Skip to main content

dfmt

Grammar-driven SQL formatting for 15 dialects with native dbt/Jinja support.

dfmt parses SQL into a full AST using the same grammar-driven parser that powers column lineage and type inference, then pretty-prints it using Wadler's algorithm. This means it understands SQL structure deeply -- not just tokens.

Key Features​

  • 15 SQL dialects: ANSI, BigQuery, PostgreSQL, DuckDB, Snowflake, Spark, Databricks, Oracle, DB2, T-SQL, MariaDB, Presto, Trino, Redshift, SQLite
  • Native dbt/Jinja: Evaluates Jinja templates, formats the SQL, and restores Jinja constructs -- not regex-based
  • Idempotent: format(format(x)) == format(x) -- always
  • Fast: ~2s for 3,500 dbt models (parallel JVM)
  • CI-ready: --check mode exits 1 if any file needs formatting
  • Configurable: keyword case, comma style, alignment, line width, and more via .dfmt.json

Quick Start​

# Create a config file
dfmt --init snowflake

# Format all SQL files
dfmt --write

# Check formatting in CI
dfmt --check

# Format files changed since main
dfmt --changed-since main --write

See Getting Started for installation and setup, or CLI Reference for the full flag list.