Project Configuration

In this section we will explain the pyproject.toml contents and why we have taken certain decisions.

Python version dependency

Since we have adopted nox as our platform to run the unit tests for different Python environments/versions, the pyproject.toml file contains dependency specifiers for Python versions. The main packages that have different dependencies based on the Python version are: numpy, pandas, and pyzmq. The dependencies are specified in the project.dependencies table in the pyproject.toml file of the cgse-common project.

Then, there is the case where the importlib.metadata entry_points used to return a dict-like object until Python 3.9, and a collection of entry points as of 3.10. The deprecated interface was completely removed as of Python 3.12. We have now implemented a check in egse.plugin that imports the import_metadata backport for Python < 3.10. This backport package is only installed when using Python 3.9 as is specified in the pyproject.toml file of the cgse-common.

Dependency Groups

Dependency groups are a way to organize optional dependencies that aren't required for basic package functionality. Dependency groups are not installed by default. They are only installed when explicitly requested.

Warning

The [dependency-groups] section is part of uv's own configuration system rather than a standard Python packaging concept. It is different from the [project.optional-dependencies] section that follows Python's packaging standards.

By default, uv includes the dev dependency group in the environment, e.g. during uv run or uv sync. You can define a group to be a default group by adding it to tool.uv.default-groups. For example, when the docs dependency group should also be a default group, add

[tool.uv]
default-groups = ["dev", "docs"]