Python_Setup_Pyenv_virtualenvwrapper.md

Python pyenv and virtualenvwrapper setup

This document will help you set up python for development on MacOS.

Install Pyenv

Pyenv allows you to have several python versions in your machine.

brew update
brew install zlib
brew install tcl-tk #otherwise we may have problems with tf package
brew install pyenv

And make sure you add the following to ~/.bash_profile or to ~/.zshrc

# For compilers to find zlib you may need to set:
export LDFLAGS="${LDFLAGS} -L/usr/local/opt/zlib/lib"
export CPPFLAGS="${CPPFLAGS} -I/usr/local/opt/zlib/include"

# For pkg-config to find zlib you may need to set:
export PKG_CONFIG_PATH="${PKG_CONFIG_PATH} /usr/local/opt/zlib/lib/pkgconfig"

Add also the lines to ~/.bashrc or ~/.zshrc

# Pyenv
if which pyenv > /dev/null; then eval "$(pyenv init -)"; fi
if which pyenv-virtualenv-init > /dev/null; then eval "$(pyenv virtualenv-init -)"; fi

This installation should build the folder $HOME/.pyenv. You can now list all the availiable python versions to download

pyenv install -l

Let's install for instance Python 3.1

pyenv install 3.1

Now your python interpreter of Python 3.1 is located in $HOME/.pyenv/versions/3.1/bin/python so you could execute a python script written in 3.1 as

~/.pyenv/versions/3.1/bin/python MyScriptIn31.py

You can set a global python interpreter for the whole system, for instance I choose the latest availiable as of now

pyenv install 3.7.2 #install required version
pyenv global 3.7.2 #make it availiable though all the system
source ~/.zshrc # to refresh the terminal by default would be source ~/.bash_profile

Now every time we type python in our shell, it will appear Python version 3.7.2. Thy this

python --version
which python

Now imagine we want to work on a specific python version (say 3.1) in a certain folder, we could go to the folder and type

pyenv local 3.1 

It creates a file named .python-version inside. Now if you check python --versionyou'll see that is 3.1 whereas gobally is 3.7.2.

Install pyenv virtualenvwrapper

Apart from the python version we need to create python environments each one specifying the versions for each package, we can use Virtualenvwrapper to do that. Virtual environments are useful becase we really don't want to have all the packages installed for all the different projects we are working on. Instead we isolate enviroments for each project.

On MacOs we run

brew install pyenv-virtualenvwrapper

And then we append to ~/.bash_profile or ~/.zshrc to

export PYENV_VIRTUALENVWRAPPER_PREFER_PYVENV="true"
export WORKON_HOME=~/.virtualenvs
pyenv virtualenvwrapper_lazy

Now all of our new environments will be created at ~/.virtualenvs. Some useful commands to work with virutualenvs

# create a new virtualenv called my_environment
mkvirtualenv my_environment

# disable virtualenv
deactivate

# list all environments
workon

# reactivate virtualenv
workon my_environment

# remove virtualenv
rmvirtualenv my_environment

In the first command you create a new environment called my_environment. You can find the interpreter in ~/.virtualenvs/my_environment/bin/python. If you use, for instance, PyCharm you only need to select this project interpreter in the preferences tab. In most of the projects, once you activate the environment you need to install all the requirements of the project (set in a file call requirements.txt) as pip install -r requirements.txt. Then you are set to develop your code!. Remember to add all the required packages in the file requirements.txt so that when you share the code with others I'll be almost plug and play.

Bonus: Run jupyter notebook using your environment

Let's say we already have a jupyter notebook on our system (mine is at /usr/local/bin/jupyter). Then, within your virtual envionment install ipykernel

pip install ipykernel
pip install ipython
pip install jupyter
pip install notebook

Yes, it would me more conveninent to add all this to your ´´´requirements.txt´´´ file.

And add the kernel to your jupyter

$(which python) -m ipykernel install --name=myenv
#ipython kernel install --user --name=myenv

You should be able to see the config running

cat /usr/local/share/jupyter/kernels/myenv/kernel.json

Now simply start a notebook and select your new kernel

jupyter notebook

To check which kernels do you have run

jupyter kernelspec list

And uninstall the recently created kernel as

jupyter kernelspec uninstall myenv

WARNING on Jupyter: If you want to add new packages doing pip install new_package then you should install again the kernel running the same command as before $(which python) -m ipykernel install --name=myenv when you are on myenv in the terminal.