Adding Shapes gives counterintuitive results

[awygle]

Shapes implement __add__, but adding them is currently not useful:

>>> from nmigen import *
>>> x = Signal(16)
>>> y = Signal(32)
>>> s = x.shape() + y.shape()
>>> s
(16, False, 32, False)

The expected result here (I argue) would be (48, False), but (probably because Shape is a named tuple under the hood) we get the not-very-useful result shown above.

Obviously in this case you can do (x+y).shape() but that's not always easy or possible (I hit a particular case involving an Enum)

[whitequark]

So there are two unrelated issues here. First, that currently you can do shape + shape and the result is never desirable. We should fix that by not using namedtuple while still allowing to destructure shapes using width, signed = shape. (Is that possible? I think we can override __iter__ to do it...)

Second, the idea of overloading the + operation on shapes to do something not really related to addition. It doesn't seem like a good idea to me, but we can discuss it in a different issue.

[awygle]

I agree it's a bad idea. We should stop implementing extraneous operators for Shape, and that's all.

[awygle]

Should we preserve comparing Shapes to tuples of (int, bool)? This currently works (and is used in the test suite) but feels like a bit of a footgun.

[whitequark]

This currently works (and is used in the test suite)

Where?

[whitequark]

Oh sorry, you mean comparing for equality, not for ordering. Yeah let's preserve that, we can always deprecate it later if it turns out to cause issues.