Reproducible database experiments behind the posts from Sheeta.
Each folder holds one experiment: the SQL, a command that rebuilds the data and reruns every query, the raw measurements, and the caveats.
Every number in a post traces to a CSV in that experiment's results/ folder.
| Folder | What it measures |
|---|---|
postgres-vs-bigquery |
Lookups, single-row writes and full-table aggregations on the same orders table in Postgres and BigQuery, at 5, 10 and 20 million rows |
query-optimization |
Composite index column order, functions on indexed columns, and warehouse table layout in Postgres and ClickHouse |
Each experiment needs Docker and uv. The folder's README lists anything else, such as a Google Cloud project for BigQuery, and the exact commands.
Found a mistake in a method or a number? Open an issue.