremental version does not always directly result in efficiency improvements because code often is written in a style unsuitable for automatic incrementalization. We show some common cases in which (small) code changes facilitating incrementality are required. We evaluate the effectiveness of the overall approach using a simple benchmark for the example, and a more extensive benchmark based on constraint-based type inference implemented with AGs.

The kennedy-warren algorithm revisited: ordering attribute grammars

Attribute Grammars (AGs) are a powerful tool for defining an executable semantics of a programming language, and thus for implementing a compiler. An execution plan for an AG determines a static evaluation order for the attributes which are defined as part of an AG specification. In building the Utrecht Haskell Compiler (UHC), a large scale AG project, we discovered that the Ordered AG approach (Kastens, 1980) for building such plans becomes impractical: the additional dependencies between attributes introduced by this algorithm too often result in grammars for which no execution plan can be generated. To avoid such problems we have implemented a refined version of the algorithm of Kennedy and Warren (1976) as part of our purely functional AG system and show how this algorithm solves the problems that surface with the Ordered AG approach. Furthermore, we present the results of applying this algorithm to the UHC code and show that this approach in some cases also has a positive effect on the runtime of the resulting program.

Visitor-based Attribute Grammars with Side Effect

The visitor design pattern is often applied to program traversal algorithms over Abstract Syntax Trees (ASTs). It defines a visitor, an object with a visit method that is executed for each node in the AST. These visitors have the advantage that the order of traversal is explicitly under control of the programmer, which is essential to deal with side-effectful computations. Unfortunately, the exchange of results between traversals is error-prone. Attribute Grammars (AGs) are an alternative way to write multi-traversal algorithms. An attribute evaluator decorates the AST with attributes in one or more traversals. The attributes form a convenient mechanism to exchange results between traversals. Unfortunately, AGs discourage the use of side effect. In this paper, we present ruler-front, a language capturing the combination of the above approaches. A ruler-front grammar can be translated to traversal algorithms in multiple languages. In this paper, we translate to the imperative, dynamically-typed language JavaScript.