Add support for OAuth 2.

This is how the OAuth2Token can be used:

[CODE]
import gdata.gauth
import gdata.client
import gdata.calendar.client
import gdata.calendar.data

token = gdata.gauth.OAuth2Token(client_id='<YOUR_CLIENT_ID>',
                                client_secret='<YOUR_CLIENT_SECRET>',
                                scope='https://www.google.com/calendar/feeds',
                                user_agent='calendar-cmdline-sample/1.0')
uri = token.generate_authorize_url()
print 'Please visit this URL to authorize the application:'
print uri
# Get the verification code from the standard input.
code = raw_input('What is the verification code? ').strip()
token.get_access_token(code)

client = gdata.calendar.client.CalendarClient(token.user_agent)
# Authorize the client.
token.authorize(client)

# Get events feed.
feed = client.GetCalendarEventFeed()
for item in feed.entry:
  print item.title.text
[/CODE]

[alainv, 2011-04-26 18:11:24]

Added some code snippet in the patch description.

Best,
Alain

[Ali Afshar, 2011-04-26 22:01:52]

I like this patch in general. Some small comments.

http://codereview.appspot.com/4440067/diff/1/src/gdata/gauth.py
File src/gdata/gauth.py (right):

http://codereview.appspot.com/4440067/diff/1/src/gdata/gauth.py#newcode146
src/gdata/gauth.py:146: def __init__(self, error_message=None):
Is error_message ever omitted to have a default of None?

http://codereview.appspot.com/4440067/diff/1/src/gdata/gauth.py#newcode1105
src/gdata/gauth.py:1105: access token.
Please describe which subset of the OAuth2 spec you are implementing, which
flow. And if other flows are planned to be added later, please make TODO.

http://codereview.appspot.com/4440067/diff/1/src/gdata/gauth.py#newcode1117
src/gdata/gauth.py:1117: Args:
Some of these arguments are specifically named as part of the OAuth2 spec,
perhaps this should be linked?

http://tools.ietf.org/html/draft-ietf-oauth-v2

http://codereview.appspot.com/4440067/diff/1/src/gdata/gauth.py#newcode1142
src/gdata/gauth.py:1142: @property
Why is this a property not a plain attribute?

http://codereview.appspot.com/4440067/diff/1/src/gdata/gauth.py#newcode1211
src/gdata/gauth.py:1211: parts[4] = urllib.urlencode(query)
None of this handles repeated query arguments, should it? I'm not sure.

http://codereview.appspot.com/4440067/diff/1/src/gdata/gauth.py#newcode1247
src/gdata/gauth.py:1247: if 'expires_in' in d:
This code is repeated from the _refresh method. Consider extracting it?

http://codereview.appspot.com/4440067/diff/1/src/gdata/gauth.py#newcode1437
src/gdata/gauth.py:1437: return OAuth2Token(parts[1], parts[2], parts[3],
parts[4], parts[5],
Any reason this is not something like (*parts[1:])

[alainv, 2011-04-26 22:58:59]

Thanks Ali!

Alain

http://codereview.appspot.com/4440067/diff/1/src/gdata/gauth.py
File src/gdata/gauth.py (right):

http://codereview.appspot.com/4440067/diff/1/src/gdata/gauth.py#newcode146
src/gdata/gauth.py:146: def __init__(self, error_message=None):
On 2011/04/26 22:01:52, Ali Afshar wrote:
> Is error_message ever omitted to have a default of None?

It is only used once, not sure what's best to do here. Do you think I should
remove the default value?

http://codereview.appspot.com/4440067/diff/1/src/gdata/gauth.py#newcode1105
src/gdata/gauth.py:1105: access token.
On 2011/04/26 22:01:52, Ali Afshar wrote:
> Please describe which subset of the OAuth2 spec you are implementing, which
> flow. And if other flows are planned to be added later, please make TODO.

Done.

http://codereview.appspot.com/4440067/diff/1/src/gdata/gauth.py#newcode1117
src/gdata/gauth.py:1117: Args:
On 2011/04/26 22:01:52, Ali Afshar wrote:
> Some of these arguments are specifically named as part of the OAuth2 spec,
> perhaps this should be linked?
> 
> http://tools.ietf.org/html/draft-ietf-oauth-v2

I added a link to the OAuth 2.0 with Google APIs above which describes the
different flows and arguments. Do you want me to add some links here again?

http://codereview.appspot.com/4440067/diff/1/src/gdata/gauth.py#newcode1142
src/gdata/gauth.py:1142: @property
On 2011/04/26 22:01:52, Ali Afshar wrote:
> Why is this a property not a plain attribute?

It's to have it read-only.

http://codereview.appspot.com/4440067/diff/1/src/gdata/gauth.py#newcode1211
src/gdata/gauth.py:1211: parts[4] = urllib.urlencode(query)
On 2011/04/26 22:01:52, Ali Afshar wrote:
> None of this handles repeated query arguments, should it? I'm not sure.

Not sure either, this was mostly inspired by what have been done on the Google
API Python client library.

http://codereview.appspot.com/4440067/diff/1/src/gdata/gauth.py#newcode1247
src/gdata/gauth.py:1247: if 'expires_in' in d:
On 2011/04/26 22:01:52, Ali Afshar wrote:
> This code is repeated from the _refresh method. Consider extracting it?

Done.

http://codereview.appspot.com/4440067/diff/1/src/gdata/gauth.py#newcode1437
src/gdata/gauth.py:1437: return OAuth2Token(parts[1], parts[2], parts[3],
parts[4], parts[5],
On 2011/04/26 22:01:52, Ali Afshar wrote:
> Any reason this is not something like (*parts[1:])

Not at all, I just followed what have been done previously. Want me to modify
the rest of calls to?

[Ali Afshar, 2011-04-27 08:34:06]

LGTM

http://codereview.appspot.com/4440067/diff/1/src/gdata/gauth.py
File src/gdata/gauth.py (right):

http://codereview.appspot.com/4440067/diff/1/src/gdata/gauth.py#newcode146
src/gdata/gauth.py:146: def __init__(self, error_message=None):
On 2011/04/26 22:58:59, alainv wrote:
> On 2011/04/26 22:01:52, Ali Afshar wrote:
> > Is error_message ever omitted to have a default of None?
> 
> It is only used once, not sure what's best to do here. Do you think I should
> remove the default value?

IMO, yes

http://codereview.appspot.com/4440067/diff/1/src/gdata/gauth.py#newcode1117
src/gdata/gauth.py:1117: Args:
On 2011/04/26 22:58:59, alainv wrote:
> On 2011/04/26 22:01:52, Ali Afshar wrote:
> > Some of these arguments are specifically named as part of the OAuth2 spec,
> > perhaps this should be linked?
> > 
> > http://tools.ietf.org/html/draft-ietf-oauth-v2
> 
> I added a link to the OAuth 2.0 with Google APIs above which describes the
> different flows and arguments. Do you want me to add some links here again?

No, not required to add again.

http://codereview.appspot.com/4440067/diff/1/src/gdata/gauth.py#newcode1142
src/gdata/gauth.py:1142: @property
On 2011/04/26 22:58:59, alainv wrote:
> On 2011/04/26 22:01:52, Ali Afshar wrote:
> > Why is this a property not a plain attribute?
> 
> It's to have it read-only.

ok

http://codereview.appspot.com/4440067/diff/1/src/gdata/gauth.py#newcode1211
src/gdata/gauth.py:1211: parts[4] = urllib.urlencode(query)
On 2011/04/26 22:58:59, alainv wrote:
> On 2011/04/26 22:01:52, Ali Afshar wrote:
> > None of this handles repeated query arguments, should it? I'm not sure.
> 
> Not sure either, this was mostly inspired by what have been done on the Google
> API Python client library.

Fair enough to leave I think, but it is a potential source of undefined behavior
if kwargs contains some keys like "scope".

http://codereview.appspot.com/4440067/diff/1/src/gdata/gauth.py#newcode1437
src/gdata/gauth.py:1437: return OAuth2Token(parts[1], parts[2], parts[3],
parts[4], parts[5],
On 2011/04/26 22:58:59, alainv wrote:
> On 2011/04/26 22:01:52, Ali Afshar wrote:
> > Any reason this is not something like (*parts[1:])
> 
> Not at all, I just followed what have been done previously. Want me to modify
> the rest of calls to?

Fine, it is consistent I guess.