six u() breaks for existing unicode strings for Python 2 about six HOT 11 OPEN

Original comment by alex (Bitbucket: alexssb, GitHub: Unknown):

Six not solve the problem below, which causes the test fails

Python 3.4 (Linux)

https://github.com/eduardtomasek/lz-string-python/blob/master/test.py#L9

Python 2.7 (Linux) - * Compatibility using six. *

https://github.com/alexsilva/lz-string-python/blob/master/test.py#L11

s = six.u('Žluťoučký kůň úpěl ďábelské ódy!')

Only works in both cases, when I use the u prefix 'u'

u'Žluťoučký kůň úpěl ďábelské ódy!'

from six.

Original comment by Wouter Bolsterlee (Bitbucket: wbolster, GitHub: wbolster):

@bfschott the file encoding is irrelevant here, and doesn't change whether string literals are bytes/text.

from six.

The best thing to do is to use Python 2 and 3.3+ and use the syntactic
u prefix.

from six.

Original comment by bfschott (Bitbucket: bfschott, GitHub: Unknown):

I'm hitting this is a third party library:

#!python
        return six.u('')

Just realized, of course, that the file is utf-8:

#!python

# -*- coding: utf-8 -*-

Not exactly sure what the proper portable way is to return a unicode literal in a return is with python. I can't do u'', just ''?

from six.

Yes, you want something like Django's smart_unicode.

from six.

Original comment by Alexander Lukanin (Bitbucket: alexanderlukanin13, GitHub: alexanderlukanin13):

You are misusing six.u. It should only be used for string literals. It's not a make-unicode-out-of-anything magic function.

from six.

Original comment by James Cleveland (Bitbucket: radiosilence, GitHub: radiosilence):

Exactly - sometimes it's difficult to make sure that a string is ascii, especially when user data etc. are involved.

from six.

Original comment by Vladimir Protasov (Bitbucket: eoranged, GitHub: eoranged):

@gutworth, It's common thing in Python 2 that function result in some conditions is str and on the other is unicode instance, so the call

result = six.u(json.dumps(data, **json_params))

may fail with

TypeError: decoding Unicode is not supported

if someone will pass ensure_ascii=False in json_params.

from six.

I don't understand. If 3rd-party libraries are doing six.u(u'mystring'), they have a bug.

from six.

Original comment by James Cleveland (Bitbucket: radiosilence, GitHub: radiosilence):

If you're using third party libraries, it's kind of a pain due to having to fork them and whatnot. Would this change actually break anything?

It's kind of interesting that this works fine:

unicode(u'hi')

Whereas this breaks:

unicode(u'hi', 'unicode_escape')

Any ideas why?

from six.

Er, how hard is it to just not pass unicode literals?

from six.