Towards Activity-Aware Tool Support for Change Tasks

Abstract

Abstract-To complete a change task, software developers perform a number of activities, such as locating and editing the relevant code. While there is a variety of approaches to support developers for change tasks, these approaches mainly focus on a single activity each. Given the wide variety of activities during a change task, a developer has to keep track of and switch between the different approaches. By knowing more about a developer's activities and in particular by knowing when she is working on which activity, we would be able to provide better and more tailored tool support, thereby reducing developer effort.In our research we investigate the characteristics of these activities, whether they can be identified, and whether we can use this additional information to improve developer support for change tasks. We conducted two exploratory studies with a total of 21 software developers collecting data on activities in the lab and field. An empirical analysis of the data shows, amongst other results, that activities comprise a consistently small amount of code elements across all developers and tasks (approx. 8.7 elements). Further analysis of the data shows, that we can automatically detect the boundaries and types of activities, and that the information on activity types can be used to improve the identification of relevant code elements.

Recognizing Developer Activity Based on Joint Modeling of Code and Command Interactions

“…The approach was also targeted towards prediction and recommendation generation of IDE commands. Kevic et al recently proposed an approach for detecting activities similar to the ones we aim to recognize in this paper, but used only source code activities and a supervised learning approach that requires training data [10]. They observed improved performance of models trained per developer to those that were trained across different developers.…”

“…The state of the art in automated developer activity detection has focused on a single dimension of the interaction data generated by a typical software developer. For example, detecting activities from source code accesses over time and recommending program elements [10] or detecting latent activities from a stream of Integrated Development Environment (IDE) commands and events and recommending the same [12], [13]. While using a single dimension of interaction data is effective, there are opportunities in leveraging multiple dimensions together.…”

“…Researchers have performed several studies to deter-mine a de facto set of activities used in software maintenance [10], [11], [14], [15]. However, the mapping between pre-determined activities by researchers and those exhibited by a set of developers in the field is likely to be imperfect, as developer work styles have been observed to have strong personal variations [10], [11]. A key strength of the approach described in this paper is that it does not require a predetermined, fixed set of activities, instead, allowing for the number of activities to be inferred based on the developers' interactions.…”

“…Another advantage of the proposed approach is that it is unsupervised and therefore does not require labeled data, which is hard to obtain at scale and from a sufficiently diverse set of developers. A supervised model would have a tendency to overfit the behavior of the individuals or groups present in the labeled training data, exemplified by the fact that such models consistently produce best results when trained and tested solely on one individual's interaction data [10]. To summarize, the contributions of this paper are:…”

Detecting Developers’ Task Switches and Types

“…Zou and Godfrey [30] replicated Coman and Sillitti's study in an industrial setting with six professional developers and found that the algorithm detects many more task switches than the ones self-reported by the participants with an error of more than 70%. Finally, on a more fine-grained level, Kevic and Fritz [31] examined the detection of activity switches and types within a change task using semantic, temporal and structural features. In two studies with 21 participants, they found that activity switches as well as the six selfidentified activity types can be predicted with more than 75% accuracy.…”