+ add example for scatter plot
+ link to ken-ichi
Figuring how to make a client side image map from a matplotlib image has stumped me more than once. Andrew Dalke does have a working example. Below, I have the minimal example.
It's simple once you get the steps right: just use mpl's transform() to convert the data into the image's coordinate system. Then flip the y-axis as required by the imagemap, then do the normal imagemap stuff and save the html. The only real gotcha, is to make sure to put the dpi in the call to savefig().
name = 'imap'
# make some fake data
xs = range(15)
ys = [random.choice(xs) for i in range(len(xs))]
xys = zip(xs, ys)
# can also use : f = pylab.subplot(121)
f, = pylab.plot(xs, ys, 'ro')
dpi = f.figure.get_dpi()
height = f.figure.get_figheight() * dpi
# convert the x,y coords into image coords.
transform = f.get_transform()
icoords = transform.transform(xys)
# the minimal 'template' to generate an image map.
tmpl = """
<img src="%s.png" usemap="#points" border="0">
# change this as needed, e.g. if not plotting points.
fmt = "<area shape='circle' coords='%i,%i,2' href='http://example.com/%i/%i' >"
# need to do height - y for the image-map
fmts = [fmt % (ix, height - iy, x, y) for (ix, iy), (x, y) in zip(icoords, xys) ]
# NOTE, this dpi is needed!
pylab.savefig('imap' + '.png', dpi=dpi)
print >> open(name + ".html", 'w'), tmpl % (name, "\n".join(fmts))
When trying to figure how to do this for a pylab.scatter plot, I found Ken-ichi had also done this for a scatter plot.
As of a matplotlib trunk revision 5711, the transform does not get set when the scatter plot is drawn. To set it, I had to set the transform explicitly:
s = pylab.scatter(xs, ys)
transformed_xys = s.get_transform().transform(zip(xs,ys))