Contributing

Contributions are welcome! savana is designed as an umbrella for independently-usable geospatial modules, new sensors, feature stacks, classification schemes, validation workflows, and utilities should be addable as their own module without breaking any existing module's API. savana.rainfall (precipitation product assessment) is the second such module, added alongside the original land-system classification pipeline following exactly this pattern, a useful reference if you're adding a third.

Reporting bugs

Please open an issue at github.com/desmond-lartey/savana/issues with:

• Your OS and Python version
• The savana version (savana.__version__)
• Which module (classification or rainfall)
• Steps to reproduce, for classification, the AOI type you used; for rainfall, the station input and product(s) involved

Development setup

git clone https://github.com/desmond-lartey/savana.git
cd savana
pip install -e ".[dev,docs,vector,rainfall]"
pytest tests/

Pull requests

• Keep new functionality in its own module where possible, mirroring the existing composites / indices / masks / classifiers split for classification-side work, or the stations / zones / ingestion / extraction / validation split for rainfall-side work
• If you're adding an entirely new module (not extending classification or rainfall), give it its own subpackage under savana/ (see savana/rainfall/ for the shape: its own __init__.py with the same PEP 562 lazy-loading pattern, its own config.py of overridable defaults, its own optional-dependency extra in pyproject.toml)
• Add or update tests in tests/
• Update the docs (docs/) if you add public API surface, everything in savana.__init__._LAZY_SYMBOL_MAP (classification) or savana.rainfall.__init__._LAZY_SYMBOL_MAP (rainfall) should have a corresponding docstring picked up in the relevant Classification API Reference / Rainfall API Reference page