On Thu, 3 Apr 2008, Daniel Hulme wrote:
On Thu, Apr 03, 2008 at 10:38:24AM -0700, david@...2429... wrote:
the answers I'm seeing to this question don't seem to match the question.
Bulia's original answer was correct and pertinent. Because not all transforms can be folded into the circle's parameters, transforms are handled in general by editing the setting the transform rather than by changing the circle's parameters. It would be extra effort to handle some transforms differently, and nobody has ever found a need for doing so.
but there is a difference between doing a transform to change the shape and simply moving the object.
otherwise you shouldn't have x,y in the object definition at all, create all objects at 0,0 and use transform to move them to where they belong.
I have seen systems that represented all circles/spheres as a unit circle/sphere about the origin with the appropriate transform, and others that folded transforms into parameters as much as possible, and ditto for rectangles/cuboids. The first method is advantageous if it is important to you that equal things have the same representation; the second is advantageous if having the transform on an object be the identity gets you extra space or time efficiency. Otherwise, there is not much to choose between them.
but what's happening here sounds like it's neither. you give the object a position and size AND you use a transform to move it elsewhere.
I agree that either one can be the right thing to do at different times, but the combination of them seems wrong.
David Lang