[Matplotlib-users] Spurious borders when setting figure size

[Christoph G. <chr...@gr...>]

Hello,

The script pasted below plots a square array.  The (very small) output
PDF is attached to this posting.  For reasons I do not understand,
there's a fine additional border immediately at the top and bottom of
the array.

If the commented-out line of code is removed, the strange border
disappears.

Now you will ask, why do I create the figure in this way and don't
simply use the commented-out line?  The answer is that this script is a
simplified version of the function "map" from
http://git.kwant-project.org/kwant/tree/kwant/plotter.py?id=v1.0.0
There, we need to create a figure also when pyplot has not been
imported.

If we would import matplotlib.pyplot, the user of our package could not
freely choose the matplotlib backend himself.

Thanks,
Christoph

****************************************************************

import numpy as np
from matplotlib import pyplot
from matplotlib.figure import Figure
from matplotlib.backends.backend_agg import FigureCanvasAgg

data = np.random.random((11, 11))

fig = Figure()
ax = fig.add_subplot(1, 1, 1, aspect='equal', adjustable='datalim')

# Uncommenting the following line makes it work.
# fig, ax = pyplot.subplots()

image = ax.imshow(data, extent=(-49.5, 49.5, -49.5, 49.5),
                  interpolation='none')
fig.canvas = FigureCanvasAgg(fig)
fig.set_size_inches(5, 5)
fig.savefig('test.pdf')