z3c.recipe.eggbasket

Install eggs from a tarball.

These details have been verified by PyPI

Maintainers

faassen jw maurits philikon

These details have not been verified by PyPI

Project links

Homepage

Project description

Code repository: ‘http://svn.zope.org/checkout/z3c.recipe.eggbasket/trunk

Change history

0.1.0 (xxxx-xx-xx)

Added tests. [maurits]

Initial implementation. [maurits, timte]

Created recipe with ZopeSkel. [Grok Team].

You have an egg (for example grok) that has a lot of dependencies. Other eggs that it depends on are found on the cheeseshop, on sourceforge, and perhaps on some more servers. When even one of these servers is down, other people (or you yourself) cannot install that egg. Or perhaps your egg depends on a specific version of another egg and that version is removed from the cheeseshop for some bad reason.

In other words: there are multiple points of failure. Interested users want to try your egg, the install fails because a server is down, they are disappointed, leave and never come back.

The goal of this recipe is to avoid having those multiple points of failure. You create a tarball containing all eggs that your egg depends on. A package like zc.sourcerelease can help here. You upload that tarball somewhere. In your buildout you point this recipe to your egg and the url of the tarball, for example like this:

[buildout]
parts = eggbasket

[eggbasket]
recipe = z3c.recipe.eggbasket
eggs = grok
url = http://grok.zope.org/releaseinfo/grok-eggs-0.12.tgz

The part using this recipe should usually be the first in line. What the recipe then does is install your egg and all its dependencies using only the eggs found in that tarball. After that you can let the rest of the buildout parts do their work.

Limitations

This approach still leaves you with multiple points of failure:
- the cheeseshop must be up so the end user can install this recipe
- the server with your tarball must be up.
Before buildout calls the install method of this recipe to do the actual work, all buildout parts are initialized. This means that all eggs and dependencies used by all recipes are installed. This can already involve a lot of eggs and also multiple points of failure. Workaround: you can first explicitly install the part that uses this recipe. So with the buildout snippet from above that would be:
```
bin/buildout install eggbasket
```

Supported options

The recipe supports the following options:

eggs: One or more eggs that you want to install with a tarball.
url: Url where we can get a tarball that contains the mentioned eggs and their dependencies.

Example usage

We have a package orange that depends on the package colour.

>>> import os.path
>>> import z3c.recipe.eggbasket.tests as testdir
>>> orange = os.path.join(os.path.dirname(testdir.__file__), 'orange')
>>> colour = os.path.join(os.path.dirname(testdir.__file__), 'colour')

We create source distributions for them in a directory:

>>> colours = tmpdir('colours')
>>> sdist(orange, colours)
>>> sdist(colour, colours)
>>> ls(colours)
-  colour-0.1.zip
-  orange-0.1.zip

We will define a buildout template that uses the recipe:

>>> buildout_template = """
... [buildout]
... parts = basket
...
... [basket]
... recipe = z3c.recipe.eggbasket
... eggs = %(eggs)s
... url = %(url)s
... """

We’ll start by creating a buildout that does not specify an egg:

>>> write('buildout.cfg', buildout_template % { 'eggs': '', 'url' : 'http://nowhere'})

In this case the recipe will do nothing. So the url does not get used. Running the buildout gives us:

>>> print system(buildout)
Installing basket.
Unused options for basket: 'url'.
<BLANKLINE>

Next we will specify an egg but not refer to a bad url:

>>> write('buildout.cfg', buildout_template % { 'eggs': 'orange', 'url' : 'http://nowhere'})
>>> print system(buildout)
Uninstalling basket.
Installing basket.
Couldn't find index page for 'orange' (maybe misspelled?)
Getting distribution for 'orange'.
basket: Distributions are not installed. A tarball will be downloaded.
basket: Downloading http://nowhere ...
basket: Url not found: http://nowhere.
<BLANKLINE>

So now we create a tar ball in a directory:

>>> import tarfile
>>> tarserver = tmpdir('tarserver')
>>> cd(tarserver)
>>> tarball = tarfile.open('colours.tgz', 'w:gz')
>>> tarball.add(colours)
>>> tarball.list(verbose=False)
tmp/tmpDlQSIQbuildoutSetUp/_TEST_/colours/
tmp/tmpDlQSIQbuildoutSetUp/_TEST_/colours/orange-0.1.zip
tmp/tmpDlQSIQbuildoutSetUp/_TEST_/colours/colour-0.1.zip
>>> tarball.close()
>>> ls(tarserver)
-  colours.tgz

We make it available on a url and use it in our buildout:

>>> cd(sample_buildout)
>>> tarball_url = 'file://' + tarserver + '/colours.tgz'
>>> write('buildout.cfg', buildout_template % { 'eggs': 'orange', 'url' : tarball_url})
>>> print system(buildout)
Uninstalling basket.
Installing basket.
Couldn't find index page for 'orange' (maybe misspelled?)
Getting distribution for 'orange'.
basket: Distributions are not installed. A tarball will be downloaded.
basket: Downloading .../tarserver/colours.tgz ...
basket: Finished downloading.
basket: Extracting tarball contents...
basket: Installing eggs to .../sample-buildout/eggs which will take a while...
Getting distribution for 'orange'.
Got orange 0.1.
Getting distribution for 'colour'.
Got colour 0.1.
<BLANKLINE>