Internship report : ExtendingML with Higher-Order Types

This thesis presents the context and nature of the work during a sixmonth internship at Gallium, a research team of the Institut National de Recherche en Informatique et en Automatique (INRIA) which is specialized in the design, formalization and implementation of programming languages and systems. My advisor was Didier Rémy, a senior member of the team, which specializes, among other things, in type systems and type inference. The internship subject was MLF, a type system developped by Didier and two of his former PhD students, Didier Le Botlan and Boris Yakobowksi. MLF is a rich research topic which, despite interest and research by other people and teams outside Gallium, has not yet been integrated in the ”mainstream” common knowledge of our discipline. A significant part of the intership time was dedicated to the study of the subject bibliography, and of the type system itself: despite very attractive properties, it is a complex system that has not yet been presented in a way making possible to thouroughly understand it in a calm afternoon of scholarly readings. The first chapter of this document will be dedicated to an informal presentation of the MLF system. I’m not trying to be formal and exhaustive (nor could I, given size limits), only to convey the necessary intuition to understand the specific problematics of my internship subject. The reader interested in a more complete description of the different part of MLF will be redirected to the growing existing litterature. The other parts of my internship dedicated to my work — Gallium team is rich in distractions; I won’t describe here the quite enjoyable pauses café(s), the very interesting research seminar, etc. —was dedicated to research and implementation. The research part was dedicated to an extension of MLF with explicit polymorphism and higher-order types, which we could name MLFω. This had already been the subject of an internship by Paolo Herm’s last year [Her09], and we hoped to extend and improve the work done on several points that I will explain in time. The second and third chapters will be dedicated respectively to the extension of MLF with explicit polymorphism (à la System F), and higherorder types (Fω). The implementation part, supposed to correspond closely to the subject of my internship, turned out to be of quite larger scope. Some of the latter parts of Boris Yakobowski’s PhD thesis had not been translated into code, and I was responsible for their integration in the current MLF typer prototype. I also tackled some other aspects of various natures — term and types printing and display, built-in types and values, interactive toplevel — that were present in the former prototypes of both Didier Le Botlan and Boris Yakobowski, but were not up-to-date anymore with the latest presentation of the MLF metatheory. I will not insist here on the software development and the interesting but delicate software engineering considerations that I had to handle duringmy internship.