In case anyone is interested, there are a few interesting projects I wanted to highlight:
Xiangpeng and I are working on improving the Parquet reader to detangle CPU and IO and improve predicate performance: - Parquet Reader Improvement Plan / Proposal - July 2025 [1] Ed is working on a custom thrift decoder to avoid the overhead of the auto-generated code from the thrift compiler: - Use custom thrift decoder to improve speed of parsing parquet metadata [2] These are some exciting times for the arrow-rs Rust Parquet implementation Andrew [1]: https://github.com/apache/arrow-rs/issues/8000 [2]: https://github.com/apache/arrow-rs/issues/5854
