Python (mandatory)

LightWin requires Python 3.12 or higher. Ensure that you have an appropriate version of Python installed on your system. If not, you can download the latest version from the official Python website.

Installation with pip (recommended)

1. Navigate to the LightWin base folder (i.e., where the pyproject.py file is located).

2. In a dedicated Python environment, run the following command:

pip install -e .

The mandatory packages with their dependencies should be automatically downloaded.

Hint

The -e flag means that the installation is editable. In other words, you can edit the source files and the changes will be taken into account when calling lightwin.

Note

If there is a problem during the installation, file an issue. Give me your OS, development environment, and the full log error. You can probably fix the issue yourself by manually installing the packages listed in pyproject.toml, taking a more recent version of the offending package(s). In any case, I’ll be happy to have the information.

Installation with conda (not recommended)

The package dependencies are listed in a single compact pyproject.toml file. This is not yet supported by conda. So the installation will require more work from your side.

1. In a dedicated environment, install all the dependencies defined in the pyproject.toml (some can be removed, see following sections).

2. Add the /path/to/LightWin/src to your PYTHONPATH environment variable. This is the folder containing the lightwin folder.

Todo

Consider providing more detailed instructions for setting the PYTHONPATH. For now, you can search for “PYTHONPATH” or “ModuleNotFoundError” online for additional guidance.

Following packages are optional

• cython - Used to speed up calculations. Check cython integration documentation.

  • Note: Installing cython prior to pymoo enable compilation of some pymoo functions for improved performance.

• pytest - To run tests and ensure everything is working as expected.

• cloudpickle - To pickle/unpickle some objects (see the util.pickling documentation).

For developers

The files are formatted with black (text width: 79 characters) and isort. The docstrings must adhere to the numpy style.

Reminders

Installation of a package

To install a package, use the appropriate method based on your environment:

• If using pip:

  pip install <package>
  

• If using conda, avoid mixing with pip to prevent potential conflicts. Instead, use:

  conda install <package>
  

Note

Since pymoo is not available on the default Anaconda channels, you should create a conda environment and use conda-forge:

conda create -n <env-name> -c conda-forge python=3.12
conda activate <env-name>
conda install cython matplotlib numpy palettable pandas scipy tkinter pymoo pytest -c conda-forge

Always specify -c conda-forge when installing or updating packages.

Warning

pip and conda are not fully compatible. Avoid using them together, or create a dedicated environment to prevent conflicts. For more details, you may refer to this video.