Get proper renderer width and height in FigureImage - #9204

This definitely needs a test adding! Could you add your example in the bug report to lib/matplotlib/tests/test_image.py?

Changig to WIP - see below, but width, height = renderer.get_canvas_width_height() doesn't seem to know about dpi in fig = plt.figure(figsize=(2,2), dpi=100)...

@dstansby

For the tests, there is already the test below, but as you can see it is only set to check png.

@image_comparison(baseline_images=['figimage-0', 'figimage-1'], extensions=['png'])
def test_figimage():
    'test the figimage method'

    for suppressComposite in False, True:
        fig = plt.figure(figsize=(2,2), dpi=100)
        fig.suppressComposite = suppressComposite
        x,y = np.ix_(np.arange(100.0)/100.0, np.arange(100.0)/100.0)
        z = np.sin(x**2 + y**2 - x*y)
        c = np.sin(20*x**2 + 50*y**2)
        img = z + c/5

        fig.figimage(img, xo=0, yo=0, origin='lower')
        fig.figimage(img[::-1,:], xo=0, yo=100, origin='lower')
        fig.figimage(img[:,::-1], xo=100, yo=0, origin='lower')
        fig.figimage(img[::-1,::-1], xo=100, yo=100, origin='lower')

Unfortunately this turns up a continued error, and I'm not quite sure how to get around it...

If I modify the above to run from the command line:

import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
import numpy as np
for suppressComposite in False, True:
    fig = plt.figure(figsize=(2,2), dpi=100)
    fig.suppressComposite = suppressComposite
    x,y = np.ix_(np.arange(100.0)/100.0, np.arange(100.0)/100.0)
    z = np.sin(x**2 + y**2 - x*y)
    c = np.sin(20*x**2 + 50*y**2)
    img = z + c/5

    fig.figimage(img, xo=0, yo=0, origin='lower')
    fig.figimage(img[::-1,:], xo=0, yo=100, origin='lower')
    fig.figimage(img[:,::-1], xo=100, yo=0, origin='lower')
    fig.figimage(img[::-1,::-1], xo=100, yo=100, origin='lower')

    if suppressComposite:
        fig.savefig('Boo.pdf')
        fig.savefig('Boo.png')
    else:
        fig.savefig('Boo2.pdf')
        fig.savefig('Boo2.png')

For pdf rendering get_canvas_width_height() is returning 144, 144 instead of 200, 200 for the above so it is ignoring the figure specified dpi and using a default value of 72. If I save as png, it assumes that dpi=300 and returns 600, 600.

So, somehow the backends are getting the figsize, but they are not getting the dpi from the fig = plt.figure(figsize=(2,2), dpi=100) command, at least as reported by width, height = renderer.get_canvas_width_height()

OK, so the matplotlib.backends.backend_mixed.MixedModeRenderer, which gets used when making the pdf and the matplotlib.backends.backend_agg.RendererAgg return very different values of width and height, and that is the problem.

For renderer.width, RenderAgg returns fig.width * fig.dpi and MixedModeRenderer simply returns fig.width. So, this PR is almost certainly wrong to use the canvas method, but the two renderers should probably be made compatible...

@dstansby the test is the existing test_image.py:test_figimage with the svg and pdf filetypes added...

Todo: remove or fix the svg test...

Thanks! backported to v2.1.x as 052b52b