Previously we used the matplotlib default behavior of setting
the colormap limits (vmin and vmax) to the limits of the plotted
data, if they are not provided by the user. Now we set the
colormap limits to the 2nd and 98th percentile of the input
data, and stretch the limits by 10% on either end in an attempt
to include all the data in the limits. If we cannot do so
we set the limits to the percentiles plus the stretch
and issue a warning.

Also, whenever the data falls outside the colorbar limits
(even when the latter are set by the user) we set the appropriate
ends of the colorbar to be pointy, indicating that the data
extends beyond the colorbar.

Closes #183

Closes #171

Decrease padding to half the bump width and add test that the
current goes to zero at the boundaries of the region

See merge request kwant/kwant!211

This checks that the padding around the counding box is sufficient.

The bump function we convolve with has finite support,
so the vector field is identically zero at distances
greater than width/2 outside of the bounding box.
Also replace this magic number with a 'padding' constant
to improve readability just a smidge.

This can always be resurrected from Git if it is really needed,
but at the moment its just 3000 lines that aren't used.

speed up kwant import from >1s to 0.3s

Closes #181

See merge request kwant/kwant!207

Previously we used the ExtensionUnavailable proxy that would
raise an exception when accessed. Now because kwant.continuum
is loaded lazily we can just raise an exception. We choose to
re-raise ImportError with a better error message.

'matplotlib', 'Figure', '_colormaps', 'mpl_available', 'mplot3d',
'has3d', 'LineCollection', 'PathCollection', 'Line3DCollection',
and 'Path3DCollection'.

We will need to issue deprecation warnings when using this with
'kwant.continuum', so we have to roll our own so we can hook into
the attribute access. When we no longer need this feature we
can replace this with a simpler implementation using
'importlib.util.LazyLoader'.

Without this the test for whether importing the plotter chooses
the matplotlib backend fails when matplotlib is not installed.

This was forgotten in !199.

Thanks to Tibor Sekera for suggesting this feature.

remove use of 'vector_factory' when calculating spectral bounds

Closes #184

See merge request kwant/kwant!201

Pablo Piskunow authored Jan 31, 2018 and Joseph Weston committed Feb 21, 2018

Instead, use a random phase vector. Now 'vector_factory' is used
exclusively to generate the vectors used in the KPM calculation
itself.

Matplotlib backends like PDF cannot display arbitrarily thin lines.
That's why kwant.plotter.streamplot simulates thinner lines by blending
them into the background.

Previously if no lines were blended, i.e. all had the same color value
1, the background color would be used for drawing them instead of the
line color.

Bas Nijholt authored Jan 10, 2018 and Anton Akhmerov committed Jan 12, 2018

Closes #182

Closes #182

version 1.3.2

This commit fixes the bug in phs_symmetrization when it fails to produce ph eigenstates
with PH^2 = +1 when one of the wave functions is mapped exactly onto another one by PH,
as in this case the PH symmetrized wf's from the two are identical. This is the case
with wfs = np.eye(2) and  particle_hole = sigma[1], the second vector returned is an
eigenstate with eigenvalue -1 not +1. In general this can be even worse, if there are
multiple such vector pairs some vectors may not be ph eigenstates at all.

The new code uses an improved version of the matrix square root method from
Applied Mathematics and Computation 234 (2014) 380-384.

Tests are modified to clear up the random unitary generation process (using kwant.rmt).
Tests are now guaranteed to feed orthonormal set of vectors to phs_symmetrization().
Add extra tests for the fully off-diagonal case where the old method fails.

Previously, in order to not to fix the matplotlib backend, we required
users to import matplotlib.pyplot before calling any Kwant plotting
functions.  This did not have the desired effect, since we did import
`matplotlib.backends` and that also fixes the backend.

Now, both backends and pyplot are imported at the last possible moment
and a warning is emitted if this fixes the backend.