The GF Grammar Compiler

GF (Grammatical Framework) is a grammar formalism based on the distinction between abstract and concrete syntax. An abstract syntax is a free algebra of trees, and a concrete syntax is a mapping from trees to nested records of strings and features. These mappings are naturally defined as functions in a functional programming language; the GF language provides the customary functional programming constructs such as algebraic data types, pattern matching, and higher-order functions, which enable productive grammar writing and linguistic generalizations. Given the seemingly transformational power of the GF language, its computational properties are not obvious. However, all grammars written in GF can be compiled into a simple and austere core language, Canonical GF (CGF). CGF is well suited for implementing parsing and generation with grammars, as well as for proving properties of GF. This paper gives a concise description of both the core and the source language, the algorithm used in compiling GF to CGF, and some back-end optimizations on CGF.