skbuild-conditional-scripts

Generic dynamic script metadata provider for scikit-build-core projects.

This micro-package provides a reusable metadata provider that can conditionally install entry point scripts based on environment variables. Originally developed for AFDKO to choose between installing a C++ binary or Python wrapper, it can be used by any scikit-build-core project that needs conditional entry point generation.

Features

• Data-driven: All script definitions live in pyproject.toml, not code
• Conditional installation: Install different scripts based on environment variables
• Generic: Can be reused by other projects with similar needs
• Simple: ~70 lines of code, easy to understand and maintain

Installation

pip install skbuild-conditional-scripts

Or include in your project locally using provider-path.

Quick Start

Add to your pyproject.toml:

[build-system]
requires = ["scikit-build-core", "skbuild-conditional-scripts"]
build-backend = "scikit_build_core.build"

[tool.scikit-build.metadata.scripts]
provider = "skbuild_conditional_scripts"

# Always installed scripts
scripts = {
    "my-command" = "mypackage:main",
    "my-tool" = "mypackage.tools:cli",
}

# Conditionally installed scripts (only when MY_ENV_VAR=ON)
conditional-scripts = {
    "my-debug-tool" = "mypackage.debug:main",
}

condition-env-var = "MY_ENV_VAR"
condition-env-value = "ON"

Then build your project:

# Normal build - only base scripts installed
pip install .

# With condition - both base and conditional scripts installed
MY_ENV_VAR=ON pip install .

Use Cases

• Binary vs Python wrapper: Install C++ binary or Python wrapper based on build configuration
• Platform-specific commands: Different entry points for different platforms (via env vars)
• Debug builds: Install additional debug tools only when requested
• Optional features: Install extra commands when feature flags are enabled

How It Works

The provider is called during the metadata generation phase (before building). It:

1. Reads configuration from [tool.scikit-build.metadata.scripts]
2. Checks the specified environment variable
3. Merges base scripts with conditional scripts if the condition matches
4. Returns the combined dictionary to scikit-build-core

Note: This is a build-time tool. End users installing from PyPI wheels don't need it.

Documentation

• WHO_NEEDS_THIS.md - When is this provider needed?
• DEPLOYMENT.md - Deployment options and strategies
• CONTRIBUTING.md - Contributing to this project
• README_SUMMARY.md - Quick navigation guide

Example: AFDKO

AFDKO uses this provider to choose between a C++ binary and Python wrapper:

[tool.scikit-build.metadata.scripts]
provider = "skbuild_conditional_scripts"

scripts = {
    # 30+ deprecated wrapper scripts always installed
    "tx" = "afdko._deprecated:tx_wrapper",
    "makeotf" = "afdko._deprecated:makeotf_wrapper",
    # ... etc
}

conditional-scripts = {
    # Python wrapper only when requested
    "afdko" = "afdko.invoker:main"
}

condition-env-var = "AFDKO_COMMAND_USE_WRAPPER"
condition-env-value = "ON"

Normal builds install the fast C++ binary. When building in environments where the C++ build fails, setting AFDKO_COMMAND_USE_WRAPPER=ON installs the Python wrapper instead.

Requirements

• Python >= 3.10
• scikit-build-core (as your build backend)

License

Apache License 2.0 - See LICENSE for details.

Support

This project is maintained by the Adobe Type team.

• Issues: GitHub Issues
• Discussions: Use issues for questions and discussions
• Contributing: See CONTRIBUTING.md

Related Projects

• scikit-build-core - Modern Python build backend for CMake
• AFDKO - Adobe Font Development Kit for OpenType