> Raster image = bitmap. As you know, bitmaps have the ability, no > matter whether you render 1 node or 1-billion nodes and edges on it, > it does not take up much more space.
It depends on the size of the bitmap :-) > But because Mondrian builds an object-model in RAM (from which it > renders), there is a definite scalability limit in terms of the > number of nodes. Rendering also takes more time due to > object-allocation, etc. In Mondrian, each object is rendered via a shape. There is a number of shapes, one of them is MOImageShape, which displays a bitmap. > The only reasons I can think to need to create a full domain model of > the graph is: > > 1) to calculate positioning > 2) for user-input; being able to drag the nodes around (thus > updating the Mondrian domain model, for subsequent display updates). I am not sure to understand. Suppose you have a domain (e.g., cars, persons and ownership). You can render this domain the way you want. Dragging nodes does not changes your models. It changes the graph built by Mondrian (i.e., instances of MONode and MOEdes). > (1) _might_ be able to be accomodated by, when generating the picture, > only keeping track of various node counts at each level and/or overall > bounds of the picture. Mondrian has a cache mechanism that does exactly that. A node may contains elements. A bitmap is created for you to speed the things up. > (2) is what I'm arguing might not be a good trade-off in all cases. > Sure, the output is now a static picture instead of something with > draggable boxes, but at the cost of speed and scale. > > Does that make sense? I find Mondrian interesting, thanks for your feedback. You can drag and drop inner nodes, the caches is reset without any particular action. What is exactly the problem you're trying to solve? Alexandre -- _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: Alexandre Bergel http://www.bergel.eu ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;.
