Buildout Egg-Installation Recipe

The egg-installation recipe installs eggs into a buildout eggs directory. It also generates scripts in a buildout bin directory with egg paths baked into them.

Change History

2.0.7 (2018-07-02)

• For the 2.0.6 change, we require zc.buildout 2.12.0. The install_requires in setup.py now also says that.

2.0.6 (2018-07-02)

• Added extra keyword argument allow_unknown_extras to support zc.buildout 2.12.0.

2.0.5 (2017-12-04)

• Fixed #429: added sorting of working set by priority of different type of paths (develop-eggs-directory, eggs-directory, other paths).

2.0.4 (2017-08-17)

• Fixed #153: buildout should cache working set environments [rafaelbco]

2.0.3 (2015-10-02)

• Releasing zc.recipe.egg as a wheel in addition to only an sdist. No functional changes. [reinout]

2.0.2 (2015-07-01)

• Fixed: In zc.recipe.egg#custom recipe's rpath support, don't assume path elements are buildout-relative if they start with one of the "special" tokens (e.g., $ORIGIN). See: . [tseaver]

2.0.1 (2013-09-05)

• Accomodated zc.buildout switch to post-merge setuptools.

2.0.0 (2013-04-02)

• Enabled 'prefer-final' option by default.

2.0.0a3 (2012-11-19)

• Added support for Python 3.2 / 3.3.
• Added 'MANIFEST.in'.
• Support non-entry-point-based scripts.
• Honor exit codes from scripts ().

2.0.0a2 (2012-05-03)

• Always unzip installed eggs.
• Switched from using 'setuptools' to 'distribute'.
• Removed multi-python support.

1.3.2 (2010-08-23)

• Bugfix for the change introduced in 1.3.1.

1.3.1 (2010-08-23)

• Support recipes that are using zc.recipe.egg by passing in a dict, rather than a zc.buildout.buildout.Options object as was expected/tested.

1.3.0 (2010-08-23)

• Small further refactorings past 1.2.3b1 to be compatible with zc.buildout 1.5.0.

1.2.3b1 (2010-04-29)

• Refactored to be used with z3c.recipe.scripts and zc.buildout 1.5.0. No new user-visible features.

1.2.2 (2009-03-18)

• Fixed a dependency information. zc.buildout >1.2.0 is required.

1.2.1 (2009-03-18)

• Refactored generation of relative egg paths to generate simpler code.

1.2.0 (2009-03-17)

• Added the dependent-scripts option. When set to true, scripts will be generated for all required eggs in addition to the eggs named specifically. This idea came from two forks of this recipe, repoze.recipe.egg and pylons_sandbox, but the option name is spelled with a dash instead of underscore and it defaults to false.
• Added a relative-paths option. When true, egg paths in scripts are generated relative to the script names.

1.1.0 (2008-07-19)

• Refactored to work honor the new buildout-level unzip option.

1.1.0b1 (2008-06-27)

• Added environment option to custom extension building options.

1.0.0 (2007-11-03)

• No code changes from last beta, just some small package meta-data improvements.

1.0.0b5 (2007-02-08)

Feature Changes

• Added support for the buildout newest option.

1.0.0b4 (2007-01-17)

Feature Changes

• Added initialization and arguments options to the scripts recipe.
• Added an eggs recipe that just installs eggs.
• Advertized the scripts recipe for creating scripts.

1.0.0b3 (2006-12-04)

Feature Changes

• Added a develop recipe for creating develop eggs.

  This is useful to:

  • Specify custom extension building options,
  • Specify a version of Python to use, and to
  • Cause develop eggs to be created after other parts.

• The develop and build recipes now return the paths created, so that created eggs or egg links are removed when a part is removed (or changed).

1.0.0b2 (2006-10-16)

Updated to work with (not get a warning from) zc.buildout 1.0.0b10.

1.0.0b1

Updated to work with zc.buildout 1.0.0b3.

1.0.0a3

• Extra path elements to be included in generated scripts can now be set via the extra-paths option.
• No longer implicitly generate "py_" scripts for each egg. There is now an interpreter option to generate a script that, when run without arguments, launches the Python interactive interpreter with the path set based on a parts eggs and extra paths. If this script is run with the name of a Python script and arguments, then the given script is run with the path set.
• You can now specify explicit entry points. This is useful for use with packages that don't declare their own entry points.
• Added Windows support.
• Now-longer implicitly generate "py_" scripts for each egg. You can now generate a script for launching a Python interpreter or for running scripts based on the eggs defined for an egg part.
• You can now specify custom entry points for packages that don't declare their entry points.
• You can now specify extra-paths to be included in generated scripts.

1.0.0a2

Added a custom recipe for building custom eggs using custom distutils build_ext arguments.

1.0.0a1

Initial public version

Detailed Documentation

Installation of distributions as eggs

The zc.recipe.egg:eggs recipe can be used to install various types if distutils distributions as eggs. It takes a number of options:

eggs
A list of eggs to install given as one or more setuptools requirement strings. Each string must be given on a separate line.

find-links
A list of URLs, files, or directories to search for distributions.

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

If not specified, the Python Package Index, , 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.

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

> >>> print(get(link_server), end='') > <html><body> <a > href="bigdemo-0.1-py2.3.egg">bigdemo-0.1-py2.3.egg</a><br> > <a > href="demo-0.1-py2.3.egg">demo-0.1-py2.3.egg</a><br> > <a > href="demo-0.2-py2.3.egg">demo-0.2-py2.3.egg</a><br> > <a > href="demo-0.3-py2.3.egg">demo-0.3-py2.3.egg</a><br> > <a > href="demo-0.4rc1-py2.3.egg">demo-0.4rc1-py2.3.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-1.2rc1.zip">demoneeded-1.2rc1.zip</a><br> > <a > href="du_zipped-1.0-pyN.N.egg">du_zipped-1.0-pyN.N.egg</a><br> > <a href="extdemo-1.4.zip">extdemo-1.4.zip</a><br> > <a href="index/">index/</a><br> <a > href="mixedcase-0.5.zip">mixedcase-0.5.zip</a><br> > <a > href="other-1.0-py2.3.egg">other-1.0-py2.3.egg</a><br> > </body></html>

We have a sample buildout. Let's update it's configuration file to install the demo package.

> >>> write(sample_buildout, 'buildout.cfg', ... """ ... > [buildout] ... parts = demo ... ... [demo] ... recipe = > zc.recipe.egg:eggs ... eggs = demo<0.3 ... find-links = %(server)s > ... index = %(server)s/index ... """ % dict(server=link_server))

In this example, we limited ourselves to revisions before 0.3. We also specified where to find distributions using the find-links option.

Let's run the buildout:

> >>> import os >>> print(system(buildout), > end='') Installing demo. Getting distribution for 'demo<0.3'. Got > demo 0.2. Getting distribution for 'demoneeded'. Got demoneeded 1.1.

Now, if we look at the buildout eggs directory:

> >>> ls(sample_buildout, 'eggs') d demo-0.2-py2.3.egg d > demoneeded-1.1-py2.3.egg - setuptools-0.7-py2.3.egg d > zc.buildout-1.0-py2.3.egg

We see that we got an egg for demo that met the requirement, as well as the egg for demoneeded, which demo requires. (We also see an egg link for the recipe in the develop-eggs directory. This egg link was actually created as part of the sample buildout setup. Normally, when using the recipe, you'll get a regular egg installation.)

Script generation

The demo egg defined a script, but we didn't get one installed:

> >>> ls(sample_buildout, 'bin') - buildout

If we want scripts provided by eggs to be installed, we should use the scripts recipe:

> >>> write(sample_buildout, 'buildout.cfg', ... """ ... > [buildout] ... parts = demo ... ... [demo] ... recipe = > zc.recipe.egg:scripts ... eggs = demo<0.3 ... find-links = > %(server)s ... index = %(server)s/index ... """ % > dict(server=link_server)) > > >>> print(system(buildout), end='') Uninstalling demo. > Installing demo. Generated script '/sample-buildout/bin/demo'.

Now we also see the script defined by the demo script:

> >>> ls(sample_buildout, 'bin') - buildout - demo

The scripts recipe defines some additional options:

entry-points
A list of entry-point identifiers of the form:

name=module:attrs

where name is a script name, module is a dotted name resolving to a module name, and attrs is a dotted name resolving to a callable object within a module.

This option is useful when working with distributions that don't declare entry points, such as distributions not written to work with setuptools.

Examples can be seen in the section "Specifying entry points" below.

scripts
Control which scripts are generated. The value should be a list of zero or more tokens. Each token is either a name, or a name followed by an '=' and a new name. Only the named scripts are generated. If no tokens are given, then script generation is disabled. If the option isn't given at all, then all scripts defined by the named eggs will be generated.

dependent-scripts
If set to the string "true", scripts will be generated for all required eggs in addition to the eggs specifically named.

interpreter
The name of a script to generate that allows access to a Python interpreter that has the path set based on the eggs installed.

extra-paths
Extra paths to include in a generated script.

initialization
Specify some Python initialization code. This is very limited. In particular, be aware that leading whitespace is stripped from the code given.

arguments
Specify some arguments to be passed to entry points as Python source.

relative-paths
If set to true, then egg paths will be generated relative to the script path. This allows a buildout to be moved without breaking egg paths. This option can be set in either the script section or in the buildout section.

Let's add an interpreter option:

> >>> write(sample_buildout, 'buildout.cfg', ... """ ... > [buildout] ... parts = demo ... ... [demo] ... recipe = > zc.recipe.egg ... eggs = demo<0.3 ... find-links = %(server)s ... > index = %(server)s/index ... interpreter = py-demo ... """ % > dict(server=link_server))

Note that we omitted the entry point name from the recipe specification. We were able to do this because the scripts recipe is the default entry point for the zc.recipe.egg egg.

> >>> print(system(buildout), end='') Uninstalling demo. > Installing demo. Generated script '/sample-buildout/bin/demo'. > Generated interpreter '/sample-buildout/bin/py-demo'.

Now we also get a py-demo script for giving us a Python prompt with the path for demo and any eggs it depends on included in sys.path. This is useful for debugging and testing.

> >>> ls(sample_buildout, 'bin') - buildout - demo - py-demo

If we run the demo script, it prints out some minimal data:

> >>> print(system(join(sample_buildout, 'bin', 'demo')), > end='') 2 1

The value it prints out happens to be some values defined in the modules installed.

We can also run the py-demo script. Here we'll just print(out) the bits if the path added to reflect the eggs:

> >>> print(system(join(sample_buildout, 'bin', > 'py-demo'), ... """import os, sys ... for p in sys.path: ... if 'demo' > in p: ... _ = sys.stdout.write(os.path.basename(p)+'\n') ... ... > """).replace('>>> ', '').replace('... ', ''), end='') ... # > doctest: +ELLIPSIS +NORMALIZE_WHITESPACE demo-0.2-py2.4.egg > demoneeded-1.1-py2.4.egg...

Egg updating

The recipe normally gets the most recent distribution that satisfies the specification. It won't do this is the buildout is either in non-newest mode or in offline mode. To see how this works, we'll remove the restriction on demo:

> >>> write(sample_buildout, 'buildout.cfg', ... """ ... > [buildout] ... parts = demo ... ... [demo] ... recipe = > zc.recipe.egg ... find-links = %(server)s ... index = %(server)s/index > ... """ % dict(server=link_server))

and run the buildout in non-newest mode:

> >>> print(system(buildout+' -N'), end='') Uninstalling > demo. Installing demo. Generated script '/sample-buildout/bin/demo'.

Note that we removed the eggs option, and the eggs defaulted to the part name. Because we removed the eggs option, the demo was reinstalled.

We'll also run the buildout in off-line mode:

> >>> print(system(buildout+' -o'), end='') Updating demo.

We didn't get an update for demo:

> >>> ls(sample_buildout, 'eggs') d demo-0.2-py2.3.egg d > demoneeded-1.1-py2.3.egg - setuptools-0.7-py2.3.egg d > zc.buildout-1.0-py2.3.egg

If we run the buildout on the default online and newest modes, we'll get an update for demo:

> >>> print(system(buildout), end='') Updating demo. > Getting distribution for 'demo'. Got demo 0.3. Generated script > '/sample-buildout/bin/demo'.

Then we'll get a new demo egg:

> >>> ls(sample_buildout, 'eggs') d demo-0.2-py2.3.egg d > demo-0.3-py2.3.egg d demoneeded-1.1-py2.3.egg - > setuptools-0.7-py2.4.egg d zc.buildout-1.0-py2.4.egg

The script is updated too:

> >>> print(system(join(sample_buildout, 'bin', 'demo')), > end='') 3 1

Controlling script generation

You can control which scripts get generated using the scripts option. For example, to suppress scripts, use the scripts option without any arguments:

> >>> write(sample_buildout, 'buildout.cfg', ... """ ... > [buildout] ... parts = demo ... ... [demo] ... recipe = > zc.recipe.egg ... find-links = %(server)s ... index = %(server)s/index > ... scripts = ... """ % dict(server=link_server)) > > >>> print(system(buildout), end='') Uninstalling demo. > Installing demo. > > >>> ls(sample_buildout, 'bin') - buildout

You can also control the name used for scripts:

> >>> write(sample_buildout, 'buildout.cfg', ... """ ... > [buildout] ... parts = demo ... ... [demo] ... recipe = > zc.recipe.egg ... find-links = %(server)s ... index = %(server)s/index > ... scripts = demo=foo ... """ % dict(server=link_server)) > > >>> print(system(buildout), end='') Uninstalling demo. > Installing demo. Generated script '/sample-buildout/bin/foo'. > > >>> ls(sample_buildout, 'bin') - buildout - foo

Specifying extra script paths

If we need to include extra paths in a script, we can use the extra-paths option:

> >>> write(sample_buildout, 'buildout.cfg', ... """ ... > [buildout] ... parts = demo ... ... [demo] ... recipe = > zc.recipe.egg ... find-links = %(server)s ... index = %(server)s/index > ... scripts = demo=foo ... extra-paths = ... /foo/bar ... > ${buildout:directory}/spam ... """ % dict(server=link_server)) > > >>> print(system(buildout), end='') Uninstalling demo. > Installing demo. Generated script '/sample-buildout/bin/foo'.

Let's look at the script that was generated:

> >>> cat(sample_buildout, 'bin', 'foo') # doctest: > +NORMALIZE_WHITESPACE #!/usr/local/bin/python2.7 <BLANKLINE> > import sys sys.path[0:0] = [ > '/sample-buildout/eggs/demo-0.3-py2.4.egg', > '/sample-buildout/eggs/demoneeded-1.1-py2.4.egg', '/foo/bar', > '/sample-buildout/spam', ] <BLANKLINE> import eggrecipedemo > <BLANKLINE> if __name__ == '__main__': > sys.exit(eggrecipedemo.main())

Relative egg paths

If the relative-paths option is specified with a true value, then paths will be generated relative to the script. This is useful when you want to be able to move a buildout directory around without breaking scripts.

> >>> write(sample_buildout, 'buildout.cfg', ... """ ... > [buildout] ... parts = demo ... ... [demo] ... recipe = > zc.recipe.egg ... find-links = %(server)s ... index = %(server)s/index > ... scripts = demo=foo ... relative-paths = true ... extra-paths = ... > /foo/bar ... ${buildout:directory}/spam ... """ % > dict(server=link_server)) > > >>> print(system(buildout), end='') Uninstalling demo. > Installing demo. Generated script '/sample-buildout/bin/foo'.

Let's look at the script that was generated:

> >>> cat(sample_buildout, 'bin', 'foo') # doctest: > +NORMALIZE_WHITESPACE #!/usr/local/bin/python2.7 <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'), '/foo/bar', join(base, 'spam'), ] > <BLANKLINE> import eggrecipedemo <BLANKLINE> if > __name__ == '__main__': sys.exit(eggrecipedemo.main())

You can specify relative paths in the buildout section, rather than in each individual script section:

> >>> write(sample_buildout, 'buildout.cfg', ... """ ... > [buildout] ... parts = demo ... relative-paths = true ... ... > [demo] ... recipe = zc.recipe.egg ... find-links = %(server)s ... > index = %(server)s/index ... scripts = demo=foo ... extra-paths = ... > /foo/bar ... ${buildout:directory}/spam ... """ % > dict(server=link_server)) > > >>> print(system(buildout), end='') Uninstalling demo. > Installing demo. Generated script '/sample-buildout/bin/foo'. > > >>> cat(sample_buildout, 'bin', 'foo') # doctest: > +NORMALIZE_WHITESPACE #!/usr/local/bin/python2.7 <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'), '/foo/bar', join(base, 'spam'), ] > <BLANKLINE> import eggrecipedemo <BLANKLINE> if > __name__ == '__main__': sys.exit(eggrecipedemo.main())

Specifying initialization code and arguments

Sometimes, we need to do more than just calling entry points. We can use the initialization and arguments options to specify extra code to be included in generated scripts:

> >>> write(sample_buildout, 'buildout.cfg', ... """ ... > [buildout] ... parts = demo ... ... [demo] ... recipe = > zc.recipe.egg ... find-links = %(server)s ... index = %(server)s/index > ... scripts = demo=foo ... extra-paths = ... /foo/bar ... > ${buildout:directory}/spam ... initialization = a = (1, 2 ... 3, 4) > ... interpreter = py ... arguments = a, 2 ... """ % > dict(server=link_server)) > > >>> print(system(buildout), end='') Uninstalling demo. > Installing demo. Generated script '/sample-buildout/bin/foo'. > Generated interpreter '/sample-buildout/bin/py'. > > >>> cat(sample_buildout, 'bin', 'foo') # doctest: > +NORMALIZE_WHITESPACE #!/usr/local/bin/python2.7 <BLANKLINE> > import sys sys.path[0:0] = [ > '/sample-buildout/eggs/demo-0.3-py2.4.egg', > '/sample-buildout/eggs/demoneeded-1.1-py2.4.egg', '/foo/bar', > '/sample-buildout/spam', ] <BLANKLINE> a = (1, 2 3, 4) > <BLANKLINE> import eggrecipedemo <BLANKLINE> if > __name__ == '__main__': sys.exit(eggrecipedemo.main(a, 2))

Here we see that the initialization code we specified was added after setting the path. Note, as mentioned above, that leading whitespace has been stripped. Similarly, the argument code we specified was added in the entry point call (to main).

Our interpreter also has the initialization code:

> >>> cat(sample_buildout, 'bin', 'py') ... # doctest: > +NORMALIZE_WHITESPACE +ELLIPSIS #!/usr/local/bin/python2.7 > <BLANKLINE> import sys <BLANKLINE> sys.path[0:0] = [ > '/sample-buildout/eggs/demo-0.3-py3.3.egg', > '/sample-buildout/eggs/demoneeded-1.1-py3.3.egg', '/foo/bar', > '/sample-buildout/spam', ] <BLANKLINE> a = (1, 2 3, 4) > <BLANKLINE> <BLANKLINE> _interactive = True ...

Specifying entry points

Scripts can be generated for entry points declared explicitly. We can declare entry points using the entry-points option:

> >>> write(sample_buildout, 'buildout.cfg', ... """ ... > [buildout] ... parts = demo ... ... [demo] ... recipe = > zc.recipe.egg ... find-links = %(server)s ... index = %(server)s/index > ... extra-paths = ... /foo/bar ... ${buildout:directory}/spam ... > entry-points = alt=eggrecipedemo:alt other=foo.bar:a.b.c ... """ % > dict(server=link_server)) > > >>> print(system(buildout), end='') Uninstalling demo. > Installing demo. Generated script '/sample-buildout/bin/demo'. > Generated script '/sample-buildout/bin/alt'. Generated script > '/sample-buildout/bin/other'. > > >>> ls(sample_buildout, 'bin') - alt - buildout - demo - > other > > >>> cat(sample_buildout, 'bin', 'other') > #!/usr/local/bin/python2.7 <BLANKLINE> import sys > sys.path[0:0] = [ '/sample-buildout/eggs/demo-0.3-py2.4.egg', > '/sample-buildout/eggs/demoneeded-1.1-py2.4.egg', '/foo/bar', > '/sample-buildout/spam', ] <BLANKLINE> import foo.bar > <BLANKLINE> if __name__ == '__main__': > sys.exit(foo.bar.a.b.c())

Generating all scripts

The bigdemo package doesn't have any scripts, but it requires the demo package, which does have a script. Specify dependent-scripts = true to generate all scripts in required packages:

> >>> write(sample_buildout, 'buildout.cfg', ... """ ... > [buildout] ... parts = bigdemo ... ... [bigdemo] ... recipe = > zc.recipe.egg ... find-links = %(server)s ... index = %(server)s/index > ... dependent-scripts = true ... """ % dict(server=link_server)) > >>> print(system(buildout+' -N'), end='') Uninstalling > demo. Installing bigdemo. Getting distribution for 'bigdemo'. Got > bigdemo 0.1. Generated script '/sample-buildout/bin/demo'.

Offline mode

If the buildout offline option is set to "true", then no attempt will be made to contact an index server:

> >>> write(sample_buildout, 'buildout.cfg', ... """ ... > [buildout] ... parts = demo ... offline = true ... ... [demo] ... > recipe = zc.recipe.egg ... index = eek! ... scripts = demo=foo ... """ > % dict(server=link_server)) > > >>> print(system(buildout), end='') Uninstalling bigdemo. > Installing demo. Generated script '/sample-buildout/bin/foo'.

Creating eggs with extensions needing custom build settings

Sometimes, It's necessary to provide extra control over how an egg is created. This is commonly true for eggs with extension modules that need to access libraries or include files.

The zc.recipe.egg:custom recipe can be used to define an egg with custom build parameters. The currently defined parameters are:

include-dirs
A new-line separated list of directories to search for include files.

library-dirs
A new-line separated list of directories to search for libraries to link with.

rpath
A new-line separated list of directories to search for dynamic libraries at run time.

define
A comma-separated list of names of C preprocessor variables to define.

undef
A comma-separated list of names of C preprocessor variables to undefine.

libraries
The name of an additional library to link with. Due to limitations in distutils and despite the option name, only a single library can be specified.

link-objects
The name of an link object to link against. Due to limitations in distutils and despite the option name, only a single link object can be specified.

debug
Compile/link with debugging information

force
Forcibly build everything (ignore file timestamps)

compiler
Specify the compiler type

swig
The path to the swig executable

swig-cpp
Make SWIG create C++ files (default is C)

swig-opts
List of SWIG command line options

In addition, the following options can be used to specify the egg:

egg
An specification for the egg to be created, to install given as a setuptools requirement string. This defaults to the part name.

find-links
A list of URLs, files, or directories to search for distributions.

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

If not specified, the Python Package Index, , 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.

environment
The name of a section with additional environment variables. The environment variables are set before the egg is built.

To illustrate this, we'll define a buildout that builds an egg for a package that has a simple extension module:

#include 
#include 

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.

The extension module is available as a source distribution, extdemo-1.4.tar.gz, on a distribution server.

We have a sample buildout that we'll add an include directory to with the necessary include file:

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

We'll also update the buildout configuration file to define a part for the egg:

> >>> write(sample_buildout, 'buildout.cfg', ... """ ... > [buildout] ... parts = extdemo ... ... [extdemo] ... recipe = > zc.recipe.egg:custom ... find-links = %(server)s ... index = > %(server)s/index ... include-dirs = include ... ... """ % > dict(server=link_server)) > > >>> print(system(buildout), end='') # doctest: +ELLIPSIS > Installing extdemo...

We got the zip_safe warning because the source distribution we used wasn't setuptools based and thus didn't set the option.

The egg is created in the develop-eggs directory not the eggs directory because it depends on buildout-specific parameters and the eggs directory can be shared across multiple buildouts.

> >>> ls(sample_buildout, 'develop-eggs') d > extdemo-1.4-py2.4-unix-i686.egg - zc.recipe.egg.egg-link

Note that no scripts or dependencies are installed. To install dependencies or scripts for a custom egg, define another part and use the zc.recipe.egg recipe, listing the custom egg as one of the eggs to be installed. The zc.recipe.egg recipe will use the installed egg.

Let's define a script that uses out ext demo:

> >>> mkdir('demo') >>> write('demo', 'demo.py', ... > """ ... import extdemo, sys ... def print(*args): ... > sys.stdout.write(' '.join(map(str, args)) + '\n') ... def main(): ... > print(extdemo.val) ... """) > > >>> write('demo', 'setup.py', ... """ ... from setuptools > import setup ... setup(name='demo') ... """) > > >>> write('buildout.cfg', ... """ ... [buildout] ... > develop = demo ... parts = extdemo demo ... ... [extdemo] ... recipe > = zc.recipe.egg:custom ... find-links = %(server)s ... index = > %(server)s/index ... include-dirs = include ... ... [demo] ... > recipe = zc.recipe.egg ... eggs = demo ... extdemo ... entry-points = > demo=demo:main ... """ % dict(server=link_server)) > > >>> print(system(buildout), end='') Develop: > '/sample-buildout/demo' Updating extdemo. Installing demo. Generated > script '/sample-buildout/bin/demo'...

When we run the script, we'll 42 printed:

> >>> print(system(join('bin', 'demo')), end='') 42

Updating

The custom recipe will normally check for new source distributions that meet the given specification. This can be suppressed using the buildout non-newest and offline modes. We'll generate a new source distribution for extdemo:

> >>> update_extdemo()

If we run the buildout in non-newest or offline modes:

> >>> print(system(buildout+' -N'), end='') Develop: > '/sample-buildout/demo' Updating extdemo. Updating demo. > > >>> print(system(buildout+' -o'), end='') Develop: > '/sample-buildout/demo' Updating extdemo. Updating demo.

We won't get an update.

> >>> ls(sample_buildout, 'develop-eggs') - demo.egg-link d > extdemo-1.4-py2.4-unix-i686.egg - zc.recipe.egg.egg-link

But if we run the buildout in the default on-line and newest modes, we will. This time we also get the test-variable message again, because the new version is imported:

> >>> print(system(buildout), end='') # doctest: +ELLIPSIS > Develop: '/sample-buildout/demo' Updating extdemo. zip_safe flag not > set; analyzing archive contents... Updating demo. ... > > >>> ls(sample_buildout, 'develop-eggs') - demo.egg-link d > extdemo-1.4-py2.4-linux-i686.egg d extdemo-1.5-py2.4-linux-i686.egg - > zc.recipe.egg.egg-link

Controlling the version used

We can specify a specific version using the egg option:

> >>> write('buildout.cfg', ... """ ... [buildout] ... > develop = demo ... parts = extdemo demo ... ... [extdemo] ... recipe > = zc.recipe.egg:custom ... egg = extdemo ==1.4 ... find-links = > %(server)s ... index = %(server)s/index ... include-dirs = include ... > ... [demo] ... recipe = zc.recipe.egg ... eggs = demo ... extdemo > ==1.4 ... entry-points = demo=demo:main ... """ % > dict(server=link_server)) > > >>> print(system(buildout+' -D'), end='') # doctest: > +ELLIPSIS Develop: '/sample-buildout/demo' ... > > >>> ls(sample_buildout, 'develop-eggs') - demo.egg-link d > extdemo-1.4-py2.4-linux-i686.egg - zc.recipe.egg.egg-link

Controlling environment variables

To set additional environment variables, the environment option is used.

Let's create a recipe which prints out environment variables. We need this to make sure the set environment variables are removed after the egg:custom recipe was run.

> >>> mkdir(sample_buildout, 'recipes') >>> > write(sample_buildout, 'recipes', 'environ.py', ... """ ... import > logging, os, zc.buildout ... ... class Environ: ... ... def > __init__(self, buildout, name, options): ... self.name = name ... > ... def install(self): ... logging.getLogger(self.name).info( ... > 'test-variable left over: %s' % ( ... 'test-variable' in os.environ)) > ... return [] ... ... def update(self): ... self.install() ... """) > >>> write(sample_buildout, 'recipes', 'setup.py', ... """ > ... from setuptools import setup ... ... setup( ... name = "recipes", > ... entry_points = {'zc.buildout': ['environ = environ:Environ']}, > ... ) ... """)

Create our buildout:

> >>> write(sample_buildout, 'buildout.cfg', ... """ ... > [buildout] ... develop = recipes ... parts = extdemo checkenv ... > ... [extdemo-env] ... test-variable = foo ... ... [extdemo] ... > recipe = zc.recipe.egg:custom ... find-links = %(server)s ... index = > %(server)s/index ... include-dirs = include ... environment = > extdemo-env ... ... [checkenv] ... recipe = recipes:environ ... ... > """ % dict(server=link_server)) >>> > print(system(buildout), end='') # doctest: +ELLIPSIS Develop: > '/sample-buildout/recipes' Uninstalling demo. Uninstalling extdemo. > Installing extdemo. Have environment test-variable: foo zip_safe flag > not set; analyzing archive contents... Installing checkenv. ...

The setup.py also printed out that we have set the environment test-variable to foo. After the buildout the variable is reset to its original value (i.e. removed).

When an environment variable has a value before zc.recipe.egg:custom is run, the original value will be restored:

> >>> import os >>> os.environ['test-variable'] = > 'bar' >>> print(system(buildout), end='') Develop: > '/sample-buildout/recipes' Updating extdemo. Updating checkenv. > checkenv: test-variable left over: True > > >>> os.environ['test-variable'] 'bar'

Sometimes it is required to prepend or append to an existing environment variable, for instance for adding something to the PATH. Therefore all variables are interpolated with os.environ before the're set:

> >>> write(sample_buildout, 'buildout.cfg', ... """ ... > [buildout] ... develop = recipes ... parts = extdemo checkenv ... > ... [extdemo-env] ... test-variable = foo:%%(test-variable)s ... ... > [extdemo] ... recipe = zc.recipe.egg:custom ... find-links = > %(server)s ... index = %(server)s/index ... include-dirs = include ... > environment = extdemo-env ... ... [checkenv] ... recipe = > recipes:environ ... ... """ % dict(server=link_server)) >>> > print(system(buildout), end='') # doctest: +ELLIPSIS Develop: > '/sample-buildout/recipes' Uninstalling extdemo. Installing extdemo. > Have environment test-variable: foo:bar zip_safe flag not set; > analyzing archive contents... Updating checkenv. ... > > >>> os.environ['test-variable'] 'bar' >>> del > os.environ['test-variable']

Create a clean buildout.cfg w/o the checkenv recipe, and delete the recipe:

> >>> write(sample_buildout, 'buildout.cfg', ... """ ... > [buildout] ... develop = recipes ... parts = extdemo ... ... > [extdemo] ... recipe = zc.recipe.egg:custom ... find-links = > %(server)s ... index = %(server)s/index ... include-dirs = include ... > ... """ % dict(server=link_server)) >>> > print(system(buildout), end='') # doctest: +ELLIPSIS Develop: > '/sample-buildout/recipes' Uninstalling checkenv. Uninstalling > extdemo. Installing extdemo... > > >>> rmdir(sample_buildout, 'recipes')

Controlling develop-egg generation

If you want to provide custom build options for a develop egg, you can use the develop recipe. The recipe has the following options:

setup
The path to a setup script or directory containing a startup script. This is required.

include-dirs
A new-line separated list of directories to search for include files.

library-dirs
A new-line separated list of directories to search for libraries to link with.

rpath
A new-line separated list of directories to search for dynamic libraries at run time.

define
A comma-separated list of names of C preprocessor variables to define.

undef
A comma-separated list of names of C preprocessor variables to undefine.

libraries
The name of an additional library to link with. Due to limitations in distutils and despite the option name, only a single library can be specified.

link-objects
The name of an link object to link against. Due to limitations in distutils and despite the option name, only a single link object can be specified.

debug
Compile/link with debugging information

force
Forcibly build everything (ignore file timestamps)

compiler
Specify the compiler type

swig
The path to the swig executable

swig-cpp
Make SWIG create C++ files (default is C)

swig-opts
List of SWIG command line options

To illustrate this, we'll use a directory containing the extdemo example from the earlier section:

> >>> ls(extdemo) - MANIFEST - MANIFEST.in - README - extdemo.c > - setup.py > > >>> write('buildout.cfg', ... """ ... [buildout] ... > develop = demo ... parts = extdemo demo ... ... [extdemo] ... setup > = %(extdemo)s ... recipe = zc.recipe.egg:develop ... include-dirs = > include ... define = TWO ... ... [demo] ... recipe = zc.recipe.egg > ... eggs = demo ... extdemo ... entry-points = demo=demo:main ... """ > % dict(extdemo=extdemo))

Note that we added a define option to cause the preprocessor variable TWO to be defined. This will cause the module-variable, 'val', to be set with a value of 2.

> >>> print(system(buildout), end='') # doctest: +ELLIPSIS > Develop: '/sample-buildout/demo' Uninstalling extdemo. Installing > extdemo. Installing demo. ...

Our develop-eggs now includes an egg link for extdemo:

> >>> ls('develop-eggs') - demo.egg-link - extdemo.egg-link - > zc.recipe.egg.egg-link

and the extdemo now has a built extension:

> >>> contents = os.listdir(extdemo) >>> bool([f for > f in contents if f.endswith('.so') or f.endswith('.pyd')]) True

Because develop eggs take precedence over non-develop eggs, the demo script will use the new develop egg:

> >>> print(system(join('bin', 'demo')), end='') 2

Egg Recipe API for other Recipes

It is common for recipes to accept a collection of egg specifications and generate scripts based on the resulting working sets. The egg recipe provides an API that other recipes can use.

A recipe can reuse the egg recipe, supporting the eggs, find-links, index, and extra-paths options. This is done by creating an egg recipe instance in a recipes's constructor. In the recipe's install script, the egg-recipe instance's working_set method is used to collect the requested eggs and working set.

To illustrate, we create a sample recipe that is a very thin layer around the egg recipe:

> >>> mkdir(sample_buildout, 'sample') >>> > write(sample_buildout, 'sample', 'sample.py', ... """ ... import > logging, os, sys ... import zc.recipe.egg ... ... def > print(*args): ... sys.stdout.write(' '.join(map(str, args)) + > '\n') ... ... class Sample: ... ... def __init__(self, buildout, > name, options): ... self.egg = zc.recipe.egg.Scripts(buildout, name, > options) ... self.name = name ... self.options = options ... ... def > install(self): ... extras = self.options['extras'].split() ... > requirements, ws = self.egg.working_set(extras) ... > print('Part:', self.name) ... print('Egg requirements:') ... > for r in requirements: ... print(r) ... print('Working set:') > ... for d in ws: ... print(d) ... print('extra paths:', > self.egg.extra_paths) ... return () ... ... update = install ... """)

Here we instantiated the egg recipe in the constructor, saving it in an attribute. This also initialized the options dictionary.

In our install method, we called the working_set method on the instance we saved. The working_set method takes an optional sequence of extra requirements to be included in the working set.

> >>> write(sample_buildout, 'sample', 'setup.py', ... """ ... > from setuptools import setup ... ... setup( ... name = "sample", ... > entry_points = {'zc.buildout': ['default = sample:Sample']}, ... > install_requires = 'zc.recipe.egg', ... ) ... """) > > >>> write(sample_buildout, 'sample', 'README.txt', " ") > > >>> write(sample_buildout, 'buildout.cfg', ... """ ... > [buildout] ... develop = sample ... parts = sample-part ... ... > [sample-part] ... recipe = sample ... eggs = demo<0.3 ... > find-links = %(server)s ... index = %(server)sindex ... extras = other > ... """ % dict(server=link_server)) > > >>> import os >>> os.chdir(sample_buildout) > >>> buildout = os.path.join(sample_buildout, 'bin', > 'buildout') >>> print(system(buildout + ' -q'), end='') > Part: sample-part Egg requirements: demo<0.3 Working set: > demoneeded 1.1 other 1.0 demo 0.2 extra paths: []

We can see that the options were augmented with additional data computed by the egg recipe by looking at .installed.cfg:

> >>> cat(sample_buildout, '.installed.cfg') [buildout] > installed_develop_eggs = > /sample-buildout/develop-eggs/sample.egg-link parts = sample-part > <BLANKLINE> [sample-part] __buildout_installed__ = > __buildout_signature__ = ... _b = /sample-buildout/bin _d = > /sample-buildout/develop-eggs _e = /sample-buildout/eggs > bin-directory = /sample-buildout/bin develop-eggs-directory = > /sample-buildout/develop-eggs eggs = demo<0.3 eggs-directory = > /sample-buildout/eggs extras = other find-links = > index = > recipe = sample

If we use the extra-paths option:

> >>> write(sample_buildout, 'buildout.cfg', ... """ ... > [buildout] ... develop = sample ... parts = sample-part ... ... > [sample-part] ... recipe = sample ... eggs = demo<0.3 ... > find-links = %(server)s ... index = %(server)sindex ... extras = other > ... extra-paths = /foo/bar ... /spam/eggs ... """ % > dict(server=link_server))

Then we'll see that reflected in the extra_paths attribute in the egg recipe instance:

> >>> print(system(buildout + ' -q'), end='') Part: > sample-part Egg requirements: demo<0.3 Working set: demo 0.2 other > 1.0 demoneeded 1.1 extra paths: ['/foo/bar', '/spam/eggs']

Working set caching

Working sets are cached, to improve speed on buildouts with multiple similar parts based on zc.recipe.egg.

The egg-recipe instance's _working_set helper method is used to make the caching easier. It does the same job as working_set() but with some differences:

• The signature is different: all information needed to build the working set is passed as parameters.
• The return value is simpler: only an instance of pkg_resources.WorkingSet is returned.

Here's an example:

> >>> from zc.buildout import testing >>> from > zc.recipe.egg.egg import Eggs >>> import os >>> > import pkg_resources >>> recipe = > Eggs(buildout=testing.Buildout(), name='fake-part', options={}) > >>> eggs_dir = os.path.join(sample_buildout, 'eggs') > >>> develop_eggs_dir = os.path.join(sample_buildout, > 'develop-eggs') >>> testing.install_develop('zc.recipe.egg', > develop_eggs_dir) >>> ws = recipe._working_set( ... > distributions=['zc.recipe.egg', 'demo<0.3'], ... > eggs_dir=eggs_dir, ... develop_eggs_dir=develop_eggs_dir, ... > index=link_server, ... ) Getting... >>> isinstance(ws, > pkg_resources.WorkingSet) True >>> sorted(dist.project_name > for dist in ws) ['demo', 'demoneeded', 'setuptools', 'zc.buildout', > 'zc.recipe.egg']

We'll monkey patch a method in the easy_install module in order to verify if the cache is working:

> >>> import zc.buildout.easy_install >>> > old_install = zc.buildout.easy_install.Installer.install > >>> def new_install(*args,kwargs): ... print('Building > working set.') ... return old_install(*args, **kwargs) >>> > zc.buildout.easy_install.Installer.install = new_install

Now we check if the caching is working by verifying if the same working set is built only once.

> >>> ws_args_1 = dict( ... distributions=['demo>=0.1'], > ... eggs_dir=eggs_dir, ... develop_eggs_dir=develop_eggs_dir, > ... offline=True, ... ) >>> ws_args_2 = dict(ws_args_1) > >>> ws_args_2['distributions'] = ['demoneeded'] > >>> recipe._working_set(**ws_args_1) Building working > set. <pkg_resources.WorkingSet object at ...> >>> > recipe._working_set(**ws_args_1) <pkg_resources.WorkingSet > object at ...> >>> recipe._working_set(**ws_args_2) > Building working set. <pkg_resources.WorkingSet object at ...> > >>> recipe._working_set(**ws_args_1) > <pkg_resources.WorkingSet object at ...> >>> > recipe._working_set(**ws_args_2) <pkg_resources.WorkingSet > object at ...>

Undo monkey patch:

> >>> zc.buildout.easy_install.Installer.install = > old_install

Since pkg_resources.WorkingSet instances are mutable, we must ensure that working_set() always returns a pristine copy. Otherwise callers would be able to modify instances inside the cache.

Let's create a working set:

> >>> ws = recipe._working_set(**ws_args_1) >>> > sorted(dist.project_name for dist in ws) ['demo', 'demoneeded']

Now we add a distribution to it:

> >>> dist = pkg_resources.get_distribution('zc.recipe.egg') > >>> ws.add(dist) >>> sorted(dist.project_name for > dist in ws) ['demo', 'demoneeded', 'zc.recipe.egg']

Let's call the working_set function again and see if the result remains valid:

> >>> ws = recipe._working_set(**ws_args_1) >>> > sorted(dist.project_name for dist in ws) ['demo', 'demoneeded']

Download