Skip to content

pip.md

Latest commit

 

History

History
164 lines (119 loc) · 6.46 KB

pip.md

File metadata and controls

164 lines (119 loc) · 6.46 KB

Pip (Pip Installs Packages)

Joseph Vantassel, The University of Texas at Austin

License

One-Minute Summary

  • pip is a command line tool for installing Python packages.
  • Basic syntax:
pip -h                           # Print help.
pip install <package>            # Install package (e.g., pip install numpy).
pip uninstall <package>          # Remove unwanted package.
pip list                         # View all currently installed packages.
  • Complex syntax:
pip install -e <path-to-setup.py>              # An "editable" install.
pip install <*.tar.gz>                         # Install from source distribution.
pip install <*.whl>                            # Install from build distribution.
pip install git+ssh://git@rmturl/usrnm/pkg.git # Install directly from GitHub.

What is pip

Pip is a command-line tool for managing Python packages. Packages are groupings of Python (and in certain cases other languages such as C or C++) codes to perform a particular task. Popular packages in science and engineering include: numpy, scipy, pandas, and matplotlib.

How Does it Work

When you use pip to install a package, it queries the Python Package Index (PyPI) for the package name. If it finds a matching package it downloads, installs, and appends it to your list of available Python packages. If pip cannot find a matching package it will return a message stating that a package by that name could not be found.

Getting Started with pip

  1. Install a modern Python interpreter if you have not done so previously.

  2. Open a terminal, either Windows Powershell (recommended) or Command Prompt on a Windows machine.

  3. Use the terminal to follow the tutorial below.

Pip Tutorial

Installing a Package

Get started with pip by installing a new package. We will use numpy as an example.

pip list                 # View the currently installed packages.
pip install numpy        # Install numpy.
pip list                 # Note the addition of numpy and its version.

Note: If you already had numpy installed, this example was probably not that illustrative as you likely received a message stating with Requirement already satisfied: .... So to better understand the process try the above procedure for a different package that you have not yet installed. Some useful packages include: scipy, pandas, and matplotlib.

Uninstall a Package

The syntax for uninstalling a package is nearly identical to installing one. We will again use numpy as an example.

pip list                 # Note numpy.
pip uninstall numpy      # Uninstall numpy. Confirm file removal if required.
pip list                 # Note numpy has been removed from the list.

If you would like to keep numpy installed for future use, simply repeat the installation process.

Installing a Python Package from Non-Traditional Sources

pip is generally used to install from PyPI, however this may not always be the case.

Creating an Editable Install

If you are developing a Python package for either private use or public distribution on PyPI, it is extremely useful to have a dynamically updated version of your package accessible on your machine. The syntax for a dynamic installation (more commonly referred to as an editable install) is simple.

pip uninstall <mypkg>             # Uninstall previous version if it exists.
pip install -e <path-to-setup.py> # Install an editable version.

After performing the editable installation, anytime you edit the source of your package, the edits will be instantly made available. This removes the need to uninstall and reinstall your package to perform testing or begin production.

Installing from a source distribution

If a package is private or simply not yet ready to be released on PyPI the author may release a non-platform specific source distribution (essentially a compressed tarball of all the necessary files to install the package). Fortunately pip makes handling source distributions simple.

pip list                        # View currently installed packages.
pip install <*.tar.gz>          # Install from source distribution.
pip list                        # See the new package installed from source.

Installing from a built distribution

A built distribution may be used in similar cases as a source distribution, however built distributions are platform specific, though they, in return, enlist shorter installation times. Regardless the pip syntax is similar in both cases.

pip list                        # View currently installed packages.
pip install <*.whl>             # Install from build distribution.
pip list                        # View the new package installed from built wheel.

Installing from Github Repository

For small open-source projects the author may not want to go through the trouble of submitting their project to PyPI and so may only provide their code on GitHub Note: Other version control systems (VCS) are supported however are not discussed here, see the pip documentation for details.

The exact syntax for the GitHub path can be tricky to get correct, but it will follow the format described below if using ssh.

pip list                        # View currently installed packages.
pip install git+ssh://git@rmturl/usrnm/pkg.git # Install directly from GitHub.
pip list                        # View the new package installed from GitHub.

Note this method of installation will not work with any GitHub repository with Python code. First, the GitHub repository must be public or you must have been granted access. Second, the author(s) of the repository must have setup the code to be "package ready" (i.e., it contains at a minimum a setup.py with the necessary installation instructions for pip). If it is not clear from the provided documentation if a package is pip install ready, ask the author and express your interest in the package, but understand that it being pip install ready is rather unlikely.

Sources

Pip Documentation

Pip and git

Stack Overflow: git-clone works pip install does not