The easiest way to install Python and Jupyter Notebook is probably with Anaconda.

Anaconda is a free and open source ditribution of Python and R programming languages particularly oriented for applications in data science. One of the major assets of Anaconda is its simplification in the management of packages and their dependencies.

Contenu de la page

  • Working environment installation on Microsoft Windows
    • Installing Python with Anaconda
    • Creating a first Jupiter Notebook
    • Installing Python packages
      • Package installation with conda (preferred)
      • Package installation with pip
      • Manually install a package (as a last resort)
    • Compose with corporate network constraints
      • Install a Python package with pip without Internet access
      • Troubleshoot Anaconda proxy issues

Pour ceux bien décidés à utiliser R en lieu et place de Python : Documentation Anaconda détaillant l’installation de R pour le Jupyter Notebook

Working environment installation on Microsoft Windows

Installing Python with Anaconda

Download the version of Anaconda supporting the latest version of Python 3: https://www.anaconda.com/download/#windows

Launch and follow the instructions of the installation executable.

Once the installation is complete, launch the anaconda console via Start menu > Anaconda3 > Anaconda Prompt.

Create a Python 3.7 conda environment. (for more information about this consult the documentation on managing environments)

conda create --name py37 python=3.7
A conda environment is a directory containing a specific collection of packages that you have decided to install. Because some packages do not support the latest version of Python, it may be appropriate to maintain an environment for a specific version of Python.

The ability to create an environment with Anaconda for a package caused me some problems, so I ended up using the Python 3.7 environment installed by default.

To activate your environment within the Anaconda console, enter:

activate py37

To leave your environment, simply enter:

deactivate

Check your version of Python in this environment:

python --version

If you have a Python 3.7 environment installed, it should be returned to you: Python 3.7.0


Create a first Jupiter Notebook

The Jupyter Notebook is an interactive web tool that allows you to create and share your data handling and analysis scripts with your explanations and numerical or visual results.

In the Anaconda console, navigate to the directory of your choice, for example: 
cd ./Documents/Python_notebooks

Launch a Jupiter notebook
Jupyter Notebook

Create a new notebook by clicking New > Python 3

In the empty dialog box, enter:

import os

curDir = os.getcwd()
print(curDir)

Run this cell either by pressing Shift+Enter or clicking Run

Its execution should return the current location of your notebook, something similar to:

C:/Users/xxxxx/Documents/Python_notebooks


Installing Python packages

Package installation with conda (preferred)

Say you want to install the Jupyter notebook extensions from the following GitHub directory: https://github.com/ipython-contrib/jupyter_contrib_nbextensions

You can do this very simply now that you have Anaconda, with the following command:

conda install -c conda-forge jupyter_contrib_nbextensions

The installable packages via conda are listed within https://anaconda.org/

With this package just installed, you will be able to activate an autocompletion wizard. To do this, access your new configuration page: http://localhost:8888/nbextensions?nbextensions=codefolding/main

And activate the Hinterland extension by checking the associated box.

You will now see in your notebook an autocompletion assistant during your keyboard entries.

Package installation with pip

If the package you are looking for is not part of Anaconda-managed packages, or you want to take advantage of the latest version of a package not yet supported by Anaconda, you can always try to find it on the Python Package Index Aka PyPi: https://pypi.org/

Package installation with pip takes the following form

pip install nom_du_package

You will potentially encounter error messages that indicate missing prerequisite packages. Yes … Anaconda takes care of everything, with pip it’s up to you to do the job.

Manually install a package (as a last resort)

As a last resort, if you need a package not available conda or pip, you can still install it manually.

  • Download the package source files
  • Unzip them if necessary
  • If there are any installation instructions (usually found within INSTALL.md or READ.md) read them and follow them OTHERWISE
  • Navigate through the Anaconda console in its directory containing the setup.py file
  • And just type and run 
    python setup.py install


Compose with corporate network constraints

Say you want to install the seaborn package.

Troubleshoot Anaconda proxy issues

When trying to install a Python package via the command-line tool conda, you might find yourself stuck with the following message

conda install -c conda-forge seaborn
Fetching package metadata:

To resolve this problem, contact your network administrator and request details about the enterprise proxy.

With this information you will edit a .condarc configuration file within your user folder (C:\Users\votre_repertoire_utilisateur).

[conda.io] The .condarc, configuration file is an optional configuration file that allows advanced users to configure different aspects of conda, such as package discovery channels, proxy configuration, and environment directories.

The .condarc file is not included by default in the Anaconda installation, but it is automatically created within your home directory (C:\Users\Username) the first time you run the command: conda config

At your command prompt, create the .condarc file by running the command

conda config

Edit your .condarc file with the information provided by the network administrator.

  • with identification information:
proxy_servers:
  http: http://user:pass@corp.com:8080
  https: https://user:pass@corp.com:8080

ssl_verify: False
  • without identification information:
proxy_servers:
  http: http://corp.com:8080
  https: https://corp.com:8080

ssl_verify: False

Install a Python package with pip without Internet access

From a machine where you have access to the Internet, download from https://pypi.org or any other relevant source the .whl file of the package you want to install.

The .whl file for seaborn can be downloaded from https://pypi.org/project/seaborn/#files

Transfer this .whl file to the machine where it is to be installed, the one that does not have Internet access or restricted access, and run the following command (replacing the name of the .whl file with the one you just get):

pip install seaborn-0.9.0-py3-none-any.whl

You will probably face some warnings in red with messages of the form:

xxxxxx xx.x.x requires xxxxxx>=xx.x.x, which is not installed.

This message indicates that you must first follow this process for each missing Python package before you can install the package that you originally wanted.

An alternative to using a .whl file is installing manually the previously processed package.

02 - Bibliothèques Python à privilégier fast_forward