Structa: A utility for analyzing repeating structures in large data sources

This looks incredible. How does it handle drastically different document structures, though?