It is more efficient to modify the graphics state, because that is
a dynamic operation. Part of the graphics state is a transform
which is applied to anything drawn into that state. When you use -
concat on an NSAffineTransform object, all you are doing is
modifying the graphics state transform; it has no immediate effect,
it just alters anything which is drawn after that by transforming
the drawing differently. On the other hand, when you apply a
transform to an NSBezierPath object, every set of coordinates
stored in that path are modified by the transform and replaced with
the new transformed coordinates. So it is more efficient to
combine all your coordinate system transformations so they happen
dynamically in the final draw operations into the graphics state,
where a transform is always going to occur anyway.
