Sptensor constructor breaks with int overflow for large tensor

[jeremy-myers]

Issue

The constructor for sptensor fails an assertion when the tensor size is large. It's important to note that the shape isn't allocated, simply checked for consistency.

Example

For size = (50000000, 50000000, 100, 50000) and nnzs = 1000000000, we get:

    X = ttb.sptenrand(sz, nonzeros=nnz)
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/jermyer/pyttb/pyttb/sptensor.py", line 3709, in sptenrand
    return ttb.sptensor.from_function(unit_uniform, shape, valid_nonzeros)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/jermyer/pyttb/pyttb/sptensor.py", line 214, in from_function
    assert False, (
AssertionError: Requested number of nonzeros must be positive and less than the total size

Checking np.prod(size) used in the assertion, np.prod(size) = -6892481975075995648 .

Solution

Use prod from the math module to get the correct result, i.e.,

>>> from math import prod
>>> sz = (50000000,50000000,100,50000)
>>> prod(sz)
12500000000000000000000

This has the added bonus that sz is a python tuple, not a numpy object, as pointed out by @dmdunla.