A curated dataset of Abstract Syntax Tree-based structural dependency graphs and object-oriented metrics for 21 open-source Java systems.

This article describes a curated dataset of structural dependency graphs and object-oriented metrics extracted from 21 open-source Java systems. Each repository was analyzed at a fixed commit using a deterministic AST-based static analysis pipeline. For every project, the dataset provides a weighted directed GraphML file encoding class-to-class structural dependencies, with node-level metrics (LCOM4, CBO, RFC, NOM, NOA) embedded as node attributes, together with execution metadata documenting repository URL, commit hash, extraction date, tool and dependency versions, and a configuration hash that makes the analysis parameters independently verifiable. The selected systems cover distinct domains, including build tools, frameworks, distributed platforms, and utility libraries, and range in size from 43 to 29,695 classes. All artifacts correspond to dataset version v1.0.1 and are publicly available through a Zenodo repository (DOI: https://doi.org/10.5281/zenodo.18744312). The dataset was generated using JASTG v1.0.0 (commit b62a943), providing version-level traceability between code and data. The dataset is intended for use in graph-based modularity analysis, community detection, and architectural structure mining on Java code representations.