Human-Centric Tools for Navigating Code

Austin Zachary Henley

Data is provided by the student.

Abstract

All software failures are fundamentally the fault of humansthe software's design was flawed. The high cost of such failures ultimately results in developers having to design, implement, and test fixes, which all take considerable time and effort, and may result in more failures. As developers work on software maintenance tasks, they must navigate enormous codebases that may comprise millions of lines of code organized across thousands of modules. However, navigating code carries with it a plethora of problems for developers. In the hopes of addressing these navigation barriers, modern code editor and development environments provide a variety of features to aid in navigation; however, they are not without their limitations. Code navigations take many forms, and in this work I focus on three key types of code navigation in modern software development: navigating the working set, navigating among versions of code, and navigating the code structure. To address the challenges of navigating code, I designed three novel software development tools, one to enhance each type of navigation. First, I designed and implemented Patchworks, a code editor interface to support developers in navigating the working set. Patchworks aims to make these more efficient by providing a fixed grid of open code fragments that developers can quickly navigate. Second, I designed and implemented Yestercode, a code editor extension to support navigating among versions of code. Yestercode does so by providing a comparison view of the current code and a previous version of the same code. Third, I designed and implemented Wandercode, a code editor extension to enable developers to efficiently navigate the structure of their code. Wandercode aims to do so by providing a visualization of the code's call graph overlayed on the code editor. My approach to designing these tools for more efficient code navigation was a human-centric onethat is, based on the needs of actual developers performing real software development tasks. Through user study evaluations, I found that these tools significantly improved developer productivity by reducing developers' time spent navigating and mental effort during software maintenance tasks.