Exploring modern Python features, performance optimizations, and emerging technologies through hands-on experiments and benchmarks.
Benchmarks demonstrating Python 3.14's experimental free-threaded build that removes the Global Interpreter Lock (GIL). Compares threading vs multiprocessing performance and explores shared memory use cases.
Key findings: 5-6x speedup for CPU-bound tasks, competitive with multiprocessing, 10x memory savings for shared state.
Benchmark comparing Arrow Flight streaming protocol against traditional Parquet file downloads served by nginx. Tests throughput, latency, and memory efficiency for columnar data transfer.
Comparison: Arrow Flight enables zero-copy streaming vs compressed file downloads. Parquet-over-HTTP represents the lower bound (fastest non-Arrow approach).
- uv - Fast Python package manager
# Install uv
curl -LsSf https://astral.sh/uv/install.sh | sh
# Navigate to an experiment
cd free-threading
# Follow the experiment's README for specific setupThis repository is a collection of hands-on experiments exploring:
- Modern Python features and performance characteristics
- Emerging technologies in the Python ecosystem
- Real-world performance comparisons and benchmarks
- Practical integration patterns and best practices
Each experiment includes runnable code, detailed explanations, and key findings.
Have an idea for a performance experiment? Open an issue or submit a pull request!
MIT - Feel free to use these benchmarks for your own testing and analysis.