I build a fun application using pier. It is still small and very experimental. It's about wine.
About wine, that's interesting
And spatial/geographical information is important for the site. Quadtrees are extremely useful for geographical applications. Usually the world is divided into a mesh of equal distant squares. This squares map onto the deepness levels of a quadtree. Tile graphics like google maps are managed in a similar way. This way round it is easy to find important spots (called POI = points of interest) fast. You just need to convert latitude/longitude in x/y dimensions and with an additional parameter of zoom a quad tree gives you everything you wanna know. For the opposite direction (finding areas of most interest without having discrete dimensions) R-Trees are pretty useful. They are not balanced by area they occupy but balanced through the amount of data that is in or near to a certain spot.
That is interesting. I would love to have a reliable and fast quadtree implementation. If there is anything I could try out, let me know. I would love to.
Cheers, Alexandre