Installation#

Both PyPI and Conda-Forge should have pre-built wheels for major distributions.

conda

conda install -c conda-forge scikit-digital-health

pip

pip install scikit-digital-health

pip from GitHub

pip install git+ssh://git@github.com/PfizerRD/scikit-digital-health.git
# install from a specific tag
pip install git+ssh://git@github.com/PfizerRD/scikit-digital-health@0.9.10
# install from a specific branch
pip install git+ssh://git@github.com/PfizerRD/scikit-digital-health@development

Run-time requirements#

numpy >=1.17.2
scipy >=1.3.1
pandas >=1.0.0
lightgbm >=2.3.0
pywavelets
scikit-learn
h5py
matplotlib
packaging
pyyaml
importlib_resources [Python < 3.7]

Windows Notes#

Windows users might need to install additional dependencies, even if not building from source.

https://wiki.python.org/moin/WindowsCompilers has more details, but generally users should expect to have to install a Microsoft Visual C++ redistributable package. For Python >3.5 this should be 14.0 or greater.

The 2015 redistributable update 3 can be found here: https://www.microsoft.com/en-us/download/details.aspx?id=53587

Building from Source#

In order to build from source, the following are required:

C compiler (currently has only been tested with GCC)
Fortran compiler (currently only been tested with GFortran)
libzip
setuptools
numpy >=1.17.2

If you are using Conda, libzip can easily be installed via:

conda install -c conda-forge libzip

C/Fortran Compilers#

NumPy only supports specific compilers on each platform. To get a list of available compilers, when in the scikit-digital-health directory, running:

# Get a list of platform supported/installed C compilers
python setup.py build_ext --help-compiler
# Get a list of platform supported/installed Fortran compilers
python setup.py build_ext --help-fcompiler

Testing the Build#

In order to test a build, you will need to clone or download the source code from GitHub, as the tests are isolated from access to the package source code. Additionally, the following are requirements to run the test suite:

pytest
psutil
tables
coverage [optional, for running test coverage]

From the top level scikit-digital-health directory, run:

pytest test/

The output should look similar to:

============================================================================
platform darwin -- Python 3.10.2, pytest-7.0.1, pluggy-1.0.0
rootdir: /Users/lukasadamowicz/Documents/Packages/scikit-digital-health
collected 434 items

test/activity/test_activity_core.py .....                             [  1%]
test/activity/test_activity_endpoints.py ................             [  4%]
# ... rest of test results ...

=========================== warnings summary ===============================
# ... warnings content ...
========= 427 passed, 7 skipped, 61 warnings in 6.30s ======================