Bytes, strings, and unicode?

[ltratt]

I see this warning in the standard library:

/tmp/pygrate2/Lib/codecs.py:506: Py3xWarning: The first string is 'str' while the second is 'unicode': mixed bytes, str and unicode operands cannot be used in string concatenation in Python 3.x: convert the operand(s) so that they are the same type.
  self.charbuffer += newchars

Since Python3 doesn't have the unicode type per se, I'm wondering if this warning is quite right? As I understand the situation (and my memory is a bit fuzzy...):

• Python2 has the types str and unicode but not a type bytes. The bytes() function returns strs.
• Python3 has the types bytes and str but not unicode.

If I've got that (roughly) correct, I wonder whether the warning at the top of this issue is correct and/or as accurate as it could be? I think that str + unicode in Python2 will always map to str + str in Python3? Or are there certain types of strings/unicode strings in Python2 that will map to bytes in Python3?

[nanjekyejoannah]

Yes, you are correct, the u prefix just exists for backward compatibility, i.e, no-op per se, its not supported in semantics to say unicode = str, that is to say no Python 3 code uses the u prefix or is expected to. Any code still using the u prefix is most likely assuming the genuine existence of the unicode type as was the case in Python 2. I think in short supporting u in python 3 is not a feature, just exists for convenience. If we look at it as a supported solely for compatibility, the warning is correct, if we assume it is a feature the warning needs fixing.

[ltratt]

I'm not sure this is entirely about u"blah" strings, but I might be wrong. The warning we're giving is really saying the + operator won't work as expected in Python3. In one sense, I suppose that's true, but I think it's not very useful, because Python3 won't have let a user construct a unicode string by definition.

Now that I think about it, I'm not even sure if the user can do much about this warning? Can they convert their string type(s) in a way that will make their code forward compatible?

[nanjekyejoannah]

By making the strings involved in the + operation of the same type. str + str or bytes + bytes , if we took u to be a feature u'str' + 'str' shouldnt lead to a warning.