Use unicode everywhere in the application.

Use unicode everywhere in the application.

My apologies for this branch, the diff is long
and the contents themself are boring: I am very
sorry. Anyway, I think this pushes quickstart
in a better place re: unicode/bytes handling.

Everything here is almost mechanical, including
the annoying problem of having to store encoded
strings in the exceptions: please let me know 
if you have any suggestions, or if I am missing 
something.

The serializers module is new stuff: it allows
for loading YAML strings as unicode and for
dumping dictionaries in the extended format
(used e.g. by juju-core in the envs.yaml file).

To QA this, just use quickstart as usual, deploying
a bundle or spinning up an environment.
Then try to break it passing exotic strings as 
options, e.g.: 
.venv/bin/python juju-quickstart --gui-charm-url ☢
(FWIW the above character is \xe2\x98\xa2).
This obviously fail, but the error should not
be unicode related.

Thank you and sorry again.

https://code.launchpad.net/~frankban/juju-quickstart/unicode-all-the-things/+merge/198112

(do not edit description out of merge proposal)

[frankban, 2013-12-06 18:30:17]

Please take a look.

[matthew.scott, 2013-12-06 18:56:03]

LGTM, QA okay.  Thanks for the work!

[rharding, 2013-12-06 19:17:34]

My tears run deep. I didn't realize it would be this messy supporting both
versions. Does it make any sense to create an alternate ValueError that can be
auto typed? 

Yay for working on getting the ball rolling. LGTM with notes below.

https://codereview.appspot.com/38500043/diff/1/quickstart/__init__.py
File quickstart/__init__.py (right):

https://codereview.appspot.com/38500043/diff/1/quickstart/__init__.py#newcode30
quickstart/__init__.py:30: return '.'.join(map(unicode, VERSION))
How does this work with py3 when there is no unicode function? I'd seen things
like
https://github.com/bookieio/breadability/blob/master/breadability/_compat.py#L12


Which creates a unicode reference to str for py3 support.

https://codereview.appspot.com/38500043/diff/1/quickstart/tests/models/test_c...
File quickstart/tests/models/test_charms.py (right):

https://codereview.appspot.com/38500043/diff/1/quickstart/tests/models/test_c...
quickstart/tests/models/test_charms.py:33: charms.parse_url('precise/juju-gui')
I was expecting these to be u'' strings now.

[frankban, 2013-12-09 10:58:53]

On 2013/12/06 19:17:34, rharding wrote:
> My tears run deep. I didn't realize it would be this messy supporting both
> versions. Does it make any sense to create an alternate ValueError that can be
> auto typed?

To be clear, with this branch we do not support both versions,
but just make the unicode handling more sane and nearer to
Python 3. Given this is not a library, I am not even sure we 
will need to support both version: we can eventually just switch
to Python3 and drop support for previous versions.

An alternate ValueError could be a nice idea: on the
other hand I liked the way we reuse already existing exception
when possible, and IMHO the more I think about being explicit
when encoding exception messages, the more it does not seem
so bad to me. When we will make the effort to port everything 
to Python3, we can remove all those unnecessary "msg.encode"
bits.

> 
> Yay for working on getting the ball rolling. LGTM with notes below.
> 
> https://codereview.appspot.com/38500043/diff/1/quickstart/__init__.py
> File quickstart/__init__.py (right):
> 
>
https://codereview.appspot.com/38500043/diff/1/quickstart/__init__.py#newcode30
> quickstart/__init__.py:30: return '.'.join(map(unicode, VERSION))
> How does this work with py3 when there is no unicode function? I'd seen things
> like
>
https://github.com/bookieio/breadability/blob/master/breadability/_compat.py#L12
> 
> Which creates a unicode reference to str for py3 support.

As mentioned above, we are still not supporting Python3 (including various
__unicode__
and __str__ methods we still have in the code with Python2 semantics).
For this reason we can avoid introducing that kind of conditionals in the code.

> 
>
https://codereview.appspot.com/38500043/diff/1/quickstart/tests/models/test_c...
> File quickstart/tests/models/test_charms.py (right):
> 
>
https://codereview.appspot.com/38500043/diff/1/quickstart/tests/models/test_c...
> quickstart/tests/models/test_charms.py:33:
charms.parse_url('precise/juju-gui')
> I was expecting these to be u'' strings now.

Are you referring to:
expected = 'charm URL has no schema: precise/juju-gui'?

If so, it is still not clear if you expected 1) the whole
string to be u'...' or 2) just the final part, e.g.:
"charm URL has no schema: u'precise/juju-gui'"

Case 1) we use unicode_literals from __future__.
Case 2) we no longer print the repr of the value in the error,
but just the value itself.

Thank you both for the reviews!

[frankban, 2013-12-09 11:02:04]

*** Submitted:

Use unicode everywhere in the application.

My apologies for this branch, the diff is long
and the contents themself are boring: I am very
sorry. Anyway, I think this pushes quickstart
in a better place re: unicode/bytes handling.

Everything here is almost mechanical, including
the annoying problem of having to store encoded
strings in the exceptions: please let me know 
if you have any suggestions, or if I am missing 
something.

The serializers module is new stuff: it allows
for loading YAML strings as unicode and for
dumping dictionaries in the extended format
(used e.g. by juju-core in the envs.yaml file).

To QA this, just use quickstart as usual, deploying
a bundle or spinning up an environment.
Then try to break it passing exotic strings as 
options, e.g.: 
.venv/bin/python juju-quickstart --gui-charm-url ☢
(FWIW the above character is \xe2\x98\xa2).
This obviously fail, but the error should not
be unicode related.

Thank you and sorry again.

R=matthew.scott, rharding
CC=
https://codereview.appspot.com/38500043