sdl2-alpha 0.2.0

Fast, correct alpha blending for SDL2 surfaces

sdl2_alpha - Fast Alpha Blending for SDL2

sdl2_alpha provides mathematically correct alpha blending to fix SDL2's broken alpha compositing behavior. Built with Rust and PyO3 for maximum performance.

Why sdl2_alpha?

SDL2's built-in alpha blending has fundamental flaws:

• Incorrect destination alpha handling
• Platform-specific premultiplication inconsistencies
• Accumulation errors that degrade transparency (the dreaded "persistent haze")

sdl2_alpha implements Porter-Duff "over" compositing with proper premultiplied alpha math, delivering visually correct results for composition-heavy applications.

🚀 Development Setup (Complete Newbie Guide)

Prerequisites

You need Rust and Python. If you don't have Rust:

# Install Rust (this installs rustc, cargo, etc.)
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
source ~/.cargo/env

# Verify Rust installation
rustc --version
cargo --version

Install maturin (Rust-Python bridge)

pip install maturin

Development Workflow

cd sdl2_alpha/

# Development build (fast compile, not optimized, includes debug info)
maturin develop

# This installs the package in development mode so you can import it immediately
# Changes to Rust code require rebuilding with maturin develop

Release Builds (for production/benchmarking)

cd sdl2_alpha/

# Release build (slow compile, fully optimized)
maturin build --release

# Install the optimized wheel
pip install target/wheels/*.whl --force-reinstall

# OR do both in one command
maturin develop --release

Testing Your Changes

cd sdl2_alpha/

# Install test dependencies
pip install -e .[test]

# Run the test suite
pytest tests/ -v

# Run just performance tests
pytest tests/test_performance.py -v -s

# Quick functionality test
python -c "
import sdl2_alpha
result = sdl2_alpha.blend_pixel((255,0,0,128), (0,255,0,255))
print(f'Test blend result: {result}')
print('✅ sdl2_alpha working!' if result[0] > 100 and result[1] > 100 else '❌ Something wrong')
"

When to Use Which Build

• Development build (maturin develop):

  • Use when actively coding/debugging
  • Much faster to compile
  • Good enough for testing functionality
  • ~3-5x slower than release

• Release build (maturin build --release):

  • Use for performance testing
  • Use for final integration
  • Takes much longer to compile
  • Full SIMD optimizations enabled

Useful Commands

# Clean everything and start over
cd sdl2_alpha/
cargo clean
rm -rf target/

# Check for Rust issues without compiling
cargo check

# See what maturin will build
maturin list-python

# Force rebuild everything
maturin develop --force

# Build specific Python version
maturin develop --python 3.10

Troubleshooting

"maturin not found":

pip install maturin

"Rust compiler errors":

cargo check  # Shows errors without building Python extension

"Import errors after rebuild":

# Sometimes Python caches the old version
python -c "import sys; print([p for p in sys.path if 'sdl2_alpha' in p])"
pip uninstall sdl2_alpha
maturin develop

"Performance is slow":

• Make sure you're using release build: maturin develop --release
• Debug builds are much slower

API Usage

import sdl2_alpha

# Blend single pixel (for testing)
result = sdl2_alpha.blend_pixel(
    (255, 128, 64, 128),  # src (r,g,b,a)
    (64, 128, 255, 255)   # dst (r,g,b,a)  
)

# Zero-copy in-place blending (FAST - the high-performance path)
sdl2_alpha.blend_rect_inplace(
    src_ptr,        # Raw pointer to source pixels
    src_width, src_height,
    src_x, src_y, src_w, src_h,  # Source rectangle (can be negative)
    dst_ptr,        # Raw pointer to destination pixels  
    dst_width, dst_height,
    dst_x, dst_y    # Destination position (can be negative)
)

# Copy-based blending (slower but easier)
blended_bytes = sdl2_alpha.blend_surface(
    src_bytes, dst_bytes, width, height
)

blended_bytes = sdl2_alpha.blend_rect(
    src_bytes, src_width, src_height, src_x, src_y, src_w, src_h,
    dst_bytes, dst_width, dst_height, dst_x, dst_y
)

Performance Characteristics

• ~72 FPS for 100 frames with 20 overlapping 50x50 elements on 800x600 surface
• ~1440 blits/sec with automatic clipping
• Zero-copy direct memory access (blend_rect_inplace)
• Automatic clipping handles negative coordinates and off-screen blits
• Fast paths for fully opaque (direct copy) and fully transparent (skip) pixels

Integration with SDL2 Applications

sdl2_alpha provides a drop-in replacement for SDL2's broken alpha blending. Use the zero-copy blend_rect_inplace() function for maximum performance in composition-heavy applications.

License

MIT