Search for nearest neighbors with bounds on distances along different axes, not the eucledian distance about rtree HOT 4 OPEN

Glad you are finding the library useful!

If the dimensions of the axes are very different then that might have negative effects on the performance of the rtree because of bias to or away from an axis. I would normally suggest translating and scaling the coordinates in this case say so they are in the 0..1 range.

Euclidean distance doesn't sound like it has a lot of meaning as you point out. I would just do a search with a rectangle (x-xDistance, y-yDistance, x+xDistance, y+yDistance) which seems to match your use case exactly. Am I missing something?

from rtree.

That's what I'm doing now (search with a rectangle), but it finds everything that intersects the query. In my case such a search might bring up a lot of unnecessary stuff (e.g. some large rectangles that are also in the tree), I thought that it could be much more efficient to do nearest() with small k because I only need the closest things to my query point.

After writing this I think I know what I'm missing -- a method like search(Geometry g, boolean fullyEncloses). Instead of intersections I'd like to be able to find all things that are completely enclosed by the query. In case of arbitrary geometries that would be full enclosure of their bounding boxes.

I don't know if such queries are possible/easy to implement with an RTree. If not, then a nearest() with different constraints for coordinates might work.

from rtree.

To get the contains seach you would do this:

RTree<Object, Rectangle> tree = ...;
List<Entry<Object,Rectangle>>  entries = 
  tree
    .search(rect)
    .filter(entry -> rect.contains(entry.geometry()))
    .toList().toBlocking().single(); 

This will be the most efficient way of doing it and any api additions would probably do this under the covers. Are you using filter with the returned observable as above?

from rtree.

No, I wasn't using filter, I thought that some direct serach solution would be faster. But since you say this is how you'd implement this type of query anyway, I just have to believe you :)

Thank you!

PS: I guess this can be added to the readme (how to perform contains queries)

from rtree.