On 2010-09-01, at 18:39, Max Carlson wrote:
> On 9/1/10 8:22 AM, P T Withington wrote:
>> This fixes the reported bug, but I'm still not convinced that the math is
>> completely correct. Unless I misunderstand completely, I feel the following
>> two assertions should also hold in the supplied test case:
>>
>> assertEquals((yellowview.x+redview.x)*blueview.xscale,
>> redview.getAttributeRelative('x', blueview), "red x in blue")
>
> There's a bug in that one - you're not taking the blueview.x into account.
> This works:
> assertEquals((yellowview.x+redview.x)*blueview.xscale + (blueview.x),
> redview.getAttributeRelative('x', blueview), "red x in blue")
Disagree.
The contract of getAttributeRelative is to answer the question: "Given the
value of an attribute in the source view, what value in the reference view is
equivalent?"
So, I'm asking, given that redview has an x value of 40, what value of x _in_
blueview would be equivalent? Blueview's `x` is relative to the canvas, and
should not be part of the equation when I am comparing redview and blueview.
To put it another way, currently it is the case that:
redview.getAttributeRelative('x', blueview)
and
redview.getAttributeRelative('x', canvas)
return the same value, which can't be right.
>> assertEquals((yellowview.x+redview.x)/(- blueview.xscale),
>> blueview.getAttributeRelative('x', redview), "blue x in red")
>
> This one still fails, and I'm not sure why. Instinct sez this has been
> broken for a long time. Digging...
I think this is related to my question as to whether a view's scale affects the
view or only it's children. We seem to have an inconsistent implementation.
You implied that a view's scale should affect it, but if that were the case,
then blueview really should be at absolute x=20, since it has x=10 and
xscale=2, but it is at absolute x=10. OTOH, blueview has an absolute width of
200, but blueview.width=100. So, we've got a view's scale affecting its
width/height, but not its x/y.
Either we have to consistently have scale affect all/none dimensions of a view,
or we have to have a much more complex implementation of getAttributeRelative
(because it will have to factor out the shared ancestor's scale if you are
asking about x/y and factor it in if you are asking about height/width).
I think this inconsistency is perhaps derived from Adam only thinking about the
scaling of a view's resource dimensions and not considering how it should or
should not affect a view's position.
