Allow vector assignments

>>>import pystencils as ps
>>>import sympy as sp
>>>a, b, c = sp.symbols("a b c")
>>>ps.Assignment(sp.Matrix([a,b,c]), sp.Matrix([1,2,3]))
(Assignment(a, 1), Assignment(b, 2), Assignment(c, 3))

Fixes #17 (closed)

changed the description

added 1 commit

• 38da1c39 - Fix #17 (closed): Allow vector assignments

Compare with previous version

It's funny that isinstance(field_access, Iterable) is true but it does not support iteration

added 1 commit

• c98ee84d - Assert same length when performing vector assignment

Compare with previous version

added 1 commit

• 1c57a059 - Assert same length when performing vector assignment

Compare with previous version

I assume this can only be used through an AssignmentCollection, which then flattens the list of lists of Assignments? So something like

a = [ps.Assignment(field.center_vector, [0,0,0])]
ps.create_kernel(a)

still wouldn't work, only

a = ps.AssignmentCollection([ps.Assignment(field.center_vector, [0,0,0])])
ps.create_kernel(a)

would. I guess all functions that accept a list of Assignments need to be made aware of vector assignments or we need to forbid the use of lists of assignments and use AssignmentCollections everywhere.

I would consider vector assignments an absolute hack. People who use it should expect them to work everywhere.

We have to many cases where either a list or an AssignmentCollection is required. I don't want to introduce more special cases. I'd deprecated the usage of list in more places.

Btw. This should work and it is also more natural if constructing an Assignment from a vector yields a tuple

a = ps.AssignmentCollection([*ps.Assignment(field.center_vector, [0,0,0])])
ps.create_kernel(a)

I only added the flatting in AssignmentCollection to make them work with dicts

AssignmentCollection({
field.center_vector: [0,0,0]
})

where you cannot unpack.

changed title from Fix #17: Allow vector assignments to Allow vector assignments

changed the description

merged

mentioned in commit 46b7b88d