This data structure optimizes storage of big condensed filled areas.
It's suboptimal for storing very sparse non-condensed data as overhead of having multiple subnodes is quite high.
It works similar to https://www.youtube.com/watch?v=NfjybO2PIq0 except it uses lazy initialization for subnodes.
Download sample app (.NET 4.5) - https://drive.google.com/open?id=0B4-tSq-u4CrecjRwZXlOc3IyZms
Coding: Vladimir Kozlov, AtomicTorch Studio http://atomictorch.com License: MIT (see License.md)