--future-unicode breaks handling of unicode() calls.

[yacc143]

trafficstars

$ cat t.py
   print(unicode("Hello World"))
$ python-modernize t.py
    -print(unicode("Hello World"))
    +from __future__ import print_function
    +from __future__ import absolute_import
    +import six
    +print((six.text_type("Hello World")))
$ python-modernize --future-unicode t.py
    -print(unicode("Hello World"))
    +from __future__ import print_function
    +from __future__ import unicode_literals
    +print((str("Hello World")))

Sadly, this translation breaks for anything requiring an unicode object in Python 2.x

[yacc143]

Same thing happens to unichr and chr => it seems like modernize "thinks" that unicode_literals change the semantics of chr/str to unichr/unicode in python 2 which it obviously does not.

[yacc143]

Interestingly 2to3 also changes unicode to str.

[yacc143]

Seems to be related that fix_unicode_future.py derives from lib2to3.fixes.fix_unicode

[takluyver]

Yes, the --future-unicode option should probably enable a fix with a higher priority that will check for unicode being called on a string literal, and eliminate the call before it gets rewritten to str (i.e. unicode('foo') --> 'foo'.

Do you want to have a go at this? There's plenty of examples of fixes and tests to copy from.

[shaib]

Do you think it would be better to go raising the priority of libmodernize.fixes.fix_unicode_type (or maybe some other fixers), or just lower that of libmodernize.fixes.fix_unicode_future ?

For future reference, changing the priority is done by adding a run_order member on the Fixer class; the default value is 5, smaller numbers are higher priorities. See https://github.com/python/cpython/blob/master/Lib/lib2to3/fixer_base.py#L33