An empirical study of constraint usage in graphical applications

The spreadsheet model of computation has simplified the task of constructing business soft-ware. This success has led researchers to apply this model to other areas of programming, especially graphical interfaces. The spreadsheet model of computation is called constraint programming in the context of general purpose programming languages. Although the technique of using constraints to specify graphical relationships among objects has enjoyed wide acceptance in the research community, commercial developers have been reluctant to incorporate this technique into commercial toolkits. This reluctance is based on the industry’s concern about the complexity of constraints resulting in debugging problems and an overall concern about the efficiency of constraints. To explore the merits of these concerns, we undertook a study that attempted to quantify data about the use of constraints, the nature of constraint networks and draw conclusions about constraint debugging and optimization from this data. The Amulet graphical library was instrumented to gather data on the structure of constraint networks as well as overall data access patterns. Seven applications, gathered from a variety of sources were profiled using this modified library. The data collected in this study led us to several conclusions. First, it was found that constraint networks tended to be narrow and short. As a result, concerns about debugging constraint based applications should be lessened. Second, because of the modular nature of constraint networks, constraints should scale well in larger applications. Lastly, the study provides a repository of data that can be used in future studies of debugging and optimizing constraints.