Data Engineering

strata

[ Open Source ][ Alpha ]

Declarative, versioned, immutable data transformations — compiles to SQL

strata is a declarative data-transformation language: write versioned, immutable models in .strata files and compile them to SQL for DuckDB, Postgres, BigQuery, and Snowflake — with column-level lineage and 3-phase contract pins. CLI covers build, compile, run, test, replay, rollback, gc, catalog, and import-dbt, plus Iceberg publication, a VS Code extension with LSP, and standalone binary/Docker distribution. Published as strata-lang 0.1.6 on PyPI, MIT-licensed.

// Features

  • Declarative .strata models — versioned, immutable transformations with typecheck, contracts, and column-level lineage (no DB needed)
  • `strata compile --dialect` emits SQL for DuckDB, Postgres, BigQuery, and Snowflake from the same source
  • Materialize and gate with `strata run --seed` and `strata test` — DuckDB by default, real Postgres/BigQuery/Snowflake warehouses
  • Iceberg publication: `run --iceberg-dir`, plus replay verify, rollback, gc, and catalog inspection with pyiceberg cross-check
  • VS Code extension (LSP diagnostics, completion, hover, go-to-definition) + standalone binary, Homebrew tap, and Docker image
  • 80% coverage gate with 744 tests — ephemeral Postgres, BigQuery emulator, and Snowflake harnesses in CI

// Technologies

// quick-start — v0.1.6

pip install strata-lang
strata build examples/daily_orders.strata
strata compile examples/daily_orders.strata --dialect postgres
strata run examples/daily_orders.strata --seed -o /tmp/demo.duckdb

Mismo fuente, 4 warehouses: DuckDB, Postgres, BigQuery, Snowflake · --iceberg-dir publica tablas Iceberg reales.