buildout.eggtractor 0.6

Auto generates zcml slugs, buildout:egg and buildout:develop entries.

buildout.eggtractor

Q: What is a buildout extension ?

A: http://pypi.python.org/pypi/zc.buildout#extensions

The problem

When developing zope/plone eggs with buildout I have to edit the buildout configuration file ( in 3 places ) each time I create/delete/rename a development egg in the src directory or in other development directories (sometime I have more than one).

I have to add/delete/rename the egg in the eggs option of the [buildout] and then add/delete/rename the egg path in the develop option of the [buildout] and in the end add/delete/rename the zcml option of the zope [instance] or in the configure.zcml file of my policy package. This is too much specially when the speed is set to development mode. I need a less boring way to develop.

Solution

buildout.eggtractor is a buildout extension that scans the src directory or a list of directories I give for eggs and picks them up automatically. So no more editing of the buildout’s configuration file.

When buildout.eggtractor finds an egg in the scanned directory it:

1. adds the egg to the ``eggs`` option of all zope instance parts or
   to a set of given parts

2. adds the egg's path in the ``develop`` option of the ``[buildout]``

3. If ``tractor-autoload-zcml`` is not given or set to other thing than
   false, scans the egg folder for ``configure.zcml``, ``meta.zcml`` and
   ``overrides.zcml`` and adds the appropriate zcml entries to the ``zcml``
   option of the zope instance parts or to a set of given parts.

This steps are done on the fly when running buildout. So I can add/delete/rename an egg and it will be picked up.

NOTE: The extension does not write to the buildout’s configuration file.

buildout.eggtractor options

tractor-src-directory:

A set of directories to scan for development eggs. Defaults to the src directory of the buildout.

tractor-target-parts:

A set of parts to update their eggs option with eggs found in the tractor-src-directory. Defaults to zope instance parts if any.

tractor-autoload-zcml(boolean):

Update the zcml option of tractor-target-parts with the eggs found in tractor-src-directory. Defaults to true

tractor-zcml-top:

A set of eggs to load their zcml files first. Defaults to an empty set.

How to use it

Using buildout.eggtractor is very simple. As said, it is a buildout extension. All I have to do is to declare it in the extensions option:

[buildout]
parts =

extensions = buildout.eggtractor

That’s all. buildout.eggtractor will scan the src directory and do its job every time I run the buildout command.

When I have other directories I want to scan I just add an tractor-src-directory option in the [buildout] and add my directories there:

[buildout]
parts =

extensions = buildout.eggtractor

tractor-src-directory =
                      dev-src1
                      dev-src2
                      src

In a few cases when the priority of loading zcml files matters. I add the egg to be loaded first in the tractor-zcml-top option in the [buildout]:

[buildout]
parts =

extensions = buildout.eggtractor

tractor-src-directory =
          dev-src1
          dev-src2
          src

tractor-zcml-top =
          plone.app.mypackage1

If I want to add the eggs found in the development directories to the eggs option of a given set of parts, I add a tractor-target-parts option and add the parts there:

[buildout]
parts = instance1 instance2 instance3

extensions = buildout.eggtractor

tractor-target-parts = instance1 instance3

This way only instance1 and instance3 will be updated.

If I have already other way to include the zcml files (ie: z3c.autoinclude) and don’t want eggtractor to generate the zcml slugs, I add an tractor-autoload-zcml option and set it to false

In most cases you will only need to add buildout.eggtractor to the extensions option of the [buildout] without any extra configuration options.

LIMITATION

The extension assumes that the egg name reflects its file system structure

example: if the egg name is com.mustap.www the extension assumes that the file system structure is one of the following:

1. com.mustap.www/src/com/mustap/www

2. com.mustap.www/com/mustap/www

This is where the extension looks for configure.zcml, meta.zcml and overrides.zcml files.

If the egg name has nothing to do with how it is structured on the system, the extension will ignore it.

XXX: I guess walking through the directory is better than this assumption.

In my case this is not a limitation as I choose my egg names that way.

Mustapha

email: mustap_at_gmail_com

web: http://www.mustap.com

Change history

0.6 (2008-10-29)

• fixed install problem on windows: http://plone.org/support/forums#nabble-f293351

• split buildout configuration values before we test if an egg was added already. using ‘in’ with a string doen’t work if an egg’s name is a substring of an already added egg (e.g. plone.app.content after plone.app.contentmenu). [csenger]

• sort output of os.listdir (it’s order is undefined) to make debugging easier. [csenger]

0.5 (2008-04-30)

• Refactoring Added tractor-target-parts option Added tractor-autoload-zcml option Updated documentation Updated tests: need more tests [mustapha]

0.4 (2008-04-27)

• Made sure the configure.zcml is added to package includes if a meta.zcml or overrides.zcml in the same packages have already been found. [hannosch]

0.3 (2008-04-27)

• Use a new line as a separator for added entries. A space makes buildout think it has to deal with a version specifier. [hannosch]

0.2 (2008-04-27)

• Added support for automatically finding multiple instances. [hannosch]

• Better use the recipe name for finding the instance, as this is less likely to conflict. [hannosch]

0.1 (2008-04-27)

• Whitespace fixes. [hannosch]

• Created recipe with ZopeSkel. [mustapha]

Detailed Documentation

Tests for buildout.eggtractor buildout extension

Let’s create a buildout configuration file:

>>> data = """
... [buildout]
... parts = zope2 instance1 instance2
... extensions = buildout.eggtractor
... eggs =
... develop =
... [instance1]
... recipe = plone.recipe.zope2instance
... zope2-location = ${zope2:location}
... user = admin:admin
... [instance2]
... recipe = plone.recipe.zope2instance
... zope2-location = ${zope2:location}
... user = admin:admin
... [zope2]
... recipe = plone.recipe.zope2install
... url = http://www.zope.org/Products/Zope/2.9.8/Zope-2.9.8-final.tgz
... """
>>> rmdir(tempdir, 'buildout.test')
>>> cd(tempdir)
>>> sh('mkdir buildout.test')
mkdir buildout.test
<BLANKLINE>
>>> cd('buildout.test')
>>> touch('buildout.cfg', data=data)
>>> ls('.')
buildout.cfg

run the buildout first time so wget our zope instances:

>>> sh('buildout bootstrap')
buildout bootstrap
Creating directory '/tmp/buildout.test/bin'.
Creating directory '/tmp/buildout.test/parts'.
Creating directory '/tmp/buildout.test/develop-eggs'.
Generated script '/tmp/buildout.test/bin/buildout'.
<BLANKLINE>
>>> sh('bin/buildout')
bin/buildout
...
Installing instance1.
Generated script '/tmp/buildout.test/bin/instance1'.
Generated script '/tmp/buildout.test/bin/repozo'.
Installing instance2.
Generated script '/tmp/buildout.test/bin/instance2'.
<BLANKLINE>
<BLANKLINE>

Now let’s create an egg in the src directory:

>>> sh("paster create --no-interactive -o src -t plone_app com.mustap.www namespace_package=com namespace_package2=mustap package=www")
paster create --no-interactive -o src -t plone_app com.mustap.www namespace_package=com namespace_package2=mustap package=www
...
...setup.py egg_info
<BLANKLINE>

Ok, so now that we have an egg, lets run the buildout in offline mode. We should get a link file in develop-egg, a zcml slugs in parts/instance1/etc/package-includes and parts/instance2/etc/package-includes and a line with the path to our egg in the bin/instance1 and bin/instance2 files.

First we check that there is nothing of the previous mentioned things:

>>> ls('develop-eggs')
>>> ls('parts/instance1/etc/package-includes')
No directory named parts/instance1/etc/package-includes

>>> ls('parts/instance2/etc/package-includes')
No directory named parts/instance2/etc/package-includes

>>> sh('grep com.mustap.www bin/instance1')
grep com.mustap.www bin/instance1
<BLANKLINE>

>>> sh('grep com.mustap.www bin/instance2')
grep com.mustap.www bin/instance2
<BLANKLINE>

OK, now run the buildout in offline mode:

>>> sh('./bin/buildout -o')
./bin/buildout -o
...

Check that we have a correct created buildout. First check that we have a link in the develop-eggs directory:

>>> ls('develop-eggs')
com.mustap.www.egg-link

Check that we have our zcml slugs in the package-includes:

>>> ls('parts', 'instance1', 'etc', 'package-includes')
001-com.mustap.www-configure.zcml

>>> ls('parts', 'instance2', 'etc', 'package-includes')
001-com.mustap.www-configure.zcml

and in the end check that there is a line in bin/instance1 and bin/instance1 that includes our egg in the path:

>>> cat('bin', 'instance1')
#!/usr/bin/python2.4
...
sys.path[0:0] = [
  '/tmp/buildout.test/src/com.mustap.www',
...
  ]
...

>>> cat('bin', 'instance2')
#!/usr/bin/python2.4
...
sys.path[0:0] = [
  '/tmp/buildout.test/src/com.mustap.www',
...
  ]
...

Let’s now try the tractor-target-parts option. We create a new buildout.cfg file with an empty tractor-target-parts:

>>> data = data.replace('eggs =', 'tractor-target-parts = \neggs = ')
>>> touch('buildout.cfg', data=data)
>>> sh('./bin/buildout -o')
./bin/buildout -o
...

We get the egg link in the develop-egg directory:

>>> ls('develop-eggs')
com.mustap.www.egg-link

But no zcml slug in the instance 1 and 2:

>>> ls('parts', 'instance1', 'etc', 'package-includes')
No directory named parts/instance1/etc/package-includes

>>> ls('parts', 'instance2', 'etc', 'package-includes')
No directory named parts/instance2/etc/package-includes

Nor a line in bin/instance1 and bin/instance2 with our egg path:

>>> code = cat('bin', 'instance1', returndata=True)
>>> code.find('com.mustap.www') == -1
True

>>> code = cat('bin', 'instance2', returndata=True)
>>> code.find('com.mustap.www') == -1
True

But if the tractor-target-parts option is not empty:

>>> data = data.replace('tractor-target-parts =', 'tractor-target-parts = instance1')
>>> touch('buildout.cfg', data=data)
>>> sh('./bin/buildout -o')
./bin/buildout -o
...

and we get a zcml slug only in the specified target:

>>> ls('parts', 'instance1', 'etc', 'package-includes')
001-com.mustap.www-configure.zcml

>>> ls('parts', 'instance2', 'etc', 'package-includes')
No directory named parts/instance2/etc/package-includes

and only the specified target’s control script is updated:

>>> code = cat('bin', 'instance1', returndata=True)
>>> code.find('com.mustap.www') == -1
False

>>> code = cat('bin', 'instance2', returndata=True)
>>> code.find('com.mustap.www') == -1
True

Contributors

• mustapha, Author

• hannosch, Minor fixes