95% of most popular Python packages support Python 3. Maybe you are lucky and get to start fresh using Python 3. However as of last year Python 2.7 still reigns supreme in pip installs and a lot of places 2.x is the only version you get to work in. I think writing Python 2 and 3 compatible code is an important skill, so lets check what it entails.
Python 2.x is legacy, Python 3.x is the present and future of the language - Python2orPython3 wiki
Summary Best Practices
The best place to start is the HOWTO: Porting Python 2 Code to Python 3 which nicely summarizes the important:
Only care about Python 2.7 ("Python 2.6 is no longer freely supported and thus is not receiving bugfixes."). If you have to care about older Python 2.x versions use six.
Learn the differences between 2 and 3, see this nice Cheat Sheet. Another nice article is: The key differences between Python 2.7.x and Python 3.x with examples.
Use existing tools: Futurize, Python-Modernize, caniusepython3. One word of caution about code translation tools: they might lead to less idiomatic or unnecessary code. In Picking a Python Version: A Manifesto we see 2to3 converting a range to list(range), you probably want a range to be 'lazy'. On the other hand, in the same example a map gets converted to a list comprehension which is more readable. The point is to always manually check any automatic conversions.
To test text versus binary, handled differently between 2 and 3, you can use mypy, an optional static type checker. String handling differences in 2 vs 3 probably warrant another article ...
The future and syntax
You can use __future__ imports in Python 2 to provide forward-compatibility, for example:
from __future__ import (absolute_import, division, print_function, unicode_literals)
Most well-known is the print statement in 2 becoming a function in 3. To use 3's input (instead of 2's raw_input), range (instead of 2's xrange), you can use builtins:
from builtins import (bytes, str, open, super, range, zip, round, input, int, pow, object)
python-future is the missing compatibility layer between Python 2 and Python 3. It allows you to use a single, clean Python 3.x-compatible codebase to support both Python 2 and Python 3 with minimal overhead.
See this overview for more info.
Use try/except on your imports (the HOWTO prefers this over version detection code):
try: import configparser except ImportError: import ConfigParser as configparser try: import simplejson as json except (ImportError, SyntaxError): import json
Write exceptions in a compatible way:
# don't: except Exception, e: # do: except Exception as e: # or just: except Exception: # don't: raise ValueError, 'Invalid value' # do: raise ValueError('Invalid value')
Things like from __future__ and try/except imports can be wrapped in a compat.py module , see Requests or Werkzeug for example. I actually learned about this technique in the 'Reading Great Code' chapter of The Hitchhiker’s Guide to Python. Armin Ronacher's Porting to Python 3 Redux provides some more examples of what you can add to your compat module, including decorators for differences in string handling, dictionaries and iterators.
The mentioned Cheat Sheet sums up all compatible idioms nicely.
Why it matters
It might take extra lines of code and be less idiomatic, but if on 2.x you probably have to migrate at some point. Python 2.7 will not be maintained past 2020.
Also if you release a package, doing a bit of extra effort might increase the amount of users of your software. Randy Olson's pip install analysis made me think.
This article only scratched the surface. Now is a good time to become familiar with Python porting. I learned some tricks writing this article, hopefully it gets you started too. The amount of resources available is impressive. One final site / book: Supporting Python3.
Good luck and let us know in the comments what imcompatible code you had to deal with, we like to hear your story ...
Keep Calm and Code in Python!
See an error in this post? Please submit a pull request on Github.