​
How Isofold Works

Isofold is a drop-in SQL proxy that optimizes queries before they reach your data warehouse. It’s designed to be transparent, verifiable, and performant. At a high level: 
Client → Isofold Proxy → Data Warehouse
Isofold rewrites queries in-flight and ensures that results remain unchanged. It supports dry-run pricing, query diffing, and output equivalence verification.

​
Core Components

​
1. Proxy Layer

The proxy receives SQL queries over HTTP, JDBC, or native Postgres wire protocol. It handles:
  • Protocol translation (if needed)
  • Authentication passthrough
  • Routing to the appropriate backend engine

​
2. Rewrite Engine

The rewrite engine is responsible for transforming SQL into semantically equivalent forms that are cheaper to execute. It performs optimizations such as:
  • Projection pruning
  • Join simplification
  • Subquery flattening
  • Predicate pushdown
You can configure aggressiveness levels per warehouse or per team.

​
3. Cost Model

For supported warehouses like BigQuery, Isofold runs a dry-run of both the original and optimized queries to compare estimated costs.

​
4. Verification Engine (Optional)

When enabled, Isofold will:
  1. Execute both the original and rewritten queries
  2. Compare result sets for equality
  3. Log any divergence or mismatch for audit purposes
export ISOFOLD_VERIFY_MODE=true

​
Example Rewrite

Original query: 
SELECT * FROM users WHERE created_at >= '2023-01-01'
Rewritten query: 
SELECT id, email, created_at FROM users WHERE created_at >= '2023-01-01'
Assuming only those columns are ever referenced, the rewrite saves scan costs.

​
Deployment Options

Isofold can be:
  • Used as a hosted service with team-specific proxy URLs
  • Deployed inside your cloud/VPC
  • Run at the edge via Fly.io
See Cloud Setup or Self-Hosted Setup to choose what works best for you.
Next: Dive deeper into the Rewrite Engine.