从3D图形获取对象的最有效方法

Question

So, I am making a game, and I have a problem. I need some way to efficiently get the objects that are located in the 3d graph.

However, the problem is that they all can cover more than one value to get. So, for example, the value 'foo' might cover the x range 400-500, y range 10-20 and the z range 30-60.

I am wondering whether there is a more efficient way to store and get these then by just storing them in a Array and searching through them, something I would prefer not to do, because of the large amount of these values.

This is for selecting biomes for a game. However, I chose to put this here versus GameDev because this seemed to be more of a data structures question. I might be wrong however, and if I was would gladly close this and open one there.

Is there any other way to do this?

public static Tile getTile(int height, int moisture, int temperature) {
    for(Tile tile : tileList){

        boolean isTile = true;

        if(!(tile.heightMax > height && tile.heightMin < height)){
            isTile = false;
        }

        if(!(tile.temperatureMax > temperature && tile.temperatureMin < temperature)){
            isTile = false;
        }

        if(!(tile.moistureMax > moisture && tile.moistureMin < moisture)){
            isTile = false;
        }

        if(isTile){
            return tile;
        }
    }
}

This is my searching method. As you can see, I am using a list to store all of the tiles, and searching for the one that fits the criteria. Is there any other way to do this with many tiles as fast as possible?

Answer 1

The way I've solved this issue in the past is to define regions that represent an area of 3D space and then store a map to each object from all regions an object's space overlaps. Then when you want to find objects at a location you first use the map to find all objects that overlap that location's region before iterating through them looking for objects at that location.

Let me give you a code view of that. To build the region map:

Map<Region, List<Item>> regionMap = new TreeMap<>();

When an item is added:

item.getOverlappingRegions().forEach(region -> {
    if (!regionMap.containsKey(region))
        regionMap.put(region, new ArrayList<>());
    regionMap.get(region).add(item)
});

Then when you're searching for an item at a location:

if (regionMap.containsKey(location.getRegion()) {
    regionMap.get(location.getRegion()).stream()
        .filter(item -> item.contains(location))
        ....
}

There's definitely overhead in this but it is very fast for searching if your regions are sufficiently small.

This is designed for 3D spaces but it can be expanded to as many dimensions as you want: it is all in the implementation of the Region and Location classes.

Answer 2

I am not entirely sure that I have understood what your problem is, but it seems very similar to a collision detection algorithm in 3D space. If this is the case, Octrees might be what you are looking for: http://www.codeproject.com/Articles/108761/Octrees