On source code completion assistants and the need of a context-aware approach
Nenhuma Miniatura disponível
Citações na Scopus
3
Tipo de produção
Artigo de evento
Data
2017-07-09
Autores
ARREBOLA, F. V.
Plinio Thomaz Aquino Junior
Plinio Thomaz Aquino Junior
Orientador
Periódico
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Título da Revista
ISSN da Revista
Título de Volume
Citação
ARREBOLA, F. V.; AQUINO JUNIOR, A. F. V. On source code completion assistants and the need of a context-aware approach. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), v.10274 LNCS, p. 191-201, July, 2017.
Texto completo (DOI)
Palavras-chave
Resumo
© Springer International Publishing AG 2017.Source code completion assistance is a popular feature in modern IDEs. However, despite their popularity, there is little research about their key characteristics and limitations. There is also little research about the way software developers interact with code completion assistants, especially when considering the different techniques assistants use to populate the list of possible completions. This paper presents a study about the features of currently available code assistants and an experiment targeting professional Java developers familiar with the Eclipse platform that aims to collect and interpret usage data of two popular code completion assistants during the execution of three programming tasks. Results indicate that half the interactions with code assistants are either dismissed, interrupted or the completion proposals displayed have no direct contribution to the completion of the programming task. In that sense, we argue that code assistants still have a long road to pursue, since they seem to diminish the importance of the ultimate goals of the task at hand and also lack the ability of identifying and exploring the concepts of context-aware computing theory. The results of this paper can drive future HCI research to the design of adaptive code completion assistants that are able to respond to end user behaviors and preferences.