demo_products
product_id · name · categoryNames and categories can change in the same source transaction as a sale.
Sales land. Refunds happen. The leaderboard moves.
Watch a real SQL join turn changing data into a live top ten.
Ranked by net revenue, descending. Product ID breaks ties. Awaiting the first committed snapshot.
The product catalog supplies names and categories. Order lines carry signed quantities and their own historical unit price. A refund has negative quantity; a canceled line is deleted. These are line items, not order counts.
product_id · name · categoryNames and categories can change in the same source transaction as a sale.
line_id · product_id · quantity
unit_price_centsNet revenue = SUM(quantity × unit_price_cents). Catalog edits never reprice historical lines.
| Product | Category | Line items | Net units | Net revenue |
|---|
Each query has its own worker, source slot, state and committed result. Snapshots can represent different source transactions. Browser notifications trigger authoritative refetches.
Independent process measurements and server history for the all-products query and the top-ten query. Sample ages and missing values stay explicit.
Process counters reset on worker restart. Latency distributions cover the preceding 60 seconds; browser delivery is excluded. Commit observers follow each query’s durable progress even if no result row changes. CPU and PUT-only request counts are not collected.
One point per minute. Missing intervals stay blank.
One point per minute. Missing intervals stay blank.