This repository contains independent stress tests, scalability benchmarks, and real-world schema evaluations for PolyXML—the high-performance, polyglot native XML data-binding engine.
| # | Experiment Name | Focus Area | Key Finding |
|---|---|---|---|
| 01 | NeTEx 304MB European Transit Benchmark | High-throughput streaming, large real-world CEN transit models, memory consumption | 845 MB/s streaming throughput (0.360s vs 88.0s predecessor xsdata) & ~75% less RAM (~1.27 GB saved). |
PolyXML provides zero-copy, high-throughput bidirectional XML serialization and deserialization across programming languages (Rust, Python, Node.js, Java, C++, and Go). In the Python ecosystem, it serves as the drop-in native backend for pyxsdata dataclasses and Pydantic models via CoreXmlParser and CoreXmlSerializer.