Hello all,
While at SciPy this year, the IPython devs began discussing dropping Python 2.4 support. Or rational is this:
-
New generator features in 2.5 would dramatically simplify our testing of our Twisted using components.
-
Being able to use absolute imports would simplify the packaging of some external deps.
-
The faster we get rid of 2.4 and 2.5 support (we are not getting rid of 2.5 yet) the faster we can transition to py3k.
But, this would mean that pylab mode for matplotlib would require either:
We know that there are people still using Python 2.4, but at this point, we feel the benefits outweight the costs. How do the matplotlib devs feel about this?
As a side note, as of IPython 0.11, the IPython threaded shells (pylab stuff) will be completely refactored. This will require matplotlib to make some moderate changes to support the new interface. Thus, even if we don’t drop 2.4 support, matplotlib will have to decide how to handle the IPython 0.10->0.11 transition.
Cheers,
Brian
My sense is that most people still using python-2.4 probably put a
premium on stability and wouldn't be interested in (or aware of) the
refactored codebase until it has seen some use and had a chance to
work out the bugs. I think it is reasonable to provide a ipy-0.10
based on the old codebase that is compatible with py-2.4 and make use
of the newer language features in the trunk.
Hopefully someone will come up with a 3to2 tool that will support
python-2.5. That might provide a faster route to ipython for python-3.
Darren
···
On Tue, Aug 25, 2009 at 4:35 PM, Brian Granger<ellisonbg.net@...149...> wrote:
Hello all,
While at SciPy this year, the IPython devs began discussing dropping Python
2.4 support. Or rational is this:
* New generator features in 2.5 would dramatically simplify our testing of
our Twisted using components.
* Being able to use absolute imports would simplify the packaging of some
external deps.
* The faster we get rid of 2.4 and 2.5 support (we are not getting rid of
2.5 yet) the faster we can transition to py3k.
But, this would mean that pylab mode for matplotlib would require either:
* Using IPython v 0.10 or below
* Using Python 2.5/2.6
We know that there are people still using Python 2.4, but at this point, we
feel the benefits outweight the costs. How do the matplotlib devs feel
about this?
As a side note, as of IPython 0.11, the IPython threaded shells (pylab
stuff) will be completely refactored. This will require matplotlib to make
some moderate changes to support the new interface. Thus, even if we don't
drop 2.4 support, matplotlib will have to decide how to handle the IPython
0.10->0.11 transition.