Proposal for Thesis Research in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

Program comprehension remains a major bottleneck for software maintenance. When a programmer must understand a large legacy program whose documentation is scarce, out-ofdate, or irrelevant, she must browse its source code to build an effective mental model of its structure and behavior. But source code is far from an ideal representation for this task. I propose a system, DR. JONES, to help a programmer understand an unfamiliar program. DR. JONES is a magic lens over the program that reveals information that usually requires effort to extract from source code. This information is broken into tightly coupled views of the program’s structure and behavior that, taken together, help the programmer build a good mental model of the program. DR. JONES presents these views as diagrams, to engage the programmer’s perception in extracting knowledge about the program. The design of DR. JONES will be motivated by user studies to find the kinds of knowledge programmers seek to extract from unfamiliar programs, and the strategies they employ to do so. DR. JONES will make use of techniques like multiscale browsing, design pattern recognition, and “software jiggling” to permit the orderly and conceptual exploration of an unfamiliar program. Let us change our traditional attitude to the construction of programs. Instead of imagining that our main task is to instruct a computerwhat to do, let us concentrate rather on explaining to human beings what we want a computer to do. Donald Knuth1 I’m like a bad penny, I always turn up. Dr. Henry “Indiana” Jones 1 Sarah Singleton’s Expedition Consider the situation of Sarah Singleton, a programmer who has just joined the software team for a hockey equipment retailer. Sarah is asked to add wish list functionality to the retailer’s Web site as soon as possible (of course). First, she must understand enough about the site’s software to ensure that her new wish list feature will interact properly with the existing code. So, she browses the code and documentation, and discusses the system with the programmers currently maintaining it. During the course of this investigation she builds a mental model of its workings: a “mental picture” of the structure (classes and APIs) and behavior (control and data flow) of the system. A similar challenge faces any programmer who must understand existing software before she is able to maintain or enhance it. If she is lucky there will be well-written documentation, both within the source code and in external documents. But if documentation is incomplete, out-ofdate, or irrelevant to the task at hand, then work remains to discover the program’s structure and behavior and build a useful mental model. Knowing where to begin and what is relevant is half the challenge, especially in programs with tens of thousands of lines of code (LoC). I propose research to develop a tool, DR. JONES, that assists a programmer in program comprehension. A guiding metaphor is that the programmer is a software archaeologist who must investigate a mysterious device from long ago to gain insight into its structure and behavior. DR. JONES acts as the archaeologist’s magic lens, allowing her to immediately see relationships and patterns in the program that would ordinarily require much more effort to deduce. DR. JONES analyzes a program2 and presents interactive, graphical views of it to the programmer. These views are designed to make explicit the kinds of concepts and patterns that programmers use to reason about and describe their programs. The programmer explores the program through DR. JONES’ magic lens to incrementally build an effective mental model of it.