Default Charm namespace should be overridable by environment variables.

If JUJU_DEFAULT_NS is specified, juju will infer that namespace.
Otherwise, 'cs:' will still be the default.

https://code.launchpad.net/~clint-fewbar/juju/namespace-from-env/+merge/94479

(do not edit description out of merge proposal)

[SpamapS, 2012-02-24 00:19:11]

Please take a look.

[SpamapS, 2012-03-07 07:14:08]

Please take a look.

[bcsaller, 2012-03-07 23:30:01]

Thanks for sticking with this. I've included some feedback.

https://codereview.appspot.com/5695056/diff/3001/juju/charm/tests/test_url.py
File juju/charm/tests/test_url.py (left):

https://codereview.appspot.com/5695056/diff/3001/juju/charm/tests/test_url.py...
juju/charm/tests/test_url.py:126: 
All the tests assume the env is set to a proper value, test the case where it is
not.

https://codereview.appspot.com/5695056/diff/3001/juju/charm/url.py
File juju/charm/url.py (right):

https://codereview.appspot.com/5695056/diff/3001/juju/charm/url.py#newcode131
juju/charm/url.py:131: default_schema, rest =
os.environ.get("JUJU_DEFAULT_NS","cs:").split(":",1)
I think this module is too low level to be interacting with the environment like
this. The juju/control layer should consume environment information and can
supply information. Those (deploy and upgrade-charm) use the repository.resolve
function which can take additional arguments for this (and handle errors in a
place where we can inform the user they set the variable incorrectly. Resolve
can take the parsed output of the environment and pass it to infer. juju/control
can even use a util method from here to help parse the argument but that parsing
has to be more solid that what is here now.

Currently a bad value, missing a ":" for example, will result in a ValueError
when trying to unpack on the left-hand-side. Need to be more defensive than
that.

[SpamapS, 2012-03-09 21:47:22]

Please take a look.

[SpamapS, 2012-03-09 22:30:58]

Please take a look.

[hazmat, 2012-03-15 18:03:43]

This code would be much simpler, and easier to reason about its correctness, if
it did all its parsing up front and just used the variables instead of
conditionally redefining them as it went with incremental parsing of 'rest'.

https://codereview.appspot.com/5695056/diff/8002/juju/charm/url.py
File juju/charm/url.py (right):

https://codereview.appspot.com/5695056/diff/8002/juju/charm/url.py#newcode155
juju/charm/url.py:155: default_infer_series, rest = rest.split('/',1)
untested

https://codereview.appspot.com/5695056/diff/8002/juju/charm/url.py#newcode182
juju/charm/url.py:182: rest = "%s/%s" % (default_infer_series, rest)
untested