On the use of visual clustering to identify landmarks in code navigation
Sandeep Reddivari, Mahesh Kotapalli · 2017
Recovering the legibility features is key to reverse engineering as the legible software systems can ease developer's code navigation and comprehension. Landmarks are important legibility features that developers use as reference points. In this paper, we leverage visual clustering to explore how landmarks can be identified via static dependencies. Besides organizing software entities with coherent patterns, visual clustering offers additional insights by rigorously rendering a holistic picture of the code base to the two-dimensional space. We contribute a couple of heuristics based on the cluster layout to identify the landmark files. Our visual exploration of Eclipse Mylyn open source Java project reveals developer's reliance on the landmarks during code navigation and shows the promise of using static dependencies to uncover the landmarks in the software space.