The architecture of the Utrecht Haskell compiler

Abstract

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.

First Class Syntax, Semantics, and Their Composition

“…We have extensive experience with attribute grammars in the construction of the Utrecht Haskell compiler [19]. The code of this compiler is completely factored out along the two axes mentioned in the introduction [20,24,18], using the notation used in Figure 4.3.…”

“…Another complication is that the interfaces of non-terminals are in general not uniform, thus necessitating all kind of tricks to change the monad at the right places, keeping information to be reused later, etc. In our generated Haskell compiler [19] we have non-terminals with more than 10 different attributes, and glueing all these together or selectively leaving some out turns out to be impossible to do by hand.…”

“…In this way we get great flexibility and there is virtually no limit to the things we may add. The Utrecht Haskell Compiler [19] has shown the effectiveness of this approach by composing a large number of attribute grammar fragments textually into a complete compiler description. This approach however is not very practical when defining relatively small language extensions; we do not want an individual user to have to generate a completely new compiler for each small extension.…”

“…Solutions to the quest for composable language description can be found at the textual level, as done by most attribute grammar systems [45,21,19], or at the semantic level, where language descriptions become first class values, which can be composed to build a complete language description.…”

“…In this way checking the well-formedness of the attribute grammar is realized through the Haskell class system. Once the language gets complicated (in our Haskell compiler UHC [19] some non-terminals have over 20 attributes), the cost of accessing attributes may become noticeable. Note that, in contrast to the weaving based approaches, this approach supports separate compilation of individual aspects: each generated fragment is individually checked to be well-typed and once compiled its source is not modifiable by other extensions.…”