Re: [Matplotlib-users] Spooky side effects of Axes3D

[Warren W. <war...@en...>]

On Sun, Mar 27, 2011 at 8:25 PM, Angus McMorland <am...@gm...> wrote:

> On 27 March 2011 20:47, Warren Weckesser <war...@en...>
> wrote:
> > I'm using matplotlib 1.0.1.  I have the following simple script to plot a
> > surface:
> >
> > -----
> > from numpy import linspace, sin, cos, meshgrid
> >
> > from matplotlib.pyplot import figure, show, xlabel, ylabel
> > from matplotlib import cm
> > from mpl_toolkits.mplot3d import Axes3D
> >
> > n = 35
> > x = linspace(-5, 5, n)
> > y = linspace(0, 10, n)
> > X, Y = meshgrid(x, y)
> > Z = X*sin(X)*cos(0.25*Y)
> >
> > fig = figure()
> > ax = fig.gca(projection='3d')
> > ax.plot_surface(X, Y, Z, rstride=1, cstride=1, cmap=cm.copper)
> > xlabel('x')
> > ylabel('y')
> > show()
> > -----
> >
> > It works fine--many thanks to all the folks working on the 3D plots!
> >
> > But notice that Axes3D is imported from matplotlib.mplot3d, but never
> > explicitly used.  If I comment out that import, however, I get the
> following
> > traceback:
> > -----
> > Traceback (most recent call last):
> >   File "surf_demo.py", line 15, in <module>
> >     ax = fig.gca(projection='3d')
> >   File
> >
> "/Library/Frameworks/Python.framework/Versions/7.0/lib/python2.7/site-packages/matplotlib/figure.py",
> > line 965, in gca
> >     return self.add_subplot(111, **kwargs)
> >   File
> >
> "/Library/Frameworks/Python.framework/Versions/7.0/lib/python2.7/site-packages/matplotlib/figure.py",
> > line 675, in add_subplot
> >     projection_class = get_projection_class(projection)
> >   File
> >
> "/Library/Frameworks/Python.framework/Versions/7.0/lib/python2.7/site-packages/matplotlib/projections/__init__.py",
> > line 61, in get_projection_class
> >     raise ValueError("Unknown projection '%s'" % projection)
> > ValueError: Unknown projection '3d'
> > -----
> >
> > Is this expected?
>
> Yes, the last lines of axes3d.py, which get called when Axes3D is imported,
> are:
>
> import matplotlib.projections as proj
> proj.projection_registry.register(Axes3D)
>
> which is what lets matplotlib know about the '3d' projection. Just
> importing the Axes3D module therefore performs the necessary
> registration of the projection to be able to use it in subsequent
> code. Code analysis tools like rope don't know about this behind the
> scenes stuff though, so it looks like a redundant import to them.
>


OK, thanks.

Warren


>
> Angus.
> --
> AJC McMorland
> Post-doctoral research fellow
> Neurobiology, University of Pittsburgh
>