Identifying the Information Needed to Understand Programs: an Experimental Approach


Authors: Francoise Balmas and Harald Wertz

Abstract :

In order to get a better understanding of what program understanding means, we are conducting experiments to observe programmers reading programs. Our primary concern is to identify which kind of information programmers are likely to search for in the code, to skip over or to miss. Secondly, we are interested to know how people construct their understanding of the program from the information they get and how they represent it, as we are convinced that this would help us to build automatic program understanding tools.
Based on first observations, we identified a preliminary list of useful information to provide programmers with and we constructed prototypical programs that compute and display this information. This process applies iteratively, as each new information view helps us to observe further how program understanding is built, therefore to construct further tools.
In this paper, we describe the experiments we did along with the programs we implemented on the basis of our observations, we outline the ways in which information and views are built, and we evaluate the information they provide as well as our approach to understand program understanding.


Ce papier est disponible sur demande fb@ai.univ-paris8.fr.