Date of Award
Master of Science
The process of software development consists of many activities, such as writing code, debugging, and navigating through code. Navigating through the code to understand or seek information for developing new code is a very time consuming and tedious task. Many tools are developed based on predictive models to help programmers in navigation. These models predict the fragments of code which might be of developer’s interest. There have been studies for comparing these models to determine their predictive accuracy. However, the models are often based on crude approximations of where a developer’s attention is. For example, prior work has both where the developer’s cursor location as well as what is on the center of the screen to approximate where he/she is looking. To address this concern, we conducted an empirical evaluation of these approximations to see how well they agree with a human evaluator’s perception of where the developer’s attention is. We conducted a replication study on 10 participants and manually coded their navigation pattern. The goals of the study was to evaluate the generalizability of prior work as well as to evaluate the prior operationalizations of navigation. The key findings of this study are: (a) The operationalization based on where the programmer clicks agreed most closely with human evaluator’s assessment and, (b) prior navigation results did not generalize well likely due to small sample size and particulars of the task content.
dissertation or thesis originally submitted to the local University of Memphis Electronic Theses & dissertation (ETD) Repository.
Singh, Alka Rani, "An Empirical Evaluation Of Predictive Models Of Programmer Navigation" (2015). Electronic Theses and Dissertations. 1288.