PyMODAlib 0.4.0b1

Library providing Python implementations of MODA's algorithms.

PyMODAlib

Introduction

PyMODAlib is a Python library containing the algorithms used by PyMODA. With PyMODAlib, you can write Python scripts to perform the same calculations as PyMODA.

Some of PyMODAlib's algorithms are MATLAB-packaged libraries, while some are Python translations of algorithms belonging to MODA.

License

You may use, distribute and modify this software under the terms of the GNU General Public License v3.0. See LICENSE.

References and citations

To cite PyMODAlib or view its references, please see the DOI at Zenodo.

User Guide

This section describes how to get started with PyMODAlib.

For a full API reference, please see PyMODAlib's ReadTheDocs page, which shows the parameters and output for every function.

Prerequisites

The following software is required to use all of PyMODAlib's functionality:

• Python 3.6 or higher.
• MATLAB Runtime, version 9.6.

Note: PyMODAlib can be used without the MATLAB Runtime, but some functions require it.

Installing PyMODAlib

PyMODAlib can be installed using pip. Open a terminal and run:

pip install pymodalib

Tip: on macOS/Linux, replace pip with the correct command for your system (e.g. pip3).

Updating PyMODAlib

PyMODAlib will be updated regularly. To update your installed version, open a terminal and run:

pip install -U pymodalib

Current status

PyMODAlib is still in development. Currently, the features implemented are:

• Wavelet transform.
• Wavelet phase coherence.
• Group coherence for one or two groups, with inter-subject surrogates.
• Detecting harmonics.
• Downsampling.

:warning: Some functions may not be fully stable, and may change before v1.0.0.

Getting started

Examples

There are examples of using PyMODAlib's functionality in the examples directory.

To download the dependencies required to run the examples, open a terminal and run:

pip install -U pymodalib matplotlib

To try the examples, download the PyMODAlib repository as a zip file or by using git clone, then run Python files from the examples subfolders.

PyMODAlib cache

The group coherence functions use a very large quantity of RAM. To mitigate this problem for machines with smaller RAM capacities, they will allocate arrays which are cached to disk. This may result in significant disk usage.

By default, PyMODAlib will use a folder named .pymodalib inside your home directory for its cache. However, it will show a RuntimeWarning unless you set the location manually. This warning is intended to make users aware of the risks of placing the cache folder on an SSD.

:warning: If the cache folder is on an SSD, it may reduce the lifespan of the SSD.

Setting the cache location

To set the location of the cache folder manually, use the PYMODALIB_CACHE environment variable. Instructions for different operating systems are below.

Windows

On Windows, press the start button and type "environment" until you can select "Edit the system environment variables". Then click "Environment variables" and click "New" in the window which appears. Name it "PYMODALIB_CACHE" and set the location by browsing for a folder.

Now restart your IDE or terminal.

Linux

Run the following commands, replacing <cache_folder> with the absolute path to your chosen folder:

echo "export PYMODALIB_CACHE=<cache_folder>" >> ~/.bashrc
source ~/.bashrc

macOS

Run the following commands, replacing <cache_folder> with the absolute path to your chosen folder:

echo "export PYMODALIB_CACHE=<cache_folder>" >> ~/.bash_profile
source ~/.bash_profile

Developer guide

This guide is aimed at developers interested in contributing to PyMODAlib.

Downloading the code

To download the code, you should fork the repository and clone your fork.

Installing the requirements

Open a terminal in the PyMODAlib folder and run:

pip install -r requirements.txt
pip install matplotlib pre-commit

Git hooks

Git hooks are used to automatically format modified Python files with black when a commit is made. This ensures that the code follows a uniform style.

To install the Git hooks, open a terminal in the PyMODAlib folder and run:

pre-commit install

Tip: If this causes an error, try python -m pre-commit install.

When you make a commit, the modified Python files will be formatted if necessary. If this occurs, you'll need to repeat your git add and git commit commands to make the commit.

Tip: You can still use PyCharm's auto-formatter while writing code. Although it sometimes disagrees with black, black will undo its changes at commit-time, so no harm is done.

Developing PyMODAlib

When developing PyMODAlib, you can test your changes by installing the library locally in "editable" mode. From the root of the repository, run:

pip install -e .

Note: After making changes to PyMODAlib, you don't need to run the pip install command again. Any Python script which imports pymodalib will reflect the changes immediately.

Switching back to the release version of PyMODAlib is simple:

pip uninstall pymodalib -y
pip install -U pymodalib

Project structure

The public-facing API is located in the algorithms package. This package contains wrappers for the actual implementations, which can be found in the implementations package.

This structure allows the implementation to be easily changed, while ensuring that the API remains backwards-compatible.

__init__.py

In pymodalib.__init__.py, many functions are imported from the algorithms package. This allows users to more easily find useful functions: for example, they can use pymodalib.wavelet_transform instead of pymodalib.algorithms.wavelet.wavelet_transform.

implementations package

The implementations package contains a matlab package and a python package. The matlab package contains wrappers for algorithms supplied by MATLAB-packaged libraries, while the python package contains algorithms implemented purely in Python.

MATLAB-packaged libraries

Currently, the MATLAB-packaged libraries are not required if functionality that depends on them is not used. MATLAB-packaged libraries are installed by downloading PyMODA and installing its dependencies.

Functions that require the MATLAB Runtime will be marked with the matlabwrapper decorator. This will check if the correct version of the MATLAB Runtime is installed.

Note: MATLAB libraries are still incompatible with Python 3.8. When MATLAB R2020a releases, Python 3.8 support will be added but the current required version of the MATLAB Runtime will no longer be supported (users will need to upgrade to the newer Runtime).

Packaging the project for PyPI

This section describes how to publish an update to PyPI.

From the documentation:

rm -r dist/
python setup.py sdist
python setup.py bdist_wheel
twine upload dist/*