skip to navigation
skip to content

Not Logged In

zc.buildout 1.7.0

System for managing development buildouts

Latest Version: 2.2.1

Contents

The Buildout project provides support for creating applications, especially Python applications. It provides tools for assembling applications from multiple parts, Python or otherwise. An application may actually contain multiple programs, processes, and configuration settings.

The word "buildout" refers to a description of a set of parts and the software to create and assemble them. It is often used informally to refer to an installed system based on a buildout definition. For example, if we are creating an application named "Foo", then "the Foo buildout" is the collection of configuration and application-specific software that allows an instance of the application to be created. We may refer to such an instance of the application informally as "a Foo buildout".

To get a feel for some of the things you might use buildouts for, see the Buildout examples.

To lean more about using buildouts, see Detailed Documentation.

To see screencasts, talks, useful links and more documentation, visit the Buildout website.

Recipes

Existing recipes include:

zc.recipe.egg
The egg recipe installes one or more eggs, with their dependencies. It installs their console-script entry points with the needed eggs included in their paths. It is suitable for use with a "clean" Python: one without packages installed in site-packages.
z3c.recipe.scripts
Like zc.recipe.egg, this recipe builds interpreter scripts and entry point scripts based on eggs. It can be used with a Python that has packages installed in site-packages, such as a system Python. The interpreter also has more features than the one offered by zc.recipe.egg.
zc.recipe.testrunner
The testrunner egg creates a test runner script for one or more eggs.
zc.recipe.zope3checkout
The zope3checkout recipe installs a Zope 3 checkout into a buildout.
zc.recipe.zope3instance
The zope3instance recipe sets up a Zope 3 instance.
zc.recipe.filestorage
The filestorage recipe sets up a ZODB file storage for use in a Zope 3 instance created by the zope3instance recipe.

Buildout examples

Here are a few examples of what you can do with buildouts. We'll present these as a set of use cases.

Try out an egg

Sometimes you want to try an egg (or eggs) that someone has released. You'd like to get a Python interpreter that lets you try things interactively or run sample scripts without having to do path manipulations. If you can and don't mind modifying your Python installation, you could use easy_install, otherwise, you could create a directory somewhere and create a buildout.cfg file in that directory containing:

[buildout]
parts = mypython

[mypython]
recipe = zc.recipe.egg
interpreter = mypython
eggs = theegg

where theegg is the name of the egg you want to try out.

Run buildout in this directory. It will create a bin subdirectory that includes a mypython script. If you run mypython without any arguments you'll get an interactive interpreter with the egg in the path. If you run it with a script and script arguments, the script will run with the egg in its path. Of course, you can specify as many eggs as you want in the eggs option.

If the egg provides any scripts (console_scripts entry points), those will be installed in your bin directory too.

Work on a package

I often work on packages that are managed separately. They don't have scripts to be installed, but I want to be able to run their tests using the zope.testing test runner. In this kind of application, the program to be installed is the test runner. A good example of this is zc.ngi.

Here I have a subversion project for the zc.ngi package. The software is in the src directory. The configuration file is very simple:

[buildout]
develop = .
parts = test

[test]
recipe = zc.recipe.testrunner
eggs = zc.ngi

I use the develop option to create a develop egg based on the current directory. I request a test script named "test" using the zc.recipe.testrunner recipe. In the section for the test script, I specify that I want to run the tests in the zc.ngi package.

When I check out this project into a new sandbox, I run bootstrap.py to get setuptools and zc.buildout and to create bin/buildout. I run bin/buildout, which installs the test script, bin/test, which I can then use to run the tests.

This is probably the most common type of buildout.

If I need to run a previous version of zc.buildout, I use the --version option of the bootstrap.py script:

$ python bootstrap.py --version 1.1.3

The zc.buildout project is a slightly more complex example of this type of buildout.

Install egg-based scripts

A variation of the Try out an egg use case is to install scripts into your ~/bin directory (on Unix, of course). My ~/bin directory is a buildout with a configuration file that looks like:

[buildout]
parts = foo bar
bin-directory = .

[foo]
...

where foo and bar are packages with scripts that I want available. As I need new scripts, I can add additional sections. The bin-directory option specified that scripts should be installed into the current directory.

Multi-program multi-machine systems

Using an older prototype version of the buildout, we've build a number of systems involving multiple programs, databases, and machines. One typical example consists of:

  • Multiple Zope instances
  • Multiple ZEO servers
  • An LDAP server
  • Cache-invalidation and Mail delivery servers
  • Dozens of add-on packages
  • Multiple test runners
  • Multiple deployment modes, including dev, stage, and prod, with prod deployment over multiple servers

Parts installed include:

  • Application software installs, including Zope, ZEO and LDAP software
  • Add-on packages
  • Bundles of configuration that define Zope, ZEO and LDAP instances
  • Utility scripts such as test runners, server-control scripts, cron jobs.

Questions and Bug Reporting

Please send questions and comments to the distutils SIG mailing list.

Report bugs using the zc.buildout Launchpad Bug Tracker.

System Python and zc.buildout 1.5

The 1.5 line of zc.buildout introduced a number of changes.

Problems

As usual, please send questions and comments to the distutils SIG mailing list. Report bugs using the zc.buildout Launchpad Bug Tracker.

If problems are keeping you from your work, here's an easy way to revert to the old code temporarily: switch to a custom "emergency" bootstrap script, available from http://svn.zope.org/repos/main/zc.buildout/branches/1.4/bootstrap/bootstrap.py .

This customized script will select zc.buildout 1.4.4 by default. zc.buildout 1.4.4 will not upgrade itself unless you explicitly specify a new version. It will also prefer older versions of zc.recipe.egg and some other common recipes. If you have trouble with other recipes, consider using a standard buildout "versions" section to specify older versions of these, as described in the Buildout documentation (http://pypi.python.org/pypi/zc.buildout#repeatable-buildouts-controlling-eggs-used).

Working with a System Python

While there are a number of new features available in zc.buildout 1.5, the biggest is that Buildout itself supports usage with a system Python. This can work if you follow a couple of simple rules.

  1. Use the new bootstrap.py (available from https://github.com/buildout/buildout/raw/1.6.x/bootstrap/bootstrap.py).
  2. Use buildout recipes that have been upgraded to work with zc.buildout 1.5 and higher. Specifically, they should use zc.buildout.easy_install.sitepackage_safe_scripts to generate their scripts, if any, rather than zc.buildout.easy_install.scripts. See the Recipes That Support a System Python section below for more details on recipes that are available as of this writing, and Updating Recipes to Support a System Python for instructions on how to update a recipe. Note that you should generally only need to update recipes that generate scripts.

You can then use include-site-packages = false and exec-sitecustomize = false buildout options to eliminate access to your Python's site packages and not execute its sitecustomize file, if it exists, respectively.

Alternately, you can use the allowed-eggs-from-site-packages buildout option as a glob-aware whitelist of eggs that may come from site-packages. This value defaults to "*", accepting all eggs.

It's important to note that recipes not upgraded for zc.buildout 1.5.0 should continue to work--just without internal support for a system Python.

Using a system Python is inherently fragile. Using a clean, freshly-installed Python without customization in site-packages is more robust and repeatable. See some of the regression tests added to the 1.5.0 line for the kinds of issues that you can encounter with a system Python, and see http://pypi.python.org/pypi/z3c.recipe.scripts#including-site-packages-and-sitecustomize for more discussion.

However, using a system Python can be very convenient, and the zc.buildout code for this feature has been tested by many users already. Moreover, it has automated tests to exercise the problems that have been encountered and fixed. Many people rely on it.

Recipes That Support a System Python

zc.recipe.egg continues to generate old-style scripts that are not safe for use with a system Python. This was done for backwards compatibility, because it is integral to so many buildouts and used as a dependency of so many other recipes.

If you want to generate new-style scripts that do support system Python usage, use z3c.recipe.scripts instead (http://pypi.python.org/pypi/z3c.recipe.scripts). z3c.recipe.scripts has the same script and interpreter generation options as zc.recipe.egg, plus a few more for the new features mentioned above. In the simplest case, you should be able to simply change recipe = zc.recipe.egg to recipe = z3c.recipe.scripts in the pertinent sections of your buildout configuration and your generated scripts will work with a system Python.

Other updated recipes include zc.recipe.testrunner 1.4.0 and z3c.recipe.tag 0.4.0. Others should be updated soon: see their change documents for details, or see Updating Recipes to Support a System Python for instructions on how to update recipes yourself.

Templates for creating Python scripts with the z3c.recipe.filetemplate recipe can be easily changed to support a system Python.

  • If you don't care about supporting relative paths, simply using a generated interpreter with the eggs you want should be sufficient, as it was before. For instance, if the interpreter is named "py", use #!${buildout:bin-directory/py} or #!/usr/bin/env ${buildout:bin-directory/py}).

  • If you do care about relative paths, (relative-paths = true in your buildout configuration), then z3c.recipe.scripts does require a bit more changes, as is usual for the relative path support in that package. First, use z3c.recipe.scripts to generate a script or interpreter with the dependencies you want. This will create a directory in parts that has a site.py and sitecustomize.py. Then, begin your script as in the snippet below. The example assumes that the z3c.recipe.scripts generated were from a Buildout configuration section labeled "scripts": adjust accordingly.

    #!${buildout:executable} -S
    ${python-relative-path-setup}
    import sys
    sys.path.insert(0, ${scripts:parts-directory|path-repr})
    import site
    

Updating Recipes to Support a System Python

You should generally only need to update recipes that generate scripts. These recipes need to change from using zc.buildout.easy_install.scripts to be using zc.buildout.easy_install.sitepackage_safe_scripts. The signatures of the two functions are different. Please compare:

def scripts(
    reqs, working_set, executable, dest,
    scripts=None,
    extra_paths=(),
    arguments='',
    interpreter=None,
    initialization='',
    relative_paths=False,
    ):

def sitepackage_safe_scripts(
    dest, working_set, executable, site_py_dest,
    reqs=(),
    scripts=None,
    interpreter=None,
    extra_paths=(),
    initialization='',
    include_site_packages=False,
    exec_sitecustomize=False,
    relative_paths=False,
    script_arguments='',
    script_initialization='',
    ):

In most cases, the arguments are merely reordered. The reqs argument is no longer required in order to make it easier to generate an interpreter alone. The arguments argument was renamed to script_arguments to clarify that it did not affect interpreter generation.

The only new required argument is site_py_dest. It must be the path to a directory in which the customized site.py and sitecustomize.py files will be written. A typical generation in a recipe will look like this.

(In the recipe's __init__ method...)

self.options = options
b_options = buildout['buildout']
options['parts-directory'] = os.path.join(
    b_options['parts-directory'], self.name)

(In the recipe's install method...)

options = self.options
generated = []
if not os.path.exists(options['parts-directory']):
    os.mkdir(options['parts-directory'])
    generated.append(options['parts-directory'])

Then options['parts-directory'] can be used for the site_py_dest value.

If you want to support the other arguments (include_site_packages, exec_sitecustomize, script_initialization, as well as the allowed-eggs-from-site-packages option), you might want to look at some of the code in https://github.com/buildout/buildout/blob/1.6.x/z3c.recipe.scripts/src/z3c/recipe/scripts/scripts.py . You might even be able to adopt some of it by subclassing or delegating. The Scripts class in that file is the closest to what you might be used to from zc.recipe.egg.

Important note for recipe authors: As of buildout 1.5.2, the code in recipes is always run with the access to the site-packages as configured in the buildout section.

virtualenv

Using virtualenv (http://pypi.python.org/pypi/virtualenv) with the --no-site-packages option already provided a simple way of using a system Python. This is intended to continue to work, and some automated tests exist to demonstrate this.

However, it is only supported to the degree that people have found it to work in the past. The existing Buildout tests for virtualenv are only for problems encountered previously. They are very far from comprehensive.

Using Buildout with a system python has at least three advantages over using Buildout in conjunction with virtualenv. They may or may not be pertinent to your desired usage.

  • Unlike virtualenv --no-site-packages, Buildout's support allows you to choose to let packages from your system Python be available to your software (see include-site-packages in http://pypi.python.org/pypi/z3c.recipe.scripts).

    You can even specify which eggs installed in your system Python can be allowed to fulfill some of your packages' dependencies (see allowed-eggs-from-site-packages in http://pypi.python.org/pypi/z3c.recipe.scripts).

    At the expense of some repeatability and platform dependency, this flexibility means that, for instance, you can rely on difficult-to-build eggs like lxml coming from your system Python.

  • Buildout's implementation has a full set of automated tests.

  • An integral Buildout implementation means fewer steps and fewer dependencies to work with a system Python.

Detailed Documentation

Buildouts

The word "buildout" refers to a description of a set of parts and the software to create and assemble them. It is often used informally to refer to an installed system based on a buildout definition. For example, if we are creating an application named "Foo", then "the Foo buildout" is the collection of configuration and application-specific software that allows an instance of the application to be created. We may refer to such an instance of the application informally as "a Foo buildout".

This document describes how to define buildouts using buildout configuration files and recipes. There are three ways to set up the buildout software and create a buildout instance:

  1. Install the zc.buildout egg with easy_install and use the buildout script installed in a Python scripts area.
  2. Use the buildout bootstrap script to create a buildout that includes both the setuptools and zc.buildout eggs. This allows you to use the buildout software without modifying a Python install. The buildout script is installed into your buildout local scripts area.
  3. Use a buildout command from an already installed buildout to bootstrap a new buildout. (See the section on bootstraping later in this document.)

Often, a software project will be managed in a software repository, such as a subversion repository, that includes some software source directories, buildout configuration files, and a copy of the buildout bootstrap script. To work on the project, one would check out the project from the repository and run the bootstrap script which installs setuptools and zc.buildout into the checkout as well as any parts defined.

We have a sample buildout that we created using the bootstrap command of an existing buildout (method 3 above). It has the absolute minimum information. We have bin, develop-eggs, eggs and parts directories, and a configuration file:

>>> ls(sample_buildout)
d  bin
-  buildout.cfg
d  develop-eggs
d  eggs
d  parts

The bin directory contains scripts.

>>> ls(sample_buildout, 'bin')
-  buildout
>>> ls(sample_buildout, 'eggs')
-  setuptools-0.6-py2.4.egg
-  zc.buildout-1.0-py2.4.egg

The develop-eggs directory is initially empty:

>>> ls(sample_buildout, 'develop-eggs')

The develop-eggs directory holds egg links for software being developed in the buildout. We separate develop-eggs and other eggs to allow eggs directories to be shared across multiple buildouts. For example, a common developer technique is to define a common eggs directory in their home that all non-develop eggs are stored in. This allows larger buildouts to be set up much more quickly and saves disk space.

The parts directory just contains some helpers for the buildout script itself.

>>> ls(sample_buildout, 'parts')
d  buildout

The parts directory provides an area where recipes can install part data. For example, if we built a custom Python, we would install it in the part directory. Part data is stored in a sub-directory of the parts directory with the same name as the part.

Buildouts are defined using configuration files. These are in the format defined by the Python ConfigParser module, with extensions that we'll describe later. By default, when a buildout is run, it looks for the file buildout.cfg in the directory where the buildout is run.

The minimal configuration file has a buildout section that defines no parts:

>>> cat(sample_buildout, 'buildout.cfg')
[buildout]
parts =

A part is simply something to be created by a buildout. It can be almost anything, such as a Python package, a program, a directory, or even a configuration file.

Recipes

A part is created by a recipe. Recipes are always installed as Python eggs. They can be downloaded from a package server, such as the Python Package Index, or they can be developed as part of a project using a "develop" egg.

A develop egg is a special kind of egg that gets installed as an "egg link" that contains the name of a source directory. Develop eggs don't have to be packaged for distribution to be used and can be modified in place, which is especially useful while they are being developed.

Let's create a recipe as part of the sample project. We'll create a recipe for creating directories. First, we'll create a recipes source directory for our local recipes:

>>> mkdir(sample_buildout, 'recipes')

and then we'll create a source file for our mkdir recipe:

>>> write(sample_buildout, 'recipes', 'mkdir.py',
... """
... import logging, os, zc.buildout
...
... class Mkdir:
...
...     def __init__(self, buildout, name, options):
...         self.name, self.options = name, options
...         options['path'] = os.path.join(
...                               buildout['buildout']['directory'],
...                               options['path'],
...                               )
...         if not os.path.isdir(os.path.dirname(options['path'])):
...             logging.getLogger(self.name).error(
...                 'Cannot create %s. %s is not a directory.',
...                 options['path'], os.path.dirname(options['path']))
...             raise zc.buildout.UserError('Invalid Path')
...
...
...     def install(self):
...         path = self.options['path']
...         logging.getLogger(self.name).info(
...             'Creating directory %s', os.path.basename(path))
...         os.mkdir(path)
...         return path
...
...     def update(self):
...         pass
... """)

Currently, recipes must define 3 methods [1]:

  • a constructor,
  • an install method, and
  • an update method.

The constructor is responsible for updating a parts options to reflect data read from other sections. The buildout system keeps track of whether a part specification has changed. A part specification has changed if it's options, after adjusting for data read from other sections, has changed, or if the recipe has changed. Only the options for the part are considered. If data are read from other sections, then that information has to be reflected in the parts options. In the Mkdir example, the given path is interpreted relative to the buildout directory, and data from the buildout directory is read. The path option is updated to reflect this. If the directory option was changed in the buildout sections, we would know to update parts created using the mkdir recipe using relative path names.

When buildout is run, it saves configuration data for installed parts in a file named ".installed.cfg". In subsequent runs, it compares part-configuration data stored in the .installed.cfg file and the part-configuration data loaded from the configuration files as modified by recipe constructors to decide if the configuration of a part has changed. If the configuration has changed, or if the recipe has changed, then the part is uninstalled and reinstalled. The buildout only looks at the part's options, so any data used to configure the part needs to be reflected in the part's options. It is the job of a recipe constructor to make sure that the options include all relevant data.

Of course, parts are also uninstalled if they are no-longer used.

The recipe defines a constructor that takes a buildout object, a part name, and an options dictionary. It saves them in instance attributes. If the path is relative, we'll interpret it as relative to the buildout directory. The buildout object passed in is a mapping from section name to a mapping of options for that section. The buildout directory is available as the directory option of the buildout section. We normalize the path and save it back into the options directory.

The install method is responsible for creating the part. In this case, we need the path of the directory to create. We'll use a path option from our options dictionary. The install method logs what it's doing using the Python logging call. We return the path that we installed. If the part is uninstalled or reinstalled, then the path returned will be removed by the buildout machinery. A recipe install method is expected to return a string, or an iterable of strings containing paths to be removed if a part is uninstalled. For most recipes, this is all of the uninstall support needed. For more complex uninstallation scenarios use Uninstall recipes.

The update method is responsible for updating an already installed part. An empty method is often provided, as in this example, if parts can't be updated. An update method can return None, a string, or an iterable of strings. If a string or iterable of strings is returned, then the saved list of paths to be uninstalled is updated with the new information by adding any new files returned by the update method.

We need to provide packaging information so that our recipe can be installed as a develop egg. The minimum information we need to specify [2] is a name. For recipes, we also need to define the names of the recipe classes as entry points. Packaging information is provided via a setup.py script:

>>> write(sample_buildout, 'recipes', 'setup.py',
... """
... from setuptools import setup
...
... setup(
...     name = "recipes",
...     entry_points = {'zc.buildout': ['mkdir = mkdir:Mkdir']},
...     )
... """)

Our setup script defines an entry point. Entry points provide a way for an egg to define the services it provides. Here we've said that we define a zc.buildout entry point named mkdir. Recipe classes must be exposed as entry points in the zc.buildout group. we give entry points names within the group.

We also need a README.txt for our recipes to avoid an annoying warning from distutils, on which setuptools and zc.buildout are based:

>>> write(sample_buildout, 'recipes', 'README.txt', " ")

Now let's update our buildout.cfg:

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... develop = recipes
... parts = data-dir
...
... [data-dir]
... recipe = recipes:mkdir
... path = mystuff
... """)

Let's go through the changes one by one:

develop = recipes

This tells the buildout to install a development egg for our recipes. Any number of paths can be listed. The paths can be relative or absolute. If relative, they are treated as relative to the buildout directory. They can be directory or file paths. If a file path is given, it should point to a Python setup script. If a directory path is given, it should point to a directory containing a setup.py file. Development eggs are installed before building any parts, as they may provide locally-defined recipes needed by the parts.

parts = data-dir

Here we've named a part to be "built". We can use any name we want except that different part names must be unique and recipes will often use the part name to decide what to do.

[data-dir]
recipe = recipes:mkdir
path = mystuff

When we name a part, we also create a section of the same name that contains part data. In this section, we'll define the recipe to be used to install the part. In this case, we also specify the path to be created.

Let's run the buildout. We do so by running the build script in the buildout:

>>> import os
>>> os.chdir(sample_buildout)
>>> buildout = os.path.join(sample_buildout, 'bin', 'buildout')
>>> print system(buildout),
Develop: '/sample-buildout/recipes'
Installing data-dir.
data-dir: Creating directory mystuff

We see that the recipe created the directory, as expected:

>>> ls(sample_buildout)
-  .installed.cfg
d  bin
-  buildout.cfg
d  develop-eggs
d  eggs
d  mystuff
d  parts
d  recipes

In addition, .installed.cfg has been created containing information about the part we installed:

>>> cat(sample_buildout, '.installed.cfg')
[buildout]
installed_develop_eggs = /sample-buildout/develop-eggs/recipes.egg-link
parts = data-dir
<BLANKLINE>
[data-dir]
__buildout_installed__ = /sample-buildout/mystuff
__buildout_signature__ = recipes-c7vHV6ekIDUPy/7fjAaYjg==
path = /sample-buildout/mystuff
recipe = recipes:mkdir

Note that the directory we installed is included in .installed.cfg. In addition, the path option includes the actual destination directory.

If we change the name of the directory in the configuration file, we'll see that the directory gets removed and recreated:

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... develop = recipes
... parts = data-dir
...
... [data-dir]
... recipe = recipes:mkdir
... path = mydata
... """)
>>> print system(buildout),
Develop: '/sample-buildout/recipes'
Uninstalling data-dir.
Installing data-dir.
data-dir: Creating directory mydata
>>> ls(sample_buildout)
-  .installed.cfg
d  bin
-  buildout.cfg
d  develop-eggs
d  eggs
d  mydata
d  parts
d  recipes

If any of the files or directories created by a recipe are removed, the part will be reinstalled:

>>> rmdir(sample_buildout, 'mydata')
>>> print system(buildout),
Develop: '/sample-buildout/recipes'
Uninstalling data-dir.
Installing data-dir.
data-dir: Creating directory mydata

Error reporting

If a user makes an error, an error needs to be printed and work needs to stop. This is accomplished by logging a detailed error message and then raising a (or an instance of a subclass of a) zc.buildout.UserError exception. Raising an error other than a UserError still displays the error, but labels it as a bug in the buildout software or recipe. In the sample above, of someone gives a non-existent directory to create the directory in:

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... develop = recipes
... parts = data-dir
...
... [data-dir]
... recipe = recipes:mkdir
... path = /xxx/mydata
... """)

We'll get a user error, not a traceback.

>>> print system(buildout),
Develop: '/sample-buildout/recipes'
data-dir: Cannot create /xxx/mydata. /xxx is not a directory.
While:
  Installing.
  Getting section data-dir.
  Initializing part data-dir.
Error: Invalid Path

Recipe Error Handling

If an error occurs during installation, it is up to the recipe to clean up any system side effects, such as files created. Let's update the mkdir recipe to support multiple paths:

>>> write(sample_buildout, 'recipes', 'mkdir.py',
... """
... import logging, os, zc.buildout
...
... class Mkdir:
...
...     def __init__(self, buildout, name, options):
...         self.name, self.options = name, options
...
...         # Normalize paths and check that their parent
...         # directories exist:
...         paths = []
...         for path in options['path'].split():
...             path = os.path.join(buildout['buildout']['directory'], path)
...             if not os.path.isdir(os.path.dirname(path)):
...                 logging.getLogger(self.name).error(
...                     'Cannot create %s. %s is not a directory.',
...                     options['path'], os.path.dirname(options['path']))
...                 raise zc.buildout.UserError('Invalid Path')
...             paths.append(path)
...         options['path'] = ' '.join(paths)
...
...     def install(self):
...         paths = self.options['path'].split()
...         for path in paths:
...             logging.getLogger(self.name).info(
...                 'Creating directory %s', os.path.basename(path))
...             os.mkdir(path)
...         return paths
...
...     def update(self):
...         pass
... """)
>>> clean_up_pyc(sample_buildout, 'recipes', 'mkdir.py')

If there is an error creating a path, the install method will exit and leave previously created paths in place:

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... develop = recipes
... parts = data-dir
...
... [data-dir]
... recipe = recipes:mkdir
... path = foo bin
... """)
>>> print system(buildout), # doctest: +ELLIPSIS
Develop: '/sample-buildout/recipes'
Uninstalling data-dir.
Installing data-dir.
data-dir: Creating directory foo
data-dir: Creating directory bin
While:
  Installing data-dir.
<BLANKLINE>
An internal error occurred due to a bug in either zc.buildout or in a
recipe being used:
Traceback (most recent call last):
  ...
OSError: [Errno 17] File exists: '/sample-buildout/bin'

We meant to create a directory bins, but typed bin. Now foo was left behind.

>>> os.path.exists('foo')
True

If we fix the typo:

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... develop = recipes
... parts = data-dir
...
... [data-dir]
... recipe = recipes:mkdir
... path = foo bins
... """)
>>> print system(buildout), # doctest: +ELLIPSIS
Develop: '/sample-buildout/recipes'
Installing data-dir.
data-dir: Creating directory foo
While:
  Installing data-dir.
<BLANKLINE>
An internal error occurred due to a bug in either zc.buildout or in a
recipe being used:
Traceback (most recent call last):
...
OSError: [Errno 17] File exists: '/sample-buildout/foo'

Now they fail because foo exists, because it was left behind.

>>> remove('foo')

Let's fix the recipe:

>>> write(sample_buildout, 'recipes', 'mkdir.py',
... """
... import logging, os, zc.buildout
...
... class Mkdir:
...
...     def __init__(self, buildout, name, options):
...         self.name, self.options = name, options
...
...         # Normalize paths and check that their parent
...         # directories exist:
...         paths = []
...         for path in options['path'].split():
...             path = os.path.join(buildout['buildout']['directory'], path)
...             if not os.path.isdir(os.path.dirname(path)):
...                 logging.getLogger(self.name).error(
...                     'Cannot create %s. %s is not a directory.',
...                     options['path'], os.path.dirname(options['path']))
...                 raise zc.buildout.UserError('Invalid Path')
...             paths.append(path)
...         options['path'] = ' '.join(paths)
...
...     def install(self):
...         paths = self.options['path'].split()
...         created = []
...         try:
...             for path in paths:
...                 logging.getLogger(self.name).info(
...                     'Creating directory %s', os.path.basename(path))
...                 os.mkdir(path)
...                 created.append(path)
...         except:
...             for d in created:
...                 os.rmdir(d)
...             raise
...
...         return paths
...
...     def update(self):
...         pass
... """)
>>> clean_up_pyc(sample_buildout, 'recipes', 'mkdir.py')

And put back the typo:

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... develop = recipes
... parts = data-dir
...
... [data-dir]
... recipe = recipes:mkdir
... path = foo bin
... """)

When we rerun the buildout:

>>> print system(buildout), # doctest: +ELLIPSIS
Develop: '/sample-buildout/recipes'
Installing data-dir.
data-dir: Creating directory foo
data-dir: Creating directory bin
While:
  Installing data-dir.
<BLANKLINE>
An internal error occurred due to a bug in either zc.buildout or in a
recipe being used:
Traceback (most recent call last):
...
OSError: [Errno 17] File exists: '/sample-buildout/bin'

we get the same error, but we don't get the directory left behind:

>>> os.path.exists('foo')
False

It's critical that recipes clean up partial effects when errors occur. Because recipes most commonly create files and directories, buildout provides a helper API for removing created files when an error occurs. Option objects have a created method that can be called to record files as they are created. If the install or update method returns with an error, then any registered paths are removed automatically. The method returns the files registered and can be used to return the files created. Let's use this API to simplify the recipe:

>>> write(sample_buildout, 'recipes', 'mkdir.py',
... """
... import logging, os, zc.buildout
...
... class Mkdir:
...
...     def __init__(self, buildout, name, options):
...         self.name, self.options = name, options
...
...         # Normalize paths and check that their parent
...         # directories exist:
...         paths = []
...         for path in options['path'].split():
...             path = os.path.join(buildout['buildout']['directory'], path)
...             if not os.path.isdir(os.path.dirname(path)):
...                 logging.getLogger(self.name).error(
...                     'Cannot create %s. %s is not a directory.',
...                     options['path'], os.path.dirname(options['path']))
...                 raise zc.buildout.UserError('Invalid Path')
...             paths.append(path)
...         options['path'] = ' '.join(paths)
...
...     def install(self):
...         paths = self.options['path'].split()
...         for path in paths:
...             logging.getLogger(self.name).info(
...                 'Creating directory %s', os.path.basename(path))
...             os.mkdir(path)
...             self.options.created(path)
...
...         return self.options.created()
...
...     def update(self):
...         pass
... """)
>>> clean_up_pyc(sample_buildout, 'recipes', 'mkdir.py')

We returned by calling created, taking advantage of the fact that it returns the registered paths. We did this for illustrative purposes. It would be simpler just to return the paths as before.

If we rerun the buildout, again, we'll get the error and no directories will be created:

>>> print system(buildout), # doctest: +ELLIPSIS
Develop: '/sample-buildout/recipes'
Installing data-dir.
data-dir: Creating directory foo
data-dir: Creating directory bin
While:
  Installing data-dir.
<BLANKLINE>
An internal error occurred due to a bug in either zc.buildout or in a
recipe being used:
Traceback (most recent call last):
...
OSError: [Errno 17] File exists: '/sample-buildout/bin'
>>> os.path.exists('foo')
False

Now, we'll fix the typo again and we'll get the directories we expect:

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... develop = recipes
... parts = data-dir
...
... [data-dir]
... recipe = recipes:mkdir
... path = foo bins
... """)
>>> print system(buildout),
Develop: '/sample-buildout/recipes'
Installing data-dir.
data-dir: Creating directory foo
data-dir: Creating directory bins
>>> os.path.exists('foo')
True
>>> os.path.exists('bins')
True

Configuration file syntax

As mentioned earlier, buildout configuration files use the format defined by the Python ConfigParser module with extensions. The extensions are:

  • option names are case sensitive
  • option values can use a substitution syntax, described below, to refer to option values in specific sections.
  • option values can be appended or removed using the - and + operators.

The ConfigParser syntax is very flexible. Section names can contain any characters other than newlines and right square braces ("]"). Option names can contain any characters other than newlines, colons, and equal signs, can not start with a space, and don't include trailing spaces.

It is likely that, in the future, some characters will be given special buildout-defined meanings. This is already true of the characters ":", "$", "%", "(", and ")". For now, it is a good idea to keep section and option names simple, sticking to alphanumeric characters, hyphens, and periods.

Annotated sections

When used with the annotate command, buildout displays annotated sections. All sections are displayed, sorted alphabetically. For each section, all key-value pairs are displayed, sorted alphabetically, along with the origin of the value (file name or COMPUTED_VALUE, DEFAULT_VALUE, COMMAND_LINE_VALUE).

>>> print system(buildout+ ' annotate'),
... # doctest: +ELLIPSIS +NORMALIZE_WHITESPACE
<BLANKLINE>
Annotated sections
==================
<BLANKLINE>
[buildout]
accept-buildout-test-releases= false
    DEFAULT_VALUE
allow-hosts= *
    DEFAULT_VALUE
allow-picked-versions= true
    DEFAULT_VALUE
allowed-eggs-from-site-packages= *
    DEFAULT_VALUE
bin-directory= bin
    DEFAULT_VALUE
develop= recipes
    /sample-buildout/buildout.cfg
develop-eggs-directory= develop-eggs
    DEFAULT_VALUE
directory= /sample-buildout
    COMPUTED_VALUE
eggs-directory= eggs
    DEFAULT_VALUE
exec-sitecustomize= true
    DEFAULT_VALUE
executable= ...
    DEFAULT_VALUE
find-links=
    DEFAULT_VALUE
include-site-packages= true
    DEFAULT_VALUE
install-from-cache= false
    DEFAULT_VALUE
installed= .installed.cfg
    DEFAULT_VALUE
log-format=
    DEFAULT_VALUE
log-level= INFO
    DEFAULT_VALUE
newest= true
    DEFAULT_VALUE
offline= false
    DEFAULT_VALUE
parts= data-dir
    /sample-buildout/buildout.cfg
parts-directory= parts
    DEFAULT_VALUE
prefer-final= false
    DEFAULT_VALUE
python= buildout
    DEFAULT_VALUE
relative-paths= false
    DEFAULT_VALUE
socket-timeout=
    DEFAULT_VALUE
unzip= false
    DEFAULT_VALUE
use-dependency-links= true
    DEFAULT_VALUE
<BLANKLINE>
[data-dir]
path= foo bins
    /sample-buildout/buildout.cfg
recipe= recipes:mkdir
    /sample-buildout/buildout.cfg
<BLANKLINE>

Variable substitutions

Buildout configuration files support variable substitution. To illustrate this, we'll create an debug recipe to allow us to see interactions with the buildout:

>>> write(sample_buildout, 'recipes', 'debug.py',
... """
... class Debug:
...
...     def __init__(self, buildout, name, options):
...         self.buildout = buildout
...         self.name = name
...         self.options = options
...
...     def install(self):
...         items = self.options.items()
...         items.sort()
...         for option, value in items:
...             print option, value
...         return ()
...
...     update = install
... """)

This recipe doesn't actually create anything. The install method doesn't return anything, because it didn't create any files or directories.

We also have to update our setup script:

>>> write(sample_buildout, 'recipes', 'setup.py',
... """
... from setuptools import setup
... entry_points = (
... '''
... [zc.buildout]
... mkdir = mkdir:Mkdir
... debug = debug:Debug
... ''')
... setup(name="recipes", entry_points=entry_points)
... """)

We've rearranged the script a bit to make the entry points easier to edit. In particular, entry points are now defined as a configuration string, rather than a dictionary.

Let's update our configuration to provide variable substitution examples:

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... develop = recipes
... parts = data-dir debug
... log-level = INFO
...
... [debug]
... recipe = recipes:debug
... File 1 = ${data-dir:path}/file
... File 2 = ${debug:File 1}/log
...
... [data-dir]
... recipe = recipes:mkdir
... path = mydata
... """)

We used a string-template substitution for File 1 and File 2. This type of substitution uses the string.Template syntax. Names substituted are qualified option names, consisting of a section name and option name joined by a colon.

Now, if we run the buildout, we'll see the options with the values substituted.

>>> print system(buildout),
Develop: '/sample-buildout/recipes'
Uninstalling data-dir.
Installing data-dir.
data-dir: Creating directory mydata
Installing debug.
File 1 /sample-buildout/mydata/file
File 2 /sample-buildout/mydata/file/log
recipe recipes:debug

Note that the substitution of the data-dir path option reflects the update to the option performed by the mkdir recipe.

It might seem surprising that mydata was created again. This is because we changed our recipes package by adding the debug module. The buildout system didn't know if this module could effect the mkdir recipe, so it assumed it could and reinstalled mydata. If we rerun the buildout:

>>> print system(buildout),
Develop: '/sample-buildout/recipes'
Updating data-dir.
Updating debug.
File 1 /sample-buildout/mydata/file
File 2 /sample-buildout/mydata/file/log
recipe recipes:debug

We can see that mydata was not recreated.

Note that, in this case, we didn't specify a log level, so we didn't get output about what the buildout was doing.

Section and option names in variable substitutions are only allowed to contain alphanumeric characters, hyphens, periods and spaces. This restriction might be relaxed in future releases.

We can ommit the section name in a variable substitution to refer to the current section. We can also use the special option, _buildout_section_name_ to get the current section name.

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... develop = recipes
... parts = data-dir debug
... log-level = INFO
...
... [debug]
... recipe = recipes:debug
... File 1 = ${data-dir:path}/file
... File 2 = ${:File 1}/log
... my_name = ${:_buildout_section_name_}
...
... [data-dir]
... recipe = recipes:mkdir
... path = mydata
... """)
>>> print system(buildout),
Develop: '/sample-buildout/recipes'
Uninstalling debug.
Updating data-dir.
Installing debug.
File 1 /sample-buildout/mydata/file
File 2 /sample-buildout/mydata/file/log
my_name debug
recipe recipes:debug

Automatic part selection and ordering

When a section with a recipe is referred to, either through variable substitution or by an initializing recipe, the section is treated as a part and added to the part list before the referencing part. For example, we can leave data-dir out of the parts list:

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... develop = recipes
... parts = debug
... log-level = INFO
...
... [debug]
... recipe = recipes:debug
... File 1 = ${data-dir:path}/file
... File 2 = ${debug:File 1}/log
...
... [data-dir]
... recipe = recipes:mkdir
... path = mydata
... """)

It will still be treated as a part:

>>> print system(buildout),
Develop: '/sample-buildout/recipes'
Uninstalling debug.
Updating data-dir.
Installing debug.
File 1 /sample-buildout/mydata/file
File 2 /sample-buildout/mydata/file/log
recipe recipes:debug
>>> cat('.installed.cfg') # doctest: +ELLIPSIS
[buildout]
installed_develop_eggs = /sample-buildout/develop-eggs/recipes.egg-link
parts = data-dir debug
...

Note that the data-dir part is included before the debug part, because the debug part refers to the data-dir part. Even if we list the data-dir part after the debug part, it will be included before:

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... develop = recipes
... parts = debug data-dir
... log-level = INFO
...
... [debug]
... recipe = recipes:debug
... File 1 = ${data-dir:path}/file
... File 2 = ${debug:File 1}/log
...
... [data-dir]
... recipe = recipes:mkdir
... path = mydata
... """)

It will still be treated as a part:

>>> print system(buildout),
Develop: '/sample-buildout/recipes'
Updating data-dir.
Updating debug.
File 1 /sample-buildout/mydata/file
File 2 /sample-buildout/mydata/file/log
recipe recipes:debug
>>> cat('.installed.cfg') # doctest: +ELLIPSIS
[buildout]
installed_develop_eggs = /sample-buildout/develop-eggs/recipes.egg-link
parts = data-dir debug
...

Extending sections (macros)

A section (other than the buildout section) can extend one or more other sections using the <= option. Options from the referenced sections are copied to the refering section before variable substitution. This, together with the ability to refer to variables of the current section allows sections to be used as macros.

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... develop = recipes
... parts = myfiles
... log-level = INFO
...
... [debug]
... recipe = recipes:debug
...
... [with_file1]
... <= debug
... file1 = ${:path}/file1
... color = red
...
... [with_file2]
... <= debug
... file2 = ${:path}/file2
... color = blue
...
... [myfiles]
... <= with_file1
...    with_file2
... path = mydata
... """)
>>> print system(buildout),
Develop: '/sample-buildout/recipes'
Uninstalling debug.
Uninstalling data-dir.
Installing myfiles.
color blue
file1 mydata/file1
file2 mydata/file2
path mydata
recipe recipes:debug

In this example, the debug, with_file1 and with_file2 sections act as macros. In particular, the variable substitutions are performed relative to the myfiles section.

Adding and removing options

We can append and remove values to an option by using the + and - operators.

This is illustrated below; first we define a base configuration.

>>> write(sample_buildout, 'base.cfg',
... """
... [buildout]
... parts = part1 part2 part3
...
... [part1]
... recipe =
... option = a1 a2
...
... [part2]
... recipe =
... option = b1
...     b2
...     b3
...     b4
...
... [part3]
... recipe =
... option = c1 c2
...
... [part4]
... recipe =
... option = d2
...     d3
...     d5
...
... """)

Extending this configuration, we can "adjust" the values set in the base configuration file.

>>> write(sample_buildout, 'extension1.cfg',
... """
... [buildout]
... extends = base.cfg
...
... # appending values
... [part1]
... option += a3 a4
...
... # removing values
... [part2]
... option -= b1
...      b2
...
... # alt. spelling
... [part3]
... option+=c3 c4 c5
...
... # combining both adding and removing
... [part4]
... option += d1
...      d4
... option -= d5
...
... # normal assignment
... [part5]
... option = h1 h2
... """)

An additional extension.

>>> write(sample_buildout, 'extension2.cfg',
... """
... [buildout]
... extends = extension1.cfg
...
... # appending values
... [part1]
... option += a5
...
... # removing values
... [part2]
... option -= b1
...     b2
...     b3
...
... """)

To verify that the options are adjusted correctly, we'll set up an extension that prints out the options.

>>> mkdir(sample_buildout, 'demo')
>>> write(sample_buildout, 'demo', 'demo.py',
... """
... def ext(buildout):
...     print [part['option'] for name, part in buildout.items() \
...           if name.startswith('part')]
... """)
>>> write(sample_buildout, 'demo', 'setup.py',
... """
... from setuptools import setup
...
... setup(
...     name="demo",
...     entry_points={'zc.buildout.extension': ['ext = demo:ext']},
...     )
... """)

Set up a buildout configuration for this extension.

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... develop = demo
... parts =
... """)
>>> os.chdir(sample_buildout)
>>> print system(os.path.join(sample_buildout, 'bin', 'buildout')), # doctest: +ELLIPSIS
Develop: '/sample-buildout/demo'
Uninstalling myfiles.
Getting distribution for 'recipes'.
zip_safe flag not set; analyzing archive contents...
Got recipes 0.0.0.
warning: install_lib: 'build/lib...' does not exist -- no Python modules to install

Verify option values.

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... develop = demo
... extensions = demo
... extends = extension2.cfg
... """)
>>> print system(os.path.join('bin', 'buildout')),
['a1 a2/na3 a4/na5', 'b4', 'c1 c2/nc3 c4 c5', 'd2/nd3/nd1/nd4', 'h1 h2']
Develop: '/sample-buildout/demo'

Annotated sections output shows which files are responsible for which operations.

>>> print system(os.path.join('bin', 'buildout') + ' annotate'),
... # doctest: +ELLIPSIS +NORMALIZE_WHITESPACE
<BLANKLINE>
Annotated sections
==================
...
<BLANKLINE>
[part1]
option= a1 a2
a3 a4
a5
    /sample-buildout/base.cfg
+=  /sample-buildout/extension1.cfg
+=  /sample-buildout/extension2.cfg
recipe=
    /sample-buildout/base.cfg
<BLANKLINE>
[part2]
option= b4
    /sample-buildout/base.cfg
-=  /sample-buildout/extension1.cfg
-=  /sample-buildout/extension2.cfg
recipe=
    /sample-buildout/base.cfg
<BLANKLINE>
[part3]
option= c1 c2
c3 c4 c5
    /sample-buildout/base.cfg
+=  /sample-buildout/extension1.cfg
recipe=
    /sample-buildout/base.cfg
<BLANKLINE>
[part4]
option= d2
d3
d1
d4
    /sample-buildout/base.cfg
+=  /sample-buildout/extension1.cfg
-=  /sample-buildout/extension1.cfg
recipe=
    /sample-buildout/base.cfg
<BLANKLINE>
[part5]
option= h1 h2
    /sample-buildout/extension1.cfg

Cleanup.

>>> os.remove(os.path.join(sample_buildout, 'base.cfg'))
>>> os.remove(os.path.join(sample_buildout, 'extension1.cfg'))
>>> os.remove(os.path.join(sample_buildout, 'extension2.cfg'))

Multiple configuration files

A configuration file can "extend" another configuration file. Options are read from the other configuration file if they aren't already defined by your configuration file.

The configuration files your file extends can extend other configuration files. The same file may be used more than once although, of course, cycles aren't allowed.

To see how this works, we use an example:

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... extends = base.cfg
...
... [debug]
... op = buildout
... """)
>>> write(sample_buildout, 'base.cfg',
... """
... [buildout]
... develop = recipes
... parts = debug
...
... [debug]
... recipe = recipes:debug
... op = base
... """)
>>> print system(buildout),
Develop: '/sample-buildout/recipes'
Installing debug.
op buildout
recipe recipes:debug

The example is pretty trivial, but the pattern it illustrates is pretty common. In a more practical example, the base buildout might represent a product and the extending buildout might be a customization.

Here is a more elaborate example.

>>> other = tmpdir('other')
>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... extends = b1.cfg b2.cfg %(b3)s
...
... [debug]
... op = buildout
... """ % dict(b3=os.path.join(other, 'b3.cfg')))
>>> write(sample_buildout, 'b1.cfg',
... """
... [buildout]
... extends = base.cfg
...
... [debug]
... op1 = b1 1
... op2 = b1 2
... """)
>>> write(sample_buildout, 'b2.cfg',
... """
... [buildout]
... extends = base.cfg
...
... [debug]
... op2 = b2 2
... op3 = b2 3
... """)
>>> write(other, 'b3.cfg',
... """
... [buildout]
... extends = b3base.cfg
...
... [debug]
... op4 = b3 4
... """)
>>> write(other, 'b3base.cfg',
... """
... [debug]
... op5 = b3base 5
... """)
>>> write(sample_buildout, 'base.cfg',
... """
... [buildout]
... develop = recipes
... parts = debug
...
... [debug]
... recipe = recipes:debug
... name = base
... """)
>>> print system(buildout),
Develop: '/sample-buildout/recipes'
Uninstalling debug.
Installing debug.
name base
op buildout
op1 b1 1
op2 b2 2
op3 b2 3
op4 b3 4
op5 b3base 5
recipe recipes:debug

There are several things to note about this example:

  • We can name multiple files in an extends option.
  • We can reference files recursively.
  • Relative file names in extended options are interpreted relative to the directory containing the referencing configuration file.

Loading Configuration from URLs

Configuration files can be loaded from URLs. To see how this works, we'll set up a web server with some configuration files.

>>> server_data = tmpdir('server_data')
>>> write(server_data, "r1.cfg",
... """
... [debug]
... op1 = r1 1
... op2 = r1 2
... """)
>>> write(server_data, "r2.cfg",
... """
... [buildout]
... extends = r1.cfg
...
... [debug]
... op2 = r2 2
... op3 = r2 3
... """)
>>> server_url = start_server(server_data)
>>> write('client.cfg',
... """
... [buildout]
... develop = recipes
... parts = debug
... extends = %(url)s/r2.cfg
...
... [debug]
... recipe = recipes:debug
... name = base
... """ % dict(url=server_url))
>>> print system(buildout+ ' -c client.cfg'),
Develop: '/sample-buildout/recipes'
Uninstalling debug.
Installing debug.
name base
op1 r1 1
op2 r2 2
op3 r2 3
recipe recipes:debug

Here we specified a URL for the file we extended. The file we downloaded, itself referred to a file on the server using a relative URL reference. Relative references are interpreted relative to the base URL when they appear in configuration files loaded via URL.

We can also specify a URL as the configuration file to be used by a buildout.

>>> os.remove('client.cfg')
>>> write(server_data, 'remote.cfg',
... """
... [buildout]
... develop = recipes
... parts = debug
... extends = r2.cfg
...
... [debug]
... recipe = recipes:debug
... name = remote
... """)
>>> print system(buildout + ' -c ' + server_url + '/remote.cfg'),
While:
  Initializing.
Error: Missing option: buildout:directory

Normally, the buildout directory defaults to directory containing a configuration file. This won't work for configuration files loaded from URLs. In this case, the buildout directory would normally be defined on the command line:

>>> print system(buildout
...              + ' -c ' + server_url + '/remote.cfg'
...              + ' buildout:directory=' + sample_buildout
...              ),
Develop: '/sample-buildout/recipes'
Uninstalling debug.
Installing debug.
name remote
op1 r1 1
op2 r2 2
op3 r2 3
recipe recipes:debug

User defaults

If the file $HOME/.buildout/default.cfg, exists, it is read before reading the configuration file. ($HOME is the value of the HOME environment variable. The '/' is replaced by the operating system file delimiter.)

>>> old_home = os.environ['HOME']
>>> home = tmpdir('home')
>>> mkdir(home, '.buildout')
>>> write(home, '.buildout', 'default.cfg',
... """
... [debug]
... op1 = 1
... op7 = 7
... """)
>>> os.environ['HOME'] = home
>>> print system(buildout),
Develop: '/sample-buildout/recipes'
Uninstalling debug.
Installing debug.
name base
op buildout
op1 b1 1
op2 b2 2
op3 b2 3
op4 b3 4
op5 b3base 5
op7 7
recipe recipes:debug

A buildout command-line argument, -U, can be used to suppress reading user defaults:

>>> print system(buildout + ' -U'),
Develop: '/sample-buildout/recipes'
Uninstalling debug.
Installing debug.
name base
op buildout
op1 b1 1
op2 b2 2
op3 b2 3
op4 b3 4
op5 b3base 5
recipe recipes:debug
>>> os.environ['HOME'] = old_home

Log level

We can control the level of logging by specifying a log level in out configuration file. For example, so suppress info messages, we can set the logging level to WARNING

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... log-level = WARNING
... extends = b1.cfg b2.cfg
... """)
>>> print system(buildout),
name base
op1 b1 1
op2 b2 2
op3 b2 3
recipe recipes:debug

Uninstall recipes

As we've seen, when parts are installed, buildout keeps track of files and directories that they create. When the parts are uninstalled these files and directories are deleted.

Sometimes more clean up is needed. For example, a recipe might add a system service by calling chkconfig --add during installation. Later during uninstallation, chkconfig --del will need to be called to remove the system service.

In order to deal with these uninstallation issues, you can register uninstall recipes. Uninstall recipes are registered using the 'zc.buildout.uninstall' entry point. Parts specify uninstall recipes using the 'uninstall' option.

In comparison to regular recipes, uninstall recipes are much simpler. They are simply callable objects that accept the name of the part to be uninstalled and the part's options dictionary. Uninstall recipes don't have access to the part itself since it maybe not be able to be instantiated at uninstallation time.

Here's a recipe that simulates installation of a system service, along with an uninstall recipe that simulates removing the service.

>>> write(sample_buildout, 'recipes', 'service.py',
... """
... class Service:
...
...     def __init__(self, buildout, name, options):
...         self.buildout = buildout
...         self.name = name
...         self.options = options
...
...     def install(self):
...         print "chkconfig --add %s" % self.options['script']
...         return ()
...
...     def update(self):
...         pass
...
...
... def uninstall_service(name, options):
...     print "chkconfig --del %s" % options['script']
... """)

To use these recipes we must register them using entry points. Make sure to use the same name for the recipe and uninstall recipe. This is required to let buildout know which uninstall recipe goes with which recipe.

>>> write(sample_buildout, 'recipes', 'setup.py',
... """
... from setuptools import setup
... entry_points = (
... '''
... [zc.buildout]
... mkdir = mkdir:Mkdir
... debug = debug:Debug
... service = service:Service
...
... [zc.buildout.uninstall]
... service = service:uninstall_service
... ''')
... setup(name="recipes", entry_points=entry_points)
... """)

Here's how these recipes could be used in a buildout:

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... develop = recipes
... parts = service
...
... [service]
... recipe = recipes:service
... script = /path/to/script
... """)

When the buildout is run the service will be installed

>>> print system(buildout)
Develop: '/sample-buildout/recipes'
Uninstalling debug.
Installing service.
chkconfig --add /path/to/script
<BLANKLINE>

The service has been installed. If the buildout is run again with no changes, the service shouldn't be changed.

>>> print system(buildout)
Develop: '/sample-buildout/recipes'
Updating service.
<BLANKLINE>

Now we change the service part to trigger uninstallation and re-installation.

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... develop = recipes
... parts = service
...
... [service]
... recipe = recipes:service
... script = /path/to/a/different/script
... """)
>>> print system(buildout)
Develop: '/sample-buildout/recipes'
Uninstalling service.
Running uninstall recipe.
chkconfig --del /path/to/script
Installing service.
chkconfig --add /path/to/a/different/script
<BLANKLINE>

Now we remove the service part, and add another part.

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... develop = recipes
... parts = debug
...
... [debug]
... recipe = recipes:debug
... """)
>>> print system(buildout)
Develop: '/sample-buildout/recipes'
Uninstalling service.
Running uninstall recipe.
chkconfig --del /path/to/a/different/script
Installing debug.
recipe recipes:debug
<BLANKLINE>

Uninstall recipes don't have to take care of removing all the files and directories created by the part. This is still done automatically, following the execution of the uninstall recipe. An upshot is that an uninstallation recipe can access files and directories created by a recipe before they are deleted.

For example, here's an uninstallation recipe that simulates backing up a directory before it is deleted. It is designed to work with the mkdir recipe introduced earlier.

>>> write(sample_buildout, 'recipes', 'backup.py',
... """
... import os
... def backup_directory(name, options):
...     path = options['path']
...     size = len(os.listdir(path))
...     print "backing up directory %s of size %s" % (path, size)
... """)

It must be registered with the zc.buildout.uninstall entry point. Notice how it is given the name 'mkdir' to associate it with the mkdir recipe.

>>> write(sample_buildout, 'recipes', 'setup.py',
... """
... from setuptools import setup
... entry_points = (
... '''
... [zc.buildout]
... mkdir = mkdir:Mkdir
... debug = debug:Debug
... service = service:Service
...
... [zc.buildout.uninstall]
... uninstall_service = service:uninstall_service
... mkdir = backup:backup_directory
... ''')
... setup(name="recipes", entry_points=entry_points)
... """)

Now we can use it with a mkdir part.

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... develop = recipes
... parts = dir debug
...
... [dir]
... recipe = recipes:mkdir
... path = my_directory
...
... [debug]
... recipe = recipes:debug
... """)

Run the buildout to install the part.

>>> print system(buildout)
Develop: '/sample-buildout/recipes'
Uninstalling debug.
Installing dir.
dir: Creating directory my_directory
Installing debug.
recipe recipes:debug
<BLANKLINE>

Now we remove the part from the configuration file.

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... develop = recipes
... parts = debug
...
... [debug]
... recipe = recipes:debug
... """)

When the buildout is run the part is removed, and the uninstall recipe is run before the directory is deleted.

>>> print system(buildout)
Develop: '/sample-buildout/recipes'
Uninstalling dir.
Running uninstall recipe.
backing up directory /sample-buildout/my_directory of size 0
Updating debug.
recipe recipes:debug
<BLANKLINE>

Now we will return the registration to normal for the benefit of the rest of the examples.

>>> write(sample_buildout, 'recipes', 'setup.py',
... """
... from setuptools import setup
... entry_points = (
... '''
... [zc.buildout]
... mkdir = mkdir:Mkdir
... debug = debug:Debug
... ''')
... setup(name="recipes", entry_points=entry_points)
... """)

Command-line usage

A number of arguments can be given on the buildout command line. The command usage is:

buildout [options and assignments] [command [command arguments]]

The following options are supported:

-h (or --help)
Print basic usage information. If this option is used, then all other options are ignored.
-c filename The -c option can be used to specify a configuration file, rather than buildout.cfg in the current directory.
-t socket_timeout
 Specify the socket timeout in seconds.
-v Increment the verbosity by 10. The verbosity is used to adjust the logging level. The verbosity is subtracted from the numeric value of the log-level option specified in the configuration file.
-q Decrement the verbosity by 10.
-U Don't read user-default configuration.
-o Run in off-line mode. This is equivalent to the assignment buildout:offline=true.
-O Run in non-off-line mode. This is equivalent to the assignment buildout:offline=false. This is the default buildout mode. The -O option would normally be used to override a true offline setting in a configuration file.
-n Run in newest mode. This is equivalent to the assignment buildout:newest=true. With this setting, which is the default, buildout will try to find the newest versions of distributions available that satisfy its requirements.
-N Run in non-newest mode. This is equivalent to the assignment buildout:newest=false. With this setting, buildout will not seek new distributions if installed distributions satisfy it's requirements.

Assignments are of the form:

section_name:option_name=value

Options and assignments can be given in any order.

Here's an example:

>>> write(sample_buildout, 'other.cfg',
... """
... [buildout]
... develop = recipes
... parts = debug
... installed = .other.cfg
... log-level = WARNING
...
... [debug]
... name = other
... recipe = recipes:debug
... """)

Note that we used the installed buildout option to specify an alternate file to store information about installed parts.

>>> print system(buildout+' -c other.cfg debug:op1=foo -v'),
Develop: '/sample-buildout/recipes'
Installing debug.
name other
op1 foo
recipe recipes:debug

Here we used the -c option to specify an alternate configuration file, and the -v option to increase the level of logging from the default, WARNING.

Options can also be combined in the usual Unix way, as in:

>>> print system(buildout+' -vcother.cfg debug:op1=foo'),
Develop: '/sample-buildout/recipes'
Updating debug.
name other
op1 foo
recipe recipes:debug

Here we combined the -v and -c options with the configuration file name. Note that the -c option has to be last, because it takes an argument.

>>> os.remove(os.path.join(sample_buildout, 'other.cfg'))
>>> os.remove(os.path.join(sample_buildout, '.other.cfg'))

The most commonly used command is 'install' and it takes a list of parts to install. if any parts are specified, only those parts are installed. To illustrate this, we'll update our configuration and run the buildout in the usual way:

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... develop = recipes
... parts = debug d1 d2 d3
...
... [d1]
... recipe = recipes:mkdir
... path = d1
...
... [d2]
... recipe = recipes:mkdir
... path = d2
...
... [d3]
... recipe = recipes:mkdir
... path = d3
...
... [debug]
... recipe = recipes:debug
... """)
>>> print system(buildout),
Develop: '/sample-buildout/recipes'
Uninstalling debug.
Installing debug.
recipe recipes:debug
Installing d1.
d1: Creating directory d1
Installing d2.
d2: Creating directory d2
Installing d3.
d3: Creating directory d3
>>> ls(sample_buildout)
-  .installed.cfg
-  b1.cfg
-  b2.cfg
-  base.cfg
d  bin
-  buildout.cfg
d  d1
d  d2
d  d3
d  demo
d  develop-eggs
d  eggs
d  parts
d  recipes
>>> cat(sample_buildout, '.installed.cfg')
... # doctest: +NORMALIZE_WHITESPACE
[buildout]
installed_develop_eggs = /sample-buildout/develop-eggs/recipes.egg-link
parts = debug d1 d2 d3
<BLANKLINE>
[debug]
__buildout_installed__ =
__buildout_signature__ = recipes-PiIFiO8ny5yNZ1S3JfT0xg==
recipe = recipes:debug
<BLANKLINE>
[d1]
__buildout_installed__ = /sample-buildout/d1
__buildout_signature__ = recipes-PiIFiO8ny5yNZ1S3JfT0xg==
path = /sample-buildout/d1
recipe = recipes:mkdir
<BLANKLINE>
[d2]
__buildout_installed__ = /sample-buildout/d2
__buildout_signature__ = recipes-PiIFiO8ny5yNZ1S3JfT0xg==
path = /sample-buildout/d2
recipe = recipes:mkdir
<BLANKLINE>
[d3]
__buildout_installed__ = /sample-buildout/d3
__buildout_signature__ = recipes-PiIFiO8ny5yNZ1S3JfT0xg==
path = /sample-buildout/d3
recipe = recipes:mkdir

Now we'll update our configuration file:

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... develop = recipes
... parts = debug d2 d3 d4
...
... [d2]
... recipe = recipes:mkdir
... path = data2
...
... [d3]
... recipe = recipes:mkdir
... path = data3
...
... [d4]
... recipe = recipes:mkdir
... path = ${d2:path}-extra
...
... [debug]
... recipe = recipes:debug
... x = 1
... """)

and run the buildout specifying just d3 and d4:

>>> print system(buildout+' install d3 d4'),
Develop: '/sample-buildout/recipes'
Uninstalling d3.
Installing d3.
d3: Creating directory data3
Installing d4.
d4: Creating directory data2-extra
>>> ls(sample_buildout)
-  .installed.cfg
-  b1.cfg
-  b2.cfg
-  base.cfg
d  bin
-  buildout.cfg
d  d1
d  d2
d  data2-extra
d  data3
d  demo
d  develop-eggs
d  eggs
d  parts
d  recipes

Only the d3 and d4 recipes ran. d3 was removed and data3 and data2-extra were created.

The .installed.cfg is only updated for the recipes that ran:

>>> cat(sample_buildout, '.installed.cfg')
... # doctest: +NORMALIZE_WHITESPACE
[buildout]
installed_develop_eggs = /sample-buildout/develop-eggs/recipes.egg-link
parts = debug d1 d2 d3 d4
<BLANKLINE>
[debug]
__buildout_installed__ =
__buildout_signature__ = recipes-PiIFiO8ny5yNZ1S3JfT0xg==
recipe = recipes:debug
<BLANKLINE>
[d1]
__buildout_installed__ = /sample-buildout/d1
__buildout_signature__ = recipes-PiIFiO8ny5yNZ1S3JfT0xg==
path = /sample-buildout/d1
recipe = recipes:mkdir
<BLANKLINE>
[d2]
__buildout_installed__ = /sample-buildout/d2
__buildout_signature__ = recipes-PiIFiO8ny5yNZ1S3JfT0xg==
path = /sample-buildout/d2
recipe = recipes:mkdir
<BLANKLINE>
[d3]
__buildout_installed__ = /sample-buildout/data3
__buildout_signature__ = recipes-PiIFiO8ny5yNZ1S3JfT0xg==
path = /sample-buildout/data3
recipe = recipes:mkdir
<BLANKLINE>
[d4]
__buildout_installed__ = /sample-buildout/data2-extra
__buildout_signature__ = recipes-PiIFiO8ny5yNZ1S3JfT0xg==
path = /sample-buildout/data2-extra
recipe = recipes:mkdir

Note that the installed data for debug, d1, and d2 haven't changed, because we didn't install those parts and that the d1 and d2 directories are still there.

Now, if we run the buildout without the install command:

>>> print system(buildout),
Develop: '/sample-buildout/recipes'
Uninstalling d2.
Uninstalling d1.
Uninstalling debug.
Installing debug.
recipe recipes:debug
x 1
Installing d2.
d2: Creating directory data2
Updating d3.
Updating d4.

We see the output of the debug recipe and that data2 was created. We also see that d1 and d2 have gone away:

>>> ls(sample_buildout)
-  .installed.cfg
-  b1.cfg
-  b2.cfg
-  base.cfg
d  bin
-  buildout.cfg
d  data2
d  data2-extra
d  data3
d  demo
d  develop-eggs
d  eggs
d  parts
d  recipes

Alternate directory and file locations

The buildout normally puts the bin, eggs, and parts directories in the directory in the directory containing the configuration file. You can provide alternate locations, and even names for these directories.

>>> alt = tmpdir('sample-alt')
>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... develop = recipes
... parts =
... develop-eggs-directory = %(developbasket)s
... eggs-directory = %(basket)s
... bin-directory = %(scripts)s
... parts-directory = %(work)s
... """ % dict(
...    developbasket = os.path.join(alt, 'developbasket'),
...    basket = os.path.join(alt, 'basket'),
...    scripts = os.path.join(alt, 'scripts'),
...    work = os.path.join(alt, 'work'),
... ))
>>> print system(buildout),
Creating directory '/sample-alt/scripts'.
Creating directory '/sample-alt/work'.
Creating directory '/sample-alt/basket'.
Creating directory '/sample-alt/developbasket'.
Develop: '/sample-buildout/recipes'
Uninstalling d4.
Uninstalling d3.
Uninstalling d2.
Uninstalling debug.
>>> ls(alt)
d  basket
d  developbasket
d  scripts
d  work
>>> ls(alt, 'developbasket')
-  recipes.egg-link

You can also specify an alternate buildout directory:

>>> rmdir(alt)
>>> alt = tmpdir('sample-alt')
>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... directory = %(alt)s
... develop = %(recipes)s
... parts =
... """ % dict(
...    alt=alt,
...    recipes=os.path.join(sample_buildout, 'recipes'),
...    ))
>>> print system(buildout),
Creating directory '/sample-alt/bin'.
Creating directory '/sample-alt/parts'.
Creating directory '/sample-alt/eggs'.
Creating directory '/sample-alt/develop-eggs'.
Develop: '/sample-buildout/recipes'
>>> ls(alt)
-  .installed.cfg
d  bin
d  develop-eggs
d  eggs
d  parts
>>> ls(alt, 'develop-eggs')
-  recipes.egg-link

Logging control

Three buildout options are used to control logging:

log-level
specifies the log level
verbosity
adjusts the log level
log-format
allows an alternate logging for mat to be specified

We've already seen the log level and verbosity. Let's look at an example of changing the format:

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... develop = recipes
... parts =
... log-level = 25
... verbosity = 5
... log-format = %(levelname)s %(message)s
... """)

Here, we've changed the format to include the log-level name, rather than the logger name.

We've also illustrated, with a contrived example, that the log level can be a numeric value and that the verbosity can be specified in the configuration file. Because the verbosity is subtracted from the log level, we get a final log level of 20, which is the INFO level.

>>> print system(buildout),
INFO Develop: '/sample-buildout/recipes'

Predefined buildout options

Buildouts have a number of predefined options that recipes can use and that users can override in their configuration files. To see these, we'll run a minimal buildout configuration with a debug logging level. One of the features of debug logging is that the configuration database is shown.

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... parts =
... """)
>>> print system(buildout+' -vv'), # doctest: +NORMALIZE_WHITESPACE
Installing 'zc.buildout >=1.9a1, <2dev', 'setuptools'.
We have a develop egg: zc.buildout X.X.
We have the best distribution that satisfies 'setuptools'.
Picked: setuptools = V.V
<BLANKLINE>
Configuration data:
[buildout]
accept-buildout-test-releases = false
allow-hosts = *
allow-picked-versions = true
allowed-eggs-from-site-packages = *
bin-directory = /sample-buildout/bin
develop-eggs-directory = /sample-buildout/develop-eggs
directory = /sample-buildout
eggs-directory = /sample-buildout/eggs
exec-sitecustomize = true
executable = python
find-links =
include-site-packages = true
install-from-cache = false
installed = /sample-buildout/.installed.cfg
log-format =
log-level = INFO
newest = true
offline = false
parts =
parts-directory = /sample-buildout/parts
prefer-final = false
python = buildout
relative-paths = false
socket-timeout =
unzip = false
use-dependency-links = true
verbosity = 20
zc.buildout-version = >=1.9a1, <2dev
<BLANKLINE>

All of these options can be overridden by configuration files or by command-line assignments. We've discussed most of these options already, but let's review them and touch on some we haven't discussed:

allowed-eggs-from-site-packages

Sometimes you need or want to control what eggs from site-packages are used. The allowed-eggs-from-site-packages option allows you to specify a whitelist of project names that may be included from site-packages. You can use globs to specify the value. It defaults to a single value of '*', indicating that any package may come from site-packages.

Here's a usage example:

[buildout]
...

allowed-eggs-from-site-packages =
    demo
    bigdemo
    zope.*

This option interacts with the include-site-packages option in the following ways.

If include-site-packages is true, then allowed-eggs-from-site-packages filters what eggs from site-packages may be chosen. Therefore, if allowed-eggs-from-site-packages is an empty list, then no eggs from site-packages are chosen, but site-packages will still be included at the end of path lists.

If include-site-packages is false, the value of allowed-eggs-from-site-packages is irrelevant.

See the include-site-packages description for more information.

bin-directory
The directory path where scripts are written. This can be a relative path, which is interpreted relative to the directory option.
develop-eggs-directory
The directory path where development egg links are created for software being created in the local project. This can be a relative path, which is interpreted relative to the directory option.
directory
The buildout directory. This is the base for other buildout file and directory locations, when relative locations are used.
eggs-directory
The directory path where downloaded eggs are put. It is common to share this directory across buildouts. Eggs in this directory should never be modified. This can be a relative path, which is interpreted relative to the directory option.
exec-sitecustomize
Normally the Python's real sitecustomize module is processed. If you do not want it to be processed in order to increase the repeatability of your buildout, set this value to 'false'. This will be honored irrespective of the setting for include-site-packages. This option will be honored by some recipes and not others. z3c.recipe.scripts honors this and zc.recipe.egg does not, for instance.
executable
The Python executable used to run the buildout. See the python option below.
include-site-packages
You can choose not to have the site-packages of the underlying Python available to your script or interpreter, in addition to the packages from your eggs. This can increase repeatability for your buildout. This option will be better used by some recipes than others. z3c.recipe.scripts honors this fully and zc.recipe.egg only partially, for instance.
installed
The file path where information about the results of the previous buildout run is written. This can be a relative path, which is interpreted relative to the directory option. This file provides an inventory of installed parts with information needed to decide which if any parts need to be uninstalled.
log-format
The format used for logging messages.
log-level
The log level before verbosity adjustment
parts
A white space separated list of parts to be installed.
parts-directory
A working directory that parts can used to store data.
python
The name of a section containing information about the default Python interpreter. Recipes that need a installation typically have options to tell them which Python installation to use. By convention, if a section-specific option isn't used, the option is looked for in the buildout section. The option must point to a section with an executable option giving the path to a Python executable. By default, the buildout section defines the default Python as the Python used to run the buildout.
relative-paths
The paths generated by zc.buildout are absolute by default, and this option is false. However, if you set this value to be true, bin/buildout will be generated with code that makes the paths relative. Some recipes, such as zc.recipe.egg and z3c.recipe.scripts, honor this value as well.
unzip
By default, zc.buildout doesn't unzip zip-safe eggs ("unzip = false"). This follows the policy followed by setuptools itself. Experience shows this policy to to be inconvenient. Zipped eggs make debugging more difficult and often import more slowly. You can include an unzip option in the buildout section to change the default unzipping policy ("unzip = true").
use-dependency-links

By default buildout will obey the setuptools dependency_links metadata when it looks for dependencies. This behavior can be controlled with the use-dependency-links buildout option:

[buildout]
...
use-dependency-links = false

The option defaults to true. If you set it to false, then dependency links are only looked for in the locations specified by find-links.

verbosity
A log-level adjustment. Typically, this is set via the -q and -v command-line options.

Creating new buildouts and bootstrapping

If zc.buildout is installed, you can use it to create a new buildout with it's own local copies of zc.buildout and setuptools and with local buildout scripts.

>>> sample_bootstrapped = tmpdir('sample-bootstrapped')
>>> print system(buildout
...              +' -c'+os.path.join(sample_bootstrapped, 'setup.cfg')
...              +' init'),
Creating '/sample-bootstrapped/setup.cfg'.
Creating directory '/sample-bootstrapped/bin'.
Creating directory '/sample-bootstrapped/parts'.
Creating directory '/sample-bootstrapped/eggs'.
Creating directory '/sample-bootstrapped/develop-eggs'.
Generated script '/sample-bootstrapped/bin/buildout'.

Note that a basic setup.cfg was created for us. This is because we provided an 'init' argument. By default, the generated setup.cfg is as minimal as it could be:

>>> cat(sample_bootstrapped, 'setup.cfg')
[buildout]
parts =

We also get other buildout artifacts:

>>> ls(sample_bootstrapped)
d  bin
d  develop-eggs
d  eggs
d  parts
-  setup.cfg
>>> ls(sample_bootstrapped, 'bin')
-  buildout
>>> _ = (ls(sample_bootstrapped, 'eggs'),
...      ls(sample_bootstrapped, 'develop-eggs'))
-  setuptools-0.6-py2.3.egg
-  zc.buildout-1.0-py2.3.egg

(We list both the eggs and develop-eggs directories because the buildout or setuptools egg could be installed in the develop-eggs directory if the original buildout had develop eggs for either buildout or setuptools.)

If relative-paths is true, the buildout script uses relative paths.

>>> write(sample_bootstrapped, 'setup.cfg',
... '''
... [buildout]
... relative-paths = true
... parts =
... ''')
>>> print system(buildout
...              +' -c'+os.path.join(sample_bootstrapped, 'setup.cfg')
...              +' bootstrap'),
Generated script '/sample-bootstrapped/bin/buildout'.
>>> buildout_script = join(sample_bootstrapped, 'bin', 'buildout')
>>> import sys
>>> if sys.platform.startswith('win'):
...     buildout_script += '-script.py'
>>> print open(buildout_script).read() # doctest: +ELLIPSIS
#!... -S
<BLANKLINE>
import os
<BLANKLINE>
join = os.path.join
base = os.path.dirname(os.path.abspath(os.path.realpath(__file__)))
base = os.path.dirname(base)
<BLANKLINE>
import sys
sys.path[0:0] = [
    join(base, 'parts/buildout'),
    ]
<BLANKLINE>
<BLANKLINE>
import os
path = sys.path[0]
if os.environ.get('PYTHONPATH'):
    path = os.pathsep.join([path, os.environ['PYTHONPATH']])
os.environ['BUILDOUT_ORIGINAL_PYTHONPATH'] = os.environ.get('PYTHONPATH', '')
os.environ['PYTHONPATH'] = path
import site # imports custom buildout-generated site.py
<BLANKLINE>
import zc.buildout.buildout
<BLANKLINE>
if __name__ == '__main__':
    sys.exit(zc.buildout.buildout.main())
<BLANKLINE>

Note that, in the above two examples, the buildout script was installed but not run. To run the buildout, we'd have to run the installed buildout script.

If we have an existing buildout that already has a buildout.cfg, we'll normally use the bootstrap command instead of init. It will complain if there isn't a configuration file:

>>> sample_bootstrapped2 = tmpdir('sample-bootstrapped2')
>>> print system(buildout
...              +' -c'+os.path.join(sample_bootstrapped2, 'setup.cfg')
...              +' bootstrap'),
While:
  Initializing.
Error: Couldn't open /sample-bootstrapped2/setup.cfg
>>> write(sample_bootstrapped2, 'setup.cfg',
... """
... [buildout]
... parts =
... """)
>>> print system(buildout
...              +' -c'+os.path.join(sample_bootstrapped2, 'setup.cfg')
...              +' bootstrap'),
Creating directory '/sample-bootstrapped2/bin'.
Creating directory '/sample-bootstrapped2/parts'.
Creating directory '/sample-bootstrapped2/eggs'.
Creating directory '/sample-bootstrapped2/develop-eggs'.
Generated script '/sample-bootstrapped2/bin/buildout'.

Similarly, if there is a configuration file and we use the init command, we'll get an error that the configuration file already exists:

>>> print system(buildout
...              +' -c'+os.path.join(sample_bootstrapped, 'setup.cfg')
...              +' init'),
While:
  Initializing.
Error: '/sample-bootstrapped/setup.cfg' already exists.

Initial eggs

When using the init command, you can specify distribution requirements or paths to use:

>>> cd(sample_bootstrapped)
>>> remove('setup.cfg')
>>> print system(buildout + ' -csetup.cfg init demo other ./src'),
Creating '/sample-bootstrapped/setup.cfg'.
Generated script '/sample-bootstrapped/bin/buildout'.
Getting distribution for 'zc.recipe.egg<2'.
Got zc.recipe.egg 1.3.3dev.
Installing py.
Getting distribution for 'demo'.
Got demo 0.4c1.
Getting distribution for 'other'.
Got other 1.0.
Getting distribution for 'demoneeded'.
Got demoneeded 1.2c1.
Generated script '/sample-bootstrapped/bin/demo'.
Generated interpreter '/sample-bootstrapped/bin/py'.

This causes a py part to be included that sets up a custom python interpreter with the given requirements or paths:

>>> cat('setup.cfg')
[buildout]
parts = py
<BLANKLINE>
[py]
recipe = zc.recipe.egg
interpreter = py
eggs =
  demo
  other
extra-paths =
  ./src

Passing requirements or paths causes the the builout to be run as part of initialization. In the example above, we got a number of distributions installed and 2 scripts generated. The first, demo, was defined by the demo project. The second, py was defined by the generated configuration. It's a "custom interpreter" that behaves like a standard Python interpeter, except that includes the specified eggs and extra paths in it's Python path.

We specified a source directory that didn't exist. Buildout created it for us:

>>> ls('.')
-  .installed.cfg
d  bin
d  develop-eggs
d  eggs
d  parts
-  setup.cfg
d  src
>>> uncd()

Newest and Offline Modes

By default buildout and recipes will try to find the newest versions of distributions needed to satisfy requirements. This can be very time consuming, especially when incrementally working on setting up a buildout or working on a recipe. The buildout newest option can be used to to suppress this. If the newest option is set to false, then new distributions won't be sought if an installed distribution meets requirements. The newest option can be set to false using the -N command-line option.

The offline option goes a bit further. If the buildout offline option is given a value of "true", the buildout and recipes that are aware of the option will avoid doing network access. This is handy when running the buildout when not connected to the internet. It also makes buildouts run much faster. This option is typically set using the buildout -o option.

Preferring Final Releases

Currently, when searching for new releases of your project's dependencies, the newest available release is used. This isn't usually ideal, as you may get a development release or alpha releases not ready to be widely used. You can request that final releases be preferred using the prefer-final option in the buildout section:

[buildout]
...
prefer-final = true

When the prefer-final option is set to true, then when searching for new releases, final releases are preferred. If there are final releases that satisfy distribution requirements, then those releases are used even if newer non-final releases are available.

In buildout version 2, all final releases will be preferred by default--that is prefer-final will also default to 'true'. You will then need to use a 'false' value for prefer-final to get the newest releases.

A separate option controls the behavior of the build system itself. When buildout looks for recipes, extensions, and for updates to itself, it does prefer final releases by default, as of the 1.5.0 release. The accept-buildout-test-releases option will let you override this behavior. However, it is typically changed by the --accept-buildout-test-releases option to the bootstrap script, since bootstrapping is the first step to selecting a buildout.

Finding distributions

By default, buildout searches the Python Package Index when looking for distributions. You can, instead, specify your own index to search using the index option:

[buildout]
...
index = http://index.example.com/

This index, or the default of http://pypi.python.org/simple/ if no index is specified, will always be searched for distributions unless running buildout with options that prevent searching for distributions. The latest version of the distribution that meets the requirements of the buildout will always be used.

You can also specify more locations to search for distributions using the find-links option. All locations specified will be searched for distributions along with the package index as described before.

Locations can be urls:

[buildout]
...
find-links = http://download.zope.org/distribution/

They can also be directories on disk:

[buildout]
...
find-links = /some/path

Finally, they can also be direct paths to distributions:

[buildout]
...
find-links = /some/path/someegg-1.0.0-py2.3.egg

Any number of locations can be specified in the find-links option:

[buildout]
...
find-links =
    http://download.zope.org/distribution/
    /some/otherpath
    /some/path/someegg-1.0.0-py2.3.egg

Controlling the installation database

The buildout installed option is used to specify the file used to save information on installed parts. This option is initialized to ".installed.cfg", but it can be overridden in the configuration file or on the command line:

>>> write('buildout.cfg',
... """
... [buildout]
... develop = recipes
... parts = debug
...
... [debug]
... recipe = recipes:debug
... """)
>>> print system(buildout+' buildout:installed=inst.cfg'),
Develop: '/sample-buildout/recipes'
Installing debug.
recipe recipes:debug
>>> ls(sample_buildout)
-  b1.cfg
-  b2.cfg
-  base.cfg
d  bin
-  buildout.cfg
d  demo
d  develop-eggs
d  eggs
-  inst.cfg
d  parts
d  recipes

The installation database can be disabled by supplying an empty buildout installed option:

>>> os.remove('inst.cfg')
>>> print system(buildout+' buildout:installed='),
Develop: '/sample-buildout/recipes'
Installing debug.
recipe recipes:debug
>>> ls(sample_buildout)
-  b1.cfg
-  b2.cfg
-  base.cfg
d  bin
-  buildout.cfg
d  demo
d  develop-eggs
d  eggs
d  parts
d  recipes

Note that there will be no installation database if there are no parts:

>>> write('buildout.cfg',
... """
... [buildout]
... parts =
... """)
>>> print system(buildout+' buildout:installed=inst.cfg'),
>>> ls(sample_buildout)
-  b1.cfg
-  b2.cfg
-  base.cfg
d  bin
-  buildout.cfg
d  demo
d  develop-eggs
d  eggs
d  parts
d  recipes

Extensions

A feature allows code to be loaded and run after configuration files have been read but before the buildout has begun any processing. The intent is to allow special plugins such as urllib2 request handlers to be loaded.

To load an extension, we use the extensions option and list one or more distribution requirements, on separate lines. The distributions named will be loaded and any zc.buildout.extension entry points found will be called with the buildout as an argument. When buildout finishes processing, any zc.buildout.unloadextension entry points found will be called with the buildout as an argument.

Let's create a sample extension in our sample buildout created in the previous section:

>>> mkdir(sample_bootstrapped, 'demo')
>>> write(sample_bootstrapped, 'demo', 'demo.py',
... """
... def ext(buildout):
...     print 'ext', list(buildout)
... def unload(buildout):
...     print 'unload', list(buildout)
... """)
>>> write(sample_bootstrapped, 'demo', 'setup.py',
... """
... from setuptools import setup
...
... setup(
...     name = "demo",
...     entry_points = {
...        'zc.buildout.extension': ['ext = demo:ext'],
...        'zc.buildout.unloadextension': ['ext = demo:unload'],
...        },
...     )
... """)

Our extension just prints out the word 'demo', and lists the sections found in the buildout passed to it.

We'll update our buildout.cfg to list the demo directory as a develop egg to be built:

>>> write(sample_bootstrapped, 'buildout.cfg',
... """
... [buildout]
... develop = demo
... parts =
... """)
>>> os.chdir(sample_bootstrapped)
>>> print system(os.path.join(sample_bootstrapped, 'bin', 'buildout')),
Develop: '/sample-bootstrapped/demo'

Now we can add the extensions option. We were a bit tricky and ran the buildout once with the demo develop egg defined but without the extension option. This is because extensions are loaded before the buildout creates develop eggs. We needed to use a separate buildout run to create the develop egg. Normally, when eggs are loaded from the network, we wouldn't need to do anything special.

>>> write(sample_bootstrapped, 'buildout.cfg',
... """
... [buildout]
... develop = demo
... extensions = demo
... parts =
... """)

We see that our extension is loaded and executed:

>>> print system(os.path.join(sample_bootstrapped, 'bin', 'buildout')),
ext ['buildout']
Develop: '/sample-bootstrapped/demo'
unload ['buildout']

Allow hosts

On some environments the links visited by zc.buildout can be forbidden by paranoiac firewalls. These URL might be on the chain of links visited by zc.buildout wheter they are defined in the find-links option, wheter they are defined by various eggs in their url, download_url, dependency_links metadata.

It is even harder to track that package_index works like a spider and might visit links and go to other location.

The allow-hosts option provides a way to prevent this, and works exactly like the one provided in easy_install.

You can provide a list of allowed host, together with wildcards:

[buildout]
...

allow-hosts =
    *.python.org
    example.com

All urls that does not match these hosts will not be visited.

[1]In the future, additional methods may be added. Older recipes with fewer methods will still be supported.
[2]If we wanted to create a distribution from this package, we would need specify much more information. See the setuptools documentation.

Always unzipping eggs

By default, zc.buildout doesn't unzip zip-safe eggs.

>>> write('buildout.cfg',
... '''
... [buildout]
... parts = eggs
... find-links = %(link_server)s
...
... [eggs]
... recipe = zc.recipe.egg
... eggs = demo
... ''' % globals())
>>> _ = system(buildout)
>>> ls('eggs')
-  demo-0.4c1-py2.4.egg
-  demoneeded-1.2c1-py2.4.egg
d  setuptools-0.6c8-py2.4.egg
-  zc.buildout.egg-link

This follows the policy followed by setuptools itself. Experience shows this policy to to be inconvenient. Zipped eggs make debugging more difficult and often import more slowly.

You can include an unzip option in the buildout section to change the default unzipping policy.

>>> write('buildout.cfg',
... '''
... [buildout]
... parts = eggs
... find-links = %(link_server)s
... unzip = true
...
... [eggs]
... recipe = zc.recipe.egg
... eggs = demo
... ''' % globals())
>>> import os
>>> for name in os.listdir('eggs'):
...     if name.startswith('demo'):
...         remove('eggs', name)
>>> _ = system(buildout)
>>> ls('eggs')
d  demo-0.4c1-py2.4.egg
d  demoneeded-1.2c1-py2.4.egg
d  setuptools-0.6c8-py2.4.egg
-  zc.buildout.egg-link

Repeatable buildouts: controlling eggs used

One of the goals of zc.buildout is to provide enough control to make buildouts repeatable. It should be possible to check the buildout configuration files for a project into a version control system and later use the checked in files to get the same buildout, subject to changes in the environment outside the buildout.

An advantage of using Python eggs is that depenencies of eggs used are automatically determined and used. The automatic inclusion of depenent distributions is at odds with the goal of repeatable buildouts.

To support repeatable buildouts, a versions section can be created with options for each distribution name whos version is to be fixed. The section can then be specified via the buildout versions option.

To see how this works, we'll create two versions of a recipe egg:

>>> mkdir('recipe')
>>> write('recipe', 'recipe.py',
... '''
... class Recipe:
...     def __init__(*a): pass
...     def install(self):
...         print 'recipe v1'
...         return ()
...     update = install
... ''')
>>> write('recipe', 'setup.py',
... '''
... from setuptools import setup
... setup(name='spam', version='1', py_modules=['recipe'],
...       entry_points={'zc.buildout': ['default = recipe:Recipe']},
...       )
... ''')
>>> write('recipe', 'README', '')
>>> print system(buildout+' setup recipe bdist_egg'), # doctest: +ELLIPSIS
Running setup script 'recipe/setup.py'.
...
>>> rmdir('recipe', 'build')
>>> write('recipe', 'recipe.py',
... '''
... class Recipe:
...     def __init__(*a): pass
...     def install(self):
...         print 'recipe v2'
...         return ()
...     update = install
... ''')
>>> write('recipe', 'setup.py',
... '''
... from setuptools import setup
... setup(name='spam', version='2', py_modules=['recipe'],
...       entry_points={'zc.buildout': ['default = recipe:Recipe']},
...       )
... ''')
>>> print system(buildout+' setup recipe bdist_egg'), # doctest: +ELLIPSIS
Running setup script 'recipe/setup.py'.
...

and we'll configure a buildout to use it:

>>> write('buildout.cfg',
... '''
... [buildout]
... parts = foo
... find-links = %s
...
... [foo]
... recipe = spam
... ''' % join('recipe', 'dist'))

If we run the buildout, it will use version 2:

>>> print system(buildout),
Getting distribution for 'spam'.
Got spam 2.
Installing foo.
recipe v2

We can specify a versions section that lists our recipe and name it in the buildout section:

>>> write('buildout.cfg',
... '''
... [buildout]
... parts = foo
... find-links = %s
... versions = release-1
...
... [release-1]
... spam = 1
... eggs = 2.2
...
... [foo]
... recipe = spam
... ''' % join('recipe', 'dist'))

Here we created a release-1 section listing the version 1 for the spam distribution. We told the buildout to use it by specifying release-1 as in the versions option.

Now, if we run the buildout, we'll use version 1 of the spam recipe:

>>> print system(buildout),
Getting distribution for 'spam==1'.
Got spam 1.
Uninstalling foo.
Installing foo.
recipe v1

Running the buildout in verbose mode will help us get information about versions used. If we run the buildout in verbose mode without specifying a versions section:

>>> print system(buildout+' buildout:versions= -v'), # doctest: +ELLIPSIS
Installing 'zc.buildout >=1.99, <2dev', 'setuptools'.
We have a develop egg: zc.buildout 1.0.0.
We have the best distribution that satisfies 'setuptools'.
Picked: setuptools = 0.6
Installing 'spam'.
We have the best distribution that satisfies 'spam'.
Picked: spam = 2.
Uninstalling foo.
Installing foo.
recipe v2

We'll get output that includes lines that tell us what versions buildout chose a for us, like:

zc.buildout.easy_install.picked: spam = 2

This allows us to discover versions that are picked dynamically, so that we can fix them in a versions section.

If we run the buildout with the versions section:

>>> print system(buildout+' -v'), # doctest: +ELLIPSIS
Installing 'zc.buildout >=1.99, <2dev', 'setuptools'.
We have a develop egg: zc.buildout 1.0.0.
We have the best distribution that satisfies 'setuptools'.
Picked: setuptools = 0.6
Installing 'spam'.
We have the distribution that satisfies 'spam==1'.
Uninstalling foo.
Installing foo.
recipe v1

We won't get output for the spam distribution, which we didn't pick, but we will get output for setuptools, which we didn't specify versions for.

You can request buildout to generate an error if it picks any versions:

>>> write('buildout.cfg',
... '''
... [buildout]
... parts = foo
... find-links = %s
... versions = release-1
... allow-picked-versions = false
...
... [release-1]
... spam = 1
... eggs = 2.2
...
... [foo]
... recipe = spam
... ''' % join('recipe', 'dist'))

Using the download utility

The zc.buildout.download module provides a download utility that handles the details of downloading files needed for a buildout run from the internet. It downloads files to the local file system, using the download cache if desired and optionally checking the downloaded files' MD5 checksum.

We setup an HTTP server that provides a file we want to download:

>>> server_data = tmpdir('sample_files')
>>> write(server_data, 'foo.txt', 'This is a foo text.')
>>> server_url = start_server(server_data)

We also use a fresh directory for temporary files in order to make sure that all temporary files have been cleaned up in the end:

>>> import tempfile
>>> old_tempdir = tempfile.tempdir
>>> tempfile.tempdir = tmpdir('tmp')

Downloading without using the cache

If no download cache should be used, the download utility is instantiated without any arguments:

>>> from zc.buildout.download import Download
>>> download = Download()
>>> print download.cache_dir
None

Downloading a file is achieved by calling the utility with the URL as an argument. A tuple is returned that consists of the path to the downloaded copy of the file and a boolean value indicating whether this is a temporary file meant to be cleaned up during the same buildout run:

>>> path, is_temp = download(server_url+'foo.txt')
>>> print path
/.../buildout-...
>>> cat(path)
This is a foo text.

As we aren't using the download cache and haven't specified a target path either, the download has ended up in a temporary file:

>>> is_temp
True
>>> import tempfile
>>> path.startswith(tempfile.gettempdir())
True

We are responsible for cleaning up temporary files behind us:

>>> remove(path)

When trying to access a file that doesn't exist, we'll get an exception:

>>> try: download(server_url+'not-there') # doctest: +ELLIPSIS
... except: print 'download error'
... else: print 'woops'
download error

Downloading a local file doesn't produce a temporary file but simply returns the local file itself:

>>> download(join(server_data, 'foo.txt'))
('/sample_files/foo.txt', False)

We can also have the downloaded file's MD5 sum checked:

>>> try: from hashlib import md5
... except ImportError: from md5 import new as md5
>>> path, is_temp = download(server_url+'foo.txt',
...                          md5('This is a foo text.').hexdigest())
>>> is_temp
True
>>> remove(path)
>>> download(server_url+'foo.txt',
...          md5('The wrong text.').hexdigest())
Traceback (most recent call last):
ChecksumError: MD5 checksum mismatch downloading 'http://localhost/foo.txt'

The error message in the event of an MD5 checksum mismatch for a local file reads somewhat differently:

>>> download(join(server_data, 'foo.txt'),
...               md5('This is a foo text.').hexdigest())
('/sample_files/foo.txt', False)
>>> download(join(server_data, 'foo.txt'),
...          md5('The wrong text.').hexdigest())
Traceback (most recent call last):
ChecksumError: MD5 checksum mismatch for local resource at '/sample_files/foo.txt'.

Finally, we can download the file to a specified place in the file system:

>>> target_dir = tmpdir('download-target')
>>> path, is_temp = download(server_url+'foo.txt',
...                          path=join(target_dir, 'downloaded.txt'))
>>> print path
/download-target/downloaded.txt
>>> cat(path)
This is a foo text.
>>> is_temp
False

Trying to download a file in offline mode will result in an error:

>>> download = Download(cache=None, offline=True)
>>> download(server_url+'foo.txt')
Traceback (most recent call last):
UserError: Couldn't download 'http://localhost/foo.txt' in offline mode.

As an exception to this rule, file system paths and URLs in the file scheme will still work:

>>> cat(download(join(server_data, 'foo.txt'))[0])
This is a foo text.
>>> cat(download('file:' + join(server_data, 'foo.txt'))[0])
This is a foo text.
>>> remove(path)

Downloading using the download cache

In order to make use of the download cache, we need to configure the download utility differently. To do this, we pass a directory path as the cache attribute upon instantiation:

>>> cache = tmpdir('download-cache')
>>> download = Download(cache=cache)
>>> print download.cache_dir
/download-cache/

Simple usage

When using the cache, a file will be stored in the cache directory when it is first downloaded. The file system path returned by the download utility points to the cached copy:

>>> ls(cache)
>>> path, is_temp = download(server_url+'foo.txt')
>>> print path
/download-cache/foo.txt
>>> cat(path)
This is a foo text.
>>> is_temp
False

Whenever the file is downloaded again, the cached copy is used. Let's change the file on the server to see this:

>>> write(server_data, 'foo.txt', 'The wrong text.')
>>> path, is_temp = download(server_url+'foo.txt')
>>> print path
/download-cache/foo.txt
>>> cat(path)
This is a foo text.

If we specify an MD5 checksum for a file that is already in the cache, the cached copy's checksum will be verified:

>>> download(server_url+'foo.txt', md5('The wrong text.').hexdigest())
Traceback (most recent call last):
ChecksumError: MD5 checksum mismatch for cached download
               from 'http://localhost/foo.txt' at '/download-cache/foo.txt'

Trying to access another file at a different URL which has the same base name will result in the cached copy being used:

>>> mkdir(server_data, 'other')
>>> write(server_data, 'other', 'foo.txt', 'The wrong text.')
>>> path, is_temp = download(server_url+'other/foo.txt')
>>> print path
/download-cache/foo.txt
>>> cat(path)
This is a foo text.

Given a target path for the download, the utility will provide a copy of the file at that location both when first downloading the file and when using a cached copy:

>>> remove(cache, 'foo.txt')
>>> ls(cache)
>>> write(server_data, 'foo.txt', 'This is a foo text.')
>>> path, is_temp = download(server_url+'foo.txt',
...                          path=join(target_dir, 'downloaded.txt'))
>>> print path
/download-target/downloaded.txt
>>> cat(path)
This is a foo text.
>>> is_temp
False
>>> ls(cache)
- foo.txt
>>> remove(path)
>>> write(server_data, 'foo.txt', 'The wrong text.')
>>> path, is_temp = download(server_url+'foo.txt',
...                          path=join(target_dir, 'downloaded.txt'))
>>> print path
/download-target/downloaded.txt
>>> cat(path)
This is a foo text.
>>> is_temp
False

In offline mode, downloads from any URL will be successful if the file is found in the cache:

>>> download = Download(cache=cache, offline=True)
>>> cat(download(server_url+'foo.txt')[0])
This is a foo text.

Local resources will be cached just like any others since download caches are sometimes used to create source distributions:

>>> remove(cache, 'foo.txt')
>>> ls(cache)
>>> write(server_data, 'foo.txt', 'This is a foo text.')
>>> download = Download(cache=cache)
>>> cat(download('file:' + join(server_data, 'foo.txt'), path=path)[0])
This is a foo text.
>>> ls(cache)
- foo.txt
>>> remove(cache, 'foo.txt')
>>> cat(download(join(server_data, 'foo.txt'), path=path)[0])
This is a foo text.
>>> ls(cache)
- foo.txt
>>> remove(cache, 'foo.txt')

However, resources with checksum mismatches will not be copied to the cache:

>>> download(server_url+'foo.txt', md5('The wrong text.').hexdigest())
Traceback (most recent call last):
ChecksumError: MD5 checksum mismatch downloading 'http://localhost/foo.txt'
>>> ls(cache)
>>> remove(path)

If the file is completely missing it should notify the user of the error:

>>> download(server_url+'bar.txt')
Traceback (most recent call last):
UserError: Error downloading extends for URL http://localhost/bar.txt: (404, 'Not Found')
>>> ls(cache)

Finally, let's see what happens if the download cache to be used doesn't exist as a directory in the file system yet:

>>> Download(cache=join(cache, 'non-existent'))(server_url+'foo.txt')
Traceback (most recent call last):
UserError: The directory:
'/download-cache/non-existent'
to be used as a download cache doesn't exist.

Using namespace sub-directories of the download cache

It is common to store cached copies of downloaded files within sub-directories of the download cache to keep some degree of order. For example, zc.buildout stores downloaded distributions in a sub-directory named "dist". Those sub-directories are also known as namespaces. So far, we haven't specified any namespaces to use, so the download utility stored files directly inside the download cache. Let's use a namespace "test" instead:

>>> download = Download(cache=cache, namespace='test')
>>> print download.cache_dir
/download-cache/test

The namespace sub-directory hasn't been created yet:

>>> ls(cache)

Downloading a file now creates the namespace sub-directory and places a copy of the file inside it:

>>> path, is_temp = download(server_url+'foo.txt')
>>> print path
/download-cache/test/foo.txt
>>> ls(cache)
d test
>>> ls(cache, 'test')
- foo.txt
>>> cat(path)
This is a foo text.
>>> is_temp
False

The next time we want to download that file, the copy from inside the cache namespace is used. To see this clearly, we put a file with the same name but different content both on the server and in the cache's root directory:

>>> write(server_data, 'foo.txt', 'The wrong text.')
>>> write(cache, 'foo.txt', 'The wrong text.')
>>> path, is_temp = download(server_url+'foo.txt')
>>> print path
/download-cache/test/foo.txt
>>> cat(path)
This is a foo text.
>>> rmdir(cache, 'test')
>>> remove(cache, 'foo.txt')
>>> write(server_data, 'foo.txt', 'This is a foo text.')

Using a hash of the URL as the filename in the cache

So far, the base name of the downloaded file read from the URL has been used for the name of the cached copy of the file. This may not be desirable in some cases, for example when downloading files from different locations that have the same base name due to some naming convention, or if the file content depends on URL parameters. In such cases, an MD5 hash of the complete URL may be used as the filename in the cache:

>>> download = Download(cache=cache, hash_name=True)
>>> path, is_temp = download(server_url+'foo.txt')
>>> print path
/download-cache/09f5793fcdc1716727f72d49519c688d
>>> cat(path)
This is a foo text.
>>> ls(cache)
- 09f5793fcdc1716727f72d49519c688d

The path was printed just to illustrate matters; we cannot know the real checksum since we don't know which port the server happens to listen at when the test is run, so we don't actually know the full URL of the file. Let's check that the checksum actually belongs to the particular URL used:

>>> path.lower() == join(cache, md5(server_url+'foo.txt').hexdigest()).lower()
True

The cached copy is used when downloading the file again:

>>> write(server_data, 'foo.txt', 'The wrong text.')
>>> (path, is_temp) == download(server_url+'foo.txt')
True
>>> cat(path)
This is a foo text.
>>> ls(cache)
- 09f5793fcdc1716727f72d49519c688d

If we change the URL, even in such a way that it keeps the base name of the file the same, the file will be downloaded again this time and put in the cache under a different name:

>>> path2, is_temp = download(server_url+'other/foo.txt')
>>> print path2
/download-cache/537b6d73267f8f4447586989af8c470e
>>> path == path2
False
>>> path2.lower() == join(cache, md5(server_url+'other/foo.txt').hexdigest()).lower()
True
>>> cat(path)
This is a foo text.
>>> cat(path2)
The wrong text.
>>> ls(cache)
- 09f5793fcdc1716727f72d49519c688d
- 537b6d73267f8f4447586989af8c470e
>>> remove(path)
>>> remove(path2)
>>> write(server_data, 'foo.txt', 'This is a foo text.')

Using the cache purely as a fall-back

Sometimes it is desirable to try downloading a file from the net if at all possible, and use the cache purely as a fall-back option when a server is down or if we are in offline mode. This mode is only in effect if a download cache is configured in the first place:

>>> download = Download(cache=cache, fallback=True)
>>> print download.cache_dir
/download-cache/

A downloaded file will be cached:

>>> ls(cache)
>>> path, is_temp = download(server_url+'foo.txt')
>>> ls(cache)
- foo.txt
>>> cat(cache, 'foo.txt')
This is a foo text.
>>> is_temp
False

If the file cannot be served, the cached copy will be used:

>>> remove(server_data, 'foo.txt')
>>> try: Download()(server_url+'foo.txt') # doctest: +ELLIPSIS
... except: print 'download error'
... else: print 'woops'
download error
>>> path, is_temp = download(server_url+'foo.txt')
>>> cat(path)
This is a foo text.
>>> is_temp
False

Similarly, if the file is served but we're in offline mode, we'll fall back to using the cache:

>>> write(server_data, 'foo.txt', 'The wrong text.')
>>> get(server_url+'foo.txt')
'The wrong text.'
>>> offline_download = Download(cache=cache, offline=True, fallback=True)
>>> path, is_temp = offline_download(server_url+'foo.txt')
>>> print path
/download-cache/foo.txt
>>> cat(path)
This is a foo text.
>>> is_temp
False

However, when downloading the file normally with the cache being used in fall-back mode, the file will be downloaded from the net and the cached copy will be replaced with the new content:

>>> cat(download(server_url+'foo.txt')[0])
The wrong text.
>>> cat(cache, 'foo.txt')
The wrong text.

When trying to download a resource whose checksum does not match, the cached copy will neither be used nor overwritten:

>>> write(server_data, 'foo.txt', 'This is a foo text.')
>>> download(server_url+'foo.txt', md5('The wrong text.').hexdigest())
Traceback (most recent call last):
ChecksumError: MD5 checksum mismatch downloading 'http://localhost/foo.txt'
>>> cat(cache, 'foo.txt')
The wrong text.

Configuring the download utility from buildout options

The configuration options explained so far derive from the build logic implemented by the calling code. Other options configure the download utility for use in a particular project or buildout run; they are read from the buildout configuration section. The latter can be passed directly as the first argument to the download utility's constructor.

The location of the download cache is specified by the download-cache option:

>>> download = Download({'download-cache': cache}, namespace='cmmi')
>>> print download.cache_dir
/download-cache/cmmi

If the download-cache option specifies a relative path, it is understood relative to the current working directory, or to the buildout directory if that is given:

>>> download = Download({'download-cache': 'relative-cache'})
>>> print download.cache_dir
/sample-buildout/relative-cache/
>>> download = Download({'directory': join(sample_buildout, 'root'),
...                      'download-cache': 'relative-cache'})
>>> print download.cache_dir
/sample-buildout/root/relative-cache/

Keyword parameters take precedence over the corresponding options:

>>> download = Download({'download-cache': cache}, cache=None)
>>> print download.cache_dir
None

Whether to assume offline mode can be inferred from either the offline or the install-from-cache option. As usual with zc.buildout, these options must assume one of the values 'true' and 'false':

>>> download = Download({'offline': 'true'})
>>> download.offline
True
>>> download = Download({'offline': 'false'})
>>> download.offline
False
>>> download = Download({'install-from-cache': 'true'})
>>> download.offline
True
>>> download = Download({'install-from-cache': 'false'})
>>> download.offline
False

These two options are combined using logical 'or':

>>> download = Download({'offline': 'true', 'install-from-cache': 'false'})
>>> download.offline
True
>>> download = Download({'offline': 'false', 'install-from-cache': 'true'})
>>> download.offline
True

The offline keyword parameter takes precedence over both the offline and install-from-cache options:

>>> download = Download({'offline': 'true'}, offline=False)
>>> download.offline
False
>>> download = Download({'install-from-cache': 'false'}, offline=True)
>>> download.offline
True

Regressions

MD5 checksum calculation needs to be reliable on all supported systems, which requires text files to be treated as binary to avoid implicit line-ending conversions:

>>> text = 'First line of text.\r\nSecond line.\r\n'
>>> f = open(join(server_data, 'foo.txt'), 'wb')
>>> f.write(text)
>>> f.close()
>>> path, is_temp = Download()(server_url+'foo.txt', md5(text).hexdigest())
>>> remove(path)

When "downloading" a directory given by file-system path or file: URL and using a download cache at the same time, the cached directory wasn't handled correctly. Consequently, the cache was defeated and an attempt to cache the directory a second time broke. This is how it should work:

>>> download = Download(cache=cache)
>>> dirpath = join(server_data, 'some_directory')
>>> mkdir(dirpath)
>>> dest, _ = download(dirpath)

If we now modify the source tree, the second download will produce the original one from the cache:

>>> mkdir(join(dirpath, 'foo'))
>>> ls(dirpath)
d foo
>>> dest, _ = download(dirpath)
>>> ls(dest)

Clean up

We should have cleaned up all temporary files created by downloading things:

>>> ls(tempfile.tempdir)

Reset the global temporary directory:

>>> tempfile.tempdir = old_tempdir

Using a download cache

Normally, when distributions are installed, if any processing is needed, they are downloaded from the internet to a temporary directory and then installed from there. A download cache can be used to avoid the download step. This can be useful to reduce network access and to create source distributions of an entire buildout.

The buildout download-cache option can be used to specify a directory to be used as a download cache.

In this example, we'll create a directory to hold the cache:

>>> cache = tmpdir('cache')

And set up a buildout that downloads some eggs:

>>> write('buildout.cfg',
... '''
... [buildout]
... parts = eggs
... download-cache = %(cache)s
... find-links = %(link_server)s
...
... [eggs]
... recipe = zc.recipe.egg
... eggs = demo ==0.2
... ''' % globals())

We specified a link server that has some distributions available for download:

>>> print get(link_server),
<html><body>
<a href="bigdemo-0.1-py2.4.egg">bigdemo-0.1-py2.4.egg</a><br>
<a href="demo-0.1-py2.4.egg">demo-0.1-py2.4.egg</a><br>
<a href="demo-0.2-py2.4.egg">demo-0.2-py2.4.egg</a><br>
<a href="demo-0.3-py2.4.egg">demo-0.3-py2.4.egg</a><br>
<a href="demo-0.4c1-py2.4.egg">demo-0.4c1-py2.4.egg</a><br>
<a href="demoneeded-1.0.zip">demoneeded-1.0.zip</a><br>
<a href="demoneeded-1.1.zip">demoneeded-1.1.zip</a><br>
<a href="demoneeded-1.2c1.zip">demoneeded-1.2c1.zip</a><br>
<a href="extdemo-1.4.zip">extdemo-1.4.zip</a><br>
<a href="index/">index/</a><br>
<a href="other-1.0-py2.4.egg">other-1.0-py2.4.egg</a><br>
</body></html>

We'll enable logging on the link server so we can see what's going on:

>>> get(link_server+'enable_server_logging')
GET 200 /enable_server_logging
''

We also specified a download cache.

If we run the buildout, we'll see the eggs installed from the link server as usual:

>>> print system(buildout),
GET 200 /
GET 200 /demo-0.2-py2.4.egg
GET 200 /demoneeded-1.2c1.zip
Installing eggs.
Getting distribution for 'demo==0.2'.
Got demo 0.2.
Getting distribution for 'demoneeded'.
Got demoneeded 1.2c1.
Generated script '/sample-buildout/bin/demo'.

We'll also get the download cache populated. The buildout doesn't put files in the cache directly. It creates an intermediate directory, dist:

>>> ls(cache)
d  dist
>>> ls(cache, 'dist')
-  demo-0.2-py2.4.egg
-  demoneeded-1.2c1.zip

If we remove the installed eggs from eggs directory and re-run the buildout:

>>> import os
>>> for  f in os.listdir('eggs'):
...     if f.startswith('demo'):
...         remove('eggs', f)
>>> print system(buildout),
GET 200 /
Updating eggs.
Getting distribution for 'demo==0.2'.
Got demo 0.2.
Getting distribution for 'demoneeded'.
Got demoneeded 1.2c1.

We see that the distributions aren't downloaded, because they're downloaded from the cache.

Installing solely from a download cache

A download cache can be used as the basis of application source releases. In an application source release, we want to distribute an application that can be built without making any network accesses. In this case, we distribute a buildout with download cache and tell the buildout to install from the download cache only, without making network accesses. The buildout install-from-cache option can be used to signal that packages should be installed only from the download cache.

Let's remove our installed eggs and run the buildout with the install-from-cache option set to true:

>>> for  f in os.listdir('eggs'):
...     if f.startswith('demo'):
...         remove('eggs', f)
>>> write('buildout.cfg',
... '''
... [buildout]
... parts = eggs
... download-cache = %(cache)s
... install-from-cache = true
... find-links = %(link_server)s
...
... [eggs]
... recipe = zc.recipe.egg
... eggs = demo
... ''' % globals())
>>> print system(buildout),
Uninstalling eggs.
Installing eggs.
Getting distribution for 'demo'.
Got demo 0.2.
Getting distribution for 'demoneeded'.
Got demoneeded 1.2c1.
Generated script '/sample-buildout/bin/demo'.

Caching extended configuration

As mentioned in the general buildout documentation, configuration files can extend each other, including the ability to download configuration being extended from a URL. If desired, zc.buildout caches downloaded configuration in order to be able to use it when run offline.

As we're going to talk about downloading things, let's start an HTTP server. Also, all of the following will take place inside the sample buildout.

>>> server_data = tmpdir('server_data')
>>> server_url = start_server(server_data)
>>> cd(sample_buildout)

We also use a fresh directory for temporary files in order to make sure that all temporary files have been cleaned up in the end:

>>> import tempfile
>>> old_tempdir = tempfile.tempdir
>>> tempfile.tempdir = tmpdir('tmp')

Basic use of the extends cache

We put some base configuration on a server and reference it from a sample buildout:

>>> write(server_data, 'base.cfg', """\
... [buildout]
... parts =
... foo = bar
... """)
>>> write('buildout.cfg', """\
... [buildout]
... extends = %sbase.cfg
... """ % server_url)

When trying to run this buildout offline, we'll find that we cannot read all of the required configuration:

>>> print system(buildout + ' -o')
While:
  Initializing.
Error: Couldn't download 'http://localhost/base.cfg' in offline mode.

Trying the same online, we can:

>>> print system(buildout)
Unused options for buildout: 'foo'.

As long as we haven't said anything about caching downloaded configuration, nothing gets cached. Offline mode will still cause the buildout to fail:

>>> print system(buildout + ' -o')
While:
  Initializing.
Error: Couldn't download 'http://localhost/base.cfg' in offline mode.

Let's now specify a cache for base configuration files. This cache is different from the download cache used by recipes for caching distributions and other files; one might, however, use a namespace subdirectory of the download cache for it. The configuration cache we specify will be created when running buildout and the base.cfg file will be put in it (with the file name being a hash of the complete URL):

>>> mkdir('cache')
>>> write('buildout.cfg', """\
... [buildout]
... extends = %sbase.cfg
... extends-cache = cache
... """ % server_url)
>>> print system(buildout)
Unused options for buildout: 'foo'.
>>> cache = join(sample_buildout, 'cache')
>>> ls(cache)
-  5aedc98d7e769290a29d654a591a3a45
>>> import os
>>> cat(cache, os.listdir(cache)[0])
[buildout]
parts =
foo = bar

We can now run buildout offline as it will read base.cfg from the cache:

>>> print system(buildout + ' -o')
Unused options for buildout: 'foo'.

The cache is being used purely as a fall-back in case we are offline or don't have access to a configuration file to be downloaded. As long as we are online, buildout attempts to download a fresh copy of each file even if a cached copy of the file exists. To see this, we put different configuration in the same place on the server and run buildout in offline mode so it takes base.cfg from the cache:

>>> write(server_data, 'base.cfg', """\
... [buildout]
... parts =
... bar = baz
... """)
>>> print system(buildout + ' -o')
Unused options for buildout: 'foo'.

In online mode, buildout will download and use the modified version:

>>> print system(buildout)
Unused options for buildout: 'bar'.

Trying offline mode again, the new version will be used as it has been put in the cache now:

>>> print system(buildout + ' -o')
Unused options for buildout: 'bar'.

Clean up:

>>> rmdir(cache)

Specifying extends cache and offline mode

Normally, the values of buildout options such as the location of a download cache or whether to use offline mode are determined by first reading the user's default configuration, updating it with the project's configuration and finally applying command-line options. User and project configuration are assembled by reading a file such as ~/.buildout/default.cfg, buildout.cfg or a URL given on the command line, recursively (depth-first) downloading any base configuration specified by the buildout:extends option read from each of those config files, and finally evaluating each config file to provide default values for options not yet read.

This works fine for all options that do not influence how configuration is downloaded in the first place. The extends-cache and offline options, however, are treated differently from the procedure described in order to make it simple and obvious to see where a particular configuration file came from under any particular circumstances.

  • Offline and extends-cache settings are read from the two root config files exclusively. Otherwise one could construct configuration files that, when read, imply that they should have been read from a different source than they have. Also, specifying the extends cache within a file that might have to be taken from the cache before being read wouldn't make a lot of sense.
  • Offline and extends-cache settings given by the user's defaults apply to the process of assembling the project's configuration. If no extends cache has been specified by the user's default configuration, the project's root config file must be available, be it from disk or from the net.
  • Offline mode turned on by the -o command line option is honoured from the beginning even though command line options are applied to the configuration last. If offline mode is not requested by the command line, it may be switched on by either the user's or the project's config root.

Extends cache

Let's see the above rules in action. We create a new home directory for our user and write user and project configuration that recursively extends online bases, using different caches:

>>> mkdir('home')
>>> mkdir('home', '.buildout')
>>> mkdir('cache')
>>> mkdir('user-cache')
>>> os.environ['HOME'] = join(sample_buildout, 'home')
>>> write('home', '.buildout', 'default.cfg', """\
... [buildout]
... extends = fancy_default.cfg
... extends-cache = user-cache
... """)
>>> write('home', '.buildout', 'fancy_default.cfg', """\
... [buildout]
... extends = %sbase_default.cfg
... """ % server_url)
>>> write(server_data, 'base_default.cfg', """\
... [buildout]
... foo = bar
... offline = false
... """)
>>> write('buildout.cfg', """\
... [buildout]
... extends = fancy.cfg
... extends-cache = cache
... """)
>>> write('fancy.cfg', """\
... [buildout]
... extends = %sbase.cfg
... """ % server_url)
>>> write(server_data, 'base.cfg', """\
... [buildout]
... parts =
... offline = false
... """)

Buildout will now assemble its configuration from all of these 6 files, defaults first. The online resources end up in the respective extends caches:

>>> print system(buildout)
Unused options for buildout: 'foo'.
>>> ls('user-cache')
-  10e772cf422123ef6c64ae770f555740
>>> cat('user-cache', os.listdir('user-cache')[0])
[buildout]
foo = bar
offline = false
>>> ls('cache')
-  c72213127e6eb2208a3e1fc1dba771a7
>>> cat('cache', os.listdir('cache')[0])
[buildout]
parts =
offline = false

If, on the other hand, the extends caches are specified in files that get extended themselves, they won't be used for assembling the configuration they belong to (user's or project's, resp.). The extends cache specified by the user's defaults does, however, apply to downloading project configuration. Let's rewrite the config files, clean out the caches and re-run buildout:

>>> write('home', '.buildout', 'default.cfg', """\
... [buildout]
... extends = fancy_default.cfg
... """)
>>> write('home', '.buildout', 'fancy_default.cfg', """\
... [buildout]
... extends = %sbase_default.cfg
... extends-cache = user-cache
... """ % server_url)
>>> write('buildout.cfg', """\
... [buildout]
... extends = fancy.cfg
... """)
>>> write('fancy.cfg', """\
... [buildout]
... extends = %sbase.cfg
... extends-cache = cache
... """ % server_url)
>>> remove('user-cache', os.listdir('user-cache')[0])
>>> remove('cache', os.listdir('cache')[0])
>>> print system(buildout)
Unused options for buildout: 'foo'.
>>> ls('user-cache')
-  0548bad6002359532de37385bb532e26
>>> cat('user-cache', os.listdir('user-cache')[0])
[buildout]
parts =
offline = false
>>> ls('cache')

Clean up:

>>> rmdir('user-cache')
>>> rmdir('cache')

Offline mode and installation from cache ----------------------------~~~~~~~~~~~~

If we run buildout in offline mode now, it will fail because it cannot get at the remote configuration file needed by the user's defaults:

>>> print system(buildout + ' -o')
While:
  Initializing.
Error: Couldn't download 'http://localhost/base_default.cfg' in offline mode.

Let's now successively turn on offline mode by different parts of the configuration and see when buildout applies this setting in each case:

>>> write('home', '.buildout', 'default.cfg', """\
... [buildout]
... extends = fancy_default.cfg
... offline = true
... """)
>>> print system(buildout)
While:
  Initializing.
Error: Couldn't download 'http://localhost/base_default.cfg' in offline mode.
>>> write('home', '.buildout', 'default.cfg', """\
... [buildout]
... extends = fancy_default.cfg
... """)
>>> write('home', '.buildout', 'fancy_default.cfg', """\
... [buildout]
... extends = %sbase_default.cfg
... offline = true
... """ % server_url)
>>> print system(buildout)
While:
  Initializing.
Error: Couldn't download 'http://localhost/base.cfg' in offline mode.
>>> write('home', '.buildout', 'fancy_default.cfg', """\
... [buildout]
... extends = %sbase_default.cfg
... """ % server_url)
>>> write('buildout.cfg', """\
... [buildout]
... extends = fancy.cfg
... offline = true
... """)
>>> print system(buildout)
While:
  Initializing.
Error: Couldn't download 'http://localhost/base.cfg' in offline mode.
>>> write('buildout.cfg', """\
... [buildout]
... extends = fancy.cfg
... """)
>>> write('fancy.cfg', """\
... [buildout]
... extends = %sbase.cfg
... offline = true
... """ % server_url)
>>> print system(buildout)
Unused options for buildout: 'foo'.

The install-from-cache option is treated accordingly:

>>> write('home', '.buildout', 'default.cfg', """\
... [buildout]
... extends = fancy_default.cfg
... install-from-cache = true
... """)
>>> print system(buildout)
While:
  Initializing.
Error: Couldn't download 'http://localhost/base_default.cfg' in offline mode.
>>> write('home', '.buildout', 'default.cfg', """\
... [buildout]
... extends = fancy_default.cfg
... """)
>>> write('home', '.buildout', 'fancy_default.cfg', """\
... [buildout]
... extends = %sbase_default.cfg
... install-from-cache = true
... """ % server_url)
>>> print system(buildout)
While:
  Initializing.
Error: Couldn't download 'http://localhost/base.cfg' in offline mode.
>>> write('home', '.buildout', 'fancy_default.cfg', """\
... [buildout]
... extends = %sbase_default.cfg
... """ % server_url)
>>> write('buildout.cfg', """\
... [buildout]
... extends = fancy.cfg
... install-from-cache = true
... """)
>>> print system(buildout)
While:
  Initializing.
Error: Couldn't download 'http://localhost/base.cfg' in offline mode.
>>> write('buildout.cfg', """\
... [buildout]
... extends = fancy.cfg
... """)
>>> write('fancy.cfg', """\
... [buildout]
... extends = %sbase.cfg
... install-from-cache = true
... """ % server_url)
>>> print system(buildout)
While:
  Installing.
  Checking for upgrades.
An internal error occurred ...
ValueError: install_from_cache set to true with no download cache
>>> rmdir('home', '.buildout')

Newest and non-newest behaviour for extends cache

While offline mode forbids network access completely, 'newest' mode determines whether to look for updated versions of a resource even if some version of it is already present locally. If we run buildout in newest mode (newest = true), the configuration files are updated with each run:

>>> mkdir("cache")
>>> write(server_data, 'base.cfg', """\
... [buildout]
... parts =
... """)
>>> write('buildout.cfg', """\
... [buildout]
... extends-cache = cache
... extends = %sbase.cfg
... """ % server_url)
>>> print system(buildout)
>>> ls('cache')
-  5aedc98d7e769290a29d654a591a3a45
>>> cat('cache', os.listdir(cache)[0])
[buildout]
parts =

A change to base.cfg is picked up on the next buildout run:

>>> write(server_data, 'base.cfg', """\
... [buildout]
... parts =
... foo = bar
... """)
>>> print system(buildout + " -n")
Unused options for buildout: 'foo'.
>>> cat('cache', os.listdir(cache)[0])
[buildout]
parts =
foo = bar

In contrast, when not using newest mode (newest = false), the files already present in the extends cache will not be updated:

>>> write(server_data, 'base.cfg', """\
... [buildout]
... parts =
... """)
>>> print system(buildout + " -N")
Unused options for buildout: 'foo'.
>>> cat('cache', os.listdir(cache)[0])
[buildout]
parts =
foo = bar

Even when updating base configuration files with a buildout run, any given configuration file will be downloaded only once during that particular run. If some base configuration file is extended more than once, its cached copy is used:

>>> write(server_data, 'baseA.cfg', """\
... [buildout]
... extends = %sbase.cfg
... foo = bar
... """ % server_url)
>>> write(server_data, 'baseB.cfg', """\
... [buildout]
... extends-cache = cache
... extends = %sbase.cfg
... bar = foo
... """ % server_url)
>>> write('buildout.cfg', """\
... [buildout]
... extends-cache = cache
... newest = true
... extends = %sbaseA.cfg %sbaseB.cfg
... """ % (server_url, server_url))
>>> print system(buildout + " -n")
Unused options for buildout: 'bar' 'foo'.

(XXX We patch download utility's API to produce readable output for the test; a better solution would utilise the logging already done by the utility.)

>>> import zc.buildout
>>> old_download = zc.buildout.download.Download.download
>>> def wrapper_download(self, url, md5sum=None, path=None):
...   print "The URL %s was downloaded." % url
...   return old_download(url, md5sum, path)
>>> zc.buildout.download.Download.download = wrapper_download
>>> zc.buildout.buildout.main([])
The URL http://localhost/baseA.cfg was downloaded.
The URL http://localhost/base.cfg was downloaded.
The URL http://localhost/baseB.cfg was downloaded.
Unused options for buildout: 'bar' 'foo'.
>>> zc.buildout.download.Download.download = old_download

The deprecated extended-by option

The buildout section used to recognise an option named extended-by that was deprecated at some point and removed in the 1.5 line. Since ignoring this option silently was considered harmful as a matter of principle, a UserError is raised if that option is encountered now:

>>> write(server_data, 'base.cfg', """\
... [buildout]
... parts =
... extended-by = foo.cfg
... """)
>>> print system(buildout)
While:
  Initializing.
Error: No-longer supported "extended-by" option found in http://localhost/base.cfg.

Clean up

We should have cleaned up all temporary files created by downloading things:

>>> ls(tempfile.tempdir)

Reset the global temporary directory:

>>> tempfile.tempdir = old_tempdir

Using zc.buildout to run setup scripts

zc buildout has a convenience command for running setup scripts. Why? There are two reasons. If a setup script doesn't import setuptools, you can't use any setuptools-provided commands, like bdist_egg. When buildout runs a setup script, it arranges to import setuptools before running the script so setuptools-provided commands are available.

If you use a squeaky-clean Python to do your development, the setup script that would import setuptools because setuptools isn't in the path. Because buildout requires setuptools and knows where it has installed a setuptools egg, it adds the setuptools egg to the Python path before running the script. To run a setup script, use the buildout setup command, passing the name of a script or a directory containing a setup script and arguments to the script. Let's look at an example:

>>> mkdir('test')
>>> cd('test')
>>> write('setup.py',
... '''
... from distutils.core import setup
... setup(name='sample')
... ''')

We've created a super simple (stupid) setup script. Note that it doesn't import setuptools. Let's try running it to create an egg. We'll use the buildout script from our sample buildout:

>>> print system(buildout+' setup'),
... # doctest: +NORMALIZE_WHITESPACE
Error: The setup command requires the path to a setup script or
directory containing a setup script, and its arguments.

Oops, we forgot to give the name of the setup script:

>>> print system(buildout+' setup setup.py bdist_egg'),
... # doctest: +ELLIPSIS
Running setup script 'setup.py'.
...
>>> ls('dist')
-  sample-0.0.0-py2.5.egg

Note that we can specify a directory name. This is often shorter and preferred by the lazy :)

>>> print system(buildout+' setup . bdist_egg'), # doctest: +ELLIPSIS
Running setup script './setup.py'.
...

Automatic Buildout Updates

When a buildout is run, one of the first steps performed is to check for updates to either zc.buildout or setuptools. To demonstrate this, we've created some "new releases" of buildout and setuptools in a new_releases folder:

>>> ls(new_releases)
d  setuptools
-  setuptools-1.99.99-py2.4.egg
d  zc.buildout
-  zc.buildout-1.100.0b1-pyN.N.egg
-  zc.buildout-1.99.99-py2.4.egg
-  zc.buildout-2.0.0-pyN.N.egg

Let's update the sample buildout.cfg to look in this area:

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... find-links = %(new_releases)s
... index = %(new_releases)s
... parts = show-versions
... develop = showversions
...
... [show-versions]
... recipe = showversions
... """ % dict(new_releases=new_releases))

We'll also include a recipe that echos the versions of setuptools and zc.buildout used:

>>> mkdir(sample_buildout, 'showversions')
>>> write(sample_buildout, 'showversions', 'showversions.py',
... """
... import pkg_resources
...
... class Recipe:
...
...     def __init__(self, buildout, name, options):
...         pass
...
...     def install(self):
...         for project in 'zc.buildout', 'setuptools':
...             req = pkg_resources.Requirement.parse(project)
...             print project, pkg_resources.working_set.find(req).version
...         return ()
...     update = install
... """)
>>> write(sample_buildout, 'showversions', 'setup.py',
... """
... from setuptools import setup
...
... setup(
...     name = "showversions",
...     entry_points = {'zc.buildout': ['default = showversions:Recipe']},
...     )
... """)

Now if we run the buildout, the buildout will upgrade itself to the new versions found in new releases:

>>> print system(buildout),
Getting distribution for 'zc.buildout>=1.99, <2dev'.
Got zc.buildout 1.99.99.
Getting distribution for 'setuptools'.
Got setuptools 1.99.99.
Upgraded:
  zc.buildout version 1.99.99,
  setuptools version 1.99.99;
restarting.
Generated script '/sample-buildout/bin/buildout'.
Develop: '/sample-buildout/showversions'
Installing show-versions.
zc.buildout 1.99.99
setuptools 1.99.99

Notice that, even though we have a newer beta version of zc.buildout available, the final "1.99.99" was selected. If you want to get non-final versions, specify a specific version in your buildout's versions section, you typically want to use the --accept-buildout-test-releases option to the bootstrap script, which internally uses the accept-buildout-test-releases = true discussed below.

Also, even thought there's a later final version, buildout won't upgrade itself past version 1.

Our buildout script's site.py has been updated to use the new eggs:

>>> cat(sample_buildout, 'parts', 'buildout', 'site.py')
... # doctest: +NORMALIZE_WHITESPACE +ELLIPSIS
"...
def addsitepackages(known_paths):
    """Add site packages, as determined by zc.buildout.
<BLANKLINE>
    See original_addsitepackages, below, for the original version."""
    setuptools_path = '/sample-buildout/eggs/setuptools-1.99.99-pyN.N.egg'
    sys.path.append(setuptools_path)
    known_paths.add(os.path.normcase(setuptools_path))
    import pkg_resources
    buildout_paths = [
        '/sample-buildout/eggs/zc.buildout-1.99.99-pyN.N.egg',
        '/sample-buildout/eggs/setuptools-1.99.99-pyN.N.egg'
        ]
    for path in buildout_paths:
        sitedir, sitedircase = makepath(path)
        if not sitedircase in known_paths and os.path.exists(sitedir):
            sys.path.append(sitedir)
            known_paths.add(sitedircase)
            pkg_resources.working_set.add_entry(sitedir)
    sys.__egginsert = len(buildout_paths) # Support setuptools.
    original_paths = [
        ...
        ]
    for path in original_paths:
        if path == setuptools_path or path not in known_paths:
            addsitedir(path, known_paths)
    return known_paths
...

Now, let's recreate the sample buildout. If we specify constraints on the versions of zc.buildout and setuptools (or distribute) to use, running the buildout will install earlier versions of these packages:

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... find-links = %(new_releases)s
... index = %(new_releases)s
... parts = show-versions
... develop = showversions
... zc.buildout-version = < 1.99
... setuptools-version = < 1.99
... distribute-version = < 1.99
...
... [show-versions]
... recipe = showversions
... """ % dict(new_releases=new_releases))

Now we can see that we actually "upgrade" to an earlier version.

>>> print system(buildout),
Upgraded:
  zc.buildout version 1.0.0,
  setuptools version 0.6;
restarting.
Develop: '/sample-buildout/showversions'
Updating show-versions.
zc.buildout 1.0.0
setuptools 0.6

There are a number of cases, described below, in which the updates don't happen.

We won't upgrade in offline mode:

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... find-links = %(new_releases)s
... index = %(new_releases)s
... parts = show-versions
... develop = showversions
...
... [show-versions]
... recipe = showversions
... """ % dict(new_releases=new_releases))
>>> print system(buildout+' -o'),
Develop: '/sample-buildout/showversions'
Updating show-versions.
zc.buildout 1.0.0
setuptools 0.6

Or in non-newest mode:

>>> print system(buildout+' -N'),
Develop: '/sample-buildout/showversions'
Updating show-versions.
zc.buildout 1.0.0
setuptools 0.6

We also won't upgrade if the buildout script being run isn't in the buildout's bin directory. To see this we'll create a new buildout directory:

>>> sample_buildout2 = tmpdir('sample_buildout2')
>>> write(sample_buildout2, 'buildout.cfg',
... """
... [buildout]
... find-links = %(new_releases)s
... index = %(new_releases)s
... parts =
... """ % dict(new_releases=new_releases))
>>> cd(sample_buildout2)
>>> print system(buildout),
Creating directory '/sample_buildout2/bin'.
Creating directory '/sample_buildout2/parts'.
Creating directory '/sample_buildout2/eggs'.
Creating directory '/sample_buildout2/develop-eggs'.
Getting distribution for 'zc.buildout>=1.99, <2dev'.
Got zc.buildout 1.99.99.
Getting distribution for 'setuptools'.
Got setuptools 1.99.99.
Not upgrading because not running a local buildout command.
>>> ls('bin')

As mentioned above, the accept-buildout-test-releases = true means that newer non-final versions of these dependencies are preferred. Typically users are not expected to actually manipulate this value. Instead, the bootstrap script creates a buildout buildout script that passes in the value as a command line override. This then results in the buildout script being rewritten to remember the decision.

We'll mimic this by passing the argument actually in the command line.

>>> cd(sample_buildout)
>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... find-links = %(new_releases)s
... index = %(new_releases)s
... parts = show-versions
... develop = showversions
...
... [show-versions]
... recipe = showversions
... """ % dict(new_releases=new_releases))
>>> print system(buildout +
...              ' buildout:accept-buildout-test-releases=true'),
... # doctest: +NORMALIZE_WHITESPACE
Getting distribution for 'zc.buildout>=1.99, <2dev'.
Got zc.buildout 1.100.0b1.
Upgraded:
  zc.buildout version 1.100.0b1,
  setuptools version 1.99.99;
restarting.
Generated script '/sample-buildout/bin/buildout'.
NOTE: Accepting early releases of build system packages.  Rerun bootstrap
      without --accept-buildout-test-releases (-t) to return to default
      behavior.
Develop: '/sample-buildout/showversions'
Updating show-versions.
zc.buildout 1.100.0b1
setuptools 1.99.99

The buildout script shows the change.

>>> buildout_script = join(sample_buildout, 'bin', 'buildout')
>>> import sys
>>> if sys.platform.startswith('win'):
...     buildout_script += '-script.py'
>>> print open(buildout_script).read() # doctest: +ELLIPSIS
#...
sys.argv.insert(1, 'buildout:accept-buildout-test-releases=true')
print ('NOTE: Accepting early releases of build system packages.  Rerun '
       'bootstrap without --accept-buildout-test-releases (-t) to return to '
       'default behavior.')
...

If the update process for buildout or setuptools fails the error should be caught (displaying a warning) and the rest of the buildout update process should continue.

>>> version = sys.version_info[0:2]
>>> egg = new_releases + '/zc.buildout-1.99.99-py%s.%s.egg' % version
>>> copy_egg = new_releases + '/zc.buildout-1.1000-py%s.%s.egg' % version
>>> import shutil
>>> shutil.copy(egg, copy_egg)

Create a broken egg

>>> mkdir(sample_buildout, 'broken')
>>> write(sample_buildout, 'broken', 'setup.py', "import broken_egg\n")
>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... find-links = %(new_releases)s
... index = %(new_releases)s
... parts = show-versions
... develop =
...   broken
...
... [broken]
... recipe = zc.recipe.egg
... eggs = broken
... """ % dict(new_releases=new_releases))
>>> import subprocess
>>> subprocess.call([buildout])
1

Debugging buildouts

Buildouts can be pretty complex. When things go wrong, it isn't always obvious why. Errors can occur due to problems in user input or due to bugs in zc.buildout or recipes. When an error occurs, Python's post-mortem debugger can be used to inspect the state of the buildout or recipe code where the error occurred. To enable this, use the -D option to the buildout. Let's create a recipe that has a bug:

>>> mkdir(sample_buildout, 'recipes')
>>> write(sample_buildout, 'recipes', 'mkdir.py',
... """
... import os, zc.buildout
...
... class Mkdir:
...
...     def __init__(self, buildout, name, options):
...         self.name, self.options = name, options
...         options['path'] = os.path.join(
...                               buildout['buildout']['directory'],
...                               options['path'],
...                               )
...
...     def install(self):
...         directory = self.options['directory']
...         os.mkdir(directory)
...         return directory
...
...     def update(self):
...         pass
... """)
>>> write(sample_buildout, 'recipes', 'setup.py',
... """
... from setuptools import setup
...
... setup(name = "recipes",
...       entry_points = {'zc.buildout': ['mkdir = mkdir:Mkdir']},
...       )
... """)

And create a buildout that uses it:

>>> write(sample_buildout, 'buildout.cfg',
... """
... [buildout]
... develop = recipes
... parts = data-dir
...
... [data-dir]
... recipe = recipes:mkdir
... path = mystuff
... """)

If we run the buildout, we'll get an error:

>>> print system(buildout),
Develop: '/sample-buildout/recipes'
Installing data-dir.
While:
  Installing data-dir.
Error: Missing option: data-dir:directory

If we want to debug the error, we can add the -D option. Here's we'll supply some input:

>>> print system(buildout+" -D", """\
... up
... p self.options.keys()
... q
... """),
Develop: '/sample-buildout/recipes'
Installing data-dir.
> /zc/buildout/buildout.py(925)__getitem__()
-> raise MissingOption("Missing option: %s:%s" % (self.name, key))
(Pdb) > /sample-buildout/recipes/mkdir.py(14)install()
-> directory = self.options['directory']
(Pdb) ['path', 'recipe']
(Pdb) While:
  Installing data-dir.
Traceback (most recent call last):
  File "/zc/buildout/buildout.py", line 1352, in main
    getattr(buildout, command)(args)
  File "/zc/buildout/buildout.py", line 383, in install
    installed_files = self[part]._call(recipe.install)
  File "/zc/buildout/buildout.py", line 961, in _call
    return f()
  File "/sample-buildout/recipes/mkdir.py", line 14, in install
    directory = self.options['directory']
  File "/zc/buildout/buildout.py", line 925, in __getitem__
    raise MissingOption("Missing option: %s:%s" % (self.name, key))
MissingOption: Missing option: data-dir:directory
<BLANKLINE>
Starting pdb:

Testing Support

The zc.buildout.testing module provides an API that can be used when writing recipe tests. This API is documented below. Many examples of using this API can be found in the zc.buildout, zc.recipe.egg, and zc.recipe.testrunner tests.

zc.buildout.testing.buildoutSetUp(test)

The buildoutSetup function can be used as a doctest setup function. It creates a sample buildout that can be used by tests, changing the current working directory to the sample_buildout. It also adds a number of names to the test namespace:

sample_buildout
This is the name of a buildout with a basic configuration.
buildout
This is the path of the buildout script in the sample buildout.
ls(*path)
List the contents of a directory. The directory path is provided as one or more strings, to be joined with os.path.join.
cat(*path)

Display the contents of a file. The file path is provided as one or more strings, to be joined with os.path.join.

On Windows, if the file doesn't exist, the function will try adding a '-script.py' suffix. This helps to work around a difference in script generation on windows.

mkdir(*path)
Create a directory. The directory path is provided as one or more strings, to be joined with os.path.join.
rmdir(*path)
Remove a directory. The directory path is provided as one or more strings, to be joined with os.path.join.
remove(*path)
Remove a directory or file. The path is provided as one or more strings, to be joined with os.path.join.
tmpdir(name)

Create a temporary directory with the given name. The directory will be automatically removed at the end of the test. The path of the created directory is returned.

Further, if the the normalize_path normlaizing substitution (see below) is used, then any paths starting with this path will be normalized to:

/name/restofpath

No two temporary directories can be created with the same name. A directory created with tmpdir can be removed with rmdir and recreated.

Note that the sample_buildout directory is created by calling this function.

write(*path_and_contents)
Create a file. The file path is provided as one or more strings, to be joined with os.path.join. The last argument is the file contents.
system(command, input='')
Execute a system command with the given input passed to the command's standard input. The output (error and regular output) from the command is returned.
get(url)
Get a web page.
cd(*path)

Change to the given directory. The directory path is provided as one or more strings, to be joined with os.path.join.

The directory will be reset at the end of the test.

uncd()
Change to the directory that was current prior to the previous call to cd. You can call cd multiple times and then uncd the same number of times to return to the same location.
join(*path)
A convenient reference to os.path.join.
register_teardown(func)
Register a tear-down function. The function will be called with no arguments at the end of the test.
start_server(path)

Start a web server on the given path. The server will be shut down at the end of the test. The server URL is returned.

You can cause the server to start and stop logging it's output using:

>>> get(server_url+'enable_server_logging')

and:

>>> get(server_url+'disable_server_logging')

This can be useful to see how buildout is interacting with a server.

sdist(setup, dest)
Create a source distribution by running the given setup file and placing the result in the given destination directory. If the setup argument is a directory, the thge setup.py file in that directory is used.
bdist_egg(setup, executable, dest)
Create an egg by running the given setup file with the given Python executable and placing the result in the given destination directory. If the setup argument is a directory, then the setup.py file in that directory is used.
find_python(version)

Find a Python executable for the given version, where version is a string like "2.4".

This function uses the following strategy to find a Python of the given version:

  • Look for an environment variable of the form PYTHON%(version)s.
  • On windows, look for Pythonm%(version)spython
  • on Unix, try running python%(version)s or just python to get the executable

zc.buildout.testing.buildoutTearDown(test)

Tear down everything set up by zc.buildout.testing.buildoutSetUp. Any functions passed to register_teardown are called as well.

install(project, destination)

Install eggs for a given project into a destination. If the destination is a test object, then the eggs directory of the sample buildout (sample_buildout) defined by the test will be used. Tests will use this to install the distributions for the packages being tested (and their dependencies) into a sample buildout. The egg to be used should already be loaded, by importing one of the modules provided, before calling this function.

install_develop(project, destination)

Like install, but a develop egg is installed even if the current egg if not a develop egg.

Output normalization

Recipe tests often generate output that is dependent on temporary file locations, operating system conventions or Python versions. To deal with these dependencies, we often use zope.testing.renormalizing.RENormalizing to normalize test output. zope.testing.renormalizing.RENormalizing takes pairs of regular expressions and substitutions. The zc.buildout.testing module provides a few helpful variables that define regular-expression/substitution pairs that you can pass to zope.testing.renormalizing.RENormalizing.

normalize_path
Converts tests paths, based on directories created with tmpdir(), to simple paths.
normalize_script
On Unix-like systems, scripts are implemented in single files without suffixes. On windows, scripts are implemented with 2 files, a -script.py file and a .exe file. This normalization converts directory listings of Windows scripts to the form generated on UNix-like systems.
normalize_egg_py
Normalize Python version and platform indicators, if specified, in egg names.

Python API for egg and script installation

The easy_install module provides some functions to provide support for egg and script installation. It provides functionality at the python level that is similar to easy_install, with a few exceptions:

  • By default, we look for new packages and the packages that they depend on. This is somewhat like (and uses) the --upgrade option of easy_install, except that we also upgrade required packages.
  • If the highest-revision package satisfying a specification is already present, then we don't try to get another one. This saves a lot of search time in the common case that packages are pegged to specific versions.
  • If there is a develop egg that satisfies a requirement, we don't look for additional distributions. We always give preference to develop eggs.
  • Distutils options for building extensions can be passed.

Distribution installation

The easy_install module provides a function, install, for installing one or more packages and their dependencies. The install function takes 2 positional arguments:

  • An iterable of setuptools requirement strings for the distributions to be installed, and
  • A destination directory to install to and to satisfy requirements from. The destination directory can be None, in which case, no new distributions are downloaded and there will be an error if the needed distributions can't be found among those already installed.

It supports a number of optional keyword arguments:

links
A sequence of URLs, file names, or directories to look for links to distributions.
index

The URL of an index server, or almost any other valid URL. :)

If not specified, the Python Package Index, http://pypi.python.org/simple/, is used. You can specify an alternate index with this option. If you use the links option and if the links point to the needed distributions, then the index can be anything and will be largely ignored. In the examples, here, we'll just point to an empty directory on our link server. This will make our examples run a little bit faster.

executable
A path to a Python executable. Distributions will be installed using this executable and will be for the matching Python version.
path
A list of additional directories to search for locally-installed distributions.
always_unzip
A flag indicating that newly-downloaded distributions should be directories even if they could be installed as zip files.
working_set
An existing working set to be augmented with additional distributions, if necessary to satisfy requirements. This allows you to call install multiple times, if necessary, to gather multiple sets of requirements.
newest
A boolean value indicating whether to search for new distributions when already-installed distributions meet the requirement. When this is true, the default, and when the destination directory is not None, then the install function will search for the newest distributions that satisfy the requirements.
versions
A dictionary mapping project names to version numbers to be used when selecting distributions. This can be used to specify a set of distribution versions independent of other requirements.
use_dependency_links
A flag indicating whether to search for dependencies using the setup dependency_links metadata or not. If true, links are searched for using dependency_links in preference to other locations. Defaults to true.
include_site_packages

A flag indicating whether Python's non-standard-library packages should be available for finding dependencies. Defaults to true.

Paths outside of Python's standard library--or more precisely, those that are not included when Python is started with the -S argument--are loosely referred to as "site-packages" here.

relative_paths
Adjust egg paths so they are relative to the script path. This allows scripts to work when scripts and eggs are moved, as long as they are both moved in the same way.

The install method returns a working set containing the distributions needed to meet the given requirements.

We have a link server that has a number of eggs:

>>> print get(link_server),
<html><body>
<a href="bigdemo-0.1-py2.4.egg">bigdemo-0.1-py2.4.egg</a><br>
<a href="demo-0.1-py2.4.egg">demo-0.1-py2.4.egg</a><br>
<a href="demo-0.2-py2.4.egg">demo-0.2-py2.4.egg</a><br>
<a href="demo-0.3-py2.4.egg">demo-0.3-py2.4.egg</a><br>
<a href="demo-0.4c1-py2.4.egg">demo-0.4c1-py2.4.egg</a><br>
<a href="demoneeded-1.0.zip">demoneeded-1.0.zip</a><br>
<a href="demoneeded-1.1.zip">demoneeded-1.1.zip</a><br>
<a href="demoneeded-1.2c1.zip">demoneeded-1.2c1.zip</a><br>
<a href="extdemo-1.4.zip">extdemo-1.4.zip</a><br>
<a href="index/">index/</a><br>
<a href="other-1.0-py2.4.egg">other-1.0-py2.4.egg</a><br>
</body></html>

Let's make a directory and install the demo egg to it, using the demo:

>>> dest = tmpdir('sample-install')
>>> import zc.buildout.easy_install
>>> ws = zc.buildout.easy_install.install(
...     ['demo==0.2'], dest,
...     links=[link_server], index=link_server+'index/')

We requested version 0.2 of the demo distribution to be installed into the destination server. We specified that we should search for links on the link server and that we should use the (empty) link server index directory as a package index.

The working set contains the distributions we retrieved.

>>> for dist in ws:
...     print dist
demo 0.2
demoneeded 1.1

We got demoneeded because it was a dependency of demo.

And the actual eggs were added to the eggs directory.

>>> ls(dest)
-  demo-0.2-py2.4.egg
-  demoneeded-1.1-py2.4.egg

If we remove the version restriction on demo, but specify a false value for newest, no new distributions will be installed:

>>> ws = zc.buildout.easy_install.install(
...     ['demo'], dest, links=[link_server], index=link_server+'index/',
...     newest=False)
>>> ls(dest)
-  demo-0.2-py2.4.egg
-  demoneeded-1.1-py2.4.egg

If we leave off the newest option, we'll get an update for demo:

>>> ws = zc.buildout.easy_install.install(
...     ['demo'], dest, links=[link_server], index=link_server+'index/')
>>> ls(dest)
-  demo-0.2-py2.4.egg
-  demo-0.3-py2.4.egg
-  demoneeded-1.1-py2.4.egg

Note that we didn't get the newest versions available. There were release candidates for newer versions of both packages. By default, final releases are preferred. We can change this behavior using the prefer_final function:

>>> zc.buildout.easy_install.prefer_final(False)
True

The old setting is returned.

>>> ws = zc.buildout.easy_install.install(
...     ['demo'], dest, links=[link_server], index=link_server+'index/')
>>> for dist in ws:
...     print dist
demo 0.4c1
demoneeded 1.2c1
>>> ls(dest)
-  demo-0.2-py2.4.egg
-  demo-0.3-py2.4.egg
-  demo-0.4c1-py2.4.egg
-  demoneeded-1.1-py2.4.egg
-  demoneeded-1.2c1-py2.4.egg

Let's put the setting back to the default.

>>> zc.buildout.easy_install.prefer_final(True)
False

We can supply additional distributions. We can also supply specifications for distributions that would normally be found via dependencies. We might do this to specify a specific version.

>>> ws = zc.buildout.easy_install.install(
...     ['demo', 'other', 'demoneeded==1.0'], dest,
...     links=[link_server], index=link_server+'index/')
>>> for dist in ws:
...     print dist
demo 0.3
other 1.0
demoneeded 1.0
>>> ls(dest)
-  demo-0.2-py2.4.egg
-  demo-0.3-py2.4.egg
-  demo-0.4c1-py2.4.egg
-  demoneeded-1.0-py2.4.egg
-  demoneeded-1.1-py2.4.egg
-  demoneeded-1.2c1-py2.4.egg
d  other-1.0-py2.4.egg

We can request that eggs be unzipped even if they are zip safe. This can be useful when debugging. (Note that Distribute will unzip eggs by default, so if you are using Distribute, most or all eggs will already be unzipped without this flag.)

>>> rmdir(dest)
>>> dest = tmpdir('sample-install')
>>> ws = zc.buildout.easy_install.install(
...     ['demo'], dest, links=[link_server], index=link_server+'index/',
...     always_unzip=True)
>>> ls(dest)
d  demo-0.3-py2.4.egg
d  demoneeded-1.1-py2.4.egg
>>> rmdir(dest)
>>> dest = tmpdir('sample-install')
>>> ws = zc.buildout.easy_install.install(
...     ['demo'], dest, links=[link_server], index=link_server+'index/',
...     always_unzip=False)
>>> ls(dest)
-  demo-0.3-py2.4.egg
-  demoneeded-1.1-py2.4.egg

We can also set a default by calling the always_unzip function:

>>> zc.buildout.easy_install.always_unzip(True)
False

The old default is returned:

>>> rmdir(dest)
>>> dest = tmpdir('sample-install')
>>> ws = zc.buildout.easy_install.install(
...     ['demo'], dest, links=[link_server], index=link_server+'index/')
>>> ls(dest)
d  demo-0.3-py2.4.egg
d  demoneeded-1.1-py2.4.egg
>>> zc.buildout.easy_install.always_unzip(False)
True
>>> rmdir(dest)
>>> dest = tmpdir('sample-install')
>>> ws = zc.buildout.easy_install.install(
...     ['demo'], dest, links=[link_server], index=link_server+'index/')
>>> ls(dest)
-  demo-0.3-py2.4.egg
-  demoneeded-1.1-py2.4.egg
>>> rmdir(dest)
>>> dest = tmpdir('sample-install')
>>> ws = zc.buildout.easy_install.install(
...     ['demo'], dest, links=[link_server], index=link_server+'index/',
...     always_unzip=True)
>>> ls(dest)
d  demo-0.3-py2.4.egg
d  demoneeded-1.1-py2.4.egg

Specifying version information independent of requirements

Sometimes it's useful to specify version information independent of normal requirements specifications. For example, a buildout may need to lock down a set of versions, without having to put put version numbers in setup files or part definitions. If a dictionary is passed to the install function, mapping project names to version numbers, then the versions numbers will be used.

>>> ws = zc.buildout.easy_install.install(
...     ['demo'], dest, links=[link_server], index=link_server+'index/',
...     versions = dict(demo='0.2', demoneeded='1.0'))
>>> [d.version for d in ws]
['0.2', '1.0']

In this example, we specified a version for demoneeded, even though we didn't define a requirement for it. The versions specified apply to dependencies as well as the specified requirements.

If we specify a version that's incompatible with a requirement, then we'll get an error:

>>> from zope.testing.loggingsupport import InstalledHandler
>>> handler = InstalledHandler('zc.buildout.easy_install')
>>> import logging
>>> logging.getLogger('zc.buildout.easy_install').propagate = False
>>> ws = zc.buildout.easy_install.install(
...     ['demo >0.2'], dest, links=[link_server],
...     index=link_server+'index/',
...     versions = dict(demo='0.2', demoneeded='1.0'))
Traceback (most recent call last):
...
IncompatibleConstraintError: Bad constraint 0.2 demo>0.2
>>> print handler
zc.buildout.easy_install DEBUG
  Installing 'demo >0.2'.
zc.buildout.easy_install ERROR
  The constraint, 0.2, is not consistent with the requirement, 'demo>0.2'.
>>> handler.clear()

If no versions are specified, a debugging message will be output reporting that a version was picked automatically:

>>> ws = zc.buildout.easy_install.install(
...     ['demo'], dest, links=[link_server], index=link_server+'index/',
...     )
>>> print handler
zc.buildout.easy_install DEBUG
  Installing 'demo'.
zc.buildout.easy_install DEBUG
  We have the best distribution that satisfies 'demo'.
zc.buildout.easy_install DEBUG
  Picked: demo = 0.3
zc.buildout.easy_install DEBUG
  Getting required 'demoneeded'
zc.buildout.easy_install DEBUG
    required by demo 0.3.
zc.buildout.easy_install DEBUG
  We have the best distribution that satisfies 'demoneeded'.
zc.buildout.easy_install DEBUG
  Picked: demoneeded = 1.1
>>> handler.uninstall()
>>> logging.getLogger('zc.buildout.easy_install').propagate = True

We can request that we get an error if versions are picked:

>>> zc.buildout.easy_install.allow_picked_versions(False)
True

(The old setting is returned.)

>>> ws = zc.buildout.easy_install.install(
...     ['demo'], dest, links=[link_server], index=link_server+'index/',
...     )
Traceback (most recent call last):
...
UserError: Picked: demo = 0.3
>>> zc.buildout.easy_install.allow_picked_versions(True)
False

The function default_versions can be used to get and set default version information to be used when no version information is passes. If called with an argument, it sets the default versions:

>>> zc.buildout.easy_install.default_versions(dict(demoneeded='1'))
{}

It always returns the previous default versions. If called without an argument, it simply returns the default versions without changing them:

>>> zc.buildout.easy_install.default_versions()
{'demoneeded': '1'}

So with the default versions set, we'll get the requested version even if the versions option isn't used:

>>> ws = zc.buildout.easy_install.install(
...     ['demo'], dest, links=[link_server], index=link_server+'index/',
...     )
>>> [d.version for d in ws]
['0.3', '1.0']

Of course, we can unset the default versions by passing an empty dictionary:

>>> zc.buildout.easy_install.default_versions({})
{'demoneeded': '1'}
>>> ws = zc.buildout.easy_install.install(
...     ['demo'], dest, links=[link_server], index=link_server+'index/',
...     )
>>> [d.version for d in ws]
['0.3', '1.1']

Dependencies in Site Packages

Paths outside of Python's standard library--or more precisely, those that are not included when Python is started with the -S argument--are loosely referred to as "site-packages" here. These site-packages are searched by default for distributions. This can be disabled, so that, for instance, a system Python can be used with buildout, cleaned of any packages installed by a user or system package manager.

The default behavior can be controlled and introspected using zc.buildout.easy_install.include_site_packages.

>>> zc.buildout.easy_install.include_site_packages()
True

Here's an example of using a Python executable that includes our dependencies.

Our "py_path" will have the "demoneeded," and "demo" packages available.

We'll simply be asking for "demoneeded" here, but without any external index or links.

>>> from zc.buildout.tests import create_sample_sys_install
>>> py_path, site_packages_path = make_py()
>>> create_sample_sys_install(site_packages_path)
>>> example_dest = tmpdir('site-packages-example-install')
>>> workingset = zc.buildout.easy_install.install(
...     ['demoneeded'], example_dest, links=[], executable=py_path,
...     index=None)
>>> [dist.project_name for dist in workingset]
['demoneeded']

That worked fine. Let's try again with site packages not allowed. We'll change the policy by changing the default. Notice that the function for changing the default value returns the previous value.

>>> zc.buildout.easy_install.include_site_packages(False)
True
>>> zc.buildout.easy_install.include_site_packages()
False
>>> zc.buildout.easy_install.clear_index_cache()
>>> rmdir(example_dest)
>>> example_dest = tmpdir('site-packages-example-install')
>>> workingset = zc.buildout.easy_install.install(
...     ['demoneeded'], example_dest, links=[], executable=py_path,
...     index=None)
Traceback (most recent call last):
    ...
MissingDistribution: Couldn't find a distribution for 'demoneeded'.
>>> zc.buildout.easy_install.clear_index_cache()

Now we'll reset the default.

>>> zc.buildout.easy_install.include_site_packages(True)
False
>>> zc.buildout.easy_install.include_site_packages()
True

Dependency links

Setuptools allows metadata that describes where to search for package dependencies. This option is called dependency_links. Buildout has its own notion of where to look for dependencies, but it also uses the setup tools dependency_links information if it's available.

Let's demo this by creating an egg that specifies dependency_links.

To begin, let's create a new egg repository. This repository hold a newer version of the 'demoneeded' egg than the sample repository does.

>>> repoloc = tmpdir('repo')
>>> from zc.buildout.tests import create_egg
>>> create_egg('demoneeded', '1.2', repoloc)
>>> link_server2 = start_server(repoloc)

Turn on logging on this server so that we can see when eggs are pulled from it.

>>> get(link_server2 + 'enable_server_logging')
GET 200 /enable_server_logging
''

Now we can create an egg that specifies that its dependencies are found on this server.

>>> repoloc = tmpdir('repo2')
>>> create_egg('hasdeps', '1.0', repoloc,
...            install_requires = "'demoneeded'",
...            dependency_links = [link_server2])

Let's add the egg to another repository.

>>> link_server3 = start_server(repoloc)

Now let's install the egg.

>>> example_dest = tmpdir('example-install')
>>> workingset = zc.buildout.easy_install.install(
...     ['hasdeps'], example_dest,
...     links=[link_server3], index=link_server3+'index/')
GET 200 /
GET 200 /demoneeded-1.2-pyN.N.egg

The server logs show that the dependency was retrieved from the server specified in the dependency_links.

Now let's see what happens if we provide two different ways to retrieve the dependencies.

>>> rmdir(example_dest)
>>> example_dest = tmpdir('example-install')
>>> workingset = zc.buildout.easy_install.install(
...     ['hasdeps'], example_dest, index=link_server+'index/',
...     links=[link_server, link_server3])
GET 200 /
GET 200 /demoneeded-1.2-pyN.N.egg

Once again the dependency is fetched from the logging server even though it is also available from the non-logging server. This is because the version on the logging server is newer and buildout normally chooses the newest egg available.

If you wish to control where dependencies come from regardless of dependency_links setup metadata use the 'use_dependency_links' option to zc.buildout.easy_install.install().

>>> rmdir(example_dest)
>>> example_dest = tmpdir('example-install')
>>> workingset = zc.buildout.easy_install.install(
...     ['hasdeps'], example_dest, index=link_server+'index/',
...     links=[link_server, link_server3],
...     use_dependency_links=False)

Notice that this time the dependency egg is not fetched from the logging server. When you specify not to use dependency_links, eggs will only be searched for using the links you explicitly provide.

Another way to control this option is with the zc.buildout.easy_install.use_dependency_links() function. This function sets the default behavior for the zc.buildout.easy_install() function.

>>> zc.buildout.easy_install.use_dependency_links(False)
True

The function returns its previous setting.

>>> rmdir(example_dest)
>>> example_dest = tmpdir('example-install')
>>> workingset = zc.buildout.easy_install.install(
...     ['hasdeps'], example_dest, index=link_server+'index/',
...     links=[link_server, link_server3])

It can be overridden by passing a keyword argument to the install function.

>>> rmdir(example_dest)
>>> example_dest = tmpdir('example-install')
>>> workingset = zc.buildout.easy_install.install(
...     ['hasdeps'], example_dest, index=link_server+'index/',
...     links=[link_server, link_server3],
...     use_dependency_links=True)
GET 200 /demoneeded-1.2-pyN.N.egg

To return the dependency_links behavior to normal call the function again.

>>> zc.buildout.easy_install.use_dependency_links(True)
False
>>> rmdir(example_dest)
>>> example_dest = tmpdir('example-install')
>>> workingset = zc.buildout.easy_install.install(
...     ['hasdeps'], example_dest, index=link_server+'index/',
...     links=[link_server, link_server3])
GET 200 /demoneeded-1.2-pyN.N.egg

Script generation

The easy_install module provides support for creating scripts from eggs. It provides two competing functions. One, scripts, is a well-established approach to generating reliable scripts with a "clean" Python--e.g., one that does not have any packages in its site-packages. The other, sitepackage_safe_scripts, is newer, a bit trickier, and is designed to work with a Python that has code in its site-packages, such as a system Python.

Both are similar to setuptools except that they provides facilities for baking a script's path into the script. This has two advantages:

  • The eggs to be used by a script are not chosen at run time, making startup faster and, more importantly, deterministic.
  • The script doesn't have to import pkg_resources because the logic that pkg_resources would execute at run time is executed at script-creation time. (There is an exception in sitepackage_safe_scripts if you want to have your Python's site packages available, as discussed below, but even in that case pkg_resources is only partially activated, which can be a significant time savings.)

The scripts function

The scripts function is the first way to generate scripts that we'll examine. It is the earlier approach that the package offered. Let's create a destination directory for it to place them in:

>>> bin = tmpdir('bin')

Now, we'll use the scripts function to generate scripts in this directory from the demo egg:

>>> import sys
>>> scripts = zc.buildout.easy_install.scripts(
...     ['demo'], ws, sys.executable, bin)

the four arguments we passed were:

  1. A sequence of distribution requirements. These are of the same form as setuptools requirements. Here we passed a single requirement, for the version 0.1 demo distribution.
  2. A working set,
  3. The Python executable to use, and
  1. The destination directory.

The bin directory now contains a generated script:

>>> ls(bin)
-  demo

The return value is a list of the scripts generated:

>>> import os, sys
>>> if sys.platform == 'win32':
...     scripts == [os.path.join(bin, 'demo.exe'),
...                 os.path.join(bin, 'demo-script.py')]
... else:
...     scripts == [os.path.join(bin, 'demo')]
True

Note that in Windows, 2 files are generated for each script. A script file, ending in '-script.py', and an exe file that allows the script to be invoked directly without having to specify the Python interpreter and without having to provide a '.py' suffix.

The demo script run the entry point defined in the demo egg:

>>> cat(bin, 'demo') # doctest: +NORMALIZE_WHITESPACE
#!/usr/local/bin/python2.4
<BLANKLINE>
import sys
sys.path[0:0] = [
  '/sample-install/demo-0.3-py2.4.egg',
  '/sample-install/demoneeded-1.1-py2.4.egg',
  ]
<BLANKLINE>
import eggrecipedemo
<BLANKLINE>
if __name__ == '__main__':
    sys.exit(eggrecipedemo.main())

Some things to note:

  • The demo and demoneeded eggs are added to the beginning of sys.path.
  • The module for the script entry point is imported and the entry point, in this case, 'main', is run.

Rather than requirement strings, you can pass tuples containing 3 strings:

  • A script name,
  • A module,
  • An attribute expression for an entry point within the module.

For example, we could have passed entry point information directly rather than passing a requirement:

>>> scripts = zc.buildout.easy_install.scripts(
...     [('demo', 'eggrecipedemo', 'main')],
...     ws, sys.executable, bin)
>>> cat(bin, 'demo') # doctest: +NORMALIZE_WHITESPACE
#!/usr/local/bin/python2.4
<BLANKLINE>
import sys
sys.path[0:0] = [
  '/sample-install/demo-0.3-py2.4.egg',
  '/sample-install/demoneeded-1.1-py2.4.egg',
  ]
<BLANKLINE>
import eggrecipedemo
<BLANKLINE>
if __name__ == '__main__':
    sys.exit(eggrecipedemo.main())

Passing entry-point information directly is handy when using eggs (or distributions) that don't declare their entry points, such as distributions that aren't based on setuptools.

The interpreter keyword argument can be used to generate a script that can be used to invoke the Python interactive interpreter with the path set based on the working set. This generated script can also be used to run other scripts with the path set on the working set:

>>> scripts = zc.buildout.easy_install.scripts(
...     ['demo'], ws, sys.executable, bin, interpreter='py')
>>> ls(bin)
-  demo
-  py
>>> if sys.platform == 'win32':
...     scripts == [os.path.join(bin, 'demo.exe'),
...                 os.path.join(bin, 'demo-script.py'),
...                 os.path.join(bin, 'py.exe'),
...                 os.path.join(bin, 'py-script.py')]
... else:
...     scripts == [os.path.join(bin, 'demo'),
...                 os.path.join(bin, 'py')]
True

The py script simply runs the Python interactive interpreter with the path set:

>>> cat(bin, 'py') # doctest: +NORMALIZE_WHITESPACE
#!/usr/local/bin/python2.4
<BLANKLINE>
import sys
<BLANKLINE>
sys.path[0:0] = [
  '/sample-install/demo-0.3-pyN.N.egg',
  '/sample-install/demoneeded-1.1-pyN.N.egg',
  ]
<BLANKLINE>
_interactive = True
if len(sys.argv) > 1:
    _options, _args = __import__("getopt").getopt(sys.argv[1:], 'ic:m:')
    _interactive = False
    for (_opt, _val) in _options:
        if _opt == '-i':
            _interactive = True
        elif _opt == '-c':
            exec _val
        elif _opt == '-m':
            sys.argv[1:] = _args
            _args = []
            __import__("runpy").run_module(
                 _val, {}, "__main__", alter_sys=True)
<BLANKLINE>
    if _args:
        sys.argv[:] = _args
        __file__ = _args[0]
        del _options, _args
        execfile(__file__)
<BLANKLINE>
if _interactive:
    del _interactive
    __import__("code").interact(banner="", local=globals())

If invoked with a script name and arguments, it will run that script, instead.

>>> write('ascript', '''
... "demo doc"
... print sys.argv
... print (__name__, __file__, __doc__)
... ''')
>>> print system(join(bin, 'py')+' ascript a b c'),
['ascript', 'a', 'b', 'c']
('__main__', 'ascript', 'demo doc')

For Python 2.5 and higher, you can also use the -m option to run a module:

>>> print system(join(bin, 'py')+' -m pdb'),
usage: pdb.py scriptfile [arg] ...
>>> print system(join(bin, 'py')+' -m pdb what'),
Error: what does not exist

An additional argument can be passed to define which scripts to install and to provide script names. The argument is a dictionary mapping original script names to new script names.

>>> bin = tmpdir('bin2')
>>> scripts = zc.buildout.easy_install.scripts(
...    ['demo'], ws, sys.executable, bin, dict(demo='run'))
>>> if sys.platform == 'win32':
...     scripts == [os.path.join(bin, 'run.exe'),
...                 os.path.join(bin, 'run-script.py')]
... else:
...     scripts == [os.path.join(bin, 'run')]
True
>>> ls(bin)
-  run
>>> print system(os.path.join(bin, 'run')),
3 1

The scripts function: Including extra paths in scripts

We can pass a keyword argument, extra paths, to cause additional paths to be included in the a generated script:

>>> foo = tmpdir('foo')
>>> scripts = zc.buildout.easy_install.scripts(
...    ['demo'], ws, sys.executable, bin, dict(demo='run'),
...    extra_paths=[foo])
>>> cat(bin, 'run') # doctest: +NORMALIZE_WHITESPACE
#!/usr/local/bin/python2.4
<BLANKLINE>
import sys
sys.path[0:0] = [
  '/sample-install/demo-0.3-py2.4.egg',
  '/sample-install/demoneeded-1.1-py2.4.egg',
  '/foo',
  ]
<BLANKLINE>
import eggrecipedemo
<BLANKLINE>
if __name__ == '__main__':
    sys.exit(eggrecipedemo.main())

The scripts function: Providing script arguments

An "argument" keyword argument can be used to pass arguments to an entry point. The value passed is a source string to be placed between the parentheses in the call:

>>> scripts = zc.buildout.easy_install.scripts(
...    ['demo'], ws, sys.executable, bin, dict(demo='run'),
...    arguments='1, 2')
>>> cat(bin, 'run') # doctest: +NORMALIZE_WHITESPACE
#!/usr/local/bin/python2.4
import sys
sys.path[0:0] = [
  '/sample-install/demo-0.3-py2.4.egg',
  '/sample-install/demoneeded-1.1-py2.4.egg',
  ]
<BLANKLINE>
import eggrecipedemo
<BLANKLINE>
if __name__ == '__main__':
    sys.exit(eggrecipedemo.main(1, 2))

The scripts function: Passing initialization code

You can also pass script initialization code:

>>> scripts = zc.buildout.easy_install.scripts(
...    ['demo'], ws, sys.executable, bin, dict(demo='run'),
...    arguments='1, 2',
...    initialization='import os\nos.chdir("foo")')
>>> cat(bin, 'run') # doctest: +NORMALIZE_WHITESPACE
#!/usr/local/bin/python2.4
import sys
sys.path[0:0] = [
  '/sample-install/demo-0.3-py2.4.egg',
  '/sample-install/demoneeded-1.1-py2.4.egg',
  ]
<BLANKLINE>
import os
os.chdir("foo")
<BLANKLINE>
import eggrecipedemo
<BLANKLINE>
if __name__ == '__main__':
    sys.exit(eggrecipedemo.main(1, 2))

The scripts function: Relative paths

Sometimes, you want to be able to move a buildout directory around and have scripts still work without having to rebuild them. We can control this using the relative_paths option to install. You need to pass a common base directory of the scripts and eggs:

>>> bo = tmpdir('bo')
>>> ba = tmpdir('ba')
>>> mkdir(bo, 'eggs')
>>> mkdir(bo, 'bin')
>>> mkdir(bo, 'other')
>>> ws = zc.buildout.easy_install.install(
...     ['demo'], join(bo, 'eggs'), links=[link_server],
...     index=link_server+'index/')
>>> scripts = zc.buildout.easy_install.scripts(
...    ['demo'], ws, sys.executable, join(bo, 'bin'), dict(demo='run'),
...    extra_paths=[ba, join(bo, 'bar')],
...    interpreter='py',
...    relative_paths=bo)
>>> cat(bo, 'bin', 'run') # doctest: +NORMALIZE_WHITESPACE
#!/usr/local/bin/python2.4
<BLANKLINE>
import os
<BLANKLINE>
join = os.path.join
base = os.path.dirname(os.path.abspath(os.path.realpath(__file__)))
base = os.path.dirname(base)
<BLANKLINE>
import sys
sys.path[0:0] = [
  join(base, 'eggs/demo-0.3-pyN.N.egg'),
  join(base, 'eggs/demoneeded-1.1-pyN.N.egg'),
  '/ba',
  join(base, 'bar'),
  ]
<BLANKLINE>
import eggrecipedemo
<BLANKLINE>
if __name__ == '__main__':
    sys.exit(eggrecipedemo.main())

Note that the extra path we specified that was outside the directory passed as relative_paths wasn't converted to a relative path.

Of course, running the script works:

>>> print system(join(bo, 'bin', 'run')),
3 1

We specified an interpreter and its paths are adjusted too:

>>> cat(bo, 'bin', 'py') # doctest: +NORMALIZE_WHITESPACE
#!/usr/local/bin/python2.4
<BLANKLINE>
import os
<BLANKLINE>
join = os.path.join
base = os.path.dirname(os.path.abspath(os.path.realpath(__file__)))
base = os.path.dirname(base)
<BLANKLINE>
import sys
<BLANKLINE>
sys.path[0:0] = [
  join(base, 'eggs/demo-0.3-pyN.N.egg'),
  join(base, 'eggs/demoneeded-1.1-pyN.N.egg'),
  '/ba',
  join(base, 'bar'),
  ]
<BLANKLINE>
_interactive = True
if len(sys.argv) > 1:
    _options, _args = __import__("getopt").getopt(sys.argv[1:], 'ic:m:')
    _interactive = False
    for (_opt, _val) in _options:
        if _opt == '-i':
            _interactive = True
        elif _opt == '-c':
            exec _val
        elif _opt == '-m':
            sys.argv[1:] = _args
            _args = []
            __import__("runpy").run_module(
                 _val, {}, "__main__", alter_sys=True)
<BLANKLINE>
    if _args:
        sys.argv[:] = _args
        __file__ = _args[0]
        del _options, _args
        execfile(__file__)
<BLANKLINE>
if _interactive:
    del _interactive
    __import__("code").interact(banner="", local=globals())

The sitepackage_safe_scripts function

The newer function for creating scripts is sitepackage_safe_scripts. It has the same basic functionality as the scripts function: it can create scripts to run arbitrary entry points, and to run a Python interpreter. The following are the differences from a user's perspective.

  • It can be used safely with a Python that has packages installed itself, such as a system-installed Python.
  • In contrast to the interpreter generated by the scripts method, which supports only a small subset of the usual Python executable's options, the interpreter generated by sitepackage_safe_scripts supports all of them. This makes it possible to use as full Python replacement for scripts that need the distributions specified in your buildout.
  • Both the interpreter and the entry point scripts allow you to include the site packages, and/or the sitecustomize, of the Python executable, if desired.

It works by creating site.py and sitecustomize.py files that set up the desired paths and initialization. These must be placed within an otherwise empty directory. Typically this is in a recipe's parts directory.

Here's the simplest example, building an interpreter script.

>>> interpreter_dir = tmpdir('interpreter')
>>> interpreter_parts_dir = os.path.join(
...     interpreter_dir, 'parts', 'interpreter')
>>> interpreter_bin_dir = os.path.join(interpreter_dir, 'bin')
>>> mkdir(interpreter_bin_dir)
>>> mkdir(interpreter_dir, 'eggs')
>>> mkdir(interpreter_dir, 'parts')
>>> mkdir(interpreter_parts_dir)
>>> ws = zc.buildout.easy_install.install(
...     ['demo'], join(interpreter_dir, 'eggs'), links=[link_server],
...     index=link_server+'index/')
>>> generated = zc.buildout.easy_install.sitepackage_safe_scripts(
...     interpreter_bin_dir, ws, sys.executable, interpreter_parts_dir,
...     interpreter='py')

Depending on whether the machine being used is running Windows or not, this produces either three or four files. In both cases, we have site.py and sitecustomize.py generated in the parts/interpreter directory. For Windows, we have py.exe and py-script.py; for other operating systems, we have py.

>>> sitecustomize_path = os.path.join(
...     interpreter_parts_dir, 'sitecustomize.py')
>>> site_path = os.path.join(interpreter_parts_dir, 'site.py')
>>> interpreter_path = os.path.join(interpreter_bin_dir, 'py')
>>> if sys.platform == 'win32':
...     py_path = os.path.join(interpreter_bin_dir, 'py-script.py')
...     expected = [sitecustomize_path,
...                 site_path,
...                 os.path.join(interpreter_bin_dir, 'py.exe'),
...                 py_path]
... else:
...     py_path = interpreter_path
...     expected = [sitecustomize_path, site_path, py_path]
...
>>> assert generated == expected, repr((generated, expected))

We didn't ask for any initialization, and we didn't ask to use the underlying sitecustomization, so sitecustomize.py is empty.

>>> cat(sitecustomize_path)

The interpreter script is simple. It puts the directory with the site.py and sitecustomize.py on the PYTHONPATH and (re)starts Python.

>>> cat(py_path)
#!/usr/bin/python -S
import os
import sys
<BLANKLINE>
argv = [sys.executable] + sys.argv[1:]
environ = os.environ.copy()
path = '/interpreter/parts/interpreter'
if environ.get('PYTHONPATH'):
    path = os.pathsep.join([path, environ['PYTHONPATH']])
environ['PYTHONPATH'] = path
os.execve(sys.executable, argv, environ)

The site.py file is a modified version of the underlying Python's site.py. The most important modification is that it has a different version of the addsitepackages function. It sets up the Python path, similarly to the behavior of the function it replaces. The following shows the part that buildout inserts, in the simplest case.

>>> sys.stdout.write('#\n'); cat(site_path)
... # doctest: +ELLIPSIS +NORMALIZE_WHITESPACE
#...
def addsitepackages(known_paths):
    """Add site packages, as determined by zc.buildout.
<BLANKLINE>
    See original_addsitepackages, below, for the original version."""
    buildout_paths = [
        '/interpreter/eggs/demo-0.3-pyN.N.egg',
        '/interpreter/eggs/demoneeded-1.1-pyN.N.egg'
        ]
    for path in buildout_paths:
        sitedir, sitedircase = makepath(path)
        if not sitedircase in known_paths and os.path.exists(sitedir):
            sys.path.append(sitedir)
            known_paths.add(sitedircase)
    return known_paths
<BLANKLINE>
def original_addsitepackages(known_paths):...

Here are some examples of the interpreter in use.

>>> print call_py(interpreter_path, "print 16+26")
42
<BLANKLINE>
>>> res = call_py(interpreter_path, "import sys; print sys.path")
>>> print res # doctest: +ELLIPSIS +NORMALIZE_WHITESPACE
['',
 '/interpreter/parts/interpreter',
 ...,
 '/interpreter/eggs/demo-0.3-pyN.N.egg',
 '/interpreter/eggs/demoneeded-1.1-pyN.N.egg']
<BLANKLINE>
>>> clean_paths = eval(res.strip()) # This is used later for comparison.

If you provide initialization, it goes in sitecustomize.py.

>>> def reset_interpreter():
...     # This is necessary because, in our tests, the timestamps of the
...     # .pyc files are not outdated when we want them to be.
...     rmdir(interpreter_bin_dir)
...     mkdir(interpreter_bin_dir)
...     rmdir(interpreter_parts_dir)
...     mkdir(interpreter_parts_dir)
...
>>> reset_interpreter()
>>> initialization_string = """\
... import os
... os.environ['FOO'] = 'bar baz bing shazam'"""
>>> generated = zc.buildout.easy_install.sitepackage_safe_scripts(
...     interpreter_bin_dir, ws, sys.executable, interpreter_parts_dir,
...     interpreter='py', initialization=initialization_string)
>>> cat(sitecustomize_path)
import os
os.environ['FOO'] = 'bar baz bing shazam'
>>> print call_py(interpreter_path, "import os; print os.environ['FOO']")
bar baz bing shazam
<BLANKLINE>

If you use relative paths, this affects the interpreter and site.py. (This is again the UNIX version; the Windows version uses subprocess instead of os.execve.)

>>> reset_interpreter()
>>> generated = zc.buildout.easy_install.sitepackage_safe_scripts(
...     interpreter_bin_dir, ws, sys.executable, interpreter_parts_dir,
...     interpreter='py', relative_paths=interpreter_dir)
>>> cat(py_path)
#!/usr/bin/python -S
import os
import sys
<BLANKLINE>
join = os.path.join
base = os.path.dirname(os.path.abspath(os.path.realpath(__file__)))
base = os.path.dirname(base)
<BLANKLINE>
argv = [sys.executable] + sys.argv[1:]
environ = os.environ.copy()
path = join(base, 'parts/interpreter')
if environ.get('PYTHONPATH'):
    path = os.pathsep.join([path, environ['PYTHONPATH']])
environ['PYTHONPATH'] = path
os.execve(sys.executable, argv, environ)

For site.py, we again show only the pertinent parts. Notice that the egg paths join a base to a path, as with the use of this argument in the scripts function.

>>> sys.stdout.write('#\n'); cat(site_path) # doctest: +ELLIPSIS
#...
def addsitepackages(known_paths):
    """Add site packages, as determined by zc.buildout.
<BLANKLINE>
    See original_addsitepackages, below, for the original version."""
    join = os.path.join
    base = os.path.dirname(os.path.abspath(os.path.realpath(__file__)))
    base = os.path.dirname(base)
    base = os.path.dirname(base)
    buildout_paths = [
        join(base, 'eggs/demo-0.3-pyN.N.egg'),
        join(base, 'eggs/demoneeded-1.1-pyN.N.egg')
        ]...

The paths resolve in practice as you would expect.

>>> print call_py(interpreter_path,
...               "import sys, pprint; pprint.pprint(sys.path)")
... # doctest: +ELLIPSIS
['',
 '/interpreter/parts/interpreter',
 ...,
 '/interpreter/eggs/demo-0.3-pyN.N.egg',
 '/interpreter/eggs/demoneeded-1.1-pyN.N.egg']
<BLANKLINE>

The extra_paths argument affects the path in site.py. Notice that /interpreter/other is added after the eggs.

>>> reset_interpreter()
>>> mkdir(interpreter_dir, 'other')
>>> generated = zc.buildout.easy_install.sitepackage_safe_scripts(
...     interpreter_bin_dir, ws, sys.executable, interpreter_parts_dir,
...     interpreter='py', extra_paths=[join(interpreter_dir, 'other')])
>>> sys.stdout.write('#\n'); cat(site_path) # doctest: +ELLIPSIS
#...
def addsitepackages(known_paths):
    """Add site packages, as determined by zc.buildout.
<BLANKLINE>
    See original_addsitepackages, below, for the original version."""
    buildout_paths = [
        '/interpreter/eggs/demo-0.3-pyN.N.egg',
        '/interpreter/eggs/demoneeded-1.1-pyN.N.egg',
        '/interpreter/other'
        ]...
>>> print call_py(interpreter_path,
...               "import sys, pprint; pprint.pprint(sys.path)")
... # doctest: +ELLIPSIS
['',
 '/interpreter/parts/interpreter',
 ...,
 '/interpreter/eggs/demo-0.3-pyN.N.egg',
 '/interpreter/eggs/demoneeded-1.1-pyN.N.egg',
 '/interpreter/other']
<BLANKLINE>

The sitepackage_safe_scripts function: using site-packages

The sitepackage_safe_scripts function supports including site packages. This has some advantages and some serious dangers.

A typical reason to include site-packages is that it is easier to install one or more dependencies in your Python than it is with buildout. Some packages, such as lxml or Python PostgreSQL integration, have dependencies that can be much easier to build and/or install using other mechanisms, such as your operating system's package manager. By installing some core packages into your Python's site-packages, this can significantly simplify some application installations.

However, doing this has a significant danger. One of the primary goals of buildout is to provide repeatability. Some packages (one of the better known Python openid packages, for instance) change their behavior depending on what packages are available. If Python curl bindings are available, these may be preferred by the library. If a certain XML package is installed, it may be preferred by the library. These hidden choices may cause small or large behavior differences. The fact that they can be rarely encountered can actually make it worse: you forget that this might be a problem, and debugging the differences can be difficult. If you allow site-packages to be included in your buildout, and the Python you use is not managed precisely by your application (for instance, it is a system Python), you open yourself up to these possibilities. Don't be unaware of the dangers.

That explained, let's see how it works. If you don't use namespace packages, this is very straightforward.

>>> reset_interpreter()
>>> generated = zc.buildout.easy_install.sitepackage_safe_scripts(
...     interpreter_bin_dir, ws, sys.executable, interpreter_parts_dir,
...     interpreter='py', include_site_packages=True)
>>> sys.stdout.write('#\n'); cat(site_path)
... # doctest: +ELLIPSIS +NORMALIZE_WHITESPACE
#...
def addsitepackages(known_paths):
    """Add site packages, as determined by zc.buildout.
<BLANKLINE>
    See original_addsitepackages, below, for the original version."""
    setuptools_path = None
    buildout_paths = [
        '/interpreter/eggs/demo-0.3-pyN.N.egg',
        '/interpreter/eggs/demoneeded-1.1-pyN.N.egg'
        ]
    for path in buildout_paths:
        sitedir, sitedircase = makepath(path)
        if not sitedircase in known_paths and os.path.exists(sitedir):
            sys.path.append(sitedir)
            known_paths.add(sitedircase)
    sys.__egginsert = len(buildout_paths) # Support distribute.
    original_paths = [
        ...
        ]
    for path in original_paths:
        if path == setuptools_path or path not in known_paths:
            addsitedir(path, known_paths)
    return known_paths
<BLANKLINE>
def original_addsitepackages(known_paths):...

It simply adds the original paths using addsitedir after the code to add the buildout paths.

Here's an example of the new script in use. Other documents and tests in this package give the feature a more thorough workout, but this should give you an idea of the feature.

>>> res = call_py(interpreter_path, "import sys; print sys.path")
>>> print res # doctest: +ELLIPSIS +NORMALIZE_WHITESPACE
['',
 '/interpreter/parts/interpreter',
 ...,
 '/interpreter/eggs/demo-0.3-py2.4.egg',
 '/interpreter/eggs/demoneeded-1.1-py2.4.egg',
 ...]
<BLANKLINE>

The clean_paths gathered earlier is a subset of this full list of paths.

>>> full_paths = eval(res.strip())
>>> len(clean_paths) < len(full_paths)
True
>>> set(os.path.normpath(p) for p in clean_paths).issubset(
...     os.path.normpath(p) for p in full_paths)
True

Unfortunately, because of how setuptools namespace packages are implemented differently for operating system packages (debs or rpms) as opposed to standard setuptools installation, there's a slightly trickier dance if you use them. To show this we'll needs some extra eggs that use namespaces. We'll use the tellmy.fortune package, which we'll need to make an initial call to another text fixture to create.

>>> from zc.buildout.tests import create_sample_namespace_eggs
>>> namespace_eggs = tmpdir('namespace_eggs')
>>> create_sample_namespace_eggs(namespace_eggs)
>>> reset_interpreter()
>>> ws = zc.buildout.easy_install.install(
...     ['demo', 'tellmy.fortune'], join(interpreter_dir, 'eggs'),
...     links=[link_server, namespace_eggs], index=link_server+'index/')
>>> generated = zc.buildout.easy_install.sitepackage_safe_scripts(
...     interpreter_bin_dir, ws, sys.executable, interpreter_parts_dir,
...     interpreter='py', include_site_packages=True)
>>> sys.stdout.write('#\n'); cat(site_path)
... # doctest: +ELLIPSIS +NORMALIZE_WHITESPACE
#...
def addsitepackages(known_paths):
    """Add site packages, as determined by zc.buildout.
<BLANKLINE>
    See original_addsitepackages, below, for the original version."""
    setuptools_path = '...setuptools...'
    sys.path.append(setuptools_path)
    known_paths.add(os.path.normcase(setuptools_path))
    import pkg_resources
    buildout_paths = [
        '/interpreter/eggs/demo-0.3-pyN.N.egg',
        '/interpreter/eggs/tellmy.fortune-1.0-pyN.N.egg',
        '...setuptools...',
        '/interpreter/eggs/demoneeded-1.1-pyN.N.egg'
        ]
    for path in buildout_paths:
        sitedir, sitedircase = makepath(path)
        if not sitedircase in known_paths and os.path.exists(sitedir):
            sys.path.append(sitedir)
            known_paths.add(sitedircase)
            pkg_resources.working_set.add_entry(sitedir)
    sys.__egginsert = len(buildout_paths) # Support distribute.
    original_paths = [
        ...
        ]
    for path in original_paths:
        if path == setuptools_path or path not in known_paths:
            addsitedir(path, known_paths)
    return known_paths
<BLANKLINE>
def original_addsitepackages(known_paths):...
>>> print call_py(interpreter_path, "import sys; print sys.path")
... # doctest: +ELLIPSIS +NORMALIZE_WHITESPACE
['',
 '/interpreter/parts/interpreter',
 ...,
 '...setuptools...',
 '/interpreter/eggs/demo-0.3-pyN.N.egg',
 '/interpreter/eggs/tellmy.fortune-1.0-pyN.N.egg',
 '/interpreter/eggs/demoneeded-1.1-pyN.N.egg',
 ...]

As you can see, the script now first imports pkg_resources. Then we need to process egg files specially to look for namespace packages there before we process process lines in .pth files that use the "import" feature--lines that might be part of the setuptools namespace package implementation for system packages, as mentioned above, and that must come after processing egg namespaces.

The most complex that this function gets is if you use namespace packages, include site-packages, and use relative paths. For completeness, we'll look at that result.

>>> reset_interpreter()
>>> generated = zc.buildout.easy_install.sitepackage_safe_scripts(
...     interpreter_bin_dir, ws, sys.executable, interpreter_parts_dir,
...     interpreter='py', include_site_packages=True,
...     relative_paths=interpreter_dir)
>>> sys.stdout.write('#\n'); cat(site_path)
... # doctest: +ELLIPSIS +NORMALIZE_WHITESPACE
#...
def addsitepackages(known_paths):
    """Add site packages, as determined by zc.buildout.
<BLANKLINE>
    See original_addsitepackages, below, for the original version."""
    join = os.path.join
    base = os.path.dirname(os.path.abspath(os.path.realpath(__file__)))
    base = os.path.dirname(base)
    base = os.path.dirname(base)
    setuptools_path = '...setuptools...'
    sys.path.append(setuptools_path)
    known_paths.add(os.path.normcase(setuptools_path))
    import pkg_resources
    buildout_paths = [
        join(base, 'eggs/demo-0.3-pyN.N.egg'),
        join(base, 'eggs/tellmy.fortune-1.0-pyN.N.egg'),
        '...setuptools...',
        join(base, 'eggs/demoneeded-1.1-pyN.N.egg')
        ]
    for path in buildout_paths:
        sitedir, sitedircase = makepath(path)
        if not sitedircase in known_paths and os.path.exists(sitedir):
            sys.path.append(sitedir)
            known_paths.add(sitedircase)
            pkg_resources.working_set.add_entry(sitedir)
    sys.__egginsert = len(buildout_paths) # Support distribute.
    original_paths = [
        ...
        ]
    for path in original_paths:
        if path == setuptools_path or path not in known_paths:
            addsitedir(path, known_paths)
    return known_paths
<BLANKLINE>
def original_addsitepackages(known_paths):...
>>> print call_py(interpreter_path, "import sys; print sys.path")
... # doctest: +ELLIPSIS +NORMALIZE_WHITESPACE
['',
 '/interpreter/parts/interpreter',
 ...,
 '...setuptools...',
 '/interpreter/eggs/demo-0.3-pyN.N.egg',
 '/interpreter/eggs/tellmy.fortune-1.0-pyN.N.egg',
 '/interpreter/eggs/demoneeded-1.1-pyN.N.egg',
 ...]

The exec_sitecustomize argument does the same thing for the sitecustomize module--it allows you to include the code from the sitecustomize module in the underlying Python if you set the argument to True. The z3c.recipe.scripts package sets up the full environment necessary to demonstrate this piece.

The sitepackage_safe_scripts function: writing scripts for entry points

All of the examples so far for this function have been creating interpreters. The function can also write scripts for entry points. They are almost identical to the scripts that we saw for the scripts function except that they import site after setting the sys.path to include our custom site.py and sitecustomize.py files. These files then initialize the Python environment as we have already seen. Let's see a simple example.

>>> reset_interpreter()
>>> ws = zc.buildout.easy_install.install(
...     ['demo'], join(interpreter_dir, 'eggs'), links=[link_server],
...     index=link_server+'index/')
>>> generated = zc.buildout.easy_install.sitepackage_safe_scripts(
...     interpreter_bin_dir, ws, sys.executable, interpreter_parts_dir,
...     reqs=['demo'])

As before, in Windows, 2 files are generated for each script. A script file, ending in '-script.py', and an exe file that allows the script to be invoked directly without having to specify the Python interpreter and without having to provide a '.py' suffix. This is in addition to the site.py and sitecustomize.py files that are generated as with our interpreter examples above.

>>> if sys.platform == 'win32':
...     demo_path = os.path.join(interpreter_bin_dir, 'demo-script.py')
...     expected = [sitecustomize_path,
...                 site_path,
...                 os.path.join(interpreter_bin_dir, 'demo.exe'),
...                 demo_path]
... else:
...     demo_path = os.path.join(interpreter_bin_dir, 'demo')
...     expected = [sitecustomize_path, site_path, demo_path]
...
>>> assert generated == expected, repr((generated, expected))

The demo script runs the entry point defined in the demo egg:

>>> cat(demo_path) # doctest: +NORMALIZE_WHITESPACE
#!/usr/local/bin/python2.4 -S
<BLANKLINE>
import sys
sys.path[0:0] = [
    '/interpreter/parts/interpreter',
    ]
<BLANKLINE>
<BLANKLINE>
import os
path = sys.path[0]
if os.environ.get('PYTHONPATH'):
    path = os.pathsep.join([path, os.environ['PYTHONPATH']])
os.environ['BUILDOUT_ORIGINAL_PYTHONPATH'] = os.environ.get('PYTHONPATH', '')
os.environ['PYTHONPATH'] = path
import site # imports custom buildout-generated site.py
<BLANKLINE>
import eggrecipedemo
<BLANKLINE>
if __name__ == '__main__':
    sys.exit(eggrecipedemo.main())
>>> demo_call = join(interpreter_bin_dir, 'demo')
>>> if sys.platform == 'win32':
...     demo_call = '"%s"' % demo_call
>>> print system(demo_call)
3 1
<BLANKLINE>

There are a few differences from the scripts function. First, the reqs argument (an iterable of string requirements or entry point tuples) is a keyword argument here. We see that in the example above. Second, the arguments argument is now named script_arguments to try and clarify that it does not affect interpreters. While the initialization argument continues to affect both the interpreters and the entry point scripts, if you have initialization that is only pertinent to the entry point scripts, you can use the script_initialization argument.

Let's see script_arguments and script_initialization in action.

>>> reset_interpreter()
>>> generated = zc.buildout.easy_install.sitepackage_safe_scripts(
...     interpreter_bin_dir, ws, sys.executable, interpreter_parts_dir,
...     reqs=['demo'], script_arguments='1, 2',
...     script_initialization='import os\nos.chdir("foo")')
>>> cat(demo_path) # doctest: +NORMALIZE_WHITESPACE
#!/usr/local/bin/python2.4 -S
import sys
sys.path[0:0] = [
  '/interpreter/parts/interpreter',
  ]
<BLANKLINE>
import os
path = sys.path[0]
if os.environ.get('PYTHONPATH'):
    path = os.pathsep.join([path, os.environ['PYTHONPATH']])
os.environ['BUILDOUT_ORIGINAL_PYTHONPATH'] = os.environ.get('PYTHONPATH', '')
os.environ['PYTHONPATH'] = path
import site # imports custom buildout-generated site.py
import os
os.chdir("foo")
<BLANKLINE>
import eggrecipedemo
<BLANKLINE>
if __name__ == '__main__':
    sys.exit(eggrecipedemo.main(1, 2))

Handling custom build options for extensions provided in source distributions

Sometimes, we need to control how extension modules are built. The build function provides this level of control. It takes a single package specification, downloads a source distribution, and builds it with specified custom build options.

The build function takes 3 positional arguments:

spec
A package specification for a source distribution
dest
A destination directory
build_ext
A dictionary of options to be passed to the distutils build_ext command when building extensions.

It supports a number of optional keyword arguments:

links
a sequence of URLs, file names, or directories to look for links to distributions,
index

The URL of an index server, or almost any other valid URL. :)

If not specified, the Python Package Index, http://pypi.python.org/simple/, is used. You can specify an alternate index with this option. If you use the links option and if the links point to the needed distributions, then the index can be anything and will be largely ignored. In the examples, here, we'll just point to an empty directory on our link server. This will make our examples run a little bit faster.

executable
A path to a Python executable. Distributions will be installed using this executable and will be for the matching Python version.
path
A list of additional directories to search for locally-installed distributions.
newest
A boolean value indicating whether to search for new distributions when already-installed distributions meet the requirement. When this is true, the default, and when the destination directory is not None, then the install function will search for the newest distributions that satisfy the requirements.
versions
A dictionary mapping project names to version numbers to be used when selecting distributions. This can be used to specify a set of distribution versions independent of other requirements.

Our link server included a source distribution that includes a simple extension, extdemo.c:

#include <Python.h>
#include <extdemo.h>

static PyMethodDef methods[] = {};

PyMODINIT_FUNC
initextdemo(void)
{
    PyObject *m;
    m = Py_InitModule3("extdemo", methods, "");
#ifdef TWO
    PyModule_AddObject(m, "val", PyInt_FromLong(2));
#else
    PyModule_AddObject(m, "val", PyInt_FromLong(EXTDEMO));
#endif
}

The extension depends on a system-dependent include file, extdemo.h, that defines a constant, EXTDEMO, that is exposed by the extension.

We'll add an include directory to our sample buildout and add the needed include file to it:

>>> mkdir('include')
>>> write('include', 'extdemo.h',
... """
... #define EXTDEMO 42
... """)

Now, we can use the build function to create an egg from the source distribution:

>>> zc.buildout.easy_install.build(
...   'extdemo', dest,
...   {'include-dirs': os.path.join(sample_buildout, 'include')},
...   links=[link_server], index=link_server+'index/')
['/sample-install/extdemo-1.4-py2.4-unix-i686.egg']

The function returns the list of eggs

Now if we look in our destination directory, we see we have an extdemo egg:

>>> ls(dest)
-  demo-0.2-py2.4.egg
d  demo-0.3-py2.4.egg
-  demoneeded-1.0-py2.4.egg
d  demoneeded-1.1-py2.4.egg
d  extdemo-1.4-py2.4-unix-i686.egg

Let's update our link server with a new version of extdemo:

>>> update_extdemo()
>>> print get(link_server),
<html><body>
<a href="bigdemo-0.1-py2.4.egg">bigdemo-0.1-py2.4.egg</a><br>
<a href="demo-0.1-py2.4.egg">demo-0.1-py2.4.egg</a><br>
<a href="demo-0.2-py2.4.egg">demo-0.2-py2.4.egg</a><br>
<a href="demo-0.3-py2.4.egg">demo-0.3-py2.4.egg</a><br>
<a href="demo-0.4c1-py2.4.egg">demo-0.4c1-py2.4.egg</a><br>
<a href="demoneeded-1.0.zip">demoneeded-1.0.zip</a><br>
<a href="demoneeded-1.1.zip">demoneeded-1.1.zip</a><br>
<a href="demoneeded-1.2c1.zip">demoneeded-1.2c1.zip</a><br>
<a href="extdemo-1.4.zip">extdemo-1.4.zip</a><br>
<a href="extdemo-1.5.zip">extdemo-1.5.zip</a><br>
<a href="index/">index/</a><br>
<a href="other-1.0-py2.4.egg">other-1.0-py2.4.egg</a><br>
</body></html>

The easy_install caches information about servers to reduce network access. To see the update, we have to call the clear_index_cache function to clear the index cache:

>>> zc.buildout.easy_install.clear_index_cache()

If we run build with newest set to False, we won't get an update:

>>> zc.buildout.easy_install.build(
...   'extdemo', dest,
...   {'include-dirs': os.path.join(sample_buildout, 'include')},
...   links=[link_server], index=link_server+'index/',
...   newest=False)
['/sample-install/extdemo-1.4-py2.4-linux-i686.egg']
>>> ls(dest)
-  demo-0.2-py2.4.egg
d  demo-0.3-py2.4.egg
-  demoneeded-1.0-py2.4.egg
d  demoneeded-1.1-py2.4.egg
d  extdemo-1.4-py2.4-unix-i686.egg

But if we run it with the default True setting for newest, then we'll get an updated egg:

>>> zc.buildout.easy_install.build(
...   'extdemo', dest,
...   {'include-dirs': os.path.join(sample_buildout, 'include')},
...   links=[link_server], index=link_server+'index/')
['/sample-install/extdemo-1.5-py2.4-unix-i686.egg']
>>> ls(dest)
-  demo-0.2-py2.4.egg
d  demo-0.3-py2.4.egg
-  demoneeded-1.0-py2.4.egg
d  demoneeded-1.1-py2.4.egg
d  extdemo-1.4-py2.4-unix-i686.egg
d  extdemo-1.5-py2.4-unix-i686.egg

The versions option also influences the versions used. For example, if we specify a version for extdemo, then that will be used, even though it isn't the newest. Let's clean out the destination directory first:

>>> import os
>>> for name in os.listdir(dest):
...     remove(dest, name)
>>> zc.buildout.easy_install.build(
...   'extdemo', dest,
...   {'include-dirs': os.path.join(sample_buildout, 'include')},
...   links=[link_server], index=link_server+'index/',
...   versions=dict(extdemo='1.4'))
['/sample-install/extdemo-1.4-py2.4-unix-i686.egg']
>>> ls(dest)
d  extdemo-1.4-py2.4-unix-i686.egg

Handling custom build options for extensions in develop eggs

The develop function is similar to the build function, except that, rather than building an egg from a source directory containing a setup.py script.

The develop function takes 2 positional arguments:

setup
The path to a setup script, typically named "setup.py", or a directory containing a setup.py script.
dest
The directory to install the egg link to

It supports some optional keyword argument:

build_ext
A dictionary of options to be passed to the distutils build_ext command when building extensions.
executable
A path to a Python executable. Distributions will be installed using this executable and will be for the matching Python version.

We have a local directory containing the extdemo source:

>>> ls(extdemo)
-  MANIFEST
-  MANIFEST.in
-  README
-  extdemo.c
-  setup.py

Now, we can use the develop function to create a develop egg from the source distribution:

>>> zc.buildout.easy_install.develop(
...   extdemo, dest,
...   {'include-dirs': os.path.join(sample_buildout, 'include')})
'/sample-install/extdemo.egg-link'

The name of the egg link created is returned.

Now if we look in our destination directory, we see we have an extdemo egg link:

>>> ls(dest)
d  extdemo-1.4-py2.4-unix-i686.egg
-  extdemo.egg-link

And that the source directory contains the compiled extension:

>>> ls(extdemo)
-  MANIFEST
-  MANIFEST.in
-  README
d  build
-  extdemo.c
d  extdemo.egg-info
-  extdemo.so
-  setup.py

Download cache

Normally, when distributions are installed, if any processing is needed, they are downloaded from the internet to a temporary directory and then installed from there. A download cache can be used to avoid the download step. This can be useful to reduce network access and to create source distributions of an entire buildout.

A download cache is specified by calling the download_cache function. The function always returns the previous setting. If no argument is passed, then the setting is unchanged. If an argument is passed, the download cache is set to the given path, which must point to an existing directory. Passing None clears the cache setting.

To see this work, we'll create a directory and set it as the cache directory:

>>> cache = tmpdir('cache')
>>> zc.buildout.easy_install.download_cache(cache)

We'll recreate our destination directory:

>>> remove(dest)
>>> dest = tmpdir('sample-install')

We'd like to see what is being fetched from the server, so we'll enable server logging:

>>> get(link_server+'enable_server_logging')
GET 200 /enable_server_logging
''

Now, if we install demo, and extdemo:

>>> ws = zc.buildout.easy_install.install(
...     ['demo==0.2'], dest,
...     links=[link_server], index=link_server+'index/',
...     always_unzip=True)
GET 200 /
GET 404 /index/demo/
GET 200 /index/
GET 200 /demo-0.2-py2.4.egg
GET 404 /index/demoneeded/
GET 200 /demoneeded-1.1.zip
>>> zc.buildout.easy_install.build(
...   'extdemo', dest,
...   {'include-dirs': os.path.join(sample_buildout, 'include')},
...   links=[link_server], index=link_server+'index/')
GET 404 /index/extdemo/
GET 200 /extdemo-1.5.zip
['/sample-install/extdemo-1.5-py2.4-linux-i686.egg']

Not only will we get eggs in our destination directory:

>>> ls(dest)
d  demo-0.2-py2.4.egg
d  demoneeded-1.1-py2.4.egg
d  extdemo-1.5-py2.4-linux-i686.egg

But we'll get distributions in the cache directory:

>>> ls(cache)
-  demo-0.2-py2.4.egg
-  demoneeded-1.1.zip
-  extdemo-1.5.zip

The cache directory contains uninstalled distributions, such as zipped eggs or source distributions.

Let's recreate our destination directory and clear the index cache:

>>> remove(dest)
>>> dest = tmpdir('sample-install')
>>> zc.buildout.easy_install.clear_index_cache()

Now when we install the distributions:

>>> ws = zc.buildout.easy_install.install(
...     ['demo==0.2'], dest,
...     links=[link_server], index=link_server+'index/',
...     always_unzip=True)
GET 200 /
GET 404 /index/demo/
GET 200 /index/
GET 404 /index/demoneeded/
>>> zc.buildout.easy_install.build(
...   'extdemo', dest,
...   {'include-dirs': os.path.join(sample_buildout, 'include')},
...   links=[link_server], index=link_server+'index/')
GET 404 /index/extdemo/
['/sample-install/extdemo-1.5-py2.4-linux-i686.egg']
>>> ls(dest)
d  demo-0.2-py2.4.egg
d  demoneeded-1.1-py2.4.egg
d  extdemo-1.5-py2.4-linux-i686.egg

Note that we didn't download the distributions from the link server.

If we remove the restriction on demo, we'll download a newer version from the link server:

>>> ws = zc.buildout.easy_install.install(
...     ['demo'], dest,
...     links=[link_server], index=link_server+'index/',
...     always_unzip=True)
GET 200 /demo-0.3-py2.4.egg

Normally, the download cache is the preferred source of downloads, but not the only one.

Installing solely from a download cache

A download cache can be used as the basis of application source releases. In an application source release, we want to distribute an application that can be built without making any network accesses. In this case, we distribute a download cache and tell the easy_install module to install from the download cache only, without making network accesses. The install_from_cache function can be used to signal that packages should be installed only from the download cache. The function always returns the previous setting. Calling it with no arguments returns the current setting without changing it:

>>> zc.buildout.easy_install.install_from_cache()
False

Calling it with a boolean value changes the setting and returns the previous setting:

>>> zc.buildout.easy_install.install_from_cache(True)
False

Let's remove demo-0.3-py2.4.egg from the cache, clear the index cache, recreate the destination directory, and reinstall demo:

>>> for  f in os.listdir(cache):
...     if f.startswith('demo-0.3-'):
...         remove(cache, f)
>>> zc.buildout.easy_install.clear_index_cache()
>>> remove(dest)
>>> dest = tmpdir('sample-install')
>>> ws = zc.buildout.easy_install.install(
...     ['demo'], dest,
...     links=[link_server], index=link_server+'index/',
...     always_unzip=True)
>>> ls(dest)
d  demo-0.2-py2.4.egg
d  demoneeded-1.1-py2.4.egg

This time, we didn't download from or even query the link server.

Distribute Support

Distribute is a drop-in replacement for Setuptools.

zc.buildout is now compatible with Distribute 0.6. To use Distribute in your buildout, you need use the --distribute option of the bootstrap.py script:

$ python bootstrap.py --distribute

This will download and install the latest Distribute 0.6 release in the eggs directory, and use this version for the scripts that are created in bin.

Notice that if you have a shared eggs directory, a buildout that uses Distribute will not interfer with other buildouts that are based on Setuptools and that are sharing the same eggs directory.

Form more information about the Distribute project, see: http://python-distribute.org

Change History

1.7.0 (2013-01-11)

  • Unless version requirements are specified, buildout won't upgrade itself past version 1.

  • Versions in versions sections can now be simple constraints, like <2.0dev in addition to being simple versions.

    This is used to prevent upgrading zc.recipe.egg and zc.recipe.testrunner past version 1.

  • If buildout is bootstrapped with a non-final release, it won't downgrade itself to a final release.

  • Fix: distribute 0.6.33 broke Python 2.4 compatibility

  • remove data_files from setup.py, which was installing README.txt in current directory during installation [Domen Kožar]

  • Windows fix: use cli-64.exe/cli.exe depending on 64/32 bit and try cli.exe if cli-64.exe is not found, fixing 9c6be7ac6d218f09e33725e07dccc4af74d8cf97

  • Windows fix: buildout init was broken, re.sub does not like a

    single backslash

  • fixed all builds on travis-ci [Domen Kožar]

  • use os._exit insted of sys.exit after ugrade forking [Domen Kožar]

  • Revert cfa0478937d16769c268bf51e60e69cd3ead50f3, it only broke a feature [Domen Kožar]

1.6.1 (2012-08-18)

  • bootstrap.py -d init would invoke buildout with arguments init bootstrap leading into installation of bootstrap package. now bootstrap.py first runs any commands passed, then tries to bootstrap. (Domen Kožar)
  • fix Python 2.4 support (Domen Kožar)
  • added travis-ci testing (Domen Kožar)

1.6.0 (2012-08-15)

  • The buildout init command now accepts distribution requirements and paths to set up a custom interpreter part that has the distributions or parts in the path. For example:

    python bootstrap.py init BeautifulSoup
    
  • Introduce a cache for the expensive buildout._dir_hash function.

  • Remove duplicate path from script's sys.path setup.

  • changed broken dash S check to pass the configuration options -S -c separately, to make zc.buildout more compatible with the PyPy interpreter, which has less flexible argument parsing than CPython.

    Note that PyPy post 1.4.0 is needed to make buildout work at all, due to missing support for the -E option, which only got added afterwards.

  • Made sure to download extended configuration files only once per buildout run even if they are referenced multiple times (patch by Rafael Monnerat).

  • Ported speedup optimization patch by Ross Patterson to 1.5.x series. Improved patch to calculate required_by packages in linear time in verbose mode (-v). Running relatively simple Buildout envornment yielded in running time improvement from 30 seconds to 10 seconds. (Domen Kožar, Ross Patterson)

  • Removed unnecessary pyc recompilation with optimization flags. Running Buildout with pre-downloaded ~300 packages that were installed in empty eggs repository yielded in running time improvement from 1126 seconds to 348 seconds. (Domen Kožar)

Bugs fixed:

  • In the download module, fixed the handling of directories that are pointed to by file-system paths and file: URLs.
  • Removed any traces of the implementation of extended-by. Raise a UserError if the option is encountered instead of ignoring it, though.
  • https://bugs.launchpad.net/bugs/697913 : Buildout doesn't honor exit code from scripts. Fixed.
  • Handle both addition and subtraction of elements (+= and -=) on the same key in the same section.

1.5.2 (2010-10-11)

  • changed metadata 'url' to pypi.python.org in order to solve a temporary outage of buildout.org

  • IMPORTANT: For better backwards compatibility with the pre-1.5 line, this release has two big changes from 1.5.0 and 1.5.1.

    • Buildout defaults to including site packages.
    • Buildout loads recipes and extensions with the same constraints to site-packages that it builds eggs, instead of never allowing access to site-packages.

    This means that the default configuration should better support pre-existing use of system Python in recipes or builds.

  • To make it easier to detect the fact that buildout has set the PYTHONPATH, BUILDOUT_ORIGINAL_PYTHONPATH is always set in the environment, even if PYTHONPATH was not originally set. BUILDOUT_ORIGINAL_PYTHONPATH will be an empty string if PYTHONPATH was not set.

1.5.1 (2010-08-29)

New features:

  • Scripts store the old PYTHONPATH in BUILDOUT_ORIGINAL_PYTHONPATH if it existed, and store nothing in the value if it did not exist. This allows code that does not want subprocesses to have the system-Python-protected site.py to set the environment of the subprocess as it was originally.

Bugs fixed:

1.5.0 (2010-08-23)

New features:

  • zc.buildout supports Python 2.7.

  • By default, Buildout and the bootstrap script now prefer final versions of Buildout, recipes, and extensions. This can be changed by using the --accept-buildout-test-releases flag (or -t for short) when calling bootstrap. This will hopefully allow beta releases of these items to be more easily and safely made in the future.

    NOTE: dependencies of your own software are not affected by this new behavior. Buildout continues to choose the newest available versions of your dependencies regardless of whether they are final releases. To prevent this, use the pre-existing switch prefer-final = true in the [buildout] section of your configuration file (see http://pypi.python.org/pypi/zc.buildout#preferring-final-releases) or pin your versions using a versions section (see http://pypi.python.org/pypi/zc.buildout#repeatable-buildouts-controlling-eggs-used).

Bugs fixed:

  • You can now again use virtualenv with Buildout. The new features to let buildout be used with a system Python are disabled in this configuration, and the previous script generation behavior (1.4.3) is used, even if the new function zc.buildout.easy_install.sitepackage_safe_scripts is used.

1.5.0b2 (2010-04-29)

This was a re-release of 1.4.3 in order to keep 1.5.0b1 release from hurting workflows that combined virtualenv with zc.buildout.

1.5.0b1 (2010-04-29)

New Features:

  • Added buildout:socket-timout option so that socket timeout can be configured both from command line and from config files. (gotcha)

  • Buildout can be safely used with a system Python (or any Python with code in site-packages), as long as you use (1) A fresh checkout, (2) the new bootstrap.py, and (3) recipes that use the new zc.buildout.easy_install.sitepackage_safe_scripts function to generate scripts and interpreters. Many recipes will need to be updated to use this new function. The scripts and interpreters generated by zc.recipe.egg will continue to use the older function, not safe with system Pythons. Use the z3c.recipe.scripts as a replacement.

    zc.recipe.egg is still a fully supported, and simpler, way of generating scripts and interpreters if you are using a "clean" Python, without code installed in site-packages. It keeps its previous behavior in order to provide backwards compatibility.

    The z3c.recipe.scripts recipe allows you to control how you use the code in site-packages. You can exclude it entirely (preferred); allow eggs in it to fulfill package dependencies declared in setup.py and buildout configuration; allow it to be available but not used to fulfill dependencies declared in setup.py or buildout configuration; or only allow certain eggs in site-packages to fulfill dependencies.

  • Added new function, zc.buildout.easy_install.sitepackage_safe_scripts, to generate scripts and interpreter. It produces a full-featured interpreter (all command-line options supported) and the ability to safely let scripts include site packages, such as with a system Python. The z3c.recipe.scripts recipe uses this new function.

  • Improve bootstrap.

    • New options let you specify where to find ez_setup.py and where to find a download cache. These options can keep bootstrap from going over the network.
    • Another new option lets you specify where to put generated eggs.
    • The buildout script generated by bootstrap honors more of the settings in the designated configuration file (e.g., buildout.cfg).
    • Correctly handle systems where pkg_resources is present but the rest of setuptools is missing (like Ubuntu installs). https://bugs.launchpad.net/zc.buildout/+bug/410528
  • You can develop zc.buildout using Distribute instead of Setuptools. Use the --distribute option on the dev.py script. (Releases should be tested with both Distribute and Setuptools.) The tests for zc.buildout pass with Setuptools and Python 2.4, 2.5, 2.6, and 2.7; and with Distribute and Python 2.5, 2.6, and 2.7. Using zc.buildout with Distribute and Python 2.4 is not recommended.

  • The distribute-version now works in the [buildout] section, mirroring the setuptools-version option (this is for consistency; using the general-purpose versions option is preferred).

Bugs fixed:

  • Using Distribute with the allow-picked-versions = false buildout option no longer causes an error.
  • The handling and documenting of default buildout options was normalized. This means, among other things, that bin/buildout -vv and bin/buildout annotate correctly list more of the options.
  • Installing a namespace package using a Python that already has a package in the same namespace (e.g., in the Python's site-packages) failed in some cases. It is now handled correctly.
  • Another variation of this error showed itself when at least two dependencies were in a shared location like site-packages, and the first one met the "versions" setting. The first dependency would be added, but subsequent dependencies from the same location (e.g., site-packages) would use the version of the package found in the shared location, ignoring the version setting. This is also now handled correctly.

1.4.3 (2009-12-10)

Bugs fixed:

  • Using pre-detected setuptools version for easy_installing tgz files. This prevents a recursion error when easy_installing an upgraded "distribute" tgz. Note that setuptools did not have this recursion problem solely because it was packaged as an .egg, which does not have to go through the easy_install step.

1.4.2 (2009-11-01)

New Feature:

  • Added a --distribute option to the bootstrap script, in order to use Distribute rather than Setuptools. By default, Setuptools is used.

Bugs fixed:

  • While checking for new versions of setuptools and buildout itself, compare requirement locations instead of requirement objects.
  • Incrementing didn't work properly when extending multiple files. https://bugs.launchpad.net/zc.buildout/+bug/421022
  • The download API computed MD5 checksums of text files wrong on Windows.

1.4.1 (2009-08-27)

New Feature:

  • Added a debug built-in recipe to make writing some tests easier.

Bugs fixed:

  • (introduced in 1.4.0) option incrementing (-=) and decrementing (-=) didn't work in the buildout section. https://bugs.launchpad.net/zc.buildout/+bug/420463
  • Option incrementing and decrementing didn't work for options specified on the command line.
  • Scripts generated with relative-paths enabled couldn't be symbolically linked to other locations and still work.
  • Scripts run using generated interpreters didn't have __file__ set correctly.
  • The standard Python -m option didn't work for custom interpreters.

1.4.0 (2009-08-26)

  • When doing variable substitutions, you can omit the section name to refer to a variable in the same section (e.g. ${:foo}).
  • When doing variable substitution, you can use the special option, _buildout_section_name_ to get the section name. This is most handy for getting the current section name (e.g. ${:_buildout_section_name_}).
  • A new special option, < allows sections to be used as macros.
  • Added annotate command for annotated sections. Displays sections key-value pairs along with the value origin.
  • Added a download API that handles the download cache, offline mode etc and is meant to be reused by recipes.
  • Used the download API to allow caching of base configurations (specified by the buildout section's 'extends' option).

1.3.1 (2009-08-12)

  • Bug fixed: extras were ignored in some cases when versions were specified.

1.3.0 (2009-06-22)

  • Better Windows compatibility in test infrastructure.
  • Now the bootstrap.py has an optional --version argument, that can be used to force zc.buildout version to use.
  • zc.buildout.testing.buildoutSetUp installs a new handler in the python root logging facility. This handler is now removed during tear down as it might disturb other packages reusing buildout's testing infrastructure.
  • fixed usage of 'relative_paths' keyword parameter on Windows
  • Added an unload entry point for extensions.
  • Fixed bug: when the relative paths option was used, relative paths could be inserted into sys.path if a relative path was used to run the generated script.

1.2.1 (2009-03-18)

  • Refactored generation of relative egg paths to generate simpler code.

1.2.0 (2009-03-17)

  • Added a relative_paths option to zc.buildout.easy_install.script to generate egg paths relative to the script they're used in.

1.1.2 (2009-03-16)

  • Added Python 2.6 support. Removed Python 2.3 support.

  • Fixed remaining deprecation warnings under Python 2.6, both when running our tests and when using the package.

  • Switched from using os.popen* to subprocess.Popen, to avoid a deprecation warning in Python 2.6. See:

    http://docs.python.org/library/subprocess.html#replacing-os-popen-os-popen2-os-popen3

  • Made sure the 'redo_pyc' function and the doctest checkers work with Python executable paths containing spaces.

  • Expand shell patterns when processing the list of paths in develop, e.g:

    [buildout]
    develop = ./local-checkouts/*
    
  • Conditionally import and use hashlib.md5 when it's available instead of md5 module, which is deprecated in Python 2.6.

  • Added Jython support for bootstrap, development bootstrap and zc.buildout support on Jython

  • Fixed a bug that would cause buildout to break while computing a directory hash if it found a broken symlink (Launchpad #250573)

1.1.1 (2008-07-28)

  • Fixed a bug that caused buildouts to fail when variable substitutions are used to name standard directories, as in:

    [buildout]
    eggs-directory = ${buildout:directory}/develop-eggs
    

1.1.0 (2008-07-19)

  • Added a buildout-level unzip option tp change the default policy for unzipping zip-safe eggs.
  • Tracebacks are now printed for internal errors (as opposed to user errors) even without the -D option.
  • pyc and pyo files are regenerated for installed eggs so that the stored path in code objects matches the the install location.

1.0.6 (2008-06-13)

1.0.5 (2008-06-10)

  • Fixed wrong split when using the += and -= syntax (mustapha)

1.0.4 (2008-06-10)

  • Added the allow-hosts option (tarek)
  • Quote the 'executable' argument when trying to detect the python version using popen4. (sidnei)
  • Quote the 'spec' argument, as in the case of installing an egg from the buildout-cache, if the filename contains spaces it would fail (sidnei)
  • Extended configuration syntax to allow -= and += operators (malthe, mustapha).

1.0.3 (2008-06-01)

  • fix for """AttributeError: Buildout instance has no attribute '_logger'""" by providing reasonable defaults within the Buildout constructor. (patch by Gottfried Ganssauge) (ajung)

1.0.2 (2008-05-13)

  • More fixes for Windows. A quoted sh-bang is now used on Windows to make the .exe files work with a Python executable in 'program files'.
  • Added "-t <timeout_in_seconds>" option for specifying the socket timeout. (ajung)

1.0.1 (2008-04-02)

  • Made easy_install.py's _get_version accept non-final releases of Python, like 2.4.4c0. (hannosch)
  • Applied various patches for Windows (patch by Gottfried Ganssauge). (ajung)
  • Applied patch fixing rmtree issues on Windows (patch by Gottfried Ganssauge). (ajung)

1.0.0 (2008-01-13)

  • Added a French translation of the buildout tutorial.

1.0.0b31 (2007-11-01)

Feature Changes

  • Added a configuration option that allows buildouts to ignore dependency_links metadata specified in setup. By default dependency_links in setup are used in addition to buildout specified find-links. This can make it hard to control where eggs come from. Here's how to tell buildout to ignore URLs in dependency_links:

    [buildout]
    use-dependency-links = false
    

    By default use-dependency-links is true, which matches the behavior of previous versions of buildout.

  • Added a configuration option that causes buildout to error if a version is picked. This is a nice safety belt when fixing all versions is intended, especially when creating releases.

Bugs Fixed

  • 151820: Develop failed if the setup.py script imported modules in the distribution directory.
  • Verbose logging of the develop command was omitting detailed output.
  • The setup command wasn't documented.
  • The setup command failed if run in a directory without specifying a configuration file.
  • The setup command raised a stupid exception if run without arguments.
  • When using a local find links or index, distributions weren't copied to the download cache.
  • When installing from source releases, a version specification (via a buildout versions section) for setuptools was ignored when deciding which setuptools to use to build an egg from the source release.

1.0.0b30 (2007-08-20)

Feature Changes

  • Changed the default policy back to what it was to avoid breakage in existing buildouts. Use:

    [buildout]
    prefer-final = true
    

    to get the new policy. The new policy will go into effect in buildout 2.

1.0.0b29 (2007-08-20)

Feature Changes

  • Now, final distributions are prefered over non-final versions. If both final and non-final versions satisfy a requirement, then the final version will be used even if it is older. The normal way to override this for specific packages is to specifically require a non-final version, either specifically or via a lower bound.

  • There is a buildout prefer-final version that can be used with a value of "false":

    prefer-final = false
    

    To prefer newer versions, regardless of whether or not they are final, buildout-wide.

  • The new simple Python index, http://cheeseshop.python.org/simple, is used as the default index. This will provide better performance than the human package index interface, http://pypi.python.org/pypi. More importantly, it lists hidden distributions, so buildouts with fixed distribution versions will be able to find old distributions even if the distributions have been hidden in the human PyPI interface.

Bugs Fixed

  • 126441: Look for default.cfg in the right place on Windows.

1.0.0b28 (2007-07-05)

Bugs Fixed

  • When requiring a specific version, buildout looked for new versions even if that single version was already installed.

1.0.0b27 (2007-06-20)

Bugs Fixed

  • Scripts were generated incorrectly on Windows. This included the buildout script itself, making buildout completely unusable.

1.0.0b26 (2007-06-19)

Feature Changes

  • Thanks to recent fixes in setuptools, I was able to change buildout to use find-link and index information when searching extensions.

    Sadly, this work, especially the timing, was motivated my the need to use alternate indexes due to performance problems in the cheese shop (http://www.python.org/pypi/). I really home we can address these performance problems soon.

1.0.0b25 (2007-05-31)

Feature Changes

  • buildout now changes to the buildout directory before running recipe install and update methods.
  • Added a new init command for creating a new buildout. This creates an empty configuration file and then bootstraps.
  • Except when using the new init command, it is now an error to run buildout without a configuration file.
  • In verbose mode, when adding distributions to fulful requirements of already-added distributions, we now show why the new distributions are being added.
  • Changed the logging format to exclude the logger name for the zc.buildout logger. This reduces noise in the output.
  • Clean up lots of messages, adding missing periods and adding quotes around requirement strings and file paths.

Bugs Fixed

  • 114614: Buildouts could take a very long time if there were dependency problems in large sets of pathologically interdependent packages.

  • 59270: Buggy recipes can cause failures in later recipes via chdir

  • 61890: file:// urls don't seem to work in find-links

    setuptools requires that file urls that point to directories must end in a "/". Added a workaround.

  • 75607: buildout should not run if it creates an empty buildout.cfg

1.0.0b24 (2007-05-09)

Feature Changes

  • Improved error reporting by showing which packages require other packages that can't be found or that cause version conflicts.
  • Added an API for use by recipe writers to clean up created files when recipe errors occur.
  • Log installed scripts.

Bugs Fixed

  • 92891: bootstrap crashes with recipe option in buildout section.
  • 113085: Buildout exited with a zero exist status when internal errors occurred.

1.0.0b23 (2007-03-19)

Feature Changes

  • Added support for download caches. A buildout can specify a cache for distribution downloads. The cache can be shared among buildouts to reduce network access and to support creating source distributions for applications allowing install without network access.
  • Log scripts created, as suggested in: https://bugs.launchpad.net/zc.buildout/+bug/71353

Bugs Fixed

  • It wasn't possible to give options on the command line for sections not defined in a configuration file.

1.0.0b22 (2007-03-15)

Feature Changes

  • Improved error reporting and debugging support:
    • Added "logical tracebacks" that show functionally what the buildout was doing when an error occurs. Don't show a Python traceback unless the -D option is used.
    • Added a -D option that causes the buildout to print a traceback and start the pdb post-mortem debugger when an error occurs.
    • Warnings are printed for unused options in the buildout section and installed-part sections. This should make it easier to catch option misspellings.
  • Changed the way the installed database (.installed.cfg) is handled to avoid database corruption when a user breaks out of a buildout with control-c.
  • Don't save an installed database if there are no installed parts or develop egg links.

1.0.0b21 (2007-03-06)

Feature Changes

  • Added support for repeatable buildouts by allowing egg versions to be specified in a versions section.

  • The easy_install module install and build functions now accept a versions argument that supplied to mapping from project name to version numbers. This can be used to fix version numbers for required distributions and their depenencies.

    When a version isn't fixed, using either a versions option or using a fixed version number in a requirement, then a debug log message is emitted indicating the version picked. This is useful for setting versions options.

    A default_versions function can be used to set a default value for this option.

  • Adjusted the output for verbosity levels. Using a single -v option no longer causes voluminous setuptools output. Uisng -vv and -vvv now triggers extra setuptools output.

  • Added a remove testing helper function that removes files or directories.

1.0.0b20 (2007-02-08)

Feature Changes

  • Added a buildout newest option, to control whether the newest distributions should be sought to meet requirements. This might also provide a hint to recipes that don't deal with distributions. For example, a recipe that manages subversion checkouts might not update a checkout if newest is set to "false".

  • Added a newest keyword parameter to the zc.buildout.easy_install.install and zc.buildout.easy_install.build functions to control whether the newest distributions that meed given requirements should be sought. If a false value is provided for this parameter and already installed eggs meet the given requirements, then no attempt will be made to search for newer distributions.

  • The recipe-testing support setUp function now adds the name buildout to the test namespace with a value that is the path to the buildout script in the sample buildout. This allows tests to use

    >>> print system(buildout),
    

    rather than:

    >>> print system(join('bin', 'buildout')),
    

Bugs Fixed

  • Paths returned from update methods replaced lists of installed files rather than augmenting them.

1.0.0b19 (2007-01-24)

Bugs Fixed

  • Explicitly specifying a Python executable failed if the output of running Python with the -V option included a 2-digit (rather than a 3-digit) version number.

1.0.0b18 (2007-01-22)

Feature Changes

  • Added documentation for some previously undocumented features of the easy_install APIs.
  • By popular demand, added a -o command-line option that is a short hand for the assignment buildout:offline=true.

Bugs Fixed

  • When deciding whether recipe develop eggs had changed, buildout incorrectly considered files in .svn and CVS directories.

1.0.0b17 (2006-12-07)

Feature Changes

  • Configuration files can now be loaded from URLs.

Bugs Fixed

1.0.0b16 (2006-12-07)

Feature Changes

  • A new command-line argument, -U, suppresses reading user defaults.
  • You can now suppress use of an installed-part database (e.g. .installed.cfg) by sprifying an empty value for the buildout installed option.

Bugs Fixed

  • When the install command is used with a list of parts, only those parts are supposed to be installed, but the buildout was also building parts that those parts depended on.

1.0.0b15 (2006-12-06)

Bugs Fixed

  • Uninstall recipes weren't loaded correctly in cases where no parts in the (new) configuration used the recipe egg.

1.0.0b14 (2006-12-05)

Feature Changes

  • Added uninstall recipes for dealing with complex uninstallation scenarios.

Bugs Fixed

  • Automatic upgrades weren't performed on Windows due to a bug that caused buildout to incorrectly determine that it wasn't running locally in a buildout.
  • Fixed some spurious test failures on Windows.

1.0.0b13 (2006-12-04)

Feature Changes

  • Variable substitutions now reflect option data written by recipes.
  • A part referenced by a part in a parts list is now added to the parts list before the referencing part. This means that you can omit parts from the parts list if they are referenced by other parts.
  • Added a develop function to the easy_install module to aid in creating develop eggs with custom build_ext options.
  • The build and develop functions in the easy_install module now return the path of the egg or egg link created.
  • Removed the limitation that parts named in the install command can only name configured parts.
  • Removed support ConfigParser-style variable substitutions (e.g. %(foo)s). Only the string-template style of variable (e.g. ${section:option}) substitutions will be supported. Supporting both violates "there's only one way to do it".
  • Deprecated the buildout-section extendedBy option.

Bugs Fixed

  • We treat setuptools as a dependency of any distribution that (declares that it) uses namespace packages, whether it declares setuptools as a dependency or not. This wasn't working for eggs intalled by virtue of being dependencies.

1.0.0b12 (2006-10-24)

Feature Changes

  • Added an initialization argument to the zc.buildout.easy_install.scripts function to include initialization code in generated scripts.

1.0.0b11 (2006-10-24)

Bugs Fixed

67737
Verbose and quite output options caused errors when the develop buildout option was used to create develop eggs.
67871
Installation failed if the source was a (local) unzipped egg.
67873
There was an error in producing an error message when part names passed to the install command weren't included in the configuration.

1.0.0b10 (2006-10-16)

Feature Changes

  • Renamed the runsetup command to setup. (The old name still works.)
  • Added a recipe update method. Now install is only called when a part is installed for the first time, or after an uninstall. Otherwise, update is called. For backward compatibility, recipes that don't define update methiods are still supported.
  • If a distribution defines namespace packages but fails to declare setuptools as one of its dependencies, we now treat setuptools as an implicit dependency. We generate a warning if the distribution is a develop egg.
  • You can now create develop eggs for setup scripts that don't use setuptools.

Bugs Fixed

  • Egg links weren't removed when corresponding entries were removed from develop sections.
  • Running a non-local buildout command (one not installed in the buildout) ket to a hang if there were new versions of zc.buildout or setuptools were available. Now we issue a warning and don't upgrade.
  • When installing zip-safe eggs from local directories, the eggs were moved, rather than copied, removing them from the source directory.

1.0.0b9 (2006-10-02)

Bugs Fixed

Non-zip-safe eggs were not unzipped when they were installed.

1.0.0b8 (2006-10-01)

Bugs Fixed

  • Installing source distributions failed when using alternate Python versions (depending on the versions of Python used.)
  • Installing eggs wasn't handled as efficiently as possible due to a bug in egg URL parsing.
  • Fixed a bug in runsetup that caused setup scripts that introspected __file__ to fail.

1.0.0b7

Added a documented testing framework for use by recipes. Refactored the buildout tests to use it.

Added a runsetup command run a setup script. This is handy if, like me, you don't install setuptools in your system Python.

1.0.0b6

Fixed https://launchpad.net/products/zc.buildout/+bug/60582 Use of extension options caused bootstrapping to fail if the eggs directory didn't already exist. We no longer use extensions for bootstrapping. There really isn't any reason to anyway.

1.0.0b5

Refactored to do more work in buildout and less work in easy_install. This makes things go a little faster, makes errors a little easier to handle, and allows extensions (like the sftp extension) to influence more of the process. This was done to fix a problem in using the sftp support.

1.0.0b4

  • Added an experimental extensions mechanism, mainly to support adding sftp support to buildouts that need it.
  • Fixed buildout self-updating on Windows.

1.0.0b3

  • Added a help option (-h, --help)

  • Increased the default level of verbosity.

  • Buildouts now automatically update themselves to new versions of zc.buildout and setuptools.

  • Added Windows support.

  • Added a recipe API for generating user errors.

  • No-longer generate a py_zc.buildout script.

  • Fixed some bugs in variable substitutions.

    The characters "-", "." and " ", weren't allowed in section or option names.

    Substitutions with invalid names were ignored, which caused missleading failures downstream.

  • Improved error handling. No longer show tracebacks for user errors.

  • Now require a recipe option (and therefore a section) for every part.

  • Expanded the easy_install module API to:

    • Allow extra paths to be provided
    • Specify explicit entry points
    • Specify entry-point arguments

1.0.0b2

Added support for specifying some build_ext options when installing eggs from source distributions.

1.0.0b1

  • Changed the bootstrapping code to only install setuptools and zc.buildout. The bootstrap code no-longer runs the buildout itself. This was to fix a bug that caused parts to be recreated unnecessarily because the recipe signature in the initial buildout reflected temporary locations for setuptools and zc.buildout.
  • Now create a minimal setup.py if it doesn't exist and issue a warning that it is being created.
  • Fixed bug in saving installed configuration data. %'s and extra spaces weren't quoted.

1.0.0a1

Initial public version

 
File Type Py Version Uploaded on Size
zc.buildout-1.7.0.tar.gz (md5) Source 2013-01-11 281KB
  • Downloads (All Versions):
  • 4230 downloads in the last day
  • 66698 downloads in the last week
  • 243326 downloads in the last month