Understanding how software developers do their work is an important first step to improving their productivity. Previous research has generally focused either on laboratory experiments or coarsely-grained industrial case studies; however, studies that seek a finegrained understanding of industrial programmers working within a realistic context remain limited. In this work, we propose to use interaction histories — that is, finely detailed records of developers’ interactions with their IDE — as our main source of information for understanding programmer’s work habits. We develop techniques to capture, mine, and analyze interaction histories, and we present two industrial case studies to show how this approach can help to better understand industrial programmers’ work at a detailed level: we explore how the basic characteristics of software maintenance task structures can be better understood, how latent dependence between program artifacts can be detected at interaction time, and show how patterns of interaction coupling can be identified. We also examine the link between programmer interactions and some of the contextual factors of software development, such as the nature of the task being performed, the design of the software system, and the expertise of the developers. In particular, we explore how task boundaries can be automatically detected from interaction histories, how system design and developer expertise may affect interaction coupling, and whether newcomer and expert developers differ in their interaction history patterns. These findings can help us to better reason about the multidimensional nature of software development, to detect potential problems concerning task, design, expertise, and other contextual factors, and to build smarter tools that exploit the inherent patterns within programmer interactions and provide improved support for task-aware and expertise-aware software development.

Understanding how software developers do their work is an important first step to improving their productivity. Previous research has generally focused either on laboratory experiments or coarsely-grained industrial case studies; however, studies that seek a finegrained understanding of industrial programmers working within a realistic context remain limited. In this work, we propose to use interaction histories — that is, finely detailed records of developers’ interactions with their IDE — as our main source of information for understanding programmer’s work habits. We develop techniques to capture, mine, and analyze interaction histories, and we present two industrial case studies to show how this approach can help to better understand industrial programmers’ work at a detailed level: we explore how the basic characteristics of software maintenance task structures can be better understood, how latent dependence between program artifacts can be detected at interaction time, and show how patterns of interaction coupling can be identified. We also examine the link between programmer interactions and some of the contextual factors of software development, such as the nature of the task being performed, the design of the software system, and the expertise of the developers. In particular, we explore how task boundaries can be automatically detected from interaction histories, how system design and developer expertise may affect interaction coupling, and whether newcomer and expert developers differ in their interaction history patterns. These findings can help us to better reason about the multidimensional nature of software development, to detect potential problems concerning task, design, expertise, and other contextual factors, and to build smarter tools that exploit the inherent patterns within programmer interactions and provide improved support for task-aware and expertise-aware software development.

[pt] Nos dias de hoje, temos acesso a dados de tamanho, dimensionalidade e complexidade sem precedentes. Para extrair informações desconhecidas e inesperadas desses dados complexos e dinâmicos, necessitamos de estratégias efetivas e eficientes. Uma dessas estratégias é usar aplicações de análise visual (visual analytics), que combinam técnicas de análise de dados e de visualização. Depois do processo de descoberta de conhecimento, um grande desafio é filtrar a informação essencial que levou à descoberta e comunicar os achados a outras pessoas. Nós propomos tirar proveito do traço deixado pela análise exploratória de dados, sob a forma do histórico da interação do usuário, para ajudar nesse processo. Com o traço, o usuário pode escolher os passos de interação desejados e criar uma narrativa, compartilhando o conhecimento adquirido com os leitores. Para atingir nosso objetivo, desenvolvemos o arcabouço BONNIE (Building Online Narratives from Noteworthy Interaction Events - Construindo Narrativas Online a partir de Eventos de Interação Relevantes). O arcabouço compreende um modelo de log para registrar os eventos de interação, código auxiliar para ajudar o(a) desenvolvedor(a) a instrumentar o seu próprio código, e um ambiente para visualizar o histórico de interação e construir narrativas. Esta tese apresenta nossa proposta para comunicar descobertas em aplicações de análise visual, o arcabouço BONNIE, e alguns estudos empíricos que realizamos para avaliar nossa solução. / [en] Nowadays, we have access to data of unprecedentedly large size, high dimensionality, and complexity. To extract unknown and unexpected information from such complex and dynamic data, we need effective and efficient strategies. One such strategy is to combine data analysis and visualization techniques, which is the essence of visual analytics applications. After the knowledge discovery process, a major challenge is to filter the essential information that led to a discovery and to communicate the findings to other people. We propose to take advantage of the trace left by the exploratory data analysis, in the form of ser interaction history, to aid in this process. With the trace, the user can choose the desired interaction steps and create a narrative, sharing the acquired knowledge with readers. To achieve our goal, we have developed the BONNIE (Building Online Narratives from Noteworthy Interaction Events) framework. The framework comprises a log model to register the interaction events, auxiliary code to help the developer instrument his or her own code, and an environment to view the user s own interaction history and build narratives. This thesis presents our proposal for communicating discoveries in visual analytics applications, the BONNIE framework, and a few empirical studies we conducted to evaluate our solution.