sqlfluff

Modular SQL linter and auto-formatter with dialect support

TLDR

Lint a SQL file with a specific dialect

$ sqlfluff lint --dialect [postgres] [path/to/query.sql]

Auto-fix lint violations in place

$ sqlfluff fix --dialect [bigquery] [path/to/queries/]

Format SQL via stdin

$ cat [query.sql] | sqlfluff format --dialect [snowflake] -

Show parsed token tree for a query

$ sqlfluff parse --dialect [mysql] [query.sql]

Lint only specific rules

$ sqlfluff lint --rules [LT01,LT02,CP01] [path/to/file.sql]

Exclude noisy rules from a run

$ sqlfluff lint --exclude-rules [L034,L036] [path/to/file.sql]

Render a templated query (expand Jinja/dbt) without linting

$ sqlfluff render --dialect [ansi] [query.sql]

SYNOPSIS

sqlfluff command [options] PATH...Commands: lint, fix, format, parse, render, rules, dialects, fluff (legacy alias)

sqlfluff is a dialect-aware linter and auto-formatter for SQL. It parses queries into an abstract syntax tree, applies a configurable set of rules, and reports or rewrites violations. Unlike text-based linters, the AST approach lets it enforce structural rules such as comma placement, keyword casing, ambiguous joins, and column-reference qualification.The tool supports a wide range of dialects (PostgreSQL, MySQL, BigQuery, Snowflake, Redshift, T-SQL, SQLite, DuckDB, Databricks, ANSI, and more) and integrates with templating engines so it can lint queries written in Jinja, dbt, or custom placeholder syntaxes without first rendering them.Operation is split into commands: lint reports violations, fix rewrites the source in place, format applies layout-only fixes, parse dumps the token tree, and render expands templating without linting. Configuration is layered: built-in defaults are overridden by .sqlfluff files discovered up the directory tree, pyproject.toml sections, then command-line flags.

PARAMETERS

--dialect NAME

SQL dialect to parse against (e.g. ansi, postgres, mysql, bigquery, snowflake, redshift, tsql, sqlite, duckdb, databricks). Run sqlfluff dialects for the full list.

--templater NAME

Templating engine to use before parsing: raw, jinja, python, placeholder, or dbt.

--rules LIST

Comma-separated list of rules (or rule groups) to enable.

--exclude-rules LIST

Comma-separated list of rules to skip.

-f, --force

For fix: apply changes without asking for confirmation.

--FIX-EVEN-UNPARSABLE

Allow fix to operate on files with templating or parse errors (dangerous).

--processes N

Number of parallel workers; 0 means all CPUs, negative means cpus + N.

-v, -vv, -vvv

Increase verbosity; stackable.

--nocolor

Disable ANSI color in output.

--format FMT

Output format: human, json, yaml, github-annotation, github-annotation-native.

--config FILE

Path to an explicit .sqlfluff configuration file.

--encoding ENC

File encoding to use when reading sources (default autodetect).

--disable-noqa

Ignore -- noqa: comments embedded in SQL.

CONFIGURATION

.sqlfluff

Per-project INI-style configuration; the search walks upward from the linted file. Example:

$ [sqlfluff]
dialect = postgres
templater = jinja
exclude_rules = L034,L036

[sqlfluff:rules:capitalisation.keywords]
capitalisation_policy = upper

[sqlfluff:indentation]
tab_space_size = 2
indent_unit = space

pyproject.toml

Settings live under **[tool.sqlfluff.*]** sections that mirror the INI layout.

~/.sqlfluff

User-level fallback applied to projects without their own config.

-- noqa: comments

Inline directives to disable rules on a line or block, e.g. **SELECT * -- noqa: L044**.

RULE GROUPS

Rules are namespaced by area: LT layout, CP capitalisation, CV convention, RF references, ST structure, AM ambiguity, AL aliasing, JJ Jinja. Use sqlfluff rules to print the complete catalog with codes, names, and descriptions.

CAVEATS

Templated SQL must parse after rendering; unresolved variables raise template errors that suppress most lint output. fix rewrites files even when the underlying SQL has subtle semantic implications (e.g. column-order changes from re-aliasing); review diffs before committing. Dialect parsers vary in coverage; very recent vendor syntax may not yet be recognized. Older numeric rule codes (L001, L002, ...) were renamed to mnemonic codes (LT01, CP01, ...) in 2.0; legacy codes are aliased but deprecated.

HISTORY

SQLFluff was started by Alan Cruickshank in 2019 as an open-source dialect-aware alternative to text-based SQL linters. Version 1.0 shipped in 2022, and 2.0 in 2023 brought a redesigned rule taxonomy with mnemonic codes and a templating-aware fix engine. The project is maintained on github.com/sqlfluff/sqlfluff under the MIT license.