matplotlib-devel Mailing List for matplotlib

Hello all,

As part of my process for updating mplot3d, I realized that transforms.py
needed to be revamped first.  In many places in transforms.py, there were
unnecessary 2-D assumptions that needed to be removed or generalized.  I
have started this effort here:

https://github.com/WeatherGod/matplotlib/compare/master...test%2Fmplot3d-ndtransforms

I still have more to do in this file, but this version passes the unit
tests, and I would welcome any comments and thoughts.  Part of the efforts
have been trying to shoehorn n-d positional arguments into various function
signatures.  For example:

def foo(self, x, y) :
    pass

becomes

def foo(self, x, y, *args) :
    pass

In some places, like shrunk_to_aspect(), I decided that I would just allow
the operation to continue for just the first two dims.  In other places, I
tried to allow for mis-matches of dimensions with reasonable defaults (e.g.,
shrunk()).

My main problem in pushing forward right now is that there are functions
that have signatures like the following:

def foo(self, x, y, ignorex=True, ignorey=True) :
    pass

Unfortunately, it is not until py3k that we can utilize keyword-only
arguments that would allow us to place a positional argument in between the
*y* and the *ignorex*, and we would run the risk of breaking function calls
like "foo(x, y, False, False)" which is valid now.  Therefore, I are going
to need to introduce some new functions to transforms.py that would allow
n-d arguments and deprecate the problematic functions.

Anyway, this is my progress so far.  I would welcome any input, thoughts,
concerns and such.

Thanks,
Ben Root