bayesflow-packaging

To structure a Python package using a src layout, place your code in a src/ directory and configure build settings in pyproject.toml. This separation enforces cleaner project organization, prevents local import errors during testing, and separates source files from configuration.

You define optional dependencies in pyproject.toml under the project section using the optional-dependencies field. This manages feature availability like calibration, allowing users to install extras only when needed and implementing guarded imports in code.

Managing API exports with __all__ involves defining a list of public symbols in your module's __init__.py. This explicitly controls public visibility, ensuring only intended classes and functions are exposed to users importing your Python package.

Implement version detection patterns by reading from package metadata or a dedicated version file. This enables dynamic versioning in pyproject.toml, ensuring your Python extension package reports its version accurately without hardcoding or duplication.

Yes, you can use pyproject.toml to configure CI for Python package development. It centralizes dependency declarations and optional features, ensuring CI environments install the correct packages for testing and building your project.

Use a src layout instead of a flat layout to prevent local import errors during testing. This structure ensures tests run against the installed package rather than the local directory, improving packaging reliability and consistency.