andrzejnovak / chillmax

MadMax Deep Leaning Boost Factors

Package management and github

It's advisable to keep all (python) packages in a virtual environment managed by conda

To install conda you can run the following and then follow the instructions. When asked to add stuff to .bashrc you should say yes

wget https://repo.continuum.io/miniconda/Miniconda3-latest-Linux-x86_64.sh
bash Miniconda3-latest-Linux-x86_64.sh

In python most real packages are published on pypi, so one can usually just run something like pip install numpy, however, our package is private so we don't want to publish it, but we can still use the pip installation.

To do this you first need to clone it from github.

git clone [e-mail]:andrzejnovak/chillmax.git

and then go into the package and install it as editable

• . means install this folder
• -e means install as editable - meaning changes to the codebase will have an immediate effect

so:

cd chillmax
pip install -e .

Then you should be able to import chillmax like any other package like numpy

Use

Installation

• Repo is packaged with some scripts being available as a regular package - notably simulation
• To install (having a conda environment is recommended, ping me if you don't):
  • -e means editable, so your package will always run what's currently in the directory

pip install -e . 

Workflow

• We should externalize useful scripts into the package, while keeping the development in jupyter notebooks
• Ideally code should be validated by tests
  • Examples can be found in test/
  • test/test_base.py::test_boost shows how to call the boost factor prediciton from the original Analytical1D.py code (absorbed)
  • CI is set up, the test will run on github, whenever new stuff is pushed. I don't think we need to do this very strictly, but as a rule, no breaking changes should be introduced