Automatic conversion to Py2/3

The future source tree includes scripts called futurize and pasteurize to aid in making Python 2 code or Python 3 code compatible with both platforms (Py2/3) using the future module. These are based on lib2to3 and use fixers from 2to3, 3to2, and python-modernize.

futurize passes Python 2 code through all the appropriate fixers to turn it into valid Python 3 code, and then adds __future__ and future package imports.

For conversions from Python 3 code to Py2/3, use the pasteurize script instead. This converts Py3-only constructs (e.g. new metaclass syntax) and adds __future__ and future imports to the top of each module.

In both cases, the result should be relatively clean Py3-style code that runs mostly unchanged on both Python 2 and Python 3.

futurize: Py2 to Py2/3

The futurize script passes Python 2 code through all the appropriate fixers to turn it into valid Python 3 code, and then adds __future__ and future package imports to re-enable compatibility with Python 2.

For example, running futurize turns this Python 2 code:

import ConfigParser                 # Py2 module name

class Upper(object):
    def __init__(self, iterable):
        self._iter = iter(iterable)
    def next(self):                 # Py2-style iterator interface
        return next(self._iter).upper()
    def __iter__(self):
        return self

itr = Upper('hello')
print next(itr),
for letter in itr:
    print letter,                   # Py2-style print statement

into this code which runs on both Py2 and Py3:

from __future__ import print_function
from future import standard_library
standard_library.install_aliases()
from future.builtins import next
from future.builtins import object
import configparser                 # Py3-style import

class Upper(object):
    def __init__(self, iterable):
        self._iter = iter(iterable)
    def __next__(self):             # Py3-style iterator interface
        return next(self._iter).upper()
    def __iter__(self):
        return self

itr = Upper('hello')
print(next(itr), end=' ')           # Py3-style print function
for letter in itr:
    print(letter, end=' ')

To write out all the changes to your Python files that futurize suggests, use the -w flag.

For complex projects, it is probably best to divide the porting into two stages. Stage 1 is for “safe” changes that modernize the code but do not break Python 2.6 compatibility or introduce a depdendency on the future package. Stage 2 is to complete the process.

Stage 1: “safe” fixes

Run the first stage of the conversion process with:

futurize --stage1 mypackage/*.py

or, if you are using zsh, recursively:

futurize --stage1 mypackage/**/*.py

This applies fixes that modernize Python 2 code without changing the effect of the code. With luck, this will not introduce any bugs into the code, or will at least be trivial to fix. The changes are those that bring the Python code up-to-date without breaking Py2 compatibility. The resulting code will be modern Python 2.6-compatible code plus __future__ imports from the following set:

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

Only those __future__ imports deemed necessary will be added unless the --all-imports command-line option is passed to futurize, in which case they are all added.

The from __future__ import unicode_literals declaration is not added unless the --unicode-literals flag is passed to futurize.

The changes include:

- except MyException, e:
+ except MyException as e:

- print >>stderr, "Blah"
+ from __future__ import print_function
+ print("Blah", stderr)

- class MyClass:
+ class MyClass(object):

- def next(self):
+ def __next__(self):

- if d.has_key(key):
+ if key in d:

Implicit relative imports fixed, e.g.:

- import mymodule
+ from __future__ import absolute_import
+ from . import mymodule

Stage 1 does not add any imports from the future package. The output of stage 1 will probably not (yet) run on Python 3.

The goal for this stage is to create most of the diff for the entire porting process, but without introducing any bugs. It should be uncontroversial and safe to apply to every Python 2 package. The subsequent patches introducing Python 3 compatibility should then be shorter and easier to review.

The complete set of fixers applied by futurize --stage1 is:

lib2to3.fixes.fix_apply
lib2to3.fixes.fix_except
lib2to3.fixes.fix_exitfunc
lib2to3.fixes.fix_funcattrs
lib2to3.fixes.fix_has_key
lib2to3.fixes.fix_idioms
lib2to3.fixes.fix_intern
lib2to3.fixes.fix_isinstance
lib2to3.fixes.fix_methodattrs
lib2to3.fixes.fix_ne
lib2to3.fixes.fix_numliterals
lib2to3.fixes.fix_paren
lib2to3.fixes.fix_reduce
lib2to3.fixes.fix_renames
lib2to3.fixes.fix_repr
lib2to3.fixes.fix_standarderror
lib2to3.fixes.fix_sys_exc
lib2to3.fixes.fix_throw
lib2to3.fixes.fix_tuple_params
lib2to3.fixes.fix_types
lib2to3.fixes.fix_ws_comma
lib2to3.fixes.fix_xreadlines
libfuturize.fixes.fix_absolute_import
libfuturize.fixes.fix_next_call
libfuturize.fixes.fix_print_with_import
libfuturize.fixes.fix_raise

The following fixers from lib2to3 are not applied:

lib2to3.fixes.fix_import

The fix_absolute_import fixer in libfuturize.fixes is applied instead of lib2to3.fixes.fix_import. The new fixer both makes implicit relative imports explicit and adds the declaration from __future__ import absolute_import at the top of each relevant module.

lib2to3.fixes.fix_next

The fix_next_call fixer in libfuturize.fixes is applied instead of fix_next in stage 1. The new fixer changes any obj.next() calls to next(obj), which is Py2/3 compatible, but doesn’t change any next method names to __next__, which would break Py2 compatibility.

fix_next is applied in stage 2.

lib2to3.fixes.fix_print

The fix_print_with_import fixer in libfuturize.fixes changes the code to use print as a function and also adds from __future__ import print_function to the top of modules using print().

In addition, it avoids adding an extra set of parentheses if these already exist. So print(x) does not become print((x)).

lib2to3.fixes.fix_raise

This fixer translates code to use the Python 3-only with_traceback() method on exceptions.

lib2to3.fixes.fix_set_literal

This converts set([1, 2, 3]) to {1, 2, 3}, breaking Python 2.6 support.

lib2to3.fixes.fix_ws_comma

This performs cosmetic changes. This is not applied by default because it does not serve to improve Python 2/3 compatibility. (In some cases it may also reduce readability: see issue #58.)

Stage 2: Py3-style code with wrappers for Py2

Run stage 2 of the conversion process with:

futurize --stage2 myfolder/*.py

This stage adds a dependency on the future package. The goal for stage 2 is to make further mostly safe changes to the Python 2 code to use Python 3-style code that then still runs on Python 2 with the help of the appropriate builtins and utilities in future.

For example:

name = raw_input('What is your name?\n')

for k, v in d.iteritems():
    assert isinstance(v, basestring)

class MyClass(object):
    def __unicode__(self):
        return u'My object'
    def __str__(self):
        return unicode(self).encode('utf-8')

would be converted by Stage 2 to this code:

from builtins import input
from builtins import str
from future.utils import iteritems, python_2_unicode_compatible

name = input('What is your name?\n')

for k, v in iteritems(d):
    assert isinstance(v, (str, bytes))

@python_2_unicode_compatible
class MyClass(object):
    def __str__(self):
        return u'My object'

Stage 2 also renames standard-library imports to their Py3 names and adds these two lines:

from future import standard_library
standard_library.install_aliases()

For example:

import ConfigParser

becomes:

from future import standard_library
standard_library.install_aliases()
import configparser

The complete list of fixers applied in Stage 2 is:

lib2to3.fixes.fix_basestring
lib2to3.fixes.fix_dict
lib2to3.fixes.fix_exec
lib2to3.fixes.fix_getcwdu
lib2to3.fixes.fix_input
lib2to3.fixes.fix_itertools
lib2to3.fixes.fix_itertools_imports
lib2to3.fixes.fix_filter
lib2to3.fixes.fix_long
lib2to3.fixes.fix_map
lib2to3.fixes.fix_nonzero
lib2to3.fixes.fix_operator
lib2to3.fixes.fix_raw_input
lib2to3.fixes.fix_zip

libfuturize.fixes.fix_cmp
libfuturize.fixes.fix_division
libfuturize.fixes.fix_execfile
libfuturize.fixes.fix_future_builtins
libfuturize.fixes.fix_future_standard_library
libfuturize.fixes.fix_future_standard_library_urllib
libfuturize.fixes.fix_metaclass
libpasteurize.fixes.fix_newstyle
libfuturize.fixes.fix_object
libfuturize.fixes.fix_unicode_keep_u
libfuturize.fixes.fix_xrange_with_import

Not applied:

lib2to3.fixes.fix_buffer    # Perhaps not safe. Test this.
lib2to3.fixes.fix_callable  # Not needed in Py3.2+
lib2to3.fixes.fix_execfile  # Some problems: see issue #37.
                            # We use the custom libfuturize.fixes.fix_execfile instead.
lib2to3.fixes.fix_future    # Removing __future__ imports is bad for Py2 compatibility!
lib2to3.fixes.fix_imports   # Called by libfuturize.fixes.fix_future_standard_library
lib2to3.fixes.fix_imports2  # We don't handle this yet (dbm)
lib2to3.fixes.fix_metaclass # Causes SyntaxError in Py2! Use the one from ``six`` instead
lib2to3.fixes.fix_unicode   # Strips off the u'' prefix, which removes a potentially
                            # helpful source of information for disambiguating
                            # unicode/byte strings.
lib2to3.fixes.fix_urllib    # Included in libfuturize.fix_future_standard_library_urllib
lib2to3.fixes.fix_xrange    # Custom one because of a bug with Py3.3's lib2to3

Fixes applied with the futurize --conservative option:

libfuturize.fixes.fix_division_safe    # instead of libfuturize.fixes.fix_division.

Separating text from bytes

After applying stage 2, the recommended step is to decide which of your Python 2 strings represent text and which represent binary data and to prefix all string literals with either b or u accordingly. Furthermore, to ensure that these types behave similarly on Python 2 as on Python 3, also wrap byte-strings or text in the bytes and str types from future. For example:

from builtins import bytes, str
b = bytes(b'\x00ABCD')
s = str(u'This is normal text')

Any unadorned string literals will then represent native platform strings (byte-strings on Py2, unicode strings on Py3).

An alternative is to pass the --unicode-literals flag:

$ futurize --unicode-literals mypython2script.py

After running this, all string literals that were not explicitly marked up as b'' will mean text (Python 3 str or Python 2 unicode).

Post-conversion

After running futurize, we recommend first running your tests on Python 3 and making further code changes until they pass on Python 3.

The next step would be manually tweaking the code to re-enable Python 2 compatibility with the help of the future package. For example, you can add the @python_2_unicode_compatible decorator to any classes that define custom __str__ methods. See What else you need to know for more info.

futurize quick-start guide

How to convert Py2 code to Py2/3 code using futurize:

Step 0: setup

Step 0 goal: set up and see the tests passing on Python 2 and failing on Python 3.

1. Clone the package from github/bitbucket. Optionally rename your repo to package-future. Examples: reportlab-future, paramiko-future, mezzanine-future.

2. Create and activate a Python 2 conda environment or virtualenv. Install the package with python setup.py install and run its test suite on Py2.7 or Py2.6 (e.g. python setup.py test or py.test or nosetests)

3. Optionally: if there is a .travis.yml file, add Python version 3.3 and remove any versions < 2.6.

4. Install Python 3.3 with e.g. sudo apt-get install python3. On other platforms, an easy way is to use Miniconda. Then e.g.:

   conda create -n py33 python=3.3 pip
   

Step 1: modern Py2 code

The goal for this step is to modernize the Python 2 code without introducing any dependencies (on future or e.g. six) at this stage.

1a. Install future into the virtualenv using:

pip install future

1b. Run futurize --stage1 -w *.py subdir1/*.py subdir2/*.py. Note that with recursive globbing in bash or zsh, you can apply stage 1 to all Python source files recursively with:

futurize --stage1 -w **/*.py

1c. Commit all changes

1d. Re-run the test suite on Py2 and fix any errors.

See Stage 1: “safe” fixes for more info.

Example error

One relatively common error after conversion is:

Traceback (most recent call last):
  ...
  File "/home/user/Install/BleedingEdge/reportlab/tests/test_encrypt.py", line 19, in <module>
    from .test_pdfencryption import parsedoc
ValueError: Attempted relative import in non-package

If you get this error, try adding an empty __init__.py file in the package directory. (In this example, in the tests/ directory.) If this doesn’t help, and if this message appears for all tests, they must be invoked differently (from the cmd line or e.g. setup.py). The way to run a module inside a package on Python 3, or on Python 2 with absolute_import in effect, is:

python -m tests.test_platypus_xref

(For more info, see PEP 328 and the PEP 8 section on absolute imports.)

Step 2: working Py3 code that still supports Py2

The goal for this step is to get the tests passing first on Py3 and then on Py2 again with the help of the future package.

2a. Run:

futurize --stage2 myfolder1/*.py myfolder2/*.py

Or, using recursive globbing with bash or zsh, you can view the stage 2 changes to all Python source files recursively with:

futurize --stage2 **/*.py

To apply the changes, add the -w argument.

This stage makes further conversions needed to support both Python 2 and 3. These will likely require imports from future on Py2 (and sometimes on Py3), such as:

from future import standard_library
standard_library.install_aliases()
# ...
from builtins import bytes
from builtins import open
from future.utils import with_metaclass

Optionally, you can use the --unicode-literals flag to add this import to the top of each module:

from __future__ import unicode_literals

All strings in the module would then be unicode on Py2 (as on Py3) unless explicitly marked with a b'' prefix.

If you would like futurize to import all the changed builtins to have their Python 3 semantics on Python 2, invoke it like this:

futurize --stage2 --all-imports myfolder/*.py

2b. Re-run your tests on Py3 now. Make changes until your tests pass on Python 3.

2c. Commit your changes! :)

2d. Now run your tests on Python 2 and notice the errors. Add wrappers from future to re-enable Python 2 compatibility. See the Cheat Sheet: Writing Python 2-3 compatible code cheat sheet and What else you need to know for more info.

After each change, re-run the tests on Py3 and Py2 to ensure they pass on both.

2e. You’re done! Celebrate! Push your code and announce to the world! Hashtags #python3 #python-future.

pasteurize: Py3 to Py2/3

Running pasteurize -w mypy3module.py turns this Python 3 code:

import configparser

class Blah:
    pass
print('Hello', end=None)

into this code which runs on both Py2 and Py3:

from __future__ import print_function
from future import standard_library
standard_library.install_hooks()

import configparser

class Blah(object):
    pass
print('Hello', end=None)

Notice that both futurize and pasteurize create explicit new-style classes that inherit from object on both Python versions, and both refer to stdlib modules (as well as builtins) under their Py3 names.

pasteurize also handles the following Python 3 features:

• keyword-only arguments
• metaclasses (using with_metaclass())
• extended tuple unpacking (PEP 3132)

To handle function annotations (PEP 3107), see Function annotations.

Known limitations

futurize and pasteurize are useful to automate much of the work of porting, particularly the boring repetitive text substitutions. They also help to flag which parts of the code require attention.

Nevertheless, futurize and pasteurize are still incomplete and make some mistakes, like 2to3, on which they are based. Please report bugs on GitHub. Contributions to the lib2to3-based fixers for futurize and pasteurize are particularly welcome! Please see Contributing.

futurize doesn’t currently make the following change automatically:

1. Strings containing \U produce a SyntaxError on Python 3. An example is:

   s = 'C:\Users'.
   

   Python 2 expands this to s = 'C:\\Users', but Python 3 requires a raw prefix (r'...'). This also applies to multi-line strings (including multi-line docstrings).

Also see the tests in future/tests/test_futurize.py marked @expectedFailure or @skip for known limitations.