Function istext sometimes gives wrong results

[ghost]

In a specific setup (did not have time to investigate the root cause or come up with a minimal example, in my specific case it happens in the json module that tries to sort keys) I end up with a list of objects (all strings), but of different types that the istext function wrongly reports to be not of text type.

It raises the following exceptions in the compare methods (__lt__, ...) because istext returns False:

exceptions.TypeError: ("unorderable types: str() and <type 'str'>",)

This can be fixed by changing the check in the (deprecated) method istext in utils/__init__.py to

return isinstance(obj, type(u'')) or isinstance(obj, str)

[azmeuk]

I could reproduce this with python2:

>>> from builtins import str
>>> "" < str("")
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File ".../env/lib/python2.7/site-packages/future/types/newstr.py", line 316, in __gt__
    raise TypeError(self.unorderable_err.format(type(other)))
TypeError: unorderable types: str() and <type 'str'>

It raises an exception with b"" < str("") too, but not with u"" < str("").

I do not know if it is a bug or a feature since it seems to be the expected behavior in the tests.

However, I would expect this to work, because without importing str from builtins, "" < str("") works with python 2 and python 3.

@edschofield Do you have a thought on this? Would you be OK for pull requests fixing this behavior?