The dataset of "GNN-CONTEXT: A GNN-based Code Context Prediction Approach for Software Development Tasks"

This research dataset accompanies our paper "GNN-CONTEXT: A GNN-based Code Context Prediction Approach for Software Development Task".The dataset comprises 3,879 code context models along with their corresponding source code, extracted from three Java open source projects: Mylyn, Platform, and PDE. Each code context model directory contains two essential files: 1. code_context_model.xml This XML file encapsulates the code context model's metadata within a <code_context_model> wrapper, including commit ID, start datetime of the first edit event, and end datetime of the last edit event. The file structure is as follows: - <graph> components represent connected components in the code context model. - Each <graph> consists of <vertices> and <edges>. - <vertex> attributes: - id: Unique identifier within the <graph> - ref_id: Reference to its source code in my_java_codes - kind: Vertex type (class, function, or variable) - label: Declared name in the source code - stereotype: Vertex stereotype derived from JStereoCode - <edge> attributes: - start: Source vertex ID - end: Destination vertex ID - label: Edge type (declares, calls, inherits, or implements)2. my_java_codes.csv This CSV file contains two columns: - id: Composite of the code context model's directory index and the vertex's ref_id - code: Corresponding source code for the code element