• CheeseShopTutorial

Here you will know how to install packages from the Python Package Index (PyPI), how to get your own distribution listed and (optionally) hosted too.

Installing Distributions from the Python Package Index

There are several ways to get packages from PyPI into Python installation:

Distutils Installation

Download distribution, extract it, get to command prompt and type:

• $ python setup.py install

The distribution will be installed into 'site-packages' directory of the Python interpreter used to run the 'setup.py install' command.

Application scripts will go into:

• Linux: the 'bin' directory of the Python interpreter
• Windows: 'Scripts' subdirectory of Python installation path

Note that Distutils doesn't install any dependencies! If a distribution uses packages from another distribution, then you need to find out the name of this distribution and install it manually.

Quite often you may see the following line in install instructions:

• $ sudo python setup.py install

Be very careful when running this command! Packages and scripts installed with 'sudo' will be placed in a operating system specific location, and this may break your operating system. Furthermore, the *distutils does not have an uninstall command*, so if you do break anything, it may be not that easy to restore it back to working condition!

There are two ways to avoid having to install as the root user:

1. Per-user site-packages

   • If using Python 2.6 or greater, a per-user 'site-packages' directory is available. This may interfere with applications supplied by your operating system, but this interference is limited to only the user account which was used to install these packages with. Other user accounts (including the root user) should be left untouched.

2. VirtualEnv

   • The VirtualEnv tool allows you to clone your system Python, so that you have a complete, isolated copy of Python available for installing into. Typically you will install the 'virtualenv' distribution as root, but it is possible to extract the virtualenv.py module from the 'virtualenv' distribution and also create virtual environments without having to use the root account at all.

Easy Installation

EasyInstall (easy_install) gives you a quick and painless way to install packages remotely by connecting to the Package Index or even other websites via HTTP. It is somewhat analogous to the CPAN and PEAR tools for Perl and PHP, respectively. Please see EasyInstall for more information.

Another alternative is pip, which allows for easy uninstalling as well as installing of packages.

Buildout Installation

Buildout is a configuration-driven build tool designed for automating application installation. It has a number of recipes that allow you to declare the working set(s) of packages required for a Python application, and automatically install them.

Please see Buildout for more information.

What the Package Index Stores

The Python Package Index stores information about packages of Python software. Each package has a name and a number of release versions. The list of release versions will increase as newer versions of the package are submitted to the Package Index.

Submitting Packages to the Package Index

If you have some Python modules or packages that you would like to share with the Python community, we'd love to have them included in the Python Package Index! First, if you haven't done so, you will want to get your project organized. You might follow the guidelines at ProjectFileAndDirectoryLayout. After that, you'll want to read the Python documentation regarding creating distributions: http://docs.python.org/distutils/index.html.

There is also an example project http://pypi.python.org/pypi/an_example_pypi_project

Your Options

There are two types of information that may be submitted to the Package Index:

1. package meta-data (name, version, description, etc), and
2. package source and binary distribution files.

Package Meta-Data

You may submit package meta-data either by:

1. writing a setup.py file and using "python setup.py register" (see docs),

2. creating a PKG-INFO file (typically generated from a setup.py file) and uploading it, or

3. using the web form and manually entering the information.

The Package Index assumes that the PKG-INFO file is either ASCII or UTF-8.

Selecting Classifiers

View the complete list of classifiers you may choose from.

See the meta-data docs for details about how to include them in your setup.py file.

PyPI will attempt to parse the "long_description" from your meta-data as ReStructuredText. If this fails, it will be presented to users as plain text (all whitespace and formatting retained).

Missing Classifier?

Is a classifier you need missing from the classifiers list?

Python frameworks with plugins or packages that target the framework can get their own category. The category should only be added after such packages exist.

Complementary packages can link to each other from their descriptions, they do not need a category to link them together; only when packages are provided by different people does a category need to be created.

To ask for a category email catalog-sig@python.org .

Package Distribution Files

Note that submitting a package distribution file automatically submits the package's meta-data. You may submit package distribution files either by:

1. appending the "upload" command to a setup.py source dist or built dist command, eg "python setup.py sdist upload", or

2. logging into the Package Index and using the package management interface to manually upload files.

The "upload" command has a number of options, including being able to sign the upload using GPG. See "python setup.py upload --help" for more information.

Exposing Multiple Releases

When a new release of a package is submitted to the Package Index, all previous releases of that package are hidden from the display. This means that listings and searches will no longer find those releases.

You may use the package admin interface to un-hide releases. This may be useful if you have both a stable and a development release active at the same time. These will be hidden again on the next submission of meta-data.

Cheesecake scores

Cheesecake is a project that tries to help Python developers via testing and scoring their packages based on some empirical factors, like, for example, whether a package can be installed using standard "python setup.py install" interface. To learn more about it, visit its homepage.

PyPI packages are no longer automatically scored.