Thanks, John! I got JJ's approach to work, but I also tried it your way
and it seems to me to generate tidier code. You don't have to deal with
clip boxes either.
Example:
[code]
aspect = 23.0 / 8.0 # (These are my figure dimensions, X / Y)
tr = fig.transFigure
w, h = 0.008, 0.02
ha = FancyArrow(0.0175, 0.03, 0.645, 0.0, width=w, head_length=h, \
transform=tr)
va = FancyArrow(0.0125, 0.045, 0.0, 0.845, width=w/aspect, \
head_length=h*aspect, transform=tr)
fig.patches.extend([ha, va])
[/code]
So it would appear that the special methods add_axes and add_subplot
exist only so that the "stateful" Matlab-like interface can be informed
as to which Axes object is "current." For any other object type, you
just add it to the list of the appropriate type in the Figure object.
···
On Sat, 2011-09-17 at 19:19 -0500, John Hunter wrote:
The artist tutorial covers drawing directly to a figure
http://matplotlib.sourceforge.net/users/artists.html#figure-container
I believe you could adapt the patches.FancyArrow to the same approach.