slowql 1.6.4

Next-generation SQL analyzer: Security, Performance, Compliance, Cost optimization

SlowQL

SlowQL is a production-focused offline SQL static analyzer that catches security vulnerabilities, performance regressions, reliability issues, compliance risks, cost inefficiencies, and code quality problems before they reach production.

It performs safe static analysis of your SQL source code with no database connection required. SlowQL ships with 282 built-in rules covering 14 SQL dialects, and is built for modern engineering teams supporting CI/CD pipelines, pre-commit hooks, GitHub Actions, SARIF output, LSP, and automated fixes.

---

---

---

---

---

Why SlowQL

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.

---

Installation

pipx (recommended)

pipx install slowql

pip

pip install slowql

Docker

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

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

---

Quick Start

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:

Analyze MyBatis mapper files:

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

---

Schema-Aware Validation

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:

Rule	Description
SCHEMA-TBL-001	Table referenced but not defined in schema
SCHEMA-COL-001	Column referenced but not present in table definition
SCHEMA-IDX-001	Missing index suggested for filtered column

---

SlowQL ships with 282 rules across six dimensions:

Dimension	Focus	Rules
Security	SQL injection, privilege escalation, credential exposure, SSRF	61
Performance	Full scans, indexing, joins, locking, sorting, pagination	73
Reliability	Data loss prevention, transactions, race conditions, idempotency	44
Quality	Naming, complexity, null handling, style, dbt, dead SQL	51
Cost	Cloud warehouse optimization, storage, compute, network	33
Compliance	GDPR, HIPAA, PCI-DSS, SOX, CCPA	18

MyBatis XML Support

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.

Supported MyBatis Tags

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

Parameter Syntax

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

Dynamic SQL Detection

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

Example

<?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.

Relevant Rules

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

Dialect‑Specific Rules

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

Dialect	Specific Rules	Examples
PostgreSQL	12	pg_sleep detection, SECURITY DEFINER without search_path, CREATE INDEX without CONCURRENTLY
MySQL	15	LOAD DATA LOCAL INFILE, utf8 vs utf8mb4, ORDER BY RAND(), MyISAM detection
T-SQL (SQL Server)	23	OPENROWSET, sp_OACreate, @@IDENTITY, MERGE without HOLDLOCK, SET NOCOUNT ON
Oracle	11	UTL_HTTP/UTL_FILE, EXECUTE IMMEDIATE injection, CONNECT BY without NOCYCLE
Snowflake	9	COPY INTO credentials, VARIANT in WHERE, CLONE without COPY GRANTS
BigQuery	6	SELECT * cost, DISTINCT on UNNEST, repeated subqueries
SQLite	6	ATTACH DATABASE file access, PRAGMA foreign_keys = OFF, AUTOINCREMENT overhead
Redshift	7	COPY with embedded credentials, COPY without MANIFEST, DISTSTYLE ALL
ClickHouse	7	url() SSRF, mutations, SELECT without FINAL, JOIN without GLOBAL
DuckDB	3	COPY without FORMAT, large IN lists, old-style casts
Presto / Trino	4	Implicit cross-joins, INSERT OVERWRITE without partition, ORDER BY without LIMIT
Spark / Databricks	5	BROADCAST on large table, UDF in WHERE, CACHE TABLE without filter

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

---

Safe Autofix

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:

Rule	Before	After
QUAL-NULL-001	WHERE x = NULL	WHERE x IS NULL
QUAL-STYLE-002	EXISTS (SELECT * FROM t)	EXISTS (SELECT 1 FROM t)
QUAL-MYSQL-003	LOCK IN SHARE MODE	FOR SHARE
QUAL-TSQL-001	SET ANSI_NULLS OFF	SET ANSI_NULLS ON
QUAL-ORA-002	SELECT 1 FROM DUAL	SELECT 1

---

Inline Suppression

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

Directive	Scope
-- slowql-disable-line RULE-ID	Current line only
-- slowql-disable-next-line RULE-ID	Next non-blank line
-- slowql-disable RULE-ID	Open block until matching enable or EOF
-- slowql-enable RULE-ID	Closes an open block
-- slowql-disable-file RULE-ID	Entire file

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 (Diff Mode)

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.

---

Git-Aware Analysis

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

---

CLI Usage

Primary Flags

--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
--non-interactive  Suppress spinners and interactive prompts
--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

Output Control

--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

Exit Codes

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

---

Configuration

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

---

CI Integration

GitHub Action (Official)

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

Direct CLI in CI

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

Pre-commit

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

---

VS Code Extension

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.

---

Query Complexity Scoring

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.

Configuration

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

Architecture

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.

---

Development

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

pytest
ruff check .
mypy src/slowql

---

License

Apache License 2.0. See LICENSE.

Issues: github.com/makroumi/slowql/issues

Discussions: github.com/makroumi/slowql/discussions

---

Back to top