brillouin_zone -> scipy Voronoi segfaults on macos 10.10.5 about pybinding HOT 3 CLOSED

Hi Miša
looks like a macos 10.10 problem, closing -- mttiw, more trouble than it's worth.

An unrelated question, model.hamiltonian seems to be always np.float32, is that intentional ?
It makes quite a difference, check |AV - VΛ|_max

from pybinding.

Hey Denis,

This indeed sounds like something that can be related to the scipy issue you pointed to. Do the suggestions in the comments help to resolve this issue?

The latest release was tested on macos 10.11, with different versions of python (3.6, 3.7, and 3.9).
What you could do is to take a look at the test configs on travis, and try to replicate the environment for example from this run. You can find more info about the scipy version bellow line 282.

conda install cmake numpy scipy matplotlib in your good install doc
didn't help -- how does that differ from just
pip install numpy scipy matplotlib pybinding ?

the two would install packages from different repositories, so even if you would specify the same version, you would install a different package.

from pybinding.

Hi Denis,

By default the precision is float. If there are no modifiers that would require different type of Hamiltonian, it will stay like that (e.g. adding magnetic field will promote to complex).

On the other hand you can choose (force) higher precision by adding some of the predefined modifiers whose only purpose is to promote the precision. You can see more here

import pybinding as pb
from pybinding.repository import graphene

model_float = pb.Model(
    graphene.monolayer_4atom(),
)
print(model_float.hamiltonian.dtype)

model_double = pb.Model(
    graphene.monolayer_4atom(),
    pb.force_double_precision()
)
print(model_double.hamiltonian.dtype)

from pybinding.