|
| setuptools |
UserPreferences |
| The PEAK Developers' Center | FrontPage | RecentChanges | TitleIndex | WordIndex | SiteNavigation | HelpContents |
setuptools is a collection of enhancements to the Python distutils (for Python 2.3.5 and up on most platforms; 64-bit platforms require a minimum of Python 2.4) that allow you to more easily build and distribute Python packages, especially ones that have dependencies on other packages.
Packages built and distributed using setuptools look to the user like ordinary Python packages based on the distutils. Your users don't need to install or even know about setuptools in order to use them, and you don't have to include the entire setuptools package in your distributions. By including just a single bootstrap module (an 8K .py file), your package will automatically download and install setuptools if the user is building your package from source and doesn't have a suitable version already installed.
Feature Highlights:
In addition to the PyPI downloads, the development version of setuptools is available from the Python SVN sandbox, and in-development versions of the 0.6 branch are available as well.
Please follow the EasyInstall Installation Instructions to install the current stable version of setuptools. In particular, be sure to read the section on Custom Installation Locations if you are installing anywhere other than Python's site-packages directory.
If you want the current in-development version of setuptools, you should first install a stable version, and then run:
ez_setup.py setuptools==dev
This will download and install the latest development (i.e. unstable) version of setuptools from the Python Subversion sandbox.
For basic use of setuptools, just import things from setuptools instead of the distutils. Here's a minimal setup script using setuptools:
from setuptools import setup, find_packages
setup(
name = "HelloWorld",
version = "0.1",
packages = find_packages(),
)
As you can see, it doesn't take much to use setuptools in a project. Just by doing the above, this project will be able to produce eggs, upload to PyPI, and automatically include all packages in the directory where the setup.py lives. See the Command Reference section below to see what commands you can give to this setup script.
Of course, before you release your project to PyPI, you'll want to add a bit more information to your setup script to help people find or learn about your project. And maybe your project will have grown by then to include a few dependencies, and perhaps some data files and scripts:
from setuptools import setup, find_packages
setup(
name = "HelloWorld",
version = "0.1",
packages = find_packages(),
scripts = ['say_hello.py'],
# Project uses reStructuredText, so ensure that the docutils get
# installed or upgraded on the target machine
install_requires = ['docutils>=0.3'],
package_data = {
# If any package contains *.txt or *.rst files, include them:
'': ['*.txt', '*.rst'],
# And include any *.msg files found in the 'hello' package, too:
'hello': ['*.msg'],
}
# metadata for upload to PyPI
author = "Me",
author_email = "me@example.com",
description = "This is an Example Package",
license = "PSF",
keywords = "hello world example examples",
url = "http://example.com/HelloWorld/", # project home page, if any
# could also include long_description, download_url, classifiers, etc.
)
In the sections that follow, we'll explain what most of these setup() arguments do (except for the metadata ones), and the various ways you might use them in your own project(s).
Setuptools can work well with most versioning schemes; there are, however, a few special things to watch out for, in order to ensure that setuptools and EasyInstall can always tell what version of your package is newer than another version. Knowing these things will also help you correctly specify what versions of other projects your project depends on.
A version consists of an alternating series of release numbers and pre-release or post-release tags. A release number is a series of digits punctuated by dots, such as 2.4 or 0.5. Each series of digits is treated numerically, so releases 2.1 and 2.1.0 are different ways to spell the same release number, denoting the first subrelease of release 2. But 2.10 is the tenth subrelease of release 2, and so is a different and newer release from 2.1 or 2.1.0. Leading zeros within a series of digits are also ignored, so 2.01 is the same as 2.1, and different from 2.0.1.
Following a release number, you can have either a pre-release or post-release tag. Pre-release tags make a version be considered older than the version they are appended to. So, revision 2.4 is newer than revision 2.4c1, which in turn is newer than 2.4b1 or 2.4a1. Postrelease tags make a version be considered newer than the version they are appended to. So, revisions like 2.4-1 and 2.4pl3 are newer than 2.4, but are older than 2.4.1 (which has a higher release number).
A pre-release tag is a series of letters that are alphabetically before "final". Some examples of prerelease tags would include alpha, beta, a, c, dev, and so on. You do not have to place a dot before the prerelease tag if it's immediately after a number, but it's okay to do so if you prefer. Thus, 2.4c1 and 2.4.c1 both represent release candidate 1 of version 2.4, and are treated as identical by setuptools.
In addition, there are three special prerelease tags that are treated as if they were the letter c: pre, preview, and rc. So, version 2.4rc1, 2.4pre1 and 2.4preview1 are all the exact same version as 2.4c1, and are treated as identical by setuptools.
A post-release tag is either a series of letters that are alphabetically greater than or equal to "final", or a dash (-). Post-release tags are generally used to separate patch numbers, port numbers, build numbers, revision numbers, or date stamps from the release number. For example, the version 2.4-r1263 might denote Subversion revision 1263 of a post-release patch of version 2.4. Or you might use 2.4-20051127 to denote a date-stamped post-release.
Notice that after each pre or post-release tag, you are free to place another release number, followed again by more pre- or post-release tags. For example, 0.6a9.dev-r41475 could denote Subversion revision 41475 of the in- development version of the ninth alpha of release 0.6. Notice that dev is a pre-release tag, so this version is a lower version number than 0.6a9, which would be the actual ninth alpha of release 0.6. But the -r41475 is a post-release tag, so this version is newer than 0.6a9.dev.
For the most part, setuptools' interpretation of version numbers is intuitive, but here are a few tips that will keep you out of trouble in the corner cases:
Don't use - or any other character than . as a separator, unless you really want a post-release. Remember that 2.1-rc2 means you've already released 2.1, whereas 2.1rc2 and 2.1.c2 are candidates you're putting out before 2.1. If you accidentally distribute copies of a post-release that you meant to be a pre-release, the only safe fix is to bump your main release number (e.g. to 2.1.1) and re-release the project.
Don't stick adjoining pre-release tags together without a dot or number between them. Version 1.9adev is the adev prerelease of 1.9, not a development pre-release of 1.9a. Use .dev instead, as in 1.9a.dev, or separate the prerelease tags with a number, as in 1.9a0dev. 1.9a.dev, 1.9a0dev, and even 1.9.a.dev are identical versions from setuptools' point of view, so you can use whatever scheme you prefer.
If you want to be certain that your chosen numbering scheme works the way you think it will, you can use the pkg_resources.parse_version() function to compare different version numbers:
>>> from pkg_resources import parse_version
>>> parse_version('1.9.a.dev') == parse_version('1.9a0dev')
True
>>> parse_version('2.1-rc2') < parse_version('2.1')
False
>>> parse_version('0.6a9dev-r41475') < parse_version('0.6a9')
True
Once you've decided on a version numbering scheme for your project, you can have setuptools automatically tag your in-development releases with various pre- or post-release tags. See the following sections for more details:
The following keyword arguments to setup() are added or changed by setuptools. All of them are optional; you do not have to supply them unless you need the associated setuptools feature.
A string or list of strings specifying what other distributions need to be present in order for the setup script to run. setuptools will attempt to obtain these (even going so far as to download them using EasyInstall) before processing the rest of the setup script or commands. This argument is needed if you are using distutils extensions as part of your build process; for example, extensions that process setup() arguments and turn them into EGG-INFO metadata files.
(Note: projects listed in setup_requires will NOT be automatically installed on the system where the setup script is being run. They are simply downloaded to the setup directory if they're not locally available already. If you want them to be installed, as well as being available when the setup script is run, you should add them to install_requires and setup_requires.)
A string naming a unittest.TestCase subclass (or a package or module containing one or more of them, or a method of such a subclass), or naming a function that can be called with no arguments and returns a unittest.TestSuite. If the named suite is a module, and the module has an additional_tests() function, it is called and the results are added to the tests to be run. If the named suite is a package, any submodules and subpackages are recursively added to the overall test suite.
Specifying this argument enables use of the test command to run the specified test suite, e.g. via setup.py test. See the section on the test command below for more details.
If you would like to use a different way of finding tests to run than what setuptools normally uses, you can specify a module name and class name in this argument. The named class must be instantiable with no arguments, and its instances must support the loadTestsFromNames() method as defined in the Python unittest module's TestLoader class. Setuptools will pass only one test "name" in the names argument: the value supplied for the test_suite argument. The loader you specify may interpret this string in any way it likes, as there are no restrictions on what may be contained in a test_suite string.
The module name and class name must be separated by a :. The default value of this argument is "setuptools.command.test:ScanningLoader". If you want to use the default unittest behavior, you can specify "unittest:TestLoader" as your test_loader argument instead. This will prevent automatic scanning of submodules and subpackages.
The module and class you specify here may be contained in another package, as long as you use the tests_require option to ensure that the package containing the loader class is available when the test command is run.
A list of strings naming resources that should be extracted together, if any of them is needed, or if any C extensions included in the project are imported. This argument is only useful if the project will be installed as a zipfile, and there is a need to have all of the listed resources be extracted to the filesystem as a unit. Resources listed here should be '/'-separated paths, relative to the source root, so to list a resource foo.png in package bar.baz, you would include the string bar/baz/foo.png in this argument.
If you only need to obtain resources one at a time, or you don't have any C extensions that access other files in the project (such as data files or shared libraries), you probably do NOT need this argument and shouldn't mess with it. For more details on how this argument works, see the section below on Automatic Resource Extraction.
For simple projects, it's usually easy enough to manually add packages to the packages argument of setup(). However, for very large projects (Twisted, PEAK, Zope, Chandler, etc.), it can be a big burden to keep the package list updated. That's what setuptools.find_packages() is for.
find_packages() takes a source directory, and a list of package names or patterns to exclude. If omitted, the source directory defaults to the same directory as the setup script. Some projects use a src or lib directory as the root of their source tree, and those projects would of course use "src" or "lib" as the first argument to find_packages(). (And such projects also need something like package_dir = {'':'src'} in their setup() arguments, but that's just a normal distutils thing.)
Anyway, find_packages() walks the target directory, and finds Python packages by looking for __init__.py files. It then filters the list of packages using the exclusion patterns.
Exclusion patterns are package names, optionally including wildcards. For example, find_packages(exclude=["*.tests"]) will exclude all packages whose last name part is tests. Or, find_packages(exclude=["*.tests", "*.tests.*"]) will also exclude any subpackages of packages named tests, but it still won't exclude a top-level tests package or the children thereof. In fact, if you really want no tests packages at all, you'll need something like this:
find_packages(exclude=["*.tests", "*.tests.*", "tests.*", "tests"])
in order to cover all the bases. Really, the exclusion patterns are intended to cover simpler use cases than this, like excluding a single, specified package and its subpackages.
Regardless of the target directory or exclusions, the find_packages() function returns a list of package names suitable for use as the packages argument to setup(), and so is usually the easiest way to set that argument in your setup script. Especially since it frees you from having to remember to modify your setup script whenever your project grows additional top-level packages or subpackages.
Packaging and installing scripts can be a bit awkward with the distutils. For one thing, there's no easy way to have a script's filename match local conventions on both Windows and POSIX platforms. For another, you often have to create a separate file just for the "main" script, when your actual "main" is a function in a module somewhere. And even in Python 2.4, using the -m option only works for actual .py files that aren't installed in a package.
setuptools fixes all of these problems by automatically generating scripts for you with the correct extension, and on Windows it will even create an .exe file so that users don't have to change their PATHEXT settings. The way to use this feature is to define "entry points" in your setup script that indicate what function the generated script should import and run. For example, to create two console scripts called foo and bar, and a GUI script called baz, you might do something like this:
setup(
# other arguments here...
entry_points = {
'console_scripts': [
'foo = my_package.some_module:main_func',
'bar = other_module:some_func',
],
'gui_scripts': [
'baz = my_package_gui.start_func',
]
}
)
When this project is installed on non-Windows platforms (using "setup.py install", "setup.py develop", or by using EasyInstall), a set of foo, bar, and baz scripts will be installed that import main_func and some_func from the specified modules. The functions you specify are called with no arguments, and their return value is passed to sys.exit(), so you can return an errorlevel or message to print to stderr.
On Windows, a set of foo.exe, bar.exe, and baz.exe launchers are created, alongside a set of foo.py, bar.py, and baz.pyw files. The .exe wrappers find and execute the right version of Python to run the .py or .pyw file.
You may define as many "console script" and "gui script" entry points as you like, and each one can optionally specify "extras" that it depends on, that will be added to sys.path when the script is run. For more information on "extras", see the section below on Declaring Extras. For more information on "entry points" in general, see the section below on Dynamic Discovery of Services and Plugins.
Occasionally, there are situations where it's desirable to make an .egg file directly executable. You can do this by including an entry point such as the following:
setup(
# other arguments here...
entry_points = {
'setuptools.installation': [
'eggsecutable = my_package.some_module:main_func',
]
}
)
Any eggs built from the above setup script will include a short excecutable prelude that imports and calls main_func() from my_package.some_module. The prelude can be run on Unix-like platforms (including Mac and Linux) by invoking the egg with /bin/sh, or by enabling execute permissions on the .egg file. For the executable prelude to run, the appropriate version of Python must be available via the PATH environment variable, under its "long" name. That is, if the egg is built for Python 2.3, there must be a python2.3 executable present in a directory on PATH.
This feature is primarily intended to support bootstrapping the installation of setuptools itself on non-Windows platforms, but may also be useful for other projects as well.
IMPORTANT NOTE: Eggs with an "eggsecutable" header cannot be renamed, or invoked via symlinks. They must be invoked using their original filename, in order to ensure that, once running, pkg_resources will know what project and version is in use. The header script will check this and exit with an error if the .egg file has been renamed or is invoked via a symlink that changes its base name.
setuptools supports automatically installing dependencies when a package is installed, and including information about dependencies in Python Eggs (so that package management tools like EasyInstall can use the information).
setuptools and pkg_resources use a common syntax for specifying a project's required dependencies. This syntax consists of a project's PyPI name, optionally followed by a comma-separated list of "extras" in square brackets, optionally followed by a comma-separated list of version specifiers. A version specifier is one of the operators <, >, <=, >=, == or !=, followed by a version identifier. Tokens may be separated by whitespace, but any whitespace or nonstandard characters within a project name or version identifier must be replaced with -.
Version specifiers for a given project are internally sorted into ascending version order, and used to establish what ranges of versions are acceptable. Adjacent redundant conditions are also consolidated (e.g. ">1, >2" becomes ">1", and "<2,<3" becomes "<3"). "!=" versions are excised from the ranges they fall within. A project's version is then checked for membership in the resulting ranges. (Note that providing conflicting conditions for the same version (e.g. "<2,>=2" or "==2,!=2") is meaningless and may therefore produce bizarre results.)
Here are some example requirement specifiers:
docutils >= 0.3
# comment lines and \ continuations are allowed in requirement strings
BazSpam ==1.1, ==1.2, ==1.3, ==1.4, ==1.5, \
==1.6, ==1.7 # and so are line-end comments
PEAK[FastCGI, reST]>=0.5a4
setuptools==0.5a7
The simplest way to include requirement specifiers is to use the install_requires argument to setup(). It takes a string or list of strings containing requirement specifiers. If you include more than one requirement in a string, each requirement must begin on a new line.
This has three effects:
Note, by the way, that if you declare your dependencies in setup.py, you do not need to use the require() function in your scripts or modules, as long as you either install the project or use setup.py develop to do development work on it. (See "Development Mode" below for more details on using setup.py develop.)
If your project depends on packages that aren't registered in PyPI, you may still be able to depend on them, as long as they are available for download as an egg, in the standard distutils sdist format, or as a single .py file. You just need to add some URLs to the dependency_links argument to setup().
The URLs must be either:
In general, it's better to link to web pages, because it is usually less complex to update a web page than to release a new version of your project. You can also use a SourceForge showfiles.php link in the case where a package you depend on is distributed via SourceForge.
If you depend on a package that's distributed as a single .py file, you must include an "#egg=project-version" suffix to the URL, to give a project name and version number. (Be sure to escape any dashes in the name or version by replacing them with underscores.) EasyInstall will recognize this suffix and automatically create a trivial setup.py to wrap the single .py file as an egg.
The dependency_links option takes the form of a list of URL strings. For example, the below will cause EasyInstall to search the specified page for eggs or source distributions, if the package's dependencies aren't already installed:
setup(
...
dependency_links = [
"http://peak.telecommunity.com/snapshots/"
],
)
Sometimes a project has "recommended" dependencies, that are not required for all uses of the project. For example, a project might offer optional PDF output if ReportLab is installed, and reStructuredText support if docutils is installed. These optional features are called "extras", and setuptools allows you to define their requirements as well. In this way, other projects that require these optional features can force the additional requirements to be installed, by naming the desired extras in their install_requires.
For example, let's say that Project A offers optional PDF and reST support:
setup(
name="Project-A",
...
extras_require = {
'PDF': ["ReportLab>=1.2", "RXP"],
'reST': ["docutils>=0.3"],
}
)
As you can see, the extras_require argument takes a dictionary mapping names of "extra" features, to strings or lists of strings describing those features' requirements. These requirements will not be automatically installed unless another package depends on them (directly or indirectly) by including the desired "extras" in square brackets after the associated project name. (Or if the extras were listed in a requirement spec on the EasyInstall command line.)
Extras can be used by a project's entry points to specify dynamic dependencies. For example, if Project A includes a "rst2pdf" script, it might declare it like this, so that the "PDF" requirements are only resolved if the "rst2pdf" script is run:
setup(
name="Project-A",
...
entry_points = {
'console_scripts':
['rst2pdf = project_a.tools.pdfgen [PDF]'],
['rst2html = project_a.tools.htmlgen'],
# more script entry points ...
}
)
Projects can also use another project's extras when specifying dependencies. For example, if project B needs "project A" with PDF support installed, it might declare the dependency like this:
setup(
name="Project-B",
install_requires = ["Project-A[PDF]"],
...
)
This will cause ReportLab to be installed along with project A, if project B is installed -- even if project A was already installed. In this way, a project can encapsulate groups of optional "downstream dependencies" under a feature name, so that packages that depend on it don't have to know what the downstream dependencies are. If a later version of Project A builds in PDF support and no longer needs ReportLab, or if it ends up needing other dependencies besides ReportLab in order to provide PDF support, Project B's setup information does not need to change, but the right packages will still be installed if needed.
Note, by the way, that if a project ends up not needing any other packages to support a feature, it should keep an empty requirements list for that feature in its extras_require argument, so that packages depending on that feature don't break (due to an invalid feature name). For example, if Project A above builds in PDF support and no longer needs ReportLab, it could change its setup to this:
setup(
name="Project-A",
...
extras_require = {
'PDF': [],
'reST': ["docutils>=0.3"],
}
)
so that Package B doesn't have to remove the [PDF] from its requirement specifier.
The distutils have traditionally allowed installation of "data files", which are placed in a platform-specific location. However, the most common use case for data files distributed with a package is for use by the package, usually by including the data files in the package directory.
Setuptools offers three ways to specify data files to be included in your packages. First, you can simply use the include_package_data keyword, e.g.:
from setuptools import setup, find_packages
setup(
...
include_package_data = True
)
This tells setuptools to install any data files it finds in your packages. The data files must be under CVS or Subversion control, or else they must be specified via the distutils' MANIFEST.in file. (They can also be tracked by another revision control system, using an appropriate plugin. See the section below on Adding Support for Other Revision Control Systems for information on how to write such plugins.)
If you want finer-grained control over what files are included (for example, if you have documentation files in your package directories and want to exclude them from installation), then you can also use the package_data keyword, e.g.:
from setuptools import setup, find_packages
setup(
...
package_data = {
# If any package contains *.txt or *.rst files, include them:
'': ['*.txt', '*.rst'],
# And include any *.msg files found in the 'hello' package, too:
'hello': ['*.msg'],
}
)
The package_data argument is a dictionary that maps from package names to lists of glob patterns. The globs may include subdirectory names, if the data files are contained in a subdirectory of the package. For example, if the package tree looks like this:
setup.py
src/
mypkg/
__init__.py
mypkg.txt
data/
somefile.dat
otherdata.dat
The setuptools setup file might look like this:
from setuptools import setup, find_packages
setup(
...
packages = find_packages('src'), # include all packages under src
package_dir = {'':'src'}, # tell distutils packages are under src
package_data = {
# If any package contains *.txt files, include them:
'': ['*.txt'],
# And include any *.dat files found in the 'data' subdirectory
# of the 'mypkg' package, also:
'mypkg': ['data/*.dat'],
}
)
Notice that if you list patterns in package_data under the empty string, these patterns are used to find files in every package, even ones that also have their own patterns listed. Thus, in the above example, the mypkg.txt file gets included even though it's not listed in the patterns for mypkg.
Also notice that if you use paths, you must use a forward slash (/) as the path separator, even if you are on Windows. Setuptools automatically converts slashes to appropriate platform-specific separators at build time.
(Note: although the package_data argument was previously only available in setuptools, it was also added to the Python distutils package as of Python 2.4; there is some documentation for the feature available on the python.org website.)
Sometimes, the include_package_data or package_data options alone aren't sufficient to precisely define what files you want included. For example, you may want to include package README files in your revision control system and source distributions, but exclude them from being installed. So, setuptools offers an exclude_package_data option as well, that allows you to do things like this:
from setuptools import setup, find_packages
setup(
...
packages = find_packages('src'), # include all packages under src
package_dir = {'':'src'}, # tell distutils packages are under src
include_package_data = True, # include everything in source control
# ...but exclude README.txt from all packages
exclude_package_data = { '': ['README.txt'] },
)
The exclude_package_data option is a dictionary mapping package names to lists of wildcard patterns, just like the package_data option. And, just as with that option, a key of '' will apply the given pattern(s) to all packages. However, any files that match these patterns will be excluded from installation, even if they were listed in package_data or were included as a result of using include_package_data.
In summary, the three options allow you to:
NOTE: Due to the way the distutils build process works, a data file that you include in your project and then stop including may be "orphaned" in your project's build directories, requiring you to run setup.py clean --all to fully remove them. This may also be important for your users and contributors if they track intermediate revisions of your project using Subversion; be sure to let them know when you make changes that remove files from inclusion so they can run setup.py clean --all.
Typically, existing programs manipulate a package's __file__ attribute in order to find the location of data files. However, this manipulation isn't compatible with PEP 302-based import hooks, including importing from zip files and Python Eggs. It is strongly recommended that, if you are using data files, you should use the Resource Management API of pkg_resources to access them. The pkg_resources module is distributed as part of setuptools, so if you're using setuptools to distribute your package, there is no reason not to use its resource management API. See also Accessing Package Resources for a quick example of converting code that uses __file__ to use pkg_resources instead.
The distutils normally install general "data files" to a platform-specific location (e.g. /usr/share). This feature intended to be used for things like documentation, example configuration files, and the like. setuptools does not install these data files in a separate location, however. They are bundled inside the egg file or directory, alongside the Python modules and packages. The data files can also be accessed using the Resource Management API, by specifying a Requirement instead of a package name:
from pkg_resources import Requirement, resource_filename
filename = resource_filename(Requirement.parse("MyProject"),"sample.conf")
The above code will obtain the filename of the "sample.conf" file in the data root of the "MyProject" distribution.
Note, by the way, that this encapsulation of data files means that you can't actually install data files to some arbitrary location on a user's machine; this is a feature, not a bug. You can always include a script in your distribution that extracts and copies your the documentation or data files to a user-specified location, at their discretion. If you put related data files in a single directory, you can use resource_filename() with the directory name to get a filesystem directory that then can be copied with the shutil module. (Even if your package is installed as a zipfile, calling resource_filename() on a directory will return an actual filesystem directory, whose contents will be that entire subtree of your distribution.)
(Of course, if you're writing a new package, you can just as easily place your data files or directories inside one of your packages, rather than using the distutils' approach. However, if you're updating an existing application, it may be simpler not to change the way it currently specifies these data files.)
If you are using tools that expect your resources to be "real" files, or your project includes non-extension native libraries or other files that your C extensions expect to be able to access, you may need to list those files in the eager_resources argument to setup(), so that the files will be extracted together, whenever a C extension in the project is imported.
This is especially important if your project includes shared libraries other than distutils-built C extensions, and those shared libraries use file extensions other than .dll, .so, or .dylib, which are the extensions that setuptools 0.6a8 and higher automatically detects as shared libraries and adds to the native_libs.txt file for you. Any shared libraries whose names do not end with one of those extensions should be listed as eager_resources, because they need to be present in the filesystem when he C extensions that link to them are used.
The pkg_resources runtime for compressed packages will automatically extract all C extensions and eager_resources at the same time, whenever any C extension or eager resource is requested via the resource_filename() API. (C extensions are imported using resource_filename() internally.) This ensures that C extensions will see all of the "real" files that they expect to see.
Note also that you can list directory resource names in eager_resources as well, in which case the directory's contents (including subdirectories) will be extracted whenever any C extension or eager resource is requested.
Please note that if you're not sure whether you need to use this argument, you don't! It's really intended to support projects with lots of non-Python dependencies and as a last resort for crufty projects that can't otherwise handle being compressed. If your package is pure Python, Python plus data files, or Python plus C, you really don't need this. You've got to be using either C or an external program that needs "real" files in your project before there's any possibility of eager_resources being relevant to your project.
setuptools supports creating libraries that "plug in" to extensible applications and frameworks, by letting you register "entry points" in your project that can be imported by the application or framework.
For example, suppose that a blogging tool wants to support plugins that provide translation for various file types to the blog's output format. The framework might define an "entry point group" called blogtool.parsers, and then allow plugins to register entry points for the file extensions they support.
This would allow people to create distributions that contain one or more parsers for different file types, and then the blogging tool would be able to find the parsers at runtime by looking up an entry point for the file extension (or mime type, or however it wants to).
Note that if the blogging tool includes parsers for certain file formats, it can register these as entry points in its own setup script, which means it doesn't have to special-case its built-in formats. They can just be treated the same as any other plugin's entry points would be.
If you're creating a project that plugs in to an existing application or framework, you'll need to know what entry points or entry point groups are defined by that application or framework. Then, you can register entry points in your setup script. Here are a few examples of ways you might register an .rst file parser entry point in the blogtool.parsers entry point group, for our hypothetical blogging tool:
setup(
# ...
entry_points = {'blogtool.parsers': '.rst = some_module:SomeClass'}
)
setup(
# ...
entry_points = {'blogtool.parsers': ['.rst = some_module:a_func']}
)
setup(
# ...
entry_points = """
[blogtool.parsers]
.rst = some.nested.module:SomeClass.some_classmethod [reST]
""",
extras_require = dict(reST = "Docutils>=0.3.5")
)
The entry_points argument to setup() accepts either a string with .ini-style sections, or a dictionary mapping entry point group names to either strings or lists of strings containing entry point specifiers. An entry point specifier consists of a name and value, separated by an = sign. The value consists of a dotted module name, optionally followed by a : and a dotted identifier naming an object within the module. It can also include a bracketed list of "extras" that are required for the entry point to be used. When the invoking application or framework requests loading of an entry point, any requirements implied by the associated extras will be passed to pkg_resources.require(), so that an appropriate error message can be displayed if the needed package(s) are missing. (Of course, the invoking app or framework can ignore such errors if it wants to make an entry point optional if a requirement isn't installed.)
Some extensible applications and frameworks may need to define their own kinds of metadata to include in eggs, which they can then access using the pkg_resources metadata APIs. Ordinarily, this is done by having plugin developers include additional files in their ProjectName.egg-info directory. However, since it can be tedious to create such files by hand, you may want to create a distutils extension that will create the necessary files from arguments to setup(), in much the same way that setuptools does for many of the setup() arguments it adds. See the section below on Creating distutils Extensions for more details, especially the subsection on Adding new EGG-INFO Files.
Under normal circumstances, the distutils assume that you are going to build a distribution of your project, not use it in its "raw" or "unbuilt" form. If you were to use the distutils that way, you would have to rebuild and reinstall your project every time you made a change to it during development.
Another problem that sometimes comes up with the distutils is that you may need to do development on two related projects at the same time. You may need to put both projects' packages in the same directory to run them, but need to keep them separate for revision control purposes. How can you do this?
Setuptools allows you to deploy your projects for use in a common directory or staging area, but without copying any files. Thus, you can edit each project's code in its checkout directory, and only need to run build commands when you change a project's C extensions or similarly compiled files. You can even deploy a project into another project's checkout directory, if that's your preferred way of working (as opposed to using a common independent staging area or the site-packages directory).
To do this, use the setup.py develop command. It works very similarly to setup.py install or the EasyInstall tool, except that it doesn't actually install anything. Instead, it creates a special .egg-link file in the deployment directory, that links to your project's source code. And, if your deployment directory is Python's site-packages directory, it will also update the easy-install.pth file to include your project's source code, thereby making it available on sys.path for all programs using that Python installation.
In addition, the develop command creates wrapper scripts in the target script directory that will run your in-development scripts after ensuring that all your install_requires packages are available on sys.path.
You can deploy the same project to multiple staging areas, e.g. if you have multiple projects on the same machine that are sharing the same project you're doing development work.
When you're done with a given development task, you can remove the project source from a staging area using setup.py develop --uninstall, specifying the desired staging area if it's not the default.
There are several options to control the precise behavior of the develop command; see the section on the develop command below for more details.
Note that you can also apply setuptools commands to non-setuptools projects, using commands like this:
python -c "import setuptools; execfile('setup.py')" develop
That is, you can simply list the normal setup commands and options following the quoted part.
Your users might not have setuptools installed on their machines, or even if they do, it might not be the right version. Fixing this is easy; just download ez_setup.py, and put it in the same directory as your setup.py script. (Be sure to add it to your revision control system, too.) Then add these two lines to the very top of your setup script, before the script imports anything from setuptools:
import ez_setup ez_setup.use_setuptools()
That's it. The ez_setup module will automatically download a matching version of setuptools from PyPI, if it isn't present on the target system. Whenever you install an updated version of setuptools, you should also update your projects' ez_setup.py files, so that a matching version gets installed on the target machine(s).
By the way, setuptools supports the new PyPI "upload" command, so you can use setup.py sdist upload or setup.py bdist_egg upload to upload your source or egg distributions respectively. Your project's current version must be registered with PyPI first, of course; you can use setup.py register to do that. Or you can do it all in one step, e.g. setup.py register sdist bdist_egg upload will register the package, build source and egg distributions, and then upload them both to PyPI, where they'll be easily found by other projects that depend on them.
(By the way, if you need to distribute a specific version of setuptools, you can specify the exact version and base download URL as parameters to the use_setuptools() function. See the function's docstring for details.)
In general, a setuptools-based project looks just like any distutils-based project -- as long as your users have an internet connection and are installing to site-packages, that is. But for some users, these conditions don't apply, and they may become frustrated if this is their first encounter with a setuptools-based project. To keep these users happy, you should review the following topics in your project's installation instructions, if they are relevant to your project and your target audience isn't already familiar with setuptools and easy_install.
If your project depends on other projects that may need to be downloaded from PyPI or elsewhere, you should list them in your installation instructions, or tell users how to find out what they are. While most users will not need this information, any users who don't have unrestricted internet access may have to find, download, and install the other projects manually. (Note, however, that they must still install those projects using easy_install, or your project will not know they are installed, and your setup script will try to download them again.)
If you want to be especially friendly to users with limited network access, you may wish to build eggs for your project and its dependencies, making them all available for download from your site, or at least create a page with links to all of the needed eggs. In this way, users with limited network access can manually download all the eggs to a single directory, then use the -f option of easy_install to specify the directory to find eggs in. Users who have full network access can just use -f with the URL of your download page, and easy_install will find all the needed eggs using your links directly. This is also useful when your target audience isn't able to compile packages (e.g. most Windows users) and your package or some of its dependencies include C code.
Users and co-developers who are tracking your in-development code using CVS, Subversion, or some other revision control system should probably read this manual's sections regarding such development. Alternately, you may wish to create a quick-reference guide containing the tips from this manual that apply to your particular situation. For example, if you recommend that people use setup.py develop when tracking your in-development code, you should let them know that this needs to be run after every update or commit.
Similarly, if you remove modules or data files from your project, you should remind them to run setup.py clean --all and delete any obsolete .pyc or .pyo. (This tip applies to the distutils in general, not just setuptools, but not everybody knows about them; be kind to your users by spelling out your project's best practices rather than leaving them guessing.)
Some users want to manage all Python packages using a single package manager, and sometimes that package manager isn't easy_install! Setuptools currently supports bdist_rpm, bdist_wininst, and bdist_dumb formats for system packaging. If a user has a locally- installed "bdist" packaging tool that internally uses the distutils install command, it should be able to work with setuptools. Some examples of "bdist" formats that this should work with include the bdist_nsi and bdist_msi formats for Windows.
However, packaging tools that build binary distributions by running setup.py install on the command line or as a subprocess will require modification to work with setuptools. They should use the --single-version-externally-managed option to the install command, combined with the standard --root or --record options. See the install command documentation below for more details. The bdist_deb command is an example of a command that currently requires this kind of patching to work with setuptools.
If you or your users have a problem building a usable system package for your project, please report the problem via the mailing list so that either the "bdist" tool in question or setuptools can be modified to resolve the issue.
If you're managing several projects that need to use ez_setup, and you are using Subversion as your revision control system, you can use the "svn:externals" property to share a single copy of ez_setup between projects, so that it will always be up-to-date whenever you check out or update an individual project, without having to manually update each project to use a new version.
However, because Subversion only supports using directories as externals, you have to turn ez_setup.py into ez_setup/__init__.py in order to do this, then create "externals" definitions that map the ez_setup directory into each project. Also, if any of your projects use find_packages() on their setup directory, you will need to exclude the resulting ez_setup package, to keep it from being included in your distributions, e.g.:
setup(
...
packages = find_packages(exclude=['ez_setup']),
)
Of course, the ez_setup package will still be included in your packages' source distributions, as it needs to be.
For your convenience, you may use the following external definition, which will track the latest version of setuptools:
ez_setup svn://svn.eby-sarna.com/svnroot/ez_setup
You can set this by executing this command in your project directory:
svn propedit svn:externals .
And then adding the line shown above to the file that comes up for editing.
For maximum performance, Python packages are best installed as zip files. Not all packages, however, are capable of running in compressed form, because they may expect to be able to access either source code or data files as normal operating system files. So, setuptools can install your project as a zipfile or a directory, and its default choice is determined by the project's zip_safe flag.
You can pass a True or False value for the zip_safe argument to the setup() function, or you can omit it. If you omit it, the bdist_egg command will analyze your project's contents to see if it can detect any conditions that would prevent it from working in a zipfile. It will output notices to the console about any such conditions that it finds.
Currently, this analysis is extremely conservative: it will consider the project unsafe if it contains any C extensions or datafiles whatsoever. This does not mean that the project can't or won't work as a zipfile! It just means that the bdist_egg authors aren't yet comfortable asserting that the project will work. If the project contains no C or data files, and does no __file__ or __path__ introspection or source code manipulation, then there is an extremely solid chance the project will work when installed as a zipfile. (And if the project uses pkg_resources for all its data file access, then C extensions and other data files shouldn't be a problem at all. See the Accessing Data Files at Runtime section above for more information.)
However, if bdist_egg can't be sure that your package will work, but you've checked over all the warnings it issued, and you are either satisfied it will work (or if you want to try it for yourself), then you should set zip_safe to True in your setup() call. If it turns out that it doesn't work, you can always change it to False, which will force setuptools to install your project as a directory rather than as a zipfile.
Of course, the end-user can still override either decision, if they are using EasyInstall to install your package. And, if you want to override for testing purposes, you can just run setup.py easy_install --zip-ok . or setup.py easy_install --always-unzip . in your project directory. to install the package as a zipfile or directory, respectively.
In the future, as we gain more experience with different packages and become more satisfied with the robustness of the pkg_resources runtime, the "zip safety" analysis may become less conservative. However, we strongly recommend that you determine for yourself whether your project functions correctly when installed as a zipfile, correct any problems if you can, and then make an explicit declaration of True or False for the zip_safe flag, so that it will not be necessary for bdist_egg or EasyInstall to try to guess whether your project can work as a zipfile.
Sometimes, a large package is more useful if distributed as a collection of smaller eggs. However, Python does not normally allow the contents of a package to be retrieved from more than one location. "Namespace packages" are a solution for this problem. When you declare a package to be a namespace package, it means that the package has no meaningful contents in its __init__.py, and that it is merely a container for modules and subpackages.
The pkg_resources runtime will then automatically ensure that the contents of namespace packages that are spread over multiple eggs or directories are combined into a single "virtual" package.
The namespace_packages argument to setup() lets you declare your project's namespace packages, so that they will be included in your project's metadata. The argument should list the namespace packages that the egg participates in. For example, the ZopeInterface project might do this:
setup(
# ...
namespace_packages = ['zope']
)
because it contains a zope.interface package that lives in the zope namespace package. Similarly, a project for a standalone zope.publisher would also declare the zope namespace package. When these projects are installed and used, Python will see them both as part of a "virtual" zope package, even though they will be installed in different locations.
Namespace packages don't have to be top-level packages. For example, Zope 3's zope.app package is a namespace package, and in the future PEAK's peak.util package will be too.
Note, by the way, that your project's source tree must include the namespace packages' __init__.py files (and the __init__.py of any parent packages), in a normal Python package layout. These __init__.py files must contain the line:
__import__('pkg_resources').declare_namespace(__name__)
This code ensures that the namespace package machinery is operating and that the current package is registered as a namespace package.
You must NOT include any other code and data in a namespace package's __init__.py. Even though it may appear to work during development, or when projects are installed as .egg files, it will not work when the projects are installed using "system" packaging tools -- in such cases the __init__.py files will not be installed, let alone executed.
You must include the declare_namespace() line in the __init__.py of every project that has contents for the namespace package in question, in order to ensure that the namespace will be declared regardless of which project's copy of __init__.py is loaded first. If the first loaded __init__.py doesn't declare it, it will never be declared, because no other copies will ever be loaded!)
Setuptools 0.6a automatically calls declare_namespace() for you at runtime, but the 0.7a versions will not. This is because the automatic declaration feature has some negative side effects, such as needing to import all namespace packages during the initialization of the pkg_resources runtime, and also the need for pkg_resources to be explicitly imported before any namespace packages work at all. Beginning with the 0.7a releases, you'll be responsible for including your own declaration lines, and the automatic declaration feature will be dropped to get rid of the negative side effects.
During the remainder of the 0.6 development cycle, therefore, setuptools will warn you about missing declare_namespace() calls in your __init__.py files, and you should correct these as soon as possible before setuptools 0.7a1 is released. Namespace packages without declaration lines will not work correctly once a user has upgraded to setuptools 0.7a1, so it's important that you make this change now in order to avoid having your code break in the field. Our apologies for the inconvenience, and thank you for your patience.
When a set of related projects are under development, it may be important to track finer-grained version increments than you would normally use for e.g. "stable" releases. While stable releases might be measured in dotted numbers with alpha/beta/etc. status codes, development versions of a project often need to be tracked by revision or build number or even build date. This is especially true when projects in development need to refer to one another, and therefore may literally need an up-to-the-minute version of something!
To support these scenarios, setuptools allows you to "tag" your source and egg distributions by adding one or more of the following to the project's "official" version identifier:
You can add these tags by adding egg_info and the desired options to the command line ahead of the sdist or bdist commands that you want to generate a daily build or snapshot for. See the section below on the egg_info command for more details.
(Also, before you release your project, be sure to see the section above on Specifying Your Project's Version for more information about how pre- and post-release tags affect how setuptools and EasyInstall interpret version numbers. This is important in order to make sure that dependency processing tools will know which versions of your project are newer than others.)
Finally, if you are creating builds frequently, and either building them in a downloadable location or are copying them to a distribution server, you should probably also check out the rotate command, which lets you automatically delete all but the N most-recently-modified distributions matching a glob pattern. So, you can use a command line like:
setup.py egg_info -rbDEV bdist_egg rotate -m.egg -k3
to build an egg whose version info includes 'DEV-rNNNN' (where NNNN is the most recent Subversion revision that affected the source tree), and then delete any egg files from the distribution directory except for the three that were built most recently.
If you have to manage automated builds for multiple packages, each with different tagging and rotation policies, you may also want to check out the alias command, which would let each package define an alias like daily that would perform the necessary tag, build, and rotate commands. Then, a simpler script or cron job could just run setup.py daily in each project directory. (And, you could also define sitewide or per-user default versions of the daily alias, so that projects that didn't define their own would use the appropriate defaults.)
setuptools enhances the distutils' default algorithm for source file selection, so that all files managed by CVS or Subversion in your project tree are included in any source distribution you build. This is a big improvement over having to manually write a MANIFEST.in file and try to keep it in sync with your project. So, if you are using CVS or Subversion, and your source distributions only need to include files that you're tracking in revision control, don't create a a MANIFEST.in file for your project. (And, if you already have one, you might consider deleting it the next time you would otherwise have to change it.)
(NOTE: other revision control systems besides CVS and Subversion can be supported using plugins; see the section below on Adding Support for Other Revision Control Systems for information on how to write such plugins.)
If you need to include automatically generated files, or files that are kept in an unsupported revision control system, you'll need to create a MANIFEST.in file to specify any files that the default file location algorithm doesn't catch. See the distutils documentation for more information on the format of the MANIFEST.in file.
But, be sure to ignore any part of the distutils documentation that deals with MANIFEST or how it's generated from MANIFEST.in; setuptools shields you from these issues and doesn't work the same way in any case. Unlike the distutils, setuptools regenerates the source distribution manifest file every time you build a source distribution, and it builds it inside the project's .egg-info directory, out of the way of your main project directory. You therefore need not worry about whether it is up-to-date or not.
Indeed, because setuptools' approach to determining the contents of a source distribution is so much simpler, its sdist command omits nearly all of the options that the distutils' more complex sdist process requires. For all practical purposes, you'll probably use only the --formats option, if you use any option at all.
(By the way, if you're using some other revision control system, you might consider creating and publishing a revision control plugin for setuptools.)
If you use the register command (setup.py register) to register your package with PyPI, that's most of the battle right there. (See the docs for the register command for more details.)
If you also use the upload command to upload actual distributions of your package, that's even better, because EasyInstall will be able to find and download them directly from your project's PyPI page.
However, there may be reasons why you don't want to upload distributions to PyPI, and just want your existing distributions (or perhaps a Subversion checkout) to be used instead.
So here's what you need to do before running the register command. There are three setup() arguments that affect EasyInstall:
A URL is considered a "primary link" if it is a link to a .tar.gz, .tgz, .zip, .egg, .egg.zip, .tar.bz2, or .exe file, or if it has an #egg=project or #egg=project-version fragment identifier attached to it. EasyInstall attempts to determine a project name and optional version number from the text of a primary link without downloading it. When it has found all the primary links, EasyInstall will select the best match based on requested version, platform compatibility, and other criteria.
So, if your url or download_url point either directly to a downloadable source distribution, or to HTML page(s) that have direct links to such, then EasyInstall will be able to locate downloads automatically. If you want to make Subversion checkouts available, then you should create links with either #egg=project or #egg=project-version added to the URL. You should replace project and version with the values they would have in an egg filename. (Be sure to actually generate an egg and then use the initial part of the filename, rather than trying to guess what the escaped form of the project name and version number will be.)
Note that Subversion checkout links are of lower precedence than other kinds of distributions, so EasyInstall will not select a Subversion checkout for downloading unless it has a version included in the #egg= suffix, and it's a higher version than EasyInstall has seen in any other links for your project.
As a result, it's a common practice to use mark checkout URLs with a version of "dev" (i.e., #egg=projectname-dev), so that users can do something like this:
easy_install --editable projectname==dev
in order to check out the in-development version of projectname.
If you expect your users to track in-development versions of your project via Subversion, there are a few additional steps you should take to ensure that things work smoothly with EasyInstall. First, you should add the following to your project's setup.cfg file:
[egg_info] tag_build = .dev tag_svn_revision = 1
This will tell setuptools to generate package version numbers like 1.0a1.dev-r1263, which will be considered to be an older release than 1.0a1. Thus, when you actually release 1.0a1, the entire egg infrastructure (including setuptools, pkg_resources and EasyInstall) will know that 1.0a1 supersedes any interim snapshots from Subversion, and handle upgrades accordingly.
(Note: the project version number you specify in setup.py should always be the next version of your software, not the last released version. Alternately, you can leave out the tag_build=.dev, and always use the last release as a version number, so that your post-1.0 builds are labelled 1.0-r1263, indicating a post-1.0 patchlevel. Most projects so far, however, seem to prefer to think of their project as being a future version still under development, rather than a past version being patched. It is of course possible for a single project to have both situations, using post-release numbering on release branches, and pre-release numbering on the trunk. But you don't have to make things this complex if you don't want to.)
Commonly, projects releasing code from Subversion will include a PyPI link to their checkout URL (as described in the previous section) with an #egg=projectname-dev suffix. This allows users to request EasyInstall to download projectname==dev in order to get the latest in-development code. Note that if your project depends on such in-progress code, you may wish to specify your install_requires (or other requirements) to include ==dev, e.g.:
install_requires = ["OtherProject>=0.2a1.dev-r143,==dev"]
The above example says, "I really want at least this particular development revision number, but feel free to follow and use an #egg=OtherProject-dev link if you find one". This avoids the need to have actual source or binary distribution snapshots of in-development code available, just to be able to depend on the latest and greatest a project has to offer.
A final note for Subversion development: if you are using SVN revision tags as described in this section, it's a good idea to run setup.py develop after each Subversion checkin or update, because your project's version number will be changing, and your script wrappers need to be updated accordingly.
Also, if the project's requirements have changed, the develop command will take care of fetching the updated dependencies, building changed extensions, etc. Be sure to also remind any of your users who check out your project from Subversion that they need to run setup.py develop after every update in order to keep their checkout completely in sync.
When you make an official release, creating source or binary distributions, you will need to override the tag settings from setup.cfg, so that you don't end up registering versions like foobar-0.7a1.dev-r34832. This is easy to do if you are developing on the trunk and using tags or branches for your releases - just make the change to setup.cfg after branching or tagging the release, so the trunk will still produce development snapshots.
Alternately, if you are not branching for releases, you can override the default version options on the command line, using something like:
python setup.py egg_info -RDb "" sdist bdist_egg register upload
The first part of this command (egg_info -RDb "") will override the configured tag information, before creating source and binary eggs, registering the project with PyPI, and uploading the files. Thus, these commands will use the plain version from your setup.py, without adding the Subversion revision number or build designation string.
Of course, if you will be doing this a lot, you may wish to create a personal alias for this operation, e.g.:
python setup.py alias -u release egg_info -RDb ""
You can then use it like this:
python setup.py release sdist bdist_egg register upload
Or of course you can create more elaborate aliases that do all of the above. See the sections below on the egg_info and alias commands for more ideas.
setuptools includes transparent support for building Pyrex extensions, as long as you define your extensions using setuptools.Extension, not distutils.Extension. You must also not import anything from Pyrex in your setup script.
If you follow these rules, you can safely list .pyx files as the source of your Extension objects in the setup script. setuptools will detect at build time whether Pyrex is installed or not. If it is, then setuptools will use it. If not, then setuptools will silently change the Extension objects to refer to the .c counterparts of the .pyx files, so that the normal distutils C compilation process will occur.
Of course, for this to work, your source distributions must include the C code generated by Pyrex, as well as your original .pyx files. This means that you will probably want to include current .c files in your revision control system, rebuilding them whenever you check changes in for the .pyx source files. This will ensure that people tracking your project in CVS or Subversion will be able to build it even if they don't have Pyrex installed, and that your source releases will be similarly usable with or without Pyrex.
Sometimes, you need to use the same commands over and over, but you can't necessarily set them as defaults. For example, if you produce both development snapshot releases and "stable" releases of a project, you may want to put the distributions in different places, or use different egg_info tagging options, etc. In these cases, it doesn't make sense to set the options in a distutils configuration file, because the values of the options changed based on what you're trying to do.
Setuptools therefore allows you to define "aliases" - shortcut names for an arbitrary string of commands and options, using setup.py alias aliasname expansion, where aliasname is the name of the new alias, and the remainder of the command line supplies its expansion. For example, this command defines a sitewide alias called "daily&quo