Fork me on GitHub

Imports

__future__ imports

To write a Python 2/3 compatible codebase, the first step is to add this line to the top of each module:

from __future__ import absolute_import, division, print_function

For guidelines about whether to import unicode_literals too, see below (Should I import unicode_literals?).

For more information about the __future__ imports, which are a standard feature of Python, see the following docs:

These are all available in Python 2.6 and up, and enabled by default in Python 3.x.

Imports of builtins

Implicit imports

If you don’t mind namespace pollution, the easiest way to provide Py2/3 compatibility for new code using future is to include the following imports at the top of every module:

from builtins import *

On Python 3, this has no effect. (It shadows builtins with globals of the same names.)

On Python 2, this import line shadows 18 builtins (listed below) to provide their Python 3 semantics.

Explicit imports

Explicit forms of the imports are often preferred and are necessary for using certain automated code-analysis tools.

The complete set of imports of builtins from future is:

from builtins import (ascii, bytes, chr, dict, filter, hex, input,
                      int, map, next, oct, open, pow, range, round,
                      str, super, zip)

These are also available under the future.builtins namespace for backward compatibility.

Importing only some of the builtins is cleaner but increases the risk of introducing Py2/3 portability bugs as your code evolves over time. For example, be aware of forgetting to import input, which could expose a security vulnerability on Python 2 if Python 3’s semantics are expected.

The internal API is currently as follows:

from future.types import bytes, dict, int, range, str
from future.builtins.misc import (ascii, chr, hex, input, next,
                                  oct, open, pow, round, super)
from future.builtins.iterators import filter, map, zip

Please note that this internal API is evolving and may not be stable between different versions of future. To understand the details of the backported builtins on Python 2, see the docs for these modules.

For more information on what the backported types provide, see What else you need to know.

Obsolete Python 2 builtins

Twelve Python 2 builtins have been removed from Python 3. To aid with porting code to Python 3 module by module, you can use the following import to cause a NameError exception to be raised on Python 2 when any of the obsolete builtins is used, just as would occur on Python 3:

from future.builtins.disabled import *

This is equivalent to:

from future.builtins.disabled import (apply, cmp, coerce, execfile,
                             file, long, raw_input, reduce, reload,
                             unicode, xrange, StandardError)

Running futurize over code that uses these Python 2 builtins does not import the disabled versions; instead, it replaces them with their equivalent Python 3 forms and then adds future imports to resurrect Python 2 support, as described in Stage 2: Py3-style code with wrappers for Py2.

Standard library imports

future supports the standard library reorganization (PEP 3108) through several mechanisms.

Direct imports

As of version 0.14, the future package comes with top-level packages for Python 2.x that provide access to the reorganized standard library modules under their Python 3.x names.

Direct imports are the preferred mechanism for accesing the renamed standard library modules in Python 2/3 compatible code. For example, the following clean Python 3 code runs unchanged on Python 2 after installing future:

>>> # Alias for future.builtins on Py2:
>>> from builtins import str, open, range, dict

>>> # Top-level packages with Py3 names provided on Py2:
>>> import queue
>>> import tkinter.dialog
>>> etc.

Notice that this code actually runs on Python 3 without the presence of the future package.

Of the 44 modules that were refactored with PEP 3108 (standard library reorganization), 29 are supported with direct imports in the above manner. The complete list is here:

### Renamed modules:

import builtins

import copyreg

import html
import html.entities
import html.parser

import http.client
import http.cookies
import http.cookiejar
import http.server

import queue

import reprlib

import socketserver

from tkinter import colorchooser
from tkinter import commondialog
from tkinter import constants
from tkinter import dialog
from tkinter import dnd
from tkinter import filedialog
from tkinter import font
from tkinter import messagebox
from tkinter import scrolledtext
from tkinter import simpledialog
from tkinter import tix
from tkinter import ttk

import winreg                    # Windows only

import xmlrpc.client
import xmlrpc.server

import _dummy_thread
import _markupbase
import _thread

Note that, as of v0.16.0, python-future no longer includes an alias for the configparser module because a full backport exists (see https://pypi.python.org/pypi/configparser).

Aliased imports

The following 14 modules were refactored or extended from Python 2.6/2.7 to 3.x but were neither renamed in Py3.x nor were the new APIs backported to Py2.x. This precludes compatibility interfaces that work out-of-the-box. Instead, the future package makes the Python 3.x APIs available on Python 2.x as follows:

from future.standard_library import install_aliases
install_aliases()

from collections import UserDict, UserList, UserString

import urllib.parse
import urllib.request
import urllib.response
import urllib.robotparser
import urllib.error

import dbm
import dbm.dumb
import dbm.gnu                # requires Python dbm support
import dbm.ndbm               # requires Python dbm support

from itertools import filterfalse, zip_longest

from subprocess import getoutput, getstatusoutput

from sys import intern

import test.support

The newly exposed urllib submodules are backports of those from Py3.x. This means, for example, that urllib.parse.unquote() now exists and takes an optional encoding argument on Py2.x as it does on Py3.x.

Limitation: Note that the http-based backports do not currently support HTTPS (as of 2015-09-11) because the SSL support changed considerably in Python 3.x. If you need HTTPS support, please use this idiom for now:

from future.moves.urllib.request import urlopen

Backports also exist of the following features from Python 3.4:

  • math.ceil returns an int on Py3
  • collections.OrderedDict (for Python 2.6)
  • collections.Counter (for Python 2.6)
  • collections.ChainMap (for all versions prior to Python 3.3)
  • itertools.count (for Python 2.6, with step parameter)
  • subprocess.check_output (for Python 2.6)
  • reprlib.recursive_repr (for Python 2.6 and 2.7)

These can then be imported on Python 2.6+ as follows:

from future.standard_library import install_aliases
install_aliases()

from math import ceil      # now returns an int
from collections import Counter, OrderedDict, ChainMap
from itertools import count
from subprocess import check_output
from reprlib import recursive_repr

External standard-library backports

Backports of the following modules from the Python 3.x standard library are available independently of the python-future project:

import enum                       # pip install enum34
import singledispatch             # pip install singledispatch
import pathlib                    # pip install pathlib

A few modules from Python 3.4 and 3.3 are also available in the backports package namespace after pip install backports.lzma etc.:

from backports import lzma
from backports import functools_lru_cache as lru_cache

The following Python 2.6 backports of standard library packages from Python 2.7+ are also available:

import argparse                   # pip install argparse
import importlib                  # pip install importlib
import unittest2 as unittest      # pip install unittest2

These are included in Python 2.7 and Python 3.x.

Included full backports

Alpha-quality full backports of the following modules from Python 3.3’s standard library to Python 2.x are also available in future.backports:

http.client
http.server
html.entities
html.parser
urllib
xmlrpc.client
xmlrpc.server

The goal for these modules, unlike the modules in the future.moves package or top-level namespace, is to backport new functionality introduced in Python 3.3.

If you need the full backport of one of these packages, please open an issue here.

Using Python 2-only dependencies on Python 3

The past module provides an experimental translation package to help with importing and using old Python 2 modules in a Python 3 environment.

This is implemented using PEP 414 import hooks together with fixers from lib2to3 and libfuturize (included with python-future) that attempt to automatically translate Python 2 code to Python 3 code with equivalent semantics upon import.

Note This feature is still in alpha and needs further development to support a full range of real-world Python 2 modules. Also be aware that the API for this package might change considerably in later versions.

Here is how to use it:

$ pip3 install plotrique==0.2.5-7 --no-compile   # to ignore SyntaxErrors
$ python3

Then pass in a whitelist of module name prefixes to the past.autotranslate() function. Example:

>>> from past import autotranslate
>>> autotranslate(['plotrique'])
>>> import plotrique

Here is another example:

>>> from past.translation import install_hooks, remove_hooks
>>> install_hooks(['mypy2module'])
>>> import mypy2module
>>> remove_hooks()

This will translate, import and run Python 2 code such as the following:

### File: mypy2module.py

# Print statements are translated transparently to functions:
print 'Hello from a print statement'

# xrange() is translated to Py3's range():
total = 0
for i in xrange(10):
    total += i
print 'Total is: %d' % total

# Dictionary methods like .keys() and .items() are supported and
# return lists as on Python 2:
d = {'a': 1, 'b': 2}
assert d.keys() == ['a', 'b']
assert isinstance(d.items(), list)

# Functions like range, reduce, map, filter also return lists:
assert isinstance(range(10), list)

# The exec statement is supported:
exec 'total += 1'
print 'Total is now: %d' % total

# Long integers are supported:
k = 1234983424324L
print 'k + 1 = %d' % k

# Most renamed standard library modules are supported:
import ConfigParser
import HTMLParser
import urllib

The attributes of the module are then accessible normally from Python 3. For example:

# This Python 3 code works
>>> type(mypy2module.d)
builtins.dict

This is a standard Python 3 data type, so, when called from Python 3 code, keys() returns a view, not a list:

>>> type(mypy2module.d.keys())
builtins.dict_keys
  • It currently requires a newline at the end of the module or it throws a ParseError.
  • This only works with pure-Python modules. C extension modules and Cython code are not supported.
  • The biggest hurdle to automatic translation is likely to be ambiguity about byte-strings and text (unicode strings) in the Python 2 code. If the past.autotranslate feature fails because of this, you could try running futurize over the code and adding a b'' or u'' prefix to the relevant string literals. To convert between byte-strings and text (unicode strings), add an .encode or .decode method call. If this succeeds, please push your patches upstream to the package maintainers.
  • Otherwise, the source translation feature offered by the past.translation package has similar limitations to the futurize script (see Known limitations). Help developing and testing this feature further would be particularly welcome.

Please report any bugs you find on the python-future bug tracker.

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

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!”

“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