Wrong results when slicing a symmetric sparse tensor in python lib

[nickirk]

Dear Edgar,

As the title suggests, I provide a minimal example to reproduce it. I am using the latest commit on master: commit c4f89dcfc6fce48b6d8da45204c9c309f3900a8f

import ctf
import numpy as np
print(ctf.__file__)
vals = [1,2,3,4,5,6]
indices = [0,1,3,4,5,7]

s = ctf.tensor([2,2,2,2], sym=[1,0,1,0], sp=1)
s.write(indices, vals)
d = ctf.tensor([2,2,2,2], sym=[1,0,1,0], sp=0)
d.write(indices, vals)

print(s)
print(d)
assert(np.array_equal(s.to_nparray(),d.to_nparray()))

d_1 = d[:,:,:,:1]
print(d_1)
s_1 = s[:,:,:,:1]
print(s_1)
assert(np.array_equal(s_1.to_nparray(),d_1.to_nparray()))

One can see that d_1 is as intended, while s_1 is wrong. One can also test that as long as we don't use symmetries in the tensor, the slicing works fine for both dense and sparse tensors:

d_ns = ctf.tensor([2,2,2,2], sym=[0,0,0,0], sp=0)
d_ns.write(indices, vals)

d_ns_1 = d_ns[:,:,:,:1]
print(d_ns)
print(d_ns_1)

s_ns = ctf.tensor([2,2,2,2], sym=[0,0,0,0], sp=1)
s_ns.write(indices, vals)

s_ns_1 = s_ns[:,:,:,:1]
print(s_ns)
print(s_ns_1)

I ran the examples on 1 core only.

If you need further information, please let me know.

Thanks in advance, Ke