Fast drilldown dashboards from a single Parquet file

A single 40MB Parquet data cube, an 18KB reader, an R2 bucket, and a few unassuming http range requests. That's all it takes to serve a fast drilldown dashboard with no database or query engine. The author rolls up 34 million NYC 311 service requests into a cube, stores it on R2, and uses Hyparquet in the browser to fetch only the needed byte ranges. The key is sorting rows so min/max statistics let the reader skip most of the file. The approach shifts complexity to the data pipeline, making it ideal for customer-facing analytics with bounded queries.