Visualising Pyttern Pattern Matches

Details

Supervisors

Mens, Kim

Faculty

Ecole polytechnique de Louvain

Degree label

Master [60] en sciences informatiques

Abstract

enIn programming courses, autograders are widely used to assess student code, especially when sizes are large. Providing targeted, high-quality feedback remains challenging in this setting. Detecting coding patterns and anti-patterns in submissions can support teachers in giving personalised feedback. Pyttern is a language that simplifies the definition and matching of such patterns in code written in Python or Java. The existing Pyttern Visualizer is a web-based tool for debugging the Pyttern language by inspecting pattern matches; it is currently limited to one file and one pattern at a time. However, real-world teaching scenarios require multiple files and composition of multiple pattern files. Teachers who want to evaluate structural quality aspects of programs from many students at the same time lack structured support to run pattern detection across many submissions simultaneously and to inspect, organize, and filter the resulting matches. This master thesis proposes an extended web-based visualization tool targeted at teachers that builds on the existing Pyttern Visualizer. The tool adds a section where users can compose patterns and analyze multiple code files at once, consult match results per file, and can open the Matcher view for any pattern–code pair. The solution is a refactoring and extension of the frontend web application of the Pyttern Visualizer that interacts with the existing Pyttern backend to run the matcher and present results in an overview, with filtering and navigation suited to teaching workflows. The extended Pyttern Visualizer supports teachers in giving pattern-based feedback at scale in programming education.