rtest 0.0.15

Python test runner built in Rust

rtest

A high-performance Python test runner built with Rust, designed as a drop-in replacement for pytest with enhanced collection resilience and built-in parallelization.

⚠️ Development Status: This project is in early development (v0.0.x). While functional, expect breaking changes and evolving features as we work toward stability.

Features

Resilient Test Collection

Unlike pytest which stops execution when collection errors occur, rtest continues running tests even when some files fail to collect:

pytest stops when collection fails:

collected 22 items / 3 errors
!!!!!!!!!!!!!!!!!!!!! Interrupted: 3 errors during collection !!!!!!!!!!!!!!!!!!!!!!!!
============================== 1 warning, 3 errors in 0.97s ==============================
# No tests run - you're stuck

rtest keeps going:

collected 22 items / 3 errors
!!!!!!!!!!!!!!!!!! Warning: 3 errors during collection !!!!!!!!!!!!!!!!!!!!!
================================== test session starts ===================================
# Your 22 working tests run while you fix the 3 broken files

Built-in Parallelization

rtest includes parallel test execution out of the box, without requiring additional plugins like pytest-xdist. Simply use the -n flag to run tests across multiple processes:

# Run tests in parallel (recommended for large test suites)
rtest -n 4                    # Use 4 processes
rtest -n auto                 # Auto-detect CPU cores
rtest --maxprocesses 8        # Limit maximum processes

Distribution Modes

Control how tests are distributed across workers with the --dist flag:

• --dist load (default): Round-robin distribution of individual tests
• --dist loadscope: Group tests by module/class scope for fixture reuse
• --dist loadfile: Group tests by file to keep related tests together
• --dist worksteal: Optimized distribution for variable test execution times
• --dist no: Sequential execution (no parallelization)

# Examples
rtest -n auto --dist loadfile      # Group by file across all CPU cores
rtest -n 4 --dist worksteal        # Work-steal optimized distribution
rtest --dist no                    # Sequential execution for debugging

Note: The loadgroup distribution mode from pytest-xdist is not yet supported. xdist_group mark parsing is planned for future releases.

Current Implementation

The current implementation focuses on enhanced test collection and parallelization, with test execution delegated to pytest for maximum compatibility.

Performance

rtest delivers significant performance improvements over pytest across popular open-source Python projects:

Test Collection Performance

Repository      pytest               rtest                Speedup
-----------     ------               -----                -------
FastAPI         5.477s ± 0.044s      0.096s ± 0.001s     56.82x
Requests        0.446s ± 0.003s      0.041s ± 0.000s     10.89x
Flask           0.479s ± 0.006s      0.045s ± 0.000s     10.60x
Click           0.367s ± 0.002s      0.042s ± 0.000s     8.64x
HTTPX           0.250s ± 0.003s      0.044s ± 0.000s     5.65x
Scikit-learn    0.225s ± 0.002s      0.226s ± 0.002s     1.00x
Pandas          0.239s ± 0.005s      0.506s ± 0.001s     0.47x

Test Execution Performance

Repository      pytest               rtest                Speedup
-----------     ------               -----                -------
Flask           1.688s ± 0.008s      0.035s ± 0.000s     48.47x
Click           1.353s ± 0.004s      0.034s ± 0.000s     40.23x
FastAPI         0.652s ± 0.004s      0.035s ± 0.001s     18.43x
Django          0.561s ± 0.016s      0.036s ± 0.001s     15.55x
HTTPX           0.252s ± 0.004s      0.035s ± 0.000s     7.19x
Pandas          0.235s ± 0.002s      0.061s ± 0.000s     3.81x
Scikit-learn    0.224s ± 0.002s      0.060s ± 0.001s     3.73x

Benchmarks performed using hyperfine with 20 runs, 3 warmup runs per measurement. Results show mean ± standard deviation across popular Python projects on Ubuntu Linux (GitHub Actions runner).

Quick Start

Installation

pip install rtest

Requires Python 3.9+

Basic Usage

# Drop-in replacement for pytest
rtest

# That's it! All your existing pytest workflows work
rtest tests/
rtest tests/test_auth.py -v
rtest -- -k "test_user" --tb=short

Advanced Usage

Environment Configuration

# Set environment variables for your tests
rtest -e DEBUG=1 -e DATABASE_URL=sqlite://test.db

# Useful for testing different configurations
rtest -e ENVIRONMENT=staging -- tests/integration/

Collection and Discovery

# See what tests would run without executing them
rtest --collect-only

# Mix `rtest` options with any pytest arguments
rtest -n 4 -- -v --tb=short -k "not slow"

Python API

from rtest import run_tests

# Programmatic test execution
run_tests()

# With custom pytest arguments
run_tests(pytest_args=["tests/unit/", "-v", "--tb=short"])

# Suitable for CI/CD pipelines and automation
result = run_tests(pytest_args=["--junitxml=results.xml"])

Command Reference

Option	Description
-n, --numprocesses N	Run tests in N parallel processes
--maxprocesses N	Maximum number of worker processes
-e, --env KEY=VALUE	Set environment variables (can be repeated)
--dist MODE	Distribution mode for parallel execution (default: load)
--collect-only	Show what tests would run without executing them
--help	Show all available options
--version	Show rtest version

Pro tip: Use -- to separate rtest options from pytest arguments:

rtest -n 4 -e DEBUG=1 -- -v -k "integration" --tb=short

Known Limitations

Parametrized Test Discovery

rtest currently discovers only the base function names for parametrized tests (created with @pytest.mark.parametrize), rather than expanding them into individual test items during collection. For example:

@pytest.mark.parametrize("value", [1, 2, 3])
def test_example(value):
    assert value > 0

pytest collection shows:

test_example[1]
test_example[2] 
test_example[3]

rtest collection shows:

test_example

However, when rtest executes tests using pytest as the executor, passing the base function name (test_example) to pytest results in identical behavior - pytest automatically runs all parametrized variants. This means test execution is functionally equivalent between the tools, but collection counts may differ.

Contributing

We welcome contributions! Check out our Contributing Guide for details on:

• Reporting bugs
• Suggesting features
• Development setup
• Documentation improvements

License

MIT - see LICENSE file for details.

---

Acknowledgments

This project takes inspiration from Astral and leverages their excellent Rust crates:

• ruff_python_ast - Python AST utilities
• ruff_python_parser - Python parser implementation

Built with Rust for the Python community