JASTG Structural Graph Benchmark

JASTG Structural Graph Benchmark is a curated dataset of AST-based structural dependency graphs extracted from 21 open-source Java projects. Each project snapshot is frozen by commit hash to ensure reproducibility. The graphs were generated with JASTG (Java AST Structural Graph), a lightweight static analysis tool that extracts explicit, typed structural relations from Java Abstract Syntax Trees (AST) without full type solving or build-system integration. For each project, the dataset provides: A structural dependency graph in GraphML format (graph_<project>.graphml), where nodes represent classes (including inner classes) and directed weighted edges represent explicit structural dependencies (e.g., inheritance/implements, field types, parameter and return types, and object instantiations). Edge weights reflect the frequency of occurrences in the analyzed source code. A JSON metadata file (metadata_<project>.json) describing the extraction parameters and snapshot information. Summary tables (benchmark_summary.csv and projects.csv) with repository URLs, licenses, commit hashes, extraction date, JASTG version, and basic graph statistics (number of classes, number of edges, density, and degree descriptors). Intended uses include benchmarking community detection and clustering methods for software modularity recovery, studying coupling and cohesion patterns, comparing structural metrics across systems, and supporting reproducible experiments in architecture and modularity analysis for Java codebases. The dataset does not redistribute any upstream project source code, only derived graph and metadata artifacts.The graphs were generated using JASTG (Java AST Structural Graph), available at: https://github.com/MarcosCordeiro/jastg How to cite:Please cite this dataset using the Zenodo DOI provided on this record. If you use the JASTG tool, also cite the JASTG software repository. Keywords: Java; static analysis; AST; structural dependency graph; software metrics; modularity; community detection; reproducibility.