Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
rt2to3: Runtime 2to3 conversion (for Python developers)
Python
branch: master

Fetching latest commit…

Cannot retrieve the latest commit at this time

Failed to load latest commit information.
.gitignore
COPYING
README.rst
rt2to3.py
setup.py

README.rst

rt2to3: Runtime 2to3 conversion (for Python developers)

Overview

The standard development process for Python 2/3 code is to write Python 2 compatible code and rely on 2to3 to convert the code to a Python 3 compatible syntax. This is cumbersome, often requiring a call to python3 setup.py build after any change.

rt2to3 avoids this by modifying the standard Python import mechanisms to optionally call 2to3 at runtime. Specifically, rt2to3 injects a path hook into sys.path_hooks. When importing a file whose path matches a predicate, often a subdirectory test, a custom file loader is used to process the module source through 2to3 before it is compiled. For speed, the results of the 2to3 conversion are cached in the __pycache__ directory.

Requirements

rt2to3 requires Python 3.2 or later.

Examples

IPython

Launch IPython from within the source directory:

$ git clone https://github.com/ipython/ipython.git
$ cd ipython
$ python3 -m rt2to3 ipython.py

Launch IPython from another directory:

$ IPY=/home/user/projects/ipython
$ python -m rt2to3 -d $IPY $IPY/ipython.py

Or specify a module to load:

$ PYTHONPATH=$IPY python3 -m rt2to3 -d $IPY -m IPython

For permanent behavior, add to your sitecustomize.py:

import sys
from rt2to3 import Runtime2to3Installer

IPY = '/home/user/projects/ipython'
nofix = ['apply', 'except', 'has_key', 'next', 'repr', 'tuple_params']

sys.path.insert(0, IPY)
Runtime2to3Installer(nofix=nofix).install(IPY)

Or use a *.pth file:

/home/user/projects/ipython
import rt2to3; rt2to3.Runtime2to3Installer(nofix=['apply', 'except', 'has_key', 'next', 'repr', 'tuple_params']).install('/home/user/projects/ipython')

Caveats

This module only affects imported code. Code which is run as a script or using execfile is not processed by 2to3. In addition, any spawned Python processes (e.g., via subprocess.Popen) will not inherit the runtime 2to3 configuration. You can work around this by using the sitecustomize.py file as suggested above.

Byte code is not cached for modules loaded using the runtime 2to3 importer. This is because the default cache tag does not know about the specific 2to3 settings (i.e., which fixers were or were not used), and so we cannot properly detect stale .pyc files.

Source inspection tools might not properly detect the 2to3 converted source, if they try to open module.__file__ directly, rather than use module.__loader__.get_data(module.__name__). The path to the 2to3 converted source is available in module.__rt2to3__.

Something went wrong with that request. Please try again.