Check Guidelines.md for the detailed formatting rules.
We use automatic yapf formatting specified in setup.cfg, and used with yapf version 0.28.0.
You can run:
./bin/setup_repository
after cloning the repository in order to set up yapf formatter
and git hook for removing non important changes from Jupyter Notebooks.
Unused functions, variables, imports, etc. can be quite reliably detected using the vulture package. To include the notebooks, make sure to first convert them to python scripts by running
jupyter nbconvert --to=python *.ipynb
before checking coverage etc. by running
vulture *.py
We use the frozen tag for each cell which should not be executed, for example because it entails some heavy computation. This tag is easily added by enabling the nbextension Freeze.
The following lines of code can be run to install and enable the extension. After installation, 3 buttons appear in the notebook, and the * button corresponds to freezing a cell.
For pip installation:
pip install jupyter_contrib_nbextensions
jupyter contrib nbextension install --user
jupyter nbextension enable Freeze
or for Conda installation:
conda install -c conda-forge jupyter_contrib_nbextensions
jupyter contrib nbextension install --sys-prefix --symlink
jupyter nbextension enable freeze/main
You can then test the core notebooks using
./bin/run_important_notebooks
or test a specific notebook using
python bin/run_notebooks.py <notebook name>
which will execute all python notebooks while skipping the frozen cells. The script will crash if any of the notebooks raise an error.
To run tests, type in terminal
pytest test/
The package cvxpy uses SCS as default solver. The problem is that for SCS to work it needs to be installed with BLAS and LAPACK support, which is not done automatically. I did not manage to set up SCS as default solver in the travis environment, so I decided to use CVXOPT instead, for which the installation is less problematic.