Setting up your laptop for Scientific Computing with Python

General Advice on Operating System

Generally, you should use either a modern Linux or *nix distribution. For most users, we recommend Ubuntu. If you have a Mac, macOS is a custom flavor of *nix OS. Please do not try to install other systems such as Ubuntu or Arch Linux on a Mac if you don’t know what you are doing.

You should avoid using Windows as much as possible. For this module, we can only offer limited help if you are using Windows.

Which Python Distribution Should I Use?

If you are a newbie to Python, you would go to the official Python website, download the installation file and install it on your computer. In fact, this may be the worst thing you can do.

So, in most modern Linux and *nix operating systems, the Python distribution is installed for managing some system services by default. The system-installed Python can be used by the user for sure. However, because this distribution manages system services and needs constant writing privileges to some directories, we strongly recommend that you do not touch the system-installed Python as a new user. Moreover, you could encounter big problems when you are trying to install packages that build from scratch.

Instead, what we need is a Python distribution that has its own environment and can be easily removed when we want to. So the answer is Miniconda, a Python distribution that is built for Scientific Computing.

Miniconda setup instructions

  1. Open a terminal and download Miniconda

     $ wget https://repo.anaconda.com/miniconda/Miniconda3-4.5.4-Linux-x86_64.sh -O miniconda.sh  # for Linux

     $ curl https://repo.anaconda.com/miniconda/Miniconda3-4.5.4-MacOSX-x86_64.sh -o miniconda.sh  # for macOS

  2. Install Miniconda

     $ bash ./miniconda.sh

    Follow the instructions and make sure Miniconda is added in your bash configuration file such as .bashrc or .zshrc.

  3. Close the current terminal and open another one (so that the bash configuration is loaded again). Type python, you should see something similar to this:

     Python 3.6.7 |Anaconda, Inc.| (default, Oct 23 2018, 19:16:44) 
 [GCC 7.3.0] on linux
 Type "help", "copyright", "credits" or "license" for more information.
 >>>

Remark: For Windows OS, please read the installation instruction here. Or, remove your Windows and install Linux.

Which Python Version Should I Use?

Python 3.6

Remark: Python 3.6 has gained full support from the scientific community. Python 2.7 was the default choice of the module. However, as Python 2’s support will be ended soon, we choose Python 3.5/3.6 as the default Python version for this edition.

Remark: Do not use Python 3.7 for now as many softwares don’t have support for this version yet.

How to Install Other Python Packages?

Generally, with Miniconda, we have three ways of installing other Python packages.

  1. Use pip. pip is the official Python packaging system that manages package installation, uninstallation, version control, custom package building, etc. You can install additional Python packages by

     $ pip install some-package-name

    If the package is available in PyPi, pip will automatically pull the software from the website and install it.

  2. Use conda. Miniconda uses the packaging system conda to manage packages and libraries installation. At heart, conda does package, dependency and environment management for any language. conda can pull and install a pre-built package from a specific server and resolve the dependency accordingly.

     $ conda install some-package-name

  3. Use setup.py. A decent Python library usually has a setup.py script. With this script, you can install the package via

     $ python setup.py install

Do I Need Anything Else?

Miniconda offers a minimum setup for Python. You will need to install some extra packages for running experiments in this module. Simply run these commands sequentially

$ pip install numpy
$ pip install scipy
$ pip install scikit-learn
$ pip install scikit-image
$ pip install opencv-contrib-python
$ pip install matplotlib
$ pip install h5py
$ pip install tensorflow

You should familiarize yourself with the software packaging system on your computer. For Debian family (including Ubuntu), that is apt-get or apt. For macOS, that is either MacPorts or homebrew (strongly recommended).

IDE (Integrated Development Environment)

There are many different IDEs for Python on the market. Here, we list some of them that generally deliver fantastic coding experience.

Say NO to Jupyter Notebook

If you have taken a Python class or Python 101, you have probably seen that instructors like Jupyter Notebook as a default presentation tool. They even walk you through the steps so that you are comfortable with this kind of coding style.

However, we are strongly against this paradigm of coding. It is true that the Jupyter Notebook offers a nice presentation style. But it is not designed for managing and engineering serious projects.

Therefore, we advise that you do not use Jupyter Notebook.

What do you need from an IDE?

Atom

Atom is a very popular text editor that is built and maintained mainly by GitHub. This editor is modern and naturally integrates some of the best software engineering practices. You can find tons of additional packages that help you configure the editor to a Python IDE.

PyCharm

PyCharm is a new Python IDE that has a beautiful interface and integrates all the features you will need for developing a Python project.

Eclipse+PyDev

If you are a Eclipse user, perhaps you would be happy to know that there is a dedicate Eclipse plugin for Python. PyDev takes advantages of the powerful Eclipse compiling and debugging framework.

Vim

For the experienced user, we recommend Vim or its community-driven fork neovim. By configuring Vim, you will be able to get a fast editor that has all the IDE features you want while being lightweight in some sense. Additionally, Vim is a very flexible editor which allows you to efficiently do more things compared to other editors.

Further Readings