Calvin Williamson wrote:
For the moment, assume they would have the same rect.
In the future, when Images are multi-tiled, then
the ImageData rect could describe some area that
might correspond to several tiles of an Image. In
that case some kind of iterator would hand back tiles,
and then there would be a scanline iterator for the
lines of each tile. Here these scanline iterators would
obviously be focused on rects smaller than the ImageData
rect. (Our current ImageIterator is sort of just this
scanline iterator)
Setting the ImageData rect basically just says
"this area will be needed" and they are computed
based on the roi that was originally passed at the root.
Am I correct then, in stating that the rect in ImageData refers to the
compRec refereced in "A Model for Efficient and Flexible Image Computing?"
If the two rects should be the same, then should I destroy the
ImageIterator associated with and ImageData when the ImageData rect changes?
But when will it change? It shouldnt change after the preprocessing
passes over the graph are made and the areas needed are computed.
Once evaluation begins it stays the same.
I didn't consider whether it would, only that it could. Since it should
never happen it doesn't hurt to destroy the iterator when the rect
changes. I'll do this since I know not what the future brings and it
seems correct.
--
Dan
