Visual Viper: a portable visualization library for streamlined scientific communications.

• Autor(a) principal: Mariana Beatriz Nunes Canelas Pais
• Data de Publicação: 2023
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositórios Científicos de Acesso Aberto de Portugal (RCAAP)
• Texto Completo: https://hdl.handle.net/10216/156029
• topic: Ciências da computação e da informação; Computer and information sciences

As the healthcare sector undergoes digital transformation, the influx of data for health professionals and researchers has surged. The increased need for data visualizations to comprehend this information led to the development of Visual Viper, a Python library aimed at automating data visualization, to streamline the often labor-intensive process of generating visualizations. Visual Viper uses Vega-Lite, a high-level grammar of interactive graphics, to create visualizations from various research data sources via a convenient application programming interface (API). This automation saves time and facilitates the consistency of science communication. The library's functionality comprises interconnected components. It begins by retrieving data from a selected source, followed by transforming this data to suit visualization requirements. Subsequently, Visual Viper renders the charts using Vega-Lite and deploys them for use in scientific communication. Implemented within a modular and extensible plugin architecture, it accommodates different data sources and visualization types. Each stage allows independent modification, enabling extensive customization based on specific use-cases without affecting the library's overall functionality. Some important paradigms used in Visual Viper's development include the application of object-oriented programming (OOP) and test-driven development (TDD). By using OOP, the library adopts a structured codebase that is easier to manage and maintain. The principles of encapsulation, inheritance, and polymorphism ensure efficiency and flexibility, while the use of classes facilitates code reuse. The 'DatasetBuilder' class is designed to fetch and preprocess data from various sources, 'ChartNotationBuilder' class is responsible for creating the chart layout and visual aesthetics based on the preprocessed data, and the 'ChartDeployer' class handles the deployment of the finished visualizations. These classes encapsulate related functions and data, reducing complexity and aiding code maintenance and extension. The TDD approach, which involves writing tests before the actual code, ensures all functions are operating as intended, thus leading to improved code quality, simplified debugging, and a faster development cycle. Its implementation ensures the library can run in various environments without significant changes (Environment Agnostic), and it can operate independently on local machines, Lambda, or as a Web API (Serverless Deployment). Future steps for Visual Viper include development of plugins including Google Sheets Dataset Builder and Figma Chart Deployer and creation of additional Vega Lite Chart Notation Builders such as Bar Chart, Survival Chart, and Forest Plot. Once these steps are complete, Visual Viper will be packaged as an importable Python package, with an efficiency evaluation to follow. In conclusion, Visual Viper provides a robust and flexible tool for data visualization, bolstering the efficiency of scientific communication in the healthcare sector.