Representing unit test data for large scale software development

  • Authors:
  • Joseph A. Cottam;Joshua Hursey;Andrew Lumsdaine

  • Affiliations:
  • Indiana University, Bloomington, IN;Indiana University, Bloomington, IN;Indiana University, Bloomington, IN

  • Venue:
  • Proceedings of the 4th ACM symposium on Software visualization
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

Large scale software projects rely on routine, automated testing to gauge progress towards its goals. The diversity and quantity of these tests grow as time and project scope increase. This is as a consequence of both experience and expanding audience. It becomes increasingly difficult to interpret testing results as the testing suites multiply and diversify. If interpretation becomes too difficult, testings results could become ignored all together. Visualization has proven to be an effective tool to aid the interpretation of large amounts of data. We have adapted visualization techniques based on small multiples to communicate the health of the software project across several levels of abstraction. The collective set of techniques we refer to as the SeeTest visualization schema. We applied this visualization technique to the Open MPI test results in order to assist developers in the software release cycle. Through the visualizations, developers found a variety of surprising mismatches between their data and their intuitions. This exploration did not involve collecting any data not already being collected, merely presenting it in manner that better supported their needs. In this paper, we detail the development of the representation we used and give more particular analysis of the insights gained by the Open MPI community. The techniques presented in this paper can be applied to other software projects.