combine sage plot with matplotlib

edit

The correct answer to this question is Trac 5128. Unfortunately, this has not yet been finished. But once this is done, what you want will be possible.

It's certainly possible in principle, but, as you have seen, it's quite annoying in practice. The main reason is that Sage plots are NOT really mpl plots until we actually save or show them as graphics files; until then they are Sage-y Python objects. Look at the source code file for some plot type, under the _render_on_subplot method for where mpl really is used.

Any contribution to this patch would be very welcome, by the way!

For some time, I was looking for a way how to generate several sage Graphics() objects and plot them on a matplotlib canvas in an arbitrary arrangement, using .matplotlib() function. It went out not being straightforward. I decided that the solution I've found may be interesting for others as well. The following code produces a plot with a nice inset:

#make some graphs
x=var('x')
g=plot(sin(x))
g_ins=plot(cos(x))
# plot main figure
from matplotlib.figure import Figure
figure = Figure() # you may set the size here
main_plot = figure.add_axes((0.2,0.2,0.7,0.7))
g.matplotlib('a.svg', figure=figure, sub=main_plot)
# plot an inset
inset = figure.add_axes((0.6,0.2,0.3,0.3))
g_ins.matplotlib('a.svg', figure=figure, sub=inset)
# display graph (note that only single sage Graphics object has to be saved )
g_ins.save('a.svg', figure=figure, sub=inset)

UPD: if figures are drawn strangely, add

aspect_ratio='automatic'

to matplotlib() parameters. By default it is 1.0, which may be undesired.

Also, you may want to draw the figure itself, not by Graphics().save() function. Replace the last line with the following:

from matplotlib.backends.backend_agg import FigureCanvasAgg
figure.set_canvas(FigureCanvasAgg(figure)) 
figure.savefig('a.svg')