The Query Transformation Pipeline

Readyset uses query transformation to convert SQL queries into dataflow graphs that continuously maintain cached results as underlying data changes, rather than re-executing queries from scratch each time. This approach reduces read latency by turning queries into simple lookups into pre-computed materialized views, though it requires SQL to be expressed in a more restrictive form with constraints like binary joins with equality predicates and no correlated execution. The key tradeoff is optimizing once at cache creation versus traditional databases that optimize query execution each time it runs.