bh.linalg.eigh is broken

[jamesavery]

The standard NumPy eigensolver bohrium.linalg.eigh isn't implemented, and it tries to use numpy. This is not the problem. However, instead of handing it gracefully off to NumPy, it crashes:

import bohrium as bh;

# Symmetric real matrix
A = bh.array([[1.,2.,3.],
              [2.,4.,5.],
              [3.,5.,6.]]);

lam,eig = bh.linalg.eigh(A)

blows up with

/home/avery/.local/lib/python3.6/site-packages/numpy/linalg/linalg.py:1456: UserWarning: Bohrium does not support ufunc `eigh_lo` it will be handled by the original NumPy.
  w, vt = gufunc(a, signature=signature, extobj=extobj)

AttributeError                            Traceback (most recent call last)
<ipython-input-28-1bc70c82985b> in <module>
     10 
     11 
---> 12 bh.linalg.eigh(Us[4])

~/.local/lib/python3.6/site-packages/numpy/linalg/linalg.py in eigh(a, UPLO)
   1454 
   1455     signature = 'D->dD' if isComplexType(t) else 'd->dd'
-> 1456     w, vt = gufunc(a, signature=signature, extobj=extobj)
   1457     w = w.astype(_realType(result_t), copy=False)
   1458     vt = vt.astype(result_t, copy=False)

bohrium/ufuncs.pyx in bohrium.ufuncs._handle__array_ufunc__()

AttributeError: module 'numpy' has no attribute 'eigh_lo'

However, just using NumPy works fine:

import numpy as np;

# Symmetric real matrix
A = np.array([[1.,2.,3.],
              [2.,4.,5.],
              [3.,5.,6.]]);

lam,eig = np.linalg.eigh(A)

and if you do a copy2numpy() first, everything also works fine:

import bohrium as bh;

# Symmetric real matrix
A = bh.array([[1.,2.,3.],
              [2.,4.,5.],
              [3.,5.,6.]]);

lam,eig = bh.linalg.eigh(A.copy2numpy())