Fix NameError in idpicker.py on Python 3.

[jamadden]

Fixes #7.

There were tests for this, they just weren't being run, so make sure to run them.

Also a number of small updates:

• Run all the doctests.
• Enable test coverage reporting.
• Add support for Python 3.6 and explicitly list PyPy in the classifiers too.
• Modernize tox.ini
• Use pip on Travis and enable Travis caching.
• Run all doctests with consistent flags and checker
• Use renormalizing.IGNORE_EXCEPTION_MODULE_IN_PYTHON2 instead of
  manually normalizing one exception
• Rename .txt to .rst and include them in the published docs. Fix
  headers to be consistent to enable this.

[tseaver]

Should be 2.2.0.dev0.

[jamadden]

Whoops. Thanks.

[tseaver]

What is the rationale for changing the RST headers in this PR?

[jamadden]

All the files need to be consistent with each other in order to be combined.

[tseaver]

How is this not going to break where text_type is not the same as str?

[jamadden]

It's not going to break mostly through luck 😄

On Python 3, text_type is str and str + str works out as expected.

>>> u'abc' + str(123)
'abc123'

On Python 2, text_type is unicode. unicode + str does unicode-promotion on the str argument; the str of an int is only going to have ASCII characters and so will always be promotable to unicode.

>>> u'abc' + str(123)
u'abc123'

FWIW, zope.container.contained.NameChooser (from which this class derives) also makes the exact same assumptions in its chooseName method, freely combining unicode and str:

        # convert to unicode and remove characters that checkName does not allow
        if isinstance(name, bytes):
            name = name.decode()
        if not isinstance(name, unicode):
            try:
                name = unicode(name)
            except:
                name = u''
        name = name.replace('/', '-').lstrip('+@')

        ...
        # for an existing name, append a number.
        # We should keep client's os.path.extsep (not ours), we assume it's '.'
        dot = name.rfind('.')
        if dot >= 0:
            suffix = name[dot:]
            name = name[:dot]
        else:
            suffix = ''

        n = name + suffix
        i = 1
        while n in container:
            i += 1
            n = name + u'-' + str(i) + suffix

[mgedmin]

"where text_type is not str" == Python 2, yes?

$ python2
>>> u'\N{SNOWMAN}-' + str(1)
u'\u2603-1'

str(number) gives pure-ASCII strings which can be automatically promoted to unicode on Python 2.

[mgedmin]

LGTM

[mgedmin]

@jamadden: you now have PyPI rights for zope.pluggableauth

[jamadden]

Thanks for the reviews!