unexpected behavior of graphics-driven assignment about quibbler

[kmaork]

@rkishony right now we don't assign two quibs from graphics when one of them depends on the other, but only to the source quib, to prevent contradictory changes to a quib.

But in cases like your example, I guess that we should not view overriding as a contradictory change.

My new approach would be to collect the overrides that will be caused by assignment made by the graphics event, and discard assignments just when they cause an override in the same quib.

The problem with that approach is that to know what overrides we might perform, we might have to ask the user to choose, which would be weird to do when some of these overrides will not be performed.

An example where this is relevant:

a = iquib(1)
b = a + 1
c = a + 2
b.allow_overriding = True
c.allow_overriding = True
plot(b, c)

Whet the plot is dragged, I'd the user chooses to override both b and c at the iquib level, their assignments will be contradictory and we will have to choose only one.

Maybe as a very nice-to-have feature in the future, if the user chooses to override the first at the iquib level, we could change the dialog for c to offer only self override or skipping. But for now I think the approach I offered is good enough.

[rkishony]

ok, but it will help if in addition you also make sure that the plot quib does not issue an override in the X position if the marker was dragged only on Y (if X didn't change) and does not issue an override in the Y position if the marker was dragged horizontally

this will allow more logical dragging along constrained curves such as in the plot(b,c) example you have

[kmaork]

I'm not sure that we should do that. If we override only one axis when the diff is in the other axis, the change might still propagate through the quibs into the "unchanged" axis, and the dot we dragged will end up in a different position from the mouse. I think when dragging a point, we logically assign to it both x and y, regardless of its previous location.