Debugging Type Errors with a Blackbox Compiler

Type error debugging can be a laborious yet necessary process for programmers of statically typed functional programming languages. Often a compiler compounds this by inaccurately reporting the location of a type error, a problem that has been a subject of research for over thirty years. However, despite its long history, the solutions proposed are often reliant on direct modifications to the compiler, often distributed in the form of patches. These patches append another level of arduous activity to the task of debugging, keeping them modernised to the ever-changing programming language they support.

This thesis investigates an additional option; the blackbox compiler. Split into three central parts, it shows the individual solutions involved in using a blackbox compiler to debug type errors in functional programming languages. First is a demonstration of how the combination of a blackbox compiler and a generic debugging algorithm can successfully locate type errors. Next tackled is a side-effect of this new combination, the introduction of extra errors, combated with a new speed boosted algorithm, evaluated with a proposed framework based on Data Science techniques to quantify the quality of a type error debugger. Lastly, the algorithms employed throughout this thesis, along with the blackbox compiler, have agnostic properties, they do not need language-specific knowledge. Thus, the final part presents utilising the agnostic abilities for an agnostic debugger to locate type errors.