Fix Numpy-related deprecation warnings

See merge request !397

Anton Akhmerov authored Aug 30, 2020 and Viacheslav Ostroukh committed Jun 22, 2021

Closes #386

Sympy 1.7 broke alias `sympy.numbers` and `sympy.core.numbers` should be used
instead. Second alias, however, can't be used in eariler Sympy version. This
fix uses one of them, dependent on what attribute is present.

Backport Matplotlib fixes from master

See merge request !395

Viacheslav Ostroukh authored Jun 14, 2021 and Viacheslav Ostroukh committed Jun 18, 2021

`renderer` argument of mplot3d's `do_3d_projection()` is deprecated in
Matplotlib 3.4. This commit switches from using (also deprecated) `renderer.M`
to `self.axes.M`.

A warning about `renderer` being deprecated is still emitted, but we still
need to keep `renderer=None` statements for reverse compatibility.

The issue is in Matplotlib. This is a temporary solution before [1] is merged.

Fixes #407.

[1] https://github.com/matplotlib/matplotlib/pull/20416

Transpose array-like in herm_conj

See merge request !387

If value passed to `herm_conj()` function can't be conjugated as an
array or a number, attempt to convert it to a tinyarray is performed.
In that way array-like list of lists/tuples can also be conjugated
correctly.

Fixes #406

If onsite/hopping value was supplied to the builder as array-like instead of
proper array, Hamiltonian could be generated silently without any Hermitian
conjugation, if required.

To expose this, `test_builder.py::test_hermitian_conjugation` is
parametrized with different types of values to ensure that Numpy arrays,
tinyarrays, lists and tuples are all conjugated properly.

See issue #406.

Merge CI fixes from master

See merge request !392

This is done instead of picking all changes that were commited to master
and not reflected in stable.

Current master revision is 8a6d320b.

Builder.substituted doesn't change the builder in place, contrary to
what the docstring suggested.

Closes #379

Joseph Weston authored Mar 16, 2020 and Anton Akhmerov committed Mar 30, 2020

Tinyarrays used to have identical hashes to tuples, but this
was not part of the official interface. While the core of
Kwant did not depend on this implementation detail, the
tests did.

update libraries for linking against Mumps in conda-forge

See merge request kwant/kwant!355

Recently conda-forge updated their packaging so that sequential
Mumps is a separate package,
[mumps-seq](https://anaconda.org/conda-forge/mumps-seq),
which names the libraries with a suffix '_seq' (contrary to how
Debian does it).
In addition we no longer need to explicitly link against Scotch,
pord or gfortran as the conda-forge built mumps libraries already
do this.
All tests still pass without explicitly linking against mpiseq,
so it is left out.

Fix behaviour of discretizer when locals contains numpy array values

See merge request !344

Incraese the timeout and number of retries when running the
link check on the documentation; this should increase the
specificity of the test.

See merge request !332
Closes #332.

The combination of longer timeout and larger number of retries should
make the linkcheck sufficiently specific again.

Retrying > once should increase the specificity of the check a lot.

Fixup/qsymm tests

See merge request !330

sympy symbols with 'real' set differently do not hash to the same
value (and possibly do not compare equal).

__eq__ for Models just uses allclose anyway.

This is more idiomatic.

Also comment other usages of this parameter.

Qsymm is now installed in the testing image directly