What's Python 3.0?

Python 3.0 is a new version of the language that is incompatible with the 2.x line of releases. The language is mostly the same, but many details, especially how built-in objects like dictionaries and strings work, have changed considerably, and a lot of deprecated features have finally been removed. The standard library has also been reorganized in a few prominent places.

The aim of Python 3.0 is to clean up the language by removing little-used features and changing the language's semantics; many such changes cannot be made without breaking compatibility with Python code written for older versions.

What's Python 3000?

Around 2000, the Python developers began to refer to "Python 3000" (or Py3K) as a mythical future version of Python that wouldn't have to be compatible with existing code. As the developers began to actually work on designing and implementing this redesigned version, it slowly was renamed into the more concrete "version 3.0".

The Python 3000 name lives on in a few places; for example, the Python 3.0 branch in Subversion is at the path "/branches/py3k/".

What changed in Python 3.0?

Read What's New in Python 3.0, which lists the changes between Python 2.6 and Python 3.0.

What will happen to the Python 2.x versions?

There will be at least one more 2.x release, Python 2.7, and a 2.8 release cannot be ruled out. 2.x will continue to be developed for at least the next year, and will probably continue to be maintained for the next three years.

As users attempt to port packages to 3.x, they will surely report bugs in the compatibility and conversion tools, and Python 2.7 will incorporate the results of this feedback. As a result, new warnings will be added to the 3.0-compatibility mode, and new fixers will be added to the 2to3 conversion tool.

Python 2.7 will also incorporate any further language changes to the Python 3.x series, as far as this is possible while remaining compatible with the existing body of 2.x code.

2.7 will probably include some new features and/or modules. Whatever these new features might be, they will certainly also be present in Python 3.1.

How should I port existing Python code to Python 3.0?

Python 2.6 and 3.0 include a tool called 2to3 that parses Python 2.x code and rewrites it to work in Python 3.0. 2to3 includes a large set of fixers, each one updating a particular language feature.

We suggest the following steps for porting a package to 3.0:

* Have a test suite that exercises significant portions of your library or application's functionality, the more the better.

* Run the test suite using Python 2.6 and specifying the -3 switch, which enables warnings about code that's incompatible with Python 3.0. Modify the code to avoid triggering the warnings.

* Run the 2to3 script over the code. If 2to3 can't perform the conversion due to code constructs that can't be automatically translated, modify your 2.x code to use 3.0-compatible features.

* Once 2to3 can convert your code without complaint, try running the test suite using the 3.0-transformed version. Examine any test failures and modify the 2.x code responsible to be 3.0-compatible until the test suite succeeds using the transformed version.

Consult the 2to3 documentation for more information about the conversion tool, and PortingPythonToPy3k for more general suggestions.

One approach to distributing your 3.0 package is to use 3.0's build_py_2to3 implementation of the build_py command. See Demo/distutils/test2to3 for an example.

You will have a single lib.py, written in 2.x. When you install in 3.0, lib2to3 will convert it to 3.x in the build area, and then install the 3.0 version.

That, of course, requires you to adjust lib.py in such a way that 2to3 will successfully and completely convert it. In two experiments, Martin von Loewis ported Django and Mark Hammond ported PythonWin to 3.0, and found that these adjustments were quite feasible. You look at what 2to3 does, find out what additional modifications need to be done, and apply them to the input of 2to3 so that

* 2to3 leaves these changes in place * they either have no effect or still work correctly when run in 2.x.

In this weblog posting, Guido van Rossum strongly suggests not changing your APIs when porting to 3.0, because API changes will make it difficult for users to tell whether test suite failures are due to the 2.6->3.0 transition or to the API changes.

Early2to3Migrations has links to some early attempts at porting packages to 3.x.

How should I port existing C extensions to Python 3.0?

Benjamin Peterson's Porting Extension Modules to 3.0 describes some of the changes to Python's C API, and gives some tips for writing code that will compile with both 2.x and 3.0. The draft in this wiki may be more complete.

How do I find out what packages have been ported to 3.0?

Packages listed in the Python Package Index can indicate whether they support Python 3.x. You can view a list of 3.x-compatible packages by searching for the "Programming Language :: Python :: 3" classifier in the package index.

My new release supports Python 3.0. How do I indicate that?

If you're using your setup.py file to upload a package to the Python Package Index, include "Programming Language :: Python :: 3" as one of the classifiers.

I want to start learning/teaching Python. What version should I use?

It depends on what your intended applications are.

If you want to learn/teach the basic ideas of programming, without much reference to external libraries, Python 3.0 is simpler. Nick Efford's paper discusses his experience.

If you're learning/teaching programming in order to accomplish a particular task, such as scientific programming or web development, your decision should probably be driven by the availability of libraries.

If you decide to stick with 2.x, check whether Python comes with your operating system by entering "python" in a command line window. If you see a response from a Python interpreter it will include a version number in its initial display. Generally any recent version will do, as Python tries to maintain backwards compatibility. The version included with your operating system probably includes extension modules and additional libraries, so you should probably stick with that version. Otherwise, you should download Python 2.6.

See the BeginnersGuide for more information on getting started with Python.

Python3000/FAQ (last edited 2008-12-08 03:13:34 by AndrewKuchling)

Unable to edit the page? See the FrontPage for instructions.