Unittest that mixes types of DataContainer, what should happen?

[paskino]

Description

This unittest mixes data types with sapyb:

CIL/Wrappers/Python/test/test_DataContainer.py

Lines 1229 to 1238 in f8f2e31

	ig = ImageGeometry(10,10)
	d1 = ig.allocate(1, dtype=np.complex64)
	d2 = ig.allocate(2, dtype=np.float32)
	a = 2.+1j
	b = -1.
	# equals to 2*[1] + -1*[2] = 0
	out = d1.sapyb(a,d2,b)
	res = np.zeros_like(d1.as_array()) + 1j
	np.testing.assert_array_equal(res, out.as_array())

As you can see we allocate a complex64 (d1) and a float32 (d2) DataContainer then do sapyb. d1 is multiplied by a complex number a and d2 by a real number b whilst executing a*d1+b*d2. In this case we are doing sapyb via numpy.

1. What should happen at line 1236?
2. Do we have to do implicit type promotion?
3. Are we expecting the unittest to pass? At line 1237 we change the result of sapyb and add the missing 1j so that the unit test pass.

With my new changes to make CIL's DataContainers agnostic on the array they wrap this test fails.

Environment

import cil, sys
print(cil.version.version, cil.version.commit_hash, sys.version, sys.platform)

[paskino]

Actually, in CIL we pass the data as is to numpy, so it's numpy which decides about type promotion etc