I'm still exploring ideas at this point, so all ideas are welcomed. basically, I want a graph that: 1. has constant memory usage 2. loads only the data needed. Running in monitor mode, the graph should only have the data for the current time block + 1 or 2 page(s) buffered. this way as the user scrolls left to view earlier results, the graph would call the collector to read additional blocks from the file. 3. the collector may write previously loaded blocks to a temp directory. this is low priority and is more of an optimization in my mind. 4. when the mouse button is released, the graph should go back to the current results when a sampler returns new results. 5. to show just the desired block, I was thinking of implementing clipping to only render the visible area. 6. the collector would provide methods like: read(blockSize), readforward(blocksize), readbackward(blockSize), readforward(size,numberOfBlocks), readbackward(size,numberOfBlocks). If we made an abstract class to handle the common stuff, and then have a default lightweight collector that should handle the normal result graphs. the new collector would extend the abstract class, providing the new methods using channels. peter lin
"BAZLEY, Sebastian" <[EMAIL PROTECTED]> wrote:Screnshot looks good. Perhaps worth considering a lightweight stand-alone data collector, thet could store the data in a file for later analysis? Also, how many XML parsers are we now using? Should/Can we rationalise these? S. --------------------------------- Do you Yahoo!? Yahoo! Search - Find what you�re looking for faster.
