UCPy: Reverse-Engineering Python

One of the recurring topics in the Python community is how to make Python programs run faster. Typically, a set of solutions is proposed which include: adding static type inference; somehow compiling programs into native code; translating Python programs into Parrot/Lisp/.net code; applying research results from dynamically-typed language implementation. Progress has been made on some of these, such as Psyco [8], but many of these proposed solutions are qualified by the caveat no one has the time/resources to work on it. In the Programming Languages Lab at the University of Calgary we have a research project underway, UCPy, whose short-term goal is to examine ways we can make Python run faster. We have learned some lessons through our design and implementation work to date, about both Python and the undertaking of such a project, which we present in this paper.