PPR Engine Benchmark

5 engines · 11 queries · 783,755 transactions · Ireland's Property Price Register

A full benchmarking of DuckDB, Polars, Pandas, Dask, and Zig against the national Irish property register — 92 MB of real transaction data from 2010 through May 2026. Stress queries include standard aggregations, self-joins, rolling windows, percentile distributions, and multi-dimensional GROUP BYs.

Final results (11 queries, 4 cores, 16 GiB LXC)

Rank Engine Ingest Queries Total
🥇Zig0.305s0.158s0.464s
🥈Polars0.096s0.929s1.025s
🥉DuckDB0.265s1.207s1.472s
Pandas1.373s3.177s4.550s
Dask1.085s10.082s11.168s

Open full interactive dashboard →

Full dashboard includes Chart.js charts, per-engine breakdowns, query stress profiles, memory safety analysis, and non-linear regression results.