Portraits of Programmer Behavior in a Frame-Based Language

The frame-based programming language Stride has the potential to simultaneously simplify and accelerate the task of coding for novices. This is facilitated through a combination of reduced cognitive load, assistance when editing and the elimination of certain syntax errors. Stride also offers the opportunity for comparison to Java, another programming language used by novice programmers, as Stride is integrated into the BlueJ (Java) development environment and user data is also captured in the Blackbox dataset. This paper sets out to determine whether there is evidence to support some of these claims. Since compiler error messages are a key mechanism for user feedback, we compare lesser-studied Stride error message data with better understood Java data. Secondly, we identify groups of Stride and Java users in order to characterise their behavior and to discover differences between frame-based and more conventional text-based programming. These groups include cross-sections of random users as well as two sets of Stride and Java programmers that appear to be engaged in similar tasks. We find that the typical Stride user is primarily a Java user and behavior patterns are similar in both languages. However, we also found a small number of Stride users whose programming time was dominated by Stride, and these users exhibit markedly different patterns for generating user-driven events. These results have implications for educators and tool designers, as well as researchers studying Stride, Java, and Blackbox.