Treffer: Estudo sobre técnicas de visualização quanto ao uso de rótulos em repositórios de software

Context: Visualizations techniques are used to analyzes large amounts of data, because they enhance human cognitive ability in the process of data exploration through the use of graphical models and visual representations. Repository mining is another widely explored area, which can transform data collected from software repositories into useful information for decisiont. By correlating these two areas it is possible to look for unidenti ed patterns in software projects. Objective: This study pourpose is use visualization techniques to analyze the use of labels in issues present in projects hosted on social development platforms. Method: The method employed was organized in ve steps: domain knowledge; data collection and preprocessing; extraction and visualization of patterns, responsible for generating the visualization with the preprocessed data, and visual analysis of it; postprocessing, which may restart the cycle already employed, searching for new patterns by using others techniques and / or parameter setting; and, nally, use of knowledge. Results:Analyzing the domain of open source projects, social software development platforms and label-centric collaboration mechanisms, labels were chosen and their use in issues in the context of the GitHub platform for the NextCloud repository. As for collection and preprocessing, we used the GitHub platform REST API and scripts, developed in Python and JavaScript. In order to characterize and analyze the use of labels, we used visualizations based on box plot, streamgraph, graph drawing and Sankey diagram techniques. Using the knowledge obtained in the previous steps, it is concluded that the analyzed project uses the labels feature and this tends to increase the number of comments on issues, improving communication between developers. As for the issues lifetime, these were shorter for issues without labels, which may indicate that they are quite simple and therefore completed quickly. Looking at the label co-occurrence graph, it is evident that in addition to several ...