Binding-Time Analysis for MetaML via Type Inference and Constraint Solving

Abstract

The two predominant program specialization techniques, partial evaluation and staged programming, take opposite approaches to automating binding-time analysis (BTA). Despite their common goal, there are no systems integrating both methods. Programmers must choose between the precision of manually placing staging annotations and the convenience of automating such annotation.

We present an automatic BTA algorithm for a subset of MetaML. Such an algorithm provides a basis for a system integrating staged programming and partial evaluation because it allows programmers to switch between automatic and manual staging. Our algorithm is based on typing algorithm coupled with arithmetic-constraint solving. The algorithm decorates each subexpression of both a program and its type with numeric variables representing staging-annotations and then generates simple arithmetic constraints that describe the space of all possible stagings of the original program. Benefits of our approach include expressive BTA specifications in the form of stage-annotated types as well as support for polyvariance.

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.