pytest-language-server 0.22.0

A blazingly fast Language Server Protocol implementation for pytest

pytest-language-server 🔥

A blazingly fast Language Server Protocol (LSP) implementation for pytest, built with Rust, with full support for fixture discovery, go to definition, code completion, find references, hover documentation, diagnostics, and more!

Table of Contents

• Features
  • Go to Definition
  • Go to Implementation
  • Call Hierarchy
  • Code Completion
  • Find References
  • Hover Documentation
  • Document Symbols
  • Workspace Symbols
  • Code Lens
  • Code Actions
  • Diagnostics & Quick Fixes
  • Performance
• Installation
• Setup
  • Neovim
  • Zed
  • VS Code
  • IntelliJ IDEA / PyCharm
  • Emacs
  • Other Editors
• Configuration
• CLI Commands
• Supported Fixture Patterns
• Fixture Priority Rules
• Supported Third-Party Fixtures
• Architecture
• Development
• Security
• Contributing
• License
• Acknowledgments

Features

Built with AI, maintained with care 🤖

This project was built with the help of AI coding agents, but carefully reviewed to ensure correctness, reliability, security and performance. If you find any issues, please open an issue or submit a pull request!

🎯 Go to Definition

Jump directly to fixture definitions from anywhere they're used:

• Local fixtures in the same file
• Fixtures in conftest.py files
• Third-party fixtures from pytest plugins (pytest-mock, pytest-asyncio, etc.)
• Respects pytest's fixture shadowing/priority rules

🔧 Go to Implementation

Jump to the yield statement in generator fixtures:

• Generator fixtures: Navigates to where yield produces the fixture value
• Teardown navigation: Useful for reviewing fixture cleanup logic
• Non-generator fallback: Falls back to definition for simple return-based fixtures

Example:

@pytest.fixture
def database():
    conn = connect()
    yield conn      # <-- Go to Implementation jumps here
    conn.close()    # Teardown code after yield

🔗 Call Hierarchy

Explore fixture dependencies with Call Hierarchy support:

• Incoming Calls: See which tests and fixtures depend on a fixture
• Outgoing Calls: See which fixtures a fixture depends on
• Works with your editor's "Show Call Hierarchy" command
• Helps understand complex fixture dependency chains

@pytest.fixture
def database():        # <-- Call Hierarchy shows:
    ...                #     Incoming: test_query, test_insert (tests using this)
                       #     Outgoing: connection (fixtures this depends on)

✨ Code Completion

Smart auto-completion for pytest fixtures:

• Context-aware: Only triggers inside test functions and fixture functions
• Hierarchy-respecting: Suggests fixtures based on pytest's priority rules (same file > conftest.py > third-party)
• Rich information: Shows fixture source file and docstring
• No duplicates: Automatically filters out shadowed fixtures
• Works everywhere: Completions available in both function parameters and function bodies
• Supports both sync and async functions

🔍 Find References

Find all usages of a fixture across your entire test suite:

• Works from fixture definitions or usage sites
• Character-position aware (distinguishes between fixture name and parameters)
• Shows references in all test files
• Correctly handles fixture overriding and hierarchies
• LSP spec compliant: Always includes the current position in results

📚 Hover Documentation

View fixture information on hover:

• Fixture signature
• Source file location
• Docstring (with proper formatting and dedenting)
• Markdown support in docstrings

📑 Document Symbols

Navigate fixtures within a file using the document outline:

• File outline view: See all fixtures defined in the current file (Cmd+Shift+O / Ctrl+Shift+O)
• Breadcrumb navigation: Shows fixture hierarchy in editor breadcrumbs
• Return type display: Shows fixture return types when available
• Sorted by position: Fixtures appear in definition order

🔎 Workspace Symbols

Search for fixtures across your entire workspace:

• Global search: Find any fixture by name (Cmd+T / Ctrl+T)
• Fuzzy matching: Case-insensitive substring search
• File context: Shows which file each fixture is defined in
• Fast lookup: Instant results from in-memory fixture database

🔢 Code Lens

See fixture usage counts directly in your editor:

• Usage count: Shows "N usages" above each fixture definition
• Click to navigate: Clicking the lens shows all references (find-references integration)
• Real-time updates: Counts update as you add/remove fixture usages
• Local fixtures only: Only shows lenses for project fixtures, not third-party

🏷️ Inlay Hints

See fixture return types inline without leaving your code:

• Type annotations: Shows return types next to fixture parameters (e.g., db: Database)
• Explicit types only: Only displays hints when fixtures have explicit return type annotations
• Generator support: Extracts yielded type from Generator[T, None, None] annotations
• Non-intrusive: Hints appear as subtle inline decorations that don't modify your code

Example:

# With a fixture defined as:
@pytest.fixture
def database() -> Database:
    return Database()

# In your test, you'll see:
def test_example(database):  # Shows ": Database" after "database"
    pass

💡 Code Actions

Intelligent code actions for pytest fixtures, covering quick fixes, type annotations, and bulk operations:

Quick Fix — Add Missing Fixture Parameter (quickfix):

• Triggered from undeclared-fixture diagnostics
• Adds the fixture as a typed parameter (e.g., db: Database instead of just db) when the fixture has a return type annotation
• Automatically inserts any import statements needed for the return type
• Smart insertion handles both empty and existing parameter lists

Add Type Annotation (source.pytest-ls):

• Cursor-based: place your cursor on an existing fixture parameter that lacks a type annotation
• Inserts : ReturnType matching the inlay-hint text (e.g., database → database: Database)
• Automatically adds any required import statements

Add All Fixture Type Annotations (source.fixAll.pytest-ls):

• File-wide: annotates every unannotated fixture parameter in the file in a single action
• Collects and deduplicates all required imports across all fixtures
• Great for bringing an entire test file up to date in one click

isort/ruff-Aware Import Insertion (best-effort):

• Attempts to place new imports into the correct isort group (stdlib vs third-party) with proper blank-line separators
• When the file already has from X import Y for the same module, attempts to merge the new name into the existing line (sorted alphabetically) instead of adding a duplicate
• Handles __future__ imports, comments between groups, and mixed import styles on a best-effort basis
• Note: Placement follows common isort conventions but does not read your project's pyproject.toml or .isort.cfg. Run ruff check --fix or isort after applying these actions to bring imports into full conformance with your project's configuration

Example — adding a type annotation:

# Before (cursor on "database"):
def test_example(database):
    pass

# After triggering "Add type annotation for fixture 'database'":
from myapp.db import Database  # ← import added automatically

def test_example(database: Database):
    pass

⚠️ Diagnostics & Quick Fixes

Detect and fix common pytest fixture issues with intelligent code actions:

Fixture Scope Validation:

• Detects when a broader-scoped fixture depends on a narrower-scoped fixture
• Example: A session-scoped fixture cannot depend on a function-scoped fixture
• Warnings show both the fixture's scope and its dependency's scope
• Prevents hard-to-debug test failures from scope violations

Circular Dependency Detection:

• Detects when fixtures form circular dependency chains (A → B → C → A)
• Reports the full cycle path for easy debugging
• Works across files (conftest.py hierarchies)

Scope mismatch example:

# ⚠️ Scope mismatch! session-scoped fixture depends on function-scoped
@pytest.fixture(scope="session")
def shared_db(temp_dir):  # temp_dir is function-scoped
    return Database(temp_dir)

@pytest.fixture  # Default is function scope
def temp_dir(tmp_path):
    return tmp_path / "test"

Undeclared Fixture Detection:

• Detects when fixtures are used in function bodies but not declared as parameters
• Line-aware scoping: Correctly handles local variables assigned later in the function
• Hierarchy-aware: Only reports fixtures that are actually available in the current file's scope
• Works in tests and fixtures: Detects undeclared usage in both test functions and fixture functions
• Excludes built-in names (self, request) and actual local variables

One-Click Quick Fixes:

• Code actions to automatically add missing fixture parameters with type annotations
• Intelligent parameter insertion (handles both empty and existing parameter lists)
• Works with both single-line and multi-line function signatures
• Automatically adds required import statements for return types
• Triggered directly from diagnostic warnings

Example:

@pytest.fixture
def user_db() -> Database:
    return Database()

def test_user(user_db):  # ✅ user_db properly declared
    user = user_db.get_user(1)
    assert user.name == "Alice"

def test_broken():  # ⚠️ Warning: 'user_db' used but not declared
    user = user_db.get_user(1)  # 💡 Quick fix: Add 'user_db: Database' parameter + import
    assert user.name == "Alice"

How to use code actions:

1. Quick fix (undeclared fixture): Place cursor on the warning squiggle → trigger code actions (Cmd+. / Ctrl+.) → select "pytest-ls: Add 'fixture_name' fixture parameter (Type)"
2. Add type annotation (single fixture): Place cursor on an unannotated fixture parameter → trigger code actions → select "pytest-ls: Add type annotation for fixture 'name'"
3. Add all type annotations (file-wide): Trigger code actions anywhere in the file → select "pytest-ls: Add all fixture type annotations (N fixtures)"

⚡️ Performance

Built with Rust for maximum performance:

• Fast workspace scanning with concurrent file processing
• Efficient AST parsing using rustpython-parser
• Lock-free data structures with DashMap
• Minimal memory footprint

Installation

Choose your preferred installation method:

📦 PyPI (Recommended)

The easiest way to install for Python projects:

# Using uv (recommended)
uv tool install pytest-language-server

# Or with pip
pip install pytest-language-server

# Or with pipx (isolated environment)
pipx install pytest-language-server

🍺 Homebrew (macOS/Linux)

Install via Homebrew for system-wide availability:

brew install bellini666/tap/pytest-language-server

To add the tap first:

brew tap bellini666/tap https://github.com/bellini666/pytest-language-server
brew install pytest-language-server

🦀 Cargo (Rust)

Install from crates.io if you have Rust installed:

cargo install pytest-language-server

📥 Pre-built Binaries

Download pre-built binaries from the GitHub Releases page.

Available for:

• Linux: x86_64, aarch64, armv7 (glibc and musl)
• macOS: Intel and Apple Silicon
• Windows: x64 and x86

🔨 From Source

Build from source for development or customization:

git clone https://github.com/bellini666/pytest-language-server
cd pytest-language-server
cargo build --release

The binary will be at target/release/pytest-language-server.

Setup

Neovim

Add this to your config:

vim.lsp.config('pytest_lsp', {
  cmd = { 'pytest-language-server' },
  filetypes = { 'python' },
  root_markers = { 'pyproject.toml', 'setup.py', 'setup.cfg', 'pytest.ini', '.git' },
})

vim.lsp.enable('pytest_lsp')

Zed

Install from the Zed Extensions Marketplace:

1. Open Zed
2. Open the command palette (Cmd+Shift+P / Ctrl+Shift+P)
3. Search for "zed: extensions"
4. Search for "pytest Language Server"
5. Click "Install"

The extension downloads platform-specific binaries from GitHub Releases. If you prefer to use your own installation (via pip, cargo, or brew), place pytest-language-server in your PATH.

After installing the extension, you need to enable the language server for Python files. Add the following to your Zed settings.json:

{
  "languages": {
    "Python": {
      "language_servers": ["pyright", "pytest-language-server", "..."]
    }
  }
}

VS Code

The extension includes pre-built binaries - no separate installation required!

Install from the Visual Studio Marketplace:

1. Open VS Code
2. Go to Extensions (Cmd+Shift+X / Ctrl+Shift+X)
3. Search for "pytest Language Server"
4. Click "Install"

Works out of the box with zero configuration!

IntelliJ IDEA / PyCharm

The plugin includes pre-built binaries - no separate installation required!

Install from the JetBrains Marketplace:

1. Open PyCharm or IntelliJ IDEA
2. Go to Settings/Preferences → Plugins
3. Search for "pytest Language Server"
4. Click "Install"

Requires PyCharm 2024.2+ or IntelliJ IDEA 2024.2+ with Python plugin.

Emacs

Add this to your config:

(use-package eglot
  :config
  (add-to-list 'eglot-server-programs
               '((python-mode python-ts-mode) . ("pytest-language-server"))))

Other Editors

Any editor with LSP support can use pytest-language-server. Configure it to run the pytest-language-server command.

Configuration

pyproject.toml

Configure pytest-language-server via your project's pyproject.toml:

[tool.pytest-language-server]
# Glob patterns for files/directories to exclude from scanning
exclude = ["build/**", "dist/**", ".tox/**"]

# Disable specific diagnostics
# Valid codes: "undeclared-fixture", "scope-mismatch", "circular-dependency"
disabled_diagnostics = ["undeclared-fixture"]

# Additional directories to scan for fixtures (planned feature)
fixture_paths = ["fixtures/", "shared/fixtures/"]

# Third-party plugins to skip when scanning venv (planned feature)
skip_plugins = ["pytest-xdist"]

Available Options:

Option	Type	Description
exclude	string[]	Glob patterns for paths to exclude from workspace scanning
disabled_diagnostics	string[]	Diagnostic codes to suppress
fixture_paths	string[]	Additional fixture directories (planned)
skip_plugins	string[]	Third-party plugins to skip (planned)

Diagnostic Codes:

• undeclared-fixture - Fixture used in function body but not declared as parameter
• scope-mismatch - Broader-scoped fixture depends on narrower-scoped fixture
• circular-dependency - Circular fixture dependency detected

Logging

Control log verbosity with the RUST_LOG environment variable:

# Minimal logging (default)
RUST_LOG=warn pytest-language-server

# Info level
RUST_LOG=info pytest-language-server

# Debug level (verbose)
RUST_LOG=debug pytest-language-server

# Trace level (very verbose)
RUST_LOG=trace pytest-language-server

Logs are written to stderr, so they won't interfere with LSP communication.

Virtual Environment Detection

The server automatically detects your Python virtual environment:

1. Checks for .venv/, venv/, or env/ in your project root
2. Falls back to $VIRTUAL_ENV environment variable
3. Scans third-party pytest plugins for fixtures

Code Actions

Code actions are available in three forms:

Kind	Trigger	What it does
quickfix	Diagnostic warning	Adds missing fixture parameter (with type + import)
source.pytest-ls	Cursor on unannotated param	Adds : ReturnType + import for one fixture
source.fixAll.pytest-ls	Anywhere in file	Adds all missing type annotations + imports

Import ordering is best-effort. Generated imports attempt to follow common isort conventions (stdlib before third-party, alphabetical within groups, merging into existing from X import lines) but the server does not read your project's pyproject.toml or .isort.cfg. After applying code actions, run ruff check --fix or isort to bring imports into full conformance with your project's configuration.

Return-type resolution precedence — when a fixture has a return type annotation (e.g. -> Path), the server resolves the import needed for that type using the following priority:

1. Builtin types (int, str, bool, dict, list, None, …) — skipped, no import needed
2. Import map lookup — if the fixture file imports the name (e.g. from pathlib import Path), the same import statement is reused (relative imports are resolved to absolute form)
3. Locally defined names — if the name is defined in the fixture file itself (e.g. a class in conftest.py) but not imported from elsewhere, an import is generated from the file's module path (e.g. from tests.conftest import MyClass)
4. Unknown names — skipped; no import is generated

Note: Module path resolution for locally defined names depends on __init__.py files on disk. The server automatically watches for __init__.py create/delete events and re-analyzes affected fixture files, so module paths stay up to date without manual intervention.

If code actions don't appear in your editor:

1. Check LSP capabilities: Ensure your editor supports code actions (most modern editors do)
2. Check code action kinds: Some editors filter by kind — ensure source.* actions are enabled alongside quickfix
3. Enable debug logging: Use RUST_LOG=info to see if actions are being created
4. Verify diagnostics: Quick-fix actions only appear where there are undeclared-fixture warnings
5. Verify fixtures have return types: Type annotation actions only appear for fixtures with explicit return type annotations
6. Trigger manually: Use your editor's code action keybinding (Cmd+. / Ctrl+.)

CLI Commands

In addition to the LSP server mode, pytest-language-server provides useful command-line tools:

Fixtures List

View all fixtures in your test suite with a hierarchical tree view:

# List all fixtures
pytest-language-server fixtures list tests/

# Skip unused fixtures
pytest-language-server fixtures list tests/ --skip-unused

# Show only unused fixtures
pytest-language-server fixtures list tests/ --only-unused

The output includes:

• Color-coded display: Files in cyan, directories in blue, used fixtures in green, unused in gray
• Usage statistics: Shows how many times each fixture is used
• Smart filtering: Hides files and directories with no matching fixtures
• Hierarchical structure: Visualizes fixture organization across conftest.py files

Example output:

Fixtures tree for: /path/to/tests

conftest.py (7 fixtures)
├── another_fixture (used 2 times)
├── cli_runner (used 7 times)
├── database (used 6 times)
├── generator_fixture (used 1 time)
├── iterator_fixture (unused)
├── sample_fixture (used 7 times)
└── shared_resource (used 5 times)
subdir/
└── conftest.py (4 fixtures)
    ├── cli_runner (used 7 times)
    ├── database (used 6 times)
    ├── local_fixture (used 4 times)
    └── sample_fixture (used 7 times)

This command is useful for:

• Auditing fixture usage across your test suite
• Finding unused fixtures that can be removed
• Understanding fixture organization and hierarchy
• Documentation - visualizing available fixtures for developers

Fixtures Unused

Find unused fixtures in your test suite, with CI-friendly exit codes:

# List unused fixtures (text format)
pytest-language-server fixtures unused tests/

# JSON output for programmatic use
pytest-language-server fixtures unused tests/ --format json

Exit codes:

• 0: All fixtures are used
• 1: Unused fixtures found

Example text output:

Found 4 unused fixture(s):

  • iterator_fixture in conftest.py
  • auto_cleanup in utils/conftest.py
  • temp_dir in utils/conftest.py
  • temp_file in utils/conftest.py

Tip: Remove unused fixtures or add tests that use them.

Example JSON output:

[
  {"file": "conftest.py", "fixture": "iterator_fixture"},
  {"file": "utils/conftest.py", "fixture": "auto_cleanup"},
  {"file": "utils/conftest.py", "fixture": "temp_dir"},
  {"file": "utils/conftest.py", "fixture": "temp_file"}
]

This command is ideal for:

• CI/CD pipelines - fail builds when unused fixtures accumulate
• Code cleanup - identify dead code in test infrastructure
• Linting - integrate with pre-commit hooks or quality gates

Supported Fixture Patterns

Decorator Style

@pytest.fixture
def my_fixture():
    """Fixture docstring."""
    return 42

Assignment Style (pytest-mock)

mocker = pytest.fixture()(_mocker)

Async Fixtures

@pytest.fixture
async def async_fixture():
    return await some_async_operation()

Fixture Dependencies

@pytest.fixture
def fixture_a():
    return "a"

@pytest.fixture
def fixture_b(fixture_a):  # Go to definition works on fixture_a
    return fixture_a + "b"

@pytest.mark.usefixtures

@pytest.mark.usefixtures("database", "cache")
class TestWithFixtures:
    def test_something(self):
        pass  # database and cache are available

@pytest.mark.parametrize with indirect

@pytest.fixture
def user(request):
    return User(name=request.param)

# All parameters treated as fixtures
@pytest.mark.parametrize("user", ["alice", "bob"], indirect=True)
def test_user(user):
    pass

# Selective indirect fixtures
@pytest.mark.parametrize("user,value", [("alice", 1)], indirect=["user"])
def test_user_value(user, value):
    pass

Imported Fixtures (from ... import *)

# conftest.py
from .pytest_fixtures import *  # Fixtures from pytest_fixtures.py are available

pytest_plugins Variable

# conftest.py
pytest_plugins = ["myapp.fixtures", "other.fixtures"]

# Also supports single strings, tuples, and annotated assignments:
# pytest_plugins = "myapp.fixtures"
# pytest_plugins = ("myapp.fixtures",)
# pytest_plugins: list[str] = ["myapp.fixtures"]

Fixtures declared in pytest_plugins modules are automatically discovered in conftest.py, test files, and plugin entry point modules. Only static string literals are supported — dynamic values are ignored.

Fixture Priority Rules

pytest-language-server correctly implements pytest's fixture shadowing rules:

1. Same file: Fixtures defined in the same file have highest priority
2. Closest conftest.py: Searches parent directories for conftest.py files
3. Virtual environment: Third-party plugin fixtures

Fixture Overriding

The LSP correctly handles complex fixture overriding scenarios:

# conftest.py (parent)
@pytest.fixture
def cli_runner():
    return "parent runner"

# tests/conftest.py (child)
@pytest.fixture
def cli_runner(cli_runner):  # Overrides parent
    return cli_runner  # Uses parent

# tests/test_example.py
def test_example(cli_runner):  # Uses child
    pass

When using find-references:

• Clicking on the function name def cli_runner(...) shows references to the child fixture
• Clicking on the parameter cli_runner(cli_runner) shows references to the parent fixture
• Character-position aware to distinguish between the two

Supported Third-Party Fixtures

Automatically discovers fixtures from 50+ popular pytest plugins, including:

• Testing frameworks: pytest-mock, pytest-asyncio, pytest-bdd, pytest-cases
• Web frameworks: pytest-flask, pytest-django, pytest-aiohttp, pytest-tornado, pytest-sanic, pytest-fastapi
• HTTP clients: pytest-httpx
• Databases: pytest-postgresql, pytest-mongodb, pytest-redis, pytest-mysql, pytest-elasticsearch
• Infrastructure: pytest-docker, pytest-kubernetes, pytest-rabbitmq, pytest-celery
• Browser testing: pytest-selenium, pytest-playwright, pytest-splinter
• Performance: pytest-benchmark, pytest-timeout
• Test data: pytest-factoryboy, pytest-freezegun, pytest-mimesis
• And many more...

The server automatically scans your virtual environment for any pytest plugin and makes their fixtures available.

Architecture

• Language: Rust 🦀
• LSP Framework: tower-lsp-server
• Parser: rustpython-parser
• Concurrency: tokio async runtime
• Data Structures: DashMap for lock-free concurrent access

Development

Prerequisites

• Rust 1.85+ (2021 edition)
• Python 3.10+ (for testing)

Building

cargo build --release

Running Tests

cargo test

Logging During Development

RUST_LOG=debug cargo run

Security

Security is a priority. This project includes:

• Automated dependency vulnerability scanning (cargo-audit)
• License compliance checking (cargo-deny)
• Daily security audits in CI/CD
• Dependency review on pull requests
• Pre-commit security hooks

See SECURITY.md for our security policy and how to report vulnerabilities.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Development Setup

1. Install pre-commit hooks:

   pre-commit install
   

2. Run security checks locally:

   cargo audit
   cargo clippy
   cargo test
   

License

MIT License - see LICENSE file for details.

Acknowledgments

Built with:

• tower-lsp-server - LSP framework
• rustpython-parser - Python AST parsing
• tokio - Async runtime

Special thanks to the pytest team for creating such an amazing testing framework.

---

Made with ❤️ and Rust. Blazingly fast 🔥

Built with AI assistance, maintained with care.