On Fri, Feb 21, 2003 at 09:50:06PM -0800, Daniel Rogers wrote: > This todo items is a little unclear to me. Why do interior node inputs > need to know anything about the inputs of the Graph? No. You are right the interior nodes dont need to know exactly, but there needs to be a better way to map the inputs to the Graph to which inputs of the interior nodes they belong to. For example right now, if one interior op (say A) has two inputs, one image (ImageData) and one scalar (ScalarData), and another interior op (say B) has just one input, an image (ImageData), then the inputs to the Graph get set like this (when the root C of the graph is set): -------------- | C | <--- graph | / \ | | A B | | / \ | | | | -------------- | | | D E F input 0 (D) to the Graph is the image input for A, input 1 (E) is the scalar input for A, input 2 (F) is the image input for B It just seems like the Graph should have an additional table or something that lets you set your inputs to whichever order you want, so that for example you can put all the image inputs at the beginning. This TODO addresses issues like that really. You are right A, B on the interior have no need to know about what order the inputs to the Graph are set up. Calvin
