Production-focused offline SQL static analyzer.

Catch security vulnerabilities, performance regressions, reliability issues, compliance risks, cost inefficiencies, and code quality problems before they reach production.

Offline-First Analysis. Catch bugs without ever connecting to a live database. SlowQL works entirely on SQL source files, making it safe to run anywhere.

Custom Rule Engine. Define your own organizational SQL conventions via YAML rules or Python plugins. Custom rules integrate seamlessly with the built-in catalog and support full reporting and suppression.

282 Built-in Rules. Covers security, performance, reliability, compliance, cost, and quality. Each rule includes impact documentation, fix guidance, and severity classification.

Dead SQL Detection. Safely identify unused database objects and redundant code. SlowQL detects unused views, stored procedures, and functions by analyzing definitions and usages across your entire project. It also flags unreachable code paths in procedures (e.g., after RETURN) and near-duplicate queries that should be consolidated.

Cross-File SQL Analysis. Detect breaking changes across multiple files. SlowQL understands relationships between DDL, views, and procedures, flagging when a schema change in one file (e.g., DROP COLUMN) breaks a query in another.

dbt & Jinja Support. Natively parses dbt models and SQL templates containing Jinja tags ({{ ref() }}, {% if %}, {% for %}). Enforces dbt best practices including missing references and hardcoded schema detection.

Migration Framework Support. Natively supports Alembic, Django migrations, Flyway, Liquibase, Prisma Migrate, and Knex. SlowQL understands the ordering, dependencies, and context of migration files to catch destructive changes before they break your existing queries.

14 SQL Dialects. Dialect-aware analysis for PostgreSQL, MySQL, SQL Server (T-SQL), Oracle, SQLite, Snowflake, BigQuery, Redshift, ClickHouse, DuckDB, Presto, Trino, Spark, and Databricks. Universal rules fire on all dialects; dialect-specific rules only fire when relevant.

Schema-Aware Validation. Optionally validate against your DDL files to catch missing tables, columns, and suggest indexes.

Safe Autofix. Conservative, exact-text-replacement fixes with FixConfidence.SAFE. No guessing, no heuristic rewrites. Preview with --diff, apply with --fix.

CI/CD Native. GitHub Actions, SARIF, pre-commit hooks, JSON/HTML/CSV exports. Exit codes based on severity thresholds.

Editor Integration. VS Code extension via slowql-vscode and foundational LSP server for other editors.

Application Code SQL Extraction. Automatically extract and analyze SQL strings embedded in Python, TypeScript/JavaScript, Java, Go, Ruby, and MyBatis XML mapper files. SlowQL uses language‑specific heuristics (AST for Python, regex for others) and a dedicated MyBatis XML parser to find SQL, flagging potential injection risks in dynamic constructions. It distinguishes safe #{param} parameterization from unsafe ${param} interpolation and marks queries using dynamic MyBatis tags (<if>, <where>, <set>, etc.) as dynamic.

pipx install slowql

pip install slowql

docker run --rm -v $(pwd):/src makroumi/slowql /src/queries.sql

Requirements: Python 3.11+, Linux / macOS / Windows.

slowql queries.sql

Analyze application code (extracts SQL strings automatically):

slowql src/app.py src/services/

Analyze with schema validation:

slowql queries.sql --schema schema.sql

Run in CI mode with failure thresholds:

slowql init --dialect postgresql --fail-on high
slowql src/ --fail-on high

slowql src/main/resources/mapper/UserMapper.xml
slowql src/main/resources/mapper/ --schema db/schema.sql

Preview and apply safe fixes:

slowql queries.sql --diff
slowql queries.sql --fix --fix-report fix-report.json

List all built-in rules dynamically or get detailed documentation for a specific rule:

slowql --list-rules
slowql --explain PERF-SCAN-001

Integrate SlowQL directly into your Python scripts with three lines of code:

import slowql
result = slowql.analyze("SELECT * FROM users")

SlowQL performs optional schema-aware validation by inspecting your DDL files. This catches structural issues that generic static analysis misses.

Tables and Columns. Detect references to non-existent tables or columns.

Index Suggestions. Identify filtered columns that lack corresponding indexes.

slowql queries.sql --schema database/schema.sql
slowql migrations/ --schema schema.sql --fail-on critical

Schema findings:

SlowQL ships with 282 rules across six dimensions:

MyBatis is a popular Java/Spring ORM framework that uses XML mapper files to define SQL statements. SlowQL now parses these mapper files and applies all existing SQL rules.

• <select>, <insert>, <update>, <delete>, <sql>
• Dynamic tags: <if>, <where>, <set>, <foreach>, <choose>, <when>, <otherwise>, <trim>

• Safe: #{param} – uses prepared‑statement style parameterization.
• Unsafe: ${param} – direct string interpolation, flagged as potential SQL injection.

Queries containing any dynamic tags are marked is_dynamic = True and are analyzed for injection and performance issues.

<?xml version="1.0" encoding="UTF-8"?>
<mapper namespace="com.example.UserMapper">
  <!-- Safe -->
  <select id="findUserById" resultType="User">
    SELECT * FROM users WHERE id = #{id}
  </select>

  <!-- Unsafe -->
  <select id="searchUsers" resultType="User">
    SELECT * FROM users WHERE name LIKE ${searchTerm}
  </select>

  <!-- Dynamic -->
  <update id="updateUser">
    UPDATE users
    <set>
      <if test="name != null">name = #{name},</if>
      <if test="email != null">email = #{email},</if>
    </set>
    WHERE id = #{id}
  </update>
</mapper>

SlowQL will extract three statements, flag SELECT * (PERF‑SCAN‑001), flag unsafe ${} (SEC‑INJ‑001), and mark the update as dynamic.

• SEC‑INJ‑001 … SEC‑INJ‑011 – injection patterns.
• PERF‑SCAN‑001 – SELECT *.
• QUAL‑DBT‑001 – hard‑coded table names.

107 rules are dialect-aware, firing only on the relevant database engine:

The remaining 175 rules are universal and fire on all dialects.

SlowQL provides conservative, zero-risk autofixes for rules where the replacement is 100% semantically equivalent:

slowql queries.sql --diff
slowql queries.sql --fix
slowql queries.sql --fix --fix-report fixes.json

Autofix principles:

1. Only exact text replacements. No schema inference, no heuristic rewrites.
2. Every fix is tagged with FixConfidence.SAFE, meaning the output is functionally identical to the input.
3. A .bak backup is always created before writing.
4. Fixes can be previewed as a unified diff before applying.

Examples of safe autofixes:

Rules can be silenced on a per-line, per-block, or per-file basis using directives written directly in SQL comments. No configuration file changes are required.

SELECT * FROM archive;  -- slowql-disable-line PERF-SCAN-001

-- slowql-disable-next-line SEC-INJ-001
SELECT id, token FROM sessions WHERE id = $1;

-- slowql-disable PERF-SCAN
SELECT * FROM event_stream;
SELECT * FROM session_log;
-- slowql-enable PERF-SCAN

-- slowql-disable-file REL-001

The rule ID may be an exact identifier, a prefix, comma-separated values, or omitted entirely to suppress all rules for that scope. Matching is case-insensitive.

Baseline Mode allows you to adopt SlowQL on an existing, chaotic codebase without drowning in thousands of initial warnings. This is similar to SonarQube's "New Code Period."

1. Create a baseline: Store all your current issues in a .slowql-baseline file.

   slowql queries/ --update-baseline

2. Run against the baseline: Now, SlowQL will only flag new issues introduced after the baseline was created.

   slowql queries/ --baseline

Because issues are fingerprinted via content hashes, standard edits like appending blank lines won't suddenly "un-suppress" your baseline issues. See the full Baseline Docs for CI/CD setup.

In CI environments, running static analysis over thousands of files on every commit is slow and unnecessary. SlowQL supports git-aware analysis to cleanly skip untouched files.

# Only analyze files that are changed, staged, or untracked
slowql . --git-diff

# Analyze files changed since branching off main
slowql . --since main

--input-file       Path to SQL file or directory
--schema           Path to DDL schema file
--baseline         Path to baseline file (suppress known issues)
--update-baseline  Update/create the baseline file
--fail-on          Failure threshold: critical, high, medium, low, info, never
--interactive      Opt-in to full interactive experience (animations, menus) when no inputs are provided
--select-dialect   Opt-in to prompt for dialect selection interactively
--non-interactive  (deprecated, now the default) Explicitly disable interactive mode
--git-diff         Only analyze files changed in the current workspace
--since            Analyze files changed since a specific git revision (e.g. main)
--cache-dir        Directory to store cache files (default: .slowql_cache)
--no-cache         Disable query result caching
--clear-cache      Clear cache directory before analysis
--jobs, -j         Number of parallel workers for analyzing multiple files. (0 = auto)
--compare          Enable query comparison mode

--format                        Primary output: console, github-actions, sarif
--export                        Export to disk: json, html, csv, sarif
--out                           Directory for exported reports
--diff                          Preview safe autofix diff
--fix                           Apply safe autofixes (single file, creates .bak)
--fix-report                    Write JSON report of fixes
--list-rules                    List all 282 rules with severity, dimension, and dialect
--list-rules --filter-dimension Filter by dimension (security, performance, etc.)
--list-rules --filter-dialect   Filter by dialect (postgresql, mysql, etc.)
--explain RULE-ID               Show full documentation for a specific rule

0    No issues found or issues below failure threshold
2    Issues found meet or exceed --fail-on threshold
3    Runtime error or tool failure

SlowQL discovers configuration from slowql.toml, .slowql.toml, slowql.yaml, .slowql.yaml, or pyproject.toml (under [tool.slowql]).

severity:
  fail_on: high
  warn_on: medium

analysis:
  dialect: postgresql
  enabled_dimensions:
    - security
    - performance
    - reliability
  disabled_rules:
    - PERF-SCAN-001
  severity_overrides:
    PERF-SCAN-001: info
    QUAL-NULL-001: critical

# Configuration is discovered from the analyzed directory or any parent.
# This allows project-specific and per-directory configuration.

schema:
  path: db/schema.sql

output:
  format: console
  verbose: false
  show_fixes: true

cost:
  cloud_provider: none

compliance:
  frameworks:
    - gdpr

- uses: slowql/slowql-action@v1
  with:
    path: "./sql/**/*.sql"
    schema: "db/schema.sql"
    fail-on: high
    format: github-actions

- name: SlowQL Analysis
  run: |
    pip install slowql
    slowql --input-file sql/ --schema db/schema.sql --fail-on high --format github-actions

repos:
  - repo: https://github.com/slowql/slowql
    rev: v1.6.2
    hooks:
      - id: slowql
        args: [--fail-on, high]

Install slowql-vscode from the VS Code Marketplace for real-time SQL analysis in your editor. The extension uses the SlowQL LSP server for diagnostics.

SlowQL now provides a numerical complexity score (0-100) for every analyzed query, helping teams enforce quality policies and track complexity trends.

• Spectral Analysis: Scores are calculated based on structural patterns like joins, subqueries, and aggregations.
• Visual Feedback: Terminal output highlights query complexity to help identify candidates for optimization.

You can enable/disable complexity scoring and set thresholds for "optimal", "complex", and "critical" queries in your .slowql.yml:

complexity:
  enabled: true
  threshold_optimal: 40
  threshold_complex: 70

SlowQL is a modular pipeline:

SQL Files → Parser (sqlglot) → AST → Analyzers → Rules → Issues → Reporters
                                 ↑                          ↓
                           Schema Inspector            AutoFixer
                           (DDL parsing)           (safe text fixes)

Parser. Uses sqlglot for multi-dialect SQL parsing. Handles statement splitting, dialect detection, and AST generation.

Engine. Orchestrates parsing, analyzer execution, schema validation, and result aggregation.

Analyzers. Six domain-specific analyzers (Security, Performance, Reliability, Compliance, Cost, Quality), each loading rules from the catalog.

Custom Rules. Dynamic plugin system that loads user-defined rules from YAML files (regex-based) or Python modules (AST-based) at runtime.

Rules. 282 detection rules implemented as PatternRule (regex), ASTRule (sqlglot AST traversal), or custom Rule subclasses.

Schema Inspector. Parses DDL files into a schema model. Enables table/column existence checks and index suggestions.

Reporters. Console (rich TUI), GitHub Actions annotations, SARIF 2.1.0, JSON, HTML, CSV.

AutoFixer. Conservative text-based fix engine. Span-based and exact-text replacements only.

git clone https://github.com/slowql/slowql.git
pip install -e ".[dev]"

pytest
ruff check .
mypy src/slowql

Apache License 2.0. See LICENSE.

Issues: github.com/slowql/slowql/issues

Discussions: github.com/slowql/slowql/discussions

Back to top