Should I import unicode_literals?

The future package can be used with or without unicode_literals imports.

In general, it is more compelling to use unicode_literals when back-porting new or existing Python 3 code to Python 2/3 than when porting existing Python 2 code to 2/3. In the latter case, explicitly marking up all unicode string literals with u'' prefixes would help to avoid unintentionally changing the existing Python 2 API. However, if changing the existing Python 2 API is not a concern, using unicode_literals may speed up the porting process.

This section summarizes the benefits and drawbacks of using unicode_literals. To avoid confusion, we recommend using unicode_literals everywhere across a code-base or not at all, instead of turning on for only some modules.

Benefits

  1. String literals are unicode on Python 3. Making them unicode on Python 2 leads to more consistency of your string types across the two runtimes. This can make it easier to understand and debug your code.

  2. Code without u'' prefixes is cleaner, one of the claimed advantages of Python 3. Even though some unicode strings would require a function call to invert them to native strings for some Python 2 APIs (see Standard library incompatibilities), the incidence of these function calls would usually be much lower than the incidence of u'' prefixes for text strings in the absence of unicode_literals.

  3. The diff when porting to a Python 2/3-compatible codebase may be smaller, less noisy, and easier to review with unicode_literals than if an explicit u'' prefix is added to every unadorned string literal.

  4. If support for Python 3.2 is required (e.g. for Ubuntu 12.04 LTS or Debian wheezy), u'' prefixes are a SyntaxError, making unicode_literals the only option for a Python 2/3 compatible codebase. [However, note that future doesn’t support Python 3.0-3.2.]

Drawbacks

  1. Adding unicode_literals to a module amounts to a “global flag day” for that module, changing the data types of all strings in the module at once. Cautious developers may prefer an incremental approach. (See here for an excellent article describing the superiority of an incremental patch-set in the the case of the Linux kernel.)

  1. Changing to unicode_literals will likely introduce regressions on Python 2 that require an initial investment of time to find and fix. The APIs may be changed in subtle ways that are not immediately obvious.

    An example on Python 2:

    ### Module: mypaths.py
    
    ...
    def unix_style_path(path):
        return path.replace('\\', '/')
    ...
    
    ### User code:
    
    >>> path1 = '\\Users\\Ed'
    >>> unix_style_path(path1)
    '/Users/ed'
    

    On Python 2, adding a unicode_literals import to mypaths.py would change the return type of the unix_style_path function from str to unicode in the user code, which is difficult to anticipate and probably unintended.

    The counter-argument is that this code is broken, in a portability sense; we see this from Python 3 raising a TypeError upon passing the function a byte-string. The code needs to be changed to make explicit whether the path argument is to be a byte string or a unicode string.

  2. With unicode_literals in effect, there is no way to specify a native string literal (str type on both platforms). This can be worked around as follows:

    >>> from __future__ import unicode_literals
    >>> ...
    >>> from future.utils import bytes_to_native_str as n
    
    >>> s = n(b'ABCD')
    >>> s
    'ABCD'  # on both Py2 and Py3
    

    although this incurs a performance penalty (a function call and, on Py3, a decode method call.)

    This is a little awkward because various Python library APIs (standard and non-standard) require a native string to be passed on both Py2 and Py3. (See Standard library incompatibilities for some examples. WSGI dictionaries are another.)

  1. If a codebase already explicitly marks up all text with u'' prefixes, and if support for Python versions 3.0-3.2 can be dropped, then removing the existing u'' prefixes and replacing these with unicode_literals imports (the porting approach Django used) would introduce more noise into the patch and make it more difficult to review. However, note that the futurize script takes advantage of PEP 414 and does not remove explicit u'' prefixes that already exist.

  2. Turning on unicode_literals converts even docstrings to unicode, but Pydoc breaks with unicode docstrings containing non-ASCII characters for Python versions < 2.7.7. (Fix committed in Jan 2014.):

    >>> def f():
    ...     u"Author: Martin von Löwis"
    
    >>> help(f)
    
    /Users/schofield/Install/anaconda/python.app/Contents/lib/python2.7/pydoc.pyc in pipepager(text, cmd)
       1376     pipe = os.popen(cmd, 'w')
       1377     try:
    -> 1378         pipe.write(text)
       1379         pipe.close()
       1380     except IOError:
    
    UnicodeEncodeError: 'ascii' codec can't encode character u'\xf6' in position 71: ordinal not in range(128)
    

See this Stack Overflow thread for other gotchas.

Others’ perspectives

In favour of unicode_literals

Django recommends importing unicode_literals as its top porting tip for migrating Django extension modules to Python 3. The following quote is from Aymeric Augustin on 23 August 2012 regarding why he chose unicode_literals for the port of Django to a Python 2/3-compatible codebase.:

“… I’d like to explain why this PEP [PEP 414, which allows explicit u'' prefixes for unicode literals on Python 3.3+] is at odds with the porting philosophy I’ve applied to Django, and why I would have vetoed taking advantage of it.

“I believe that aiming for a Python 2 codebase with Python 3 compatibility hacks is a counter-productive way to port a project. You end up with all the drawbacks of Python 2 (including the legacy u prefixes) and none of the advantages Python 3 (especially the sane string handling).

“Working to write Python 3 code, with legacy compatibility for Python 2, is much more rewarding. Of course it takes more effort, but the results are much cleaner and much more maintainable. It’s really about looking towards the future or towards the past.

“I understand the reasons why PEP 414 was proposed and why it was accepted. It makes sense for legacy software that is minimally maintained. I hope nobody puts Django in this category!”

Against unicode_literals

“There are so many subtle problems that unicode_literals causes. For instance lots of people accidentally introduce unicode into filenames and that seems to work, until they are using it on a system where there are unicode characters in the filesystem path.”

—Armin Ronacher

“+1 from me for avoiding the unicode_literals future, as it can have very strange side effects in Python 2…. This is one of the key reasons I backed Armin’s PEP 414.”

—Nick Coghlan

“Yeah, one of the nuisances of the WSGI spec is that the header values IIRC are the str or StringType on both py2 and py3. With unicode_literals this causes hard-to-spot bugs, as some WSGI servers might be more tolerant than others, but usually using unicode in python 2 for WSGI headers will cause the response to fail.”

—Antti Haapala