Increased speed by caching all possible selections #1
Conversation
|
Thanks! I'm just about to go out of town and need to look at this more closely: part of my goal with this code is to keep it very simple for a talk I'm giving. But I like the idea of caching, will give it proper attention in about 10 days. |
|
Felt more productive to share this quick experiment than to simply throw it out. This is just one of many different ways to increase speed (DOM tree would be another I guess). |
|
Hey, I finally had time to look at this code. Thanks so much for contributing it! I'm not going to merge it though, because I chose to improve performance a different way. Your way is actually faster, but the main purpose of this code is a tutorial for D3 and I want to keep it as simple as possible. FWIW, it turns out the source of slowness in my code was calling .style("fill", ...) on all 40,000 dots. Your code only sets the fill on the dots that are actually changed, which is why it's so much better. The route I took in f5f3f92 was to use .attr("class", ...) on all 40,000 dots to set the colors instead of .style(). Some more discussion here: http://nelsonslog.wordpress.com/2013/03/16/optimizing-zipdecode/ |
|
I agree that G tree and styles is the most structural way to do this and much better suited for a tutorial. |
This quick hack avoids having to go through all the rectangles and focuses on two groups of nodes: the last one selected (need to turn color back off) and the new selection. All possible selections have been cached in the node object.