Am 30.12.2008, 19:39 Uhr, schrieb Christopher Barker <[email protected]>:
> hmm. I wonder if that's the best approach, or if I should create a Null > bounding box object.... I faced the same problem in fc2 and my temporary solution is to disable the use of empty groups (that is, no children) by raising an exception. Nodes with no bounding boxes need special treatments in different places, e.g. if you're currently doing something like node.boundingBox.center + ( 10, 0 ) you will run into an error with the proposed new NoBoundingBox class. You cannot return (0,0) for noBoundingBox.center, because if you merge a bounding box with center (0,0) and size (0,0) with another bounding box, the resulting bounding box will now probably be enlarged and contains (0,0). Since objects with no bounding box need different treatment in a variety of situations I just disabled them for now instead of writing the NoBoundingBox class (or removing the boundingBox property from such nodes altogether). In fc2 all nodes classes are subclasses of from NodeWithBounds which specifically assumes that a node has bounds. I think nodes without bounds should have a different superclass. Then the rendering/culling/... other code that specifically relies on the node to have a valid bounding box ... can work only on the nodes which actually have a bounding box. While all this can be done it didn't seem to justify the extra complexity in implementation and testing for the time being. Objects with no bounding box are not really useful (except during their construction; fc2 allows for empty groups and only barks when such objects are used in rendering or culling ). -Matthias _______________________________________________ FloatCanvas mailing list [email protected] http://mail.mithis.com/cgi-bin/mailman/listinfo/floatcanvas
