Hi all,
I’d like to introduce GitHub - idlab-discover/pyroquet · GitHub , a mojo-native reader / writer for parquet files. It adheres to the specification File Format | Parquet. It does not fully cover the specification yet but all of the fundamentals and some of the bells and whistles are present in the library.
This first release version is a full rewrite of an earlier attempt at porting fastparquet to Mojo. Back then, all data types were backed by custom column implementations, but upon stabilization of Mojo to 1.0.0 I decided that I’d rather start to build on existing libraries.
Numeric columns are now backed by NuMojo’s NDArray type. String columns and some other bits have custom data back-ends.
The original library that served as inspiration was fastparquet, but the library was retired in 2026. Pyarrow has excellent support for Apache parquet. It was also more performant than fastparquet so pyarrow’s parquet engine became the new reference.
Verification of the project was done against fastparquet, pyarrow and duckdb (which has great native support for parquet).
Scope-wise, this project is much bigger than the other contributions I have posted here (Balsa and mojo-snappy), but I have used pyroquet to build mojo-native data pipelines in intrusion detection and it works well so I feel comfortable releasing a 0.1.0.