parq is homebrew for data.
- Its input is wild data. Any csv, xls, json, xml source.
- Its output is a reproducible parquet dataset saved in an OCI registry, like docker images or [homebrew] bottles.
It's a low co-ordination way to let anyone who cares about a dataset publish a tidier version of it in parquet: a format designed to be nice to query locally or over http. (vs e.g. xml, a format designed for modelling hierarchies)