NumPy Introduction

From WikiOD

Remarks[edit | edit source]

NumPy (pronounced “numb pie” or sometimes “numb pea”) is an extension to the Python programming language that adds support for large, multi-dimensional arrays, along with an extensive library of high-level mathematical functions to operate on these arrays.

Versions[edit | edit source]

Version Release Date
1.3.0 2009-03-20
1.4.0 2010-07-21
1.5.0 2010-11-18
1.6.0 2011-05-15
1.6.1 2011-07-24
1.6.2 2012-05-20
1.7.0 2013-02-12
1.7.1 2013-04-07
1.7.2 2013-12-31
1.8.0 2013-11-10
1.8.1 2014-03-26
1.8.2 2014-08-09
1.9.0 2014-09-07
1.9.1 2014-11-02
1.9.2 2015-03-01
1.10.0 2015-10-07
1.10.1 2015-10-12
1.10.2 2015-12-14
1.10.4* 2016-01-07
1.11.0 2016-05-29

Basic Import[edit | edit source]

Import the numpy module to use any part of it.

import numpy as np

Most examples will use np as shorthand for numpy. Assume "np" means "numpy" in code examples.

x = np.array([1,2,3,4])

Installation on Linux[edit | edit source]

NumPy is available in the default repositories of most popular Linux distributions and can be installed in the same way that packages in a Linux distribution are usually installed.

Some Linux distributions have different NumPy packages for Python 2.x and Python 3.x. In Ubuntu and Debian, install numpy at the system level using the APT package manager:

sudo apt-get install python-numpy  
sudo apt-get install python3-numpy

For other distributions, use their package managers, like zypper (Suse), yum (Fedora) etc.

numpy can also be installed with Python's package manager pip for Python 2 and with pip3 for Python 3:

pip install numpy  # install numpy for Python 2
pip3 install numpy  # install numpy for Python 3

pip is available in the default repositories of most popular Linux distributions and can be installed for Python 2 and Python 3 using:

sudo apt-get install python-pip  # pip for Python 2
sudo apt-get install python3-pip  # pip for Python 3

After installation, use pip for Python 2 and pip3 for Python 3 to use pip for installing Python packages. But note that you might need to install many dependencies, which are required to build numpy from source (including development-packages, compilers, fortran etc).

Besides installing numpy at the system level, it is also common (perhaps even highly recommended) to install numpy in virtual environments using popular Python packages such as virtualenv. In Ubuntu, virtualenv can be installed using:

sudo apt-get install virtualenv

Then, create and activate a virtualenv for either Python 2 or Python 3 and then use pip to install numpy:

virtualenv venv  # create virtualenv named venv for Python 2
virtualenv venv -p python3  # create virtualenv named venv for Python 3
source venv/bin/activate  # activate virtualenv named venv
pip install numpy  # use pip for Python 2 and Python 3; do not use pip3 for Python3

Installation on Windows[edit | edit source]

Numpy installation through pypi (the default package index used by pip) generally fails on Windows computers. The easiest way to install on Windows is by using precompiled binaries.

One source for precompiled wheels of many packages is Christopher Gohkle's site. Choose a version according to your Python version and system. An example for Python 3.5 on a 64 bit system:

  1. Download numpy-1.11.1+mkl-cp35-cp35m-win_amd64.whl from here
  2. Open a Windows terminal (cmd or powershell)
  3. Type the command pip install C:\path_to_download\numpy-1.11.1+mkl-cp35-cp35m-win_amd64.whl

If you don't want to mess around with single packages, you can use the Winpython distribution which bundles most packages together and provides a confined environment to work with. Similarly, the Anaconda Python distrubution comes pre-installed with numpy and numerous other common packages.

Another popular source is the conda package manager, which also supports virtual environments.

  1. Download and install conda.
  2. Open a Windows terminal.
  3. Type the command conda install numpy

Installation on Mac[edit | edit source]

The easiest way to set up NumPy on Mac is with pip

pip install numpy

Installation using Conda.

Conda available for Windows, Mac, and Linux

  1. Install Conda. There are two ways to install Conda, either with Anaconda (Full package, include numpy) or Miniconda (only Conda,Python, and the packages they depend on, without any additional package). Both Anaconda & Miniconda install the same Conda.
  2. Additional command for Miniconda, type the command conda install numpy

Temporary Jupyter Notebook hosted by Rackspace[edit | edit source]

Jupyter Notebooks are an interactive, browser-based development environment. They were originally developed to run computation python and as such play very well with numpy. To try numpy in a Jupyter notebook without fully installing either on one's local system Rackspace provides free temporary notebooks at

Note: that this is not a proprietary service with any sort of upsells. Jupyter is a wholly open-sourced technology developed by UC Berkeley and Cal Poly San Luis Obispo. Rackspace donates this service as part of the development process.

To try numpy at

  1. visit
  2. either select Welcome to Python.ipynb or
  3. New >> Python 2 or
  4. New >> Python 3