kaze/orx
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[orx-kdtree] Refactor KDTree interface to improve ease of use

Browse Source
This commit is contained in:
Edwin Jakobs
2022-01-01 13:46:52 +01:00
parent 7cba88143d
commit e7b143493c
4 changed files with 129 additions and 89 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
orx-kdtree/src/demo/kotlin/DemoKNearestNeighbour01.kt
View File

@@ -2,13 +2,13 @@ import org.openrndr.application
import org.openrndr.color.ColorRGBa
import org.openrndr.extra.kdtree.buildKDTree
import org.openrndr.extra.kdtree.findKNearest
import org.openrndr.extra.kdtree.kdTree
import org.openrndr.extra.kdtree.vector2Mapper
import org.openrndr.math.Vector2
import org.openrndr.shape.LineSegment
fun main() {
application {
configure {
width = 1080
height = 720
@@ -18,12 +18,12 @@ fun main() {
val points = MutableList(1000) {
Vector2(Math.random() * width, Math.random() * height)
}
val tree = buildKDTree(points, 2, ::vector2Mapper)
val tree = points.kdTree()
extend {
drawer.circles(points, 5.0)
val kNearest = findKNearest(tree, mouse.position, k=7, dimensions = 2, ::vector2Mapper)
val kNearest = tree.findKNearest(mouse.position, k = 7)
drawer.fill = ColorRGBa.RED
drawer.stroke = ColorRGBa.RED
drawer.strokeWeight = 2.0

9
orx-kdtree/src/demo/kotlin/DemoNearestNeighbour01.kt
View File

@@ -1,8 +1,5 @@
import org.openrndr.application
import org.openrndr.extensions.SingleScreenshot
import org.openrndr.extra.kdtree.buildKDTree
import org.openrndr.extra.kdtree.findNearest
import org.openrndr.extra.kdtree.vector2Mapper
import org.openrndr.extra.kdtree.kdTree
import org.openrndr.math.Vector2
fun main() {
@@ -15,10 +12,10 @@ fun main() {
val points = MutableList(1000) {
Vector2(Math.random() * width, Math.random() * height)
}
val tree = buildKDTree(points, 2, ::vector2Mapper)
val tree = points.kdTree()
extend {
drawer.circles(points, 5.0)
val nearest = findNearest(tree, mouse.position, 2, ::vector2Mapper)
val nearest = tree.findNearest(mouse.position)
nearest?.let {
drawer.circle(it.x, it.y, 20.0)
}

8
orx-kdtree/src/demo/kotlin/DemoRangeQuery01.kt
View File

@@ -1,8 +1,6 @@
import org.openrndr.application
import org.openrndr.color.ColorRGBa
import org.openrndr.extra.kdtree.buildKDTree
import org.openrndr.extra.kdtree.findAllInRange
import org.openrndr.extra.kdtree.vector2Mapper
import org.openrndr.extra.kdtree.kdTree
import org.openrndr.math.Vector2
@@ -18,13 +16,13 @@ fun main() {
val points = MutableList(1000) {
Vector2(Math.random() * width, Math.random() * height)
}
val tree = buildKDTree(points, 2, ::vector2Mapper)
val tree = points.kdTree()
val radius = 50.0
extend {
drawer.circles(points, 5.0)
val allInRange = findAllInRange(tree, mouse.position, maxDistance = radius, dimensions = 2, ::vector2Mapper)
val allInRange = tree.findAllInRadius(mouse.position, radius = radius)
drawer.fill = ColorRGBa.PINK
drawer.stroke = ColorRGBa.PINK
drawer.strokeWeight = 2.0

195
orx-kdtree/src/main/kotlin/KDTree.kt
View File

@@ -4,10 +4,7 @@ import kotlinx.coroutines.CoroutineScope
import kotlinx.coroutines.GlobalScope
import kotlinx.coroutines.launch
import kotlinx.coroutines.runBlocking
import org.openrndr.math.IntVector2
import org.openrndr.math.Vector2
import org.openrndr.math.Vector3
import org.openrndr.math.Vector4
import org.openrndr.math.*
import java.util.*
import kotlin.IllegalStateException
import kotlin.math.abs
@@ -47,7 +44,7 @@ fun vector4Mapper(v: Vector4, dimension: Int): Double {
}
}
class KDTreeNode<T> {
class KDTreeNode<T>(val dimensions: Int, val mapper: (T, Int) -> Double) {
var parent: KDTreeNode<T>? = null
var median: Double = 0.0
var dimension: Int = 0
@@ -57,6 +54,26 @@ class KDTreeNode<T> {
internal val isLeaf: Boolean
get() = children[0] == null && children[1] == null
fun insert(item: T): KDTreeNode<T> {
return insert(this, item, dimensions, mapper)
}
fun remove(node: KDTreeNode<T>): KDTreeNode<T>? {
return org.openrndr.extra.kdtree.remove(node, mapper)
}
fun findNearest(query: T, includeQuery: Boolean = false): T? = findNearest(this, query, includeQuery)
fun findKNearest(query: T, k: Int, includeQuery: Boolean = false): List<T> {
return findKNearest(this, query, k, includeQuery)
}
fun findAllInRadius(query: T, radius: Double, includeQuery: Boolean = false): List<T> {
return findAllInRadius(this, query, radius, includeQuery)
}
override fun toString(): String {
return "KDTreeNode{" +
"median=" + median +
@@ -68,24 +85,32 @@ class KDTreeNode<T> {
}
}
fun <T> insertItem(root: KDTreeNode<T>, item: T, mapper: (T, Int) -> Double): KDTreeNode<T> {
private fun <T> insertItem(root: KDTreeNode<T>, item: T): KDTreeNode<T> {
return if (root.isLeaf) {
root.item = item
root
} else {
if (mapper(item, root.dimension) < root.median) {
insertItem(root.children[0] ?: throw IllegalStateException("left is null"), item, mapper)
if (root.mapper(item, root.dimension) < root.median) {
insertItem(root.children[0] ?: throw IllegalStateException("left is null"), item)
} else {
insertItem(root.children[1] ?: throw IllegalStateException("right is null"), item, mapper)
insertItem(root.children[1] ?: throw IllegalStateException("right is null"), item)
}
}
}
fun <T> buildKDTree(items: MutableList<T>, dimensions: Int, mapper: (T, Int) -> Double): KDTreeNode<T> {
val root = KDTreeNode<T>()
val root = KDTreeNode<T>(dimensions, mapper)
val start = System.currentTimeMillis()
fun <T> buildTreeTask(scope: CoroutineScope, node: KDTreeNode<T>, items: MutableList<T>, dimensions: Int, levels: Int, mapper: (T, Int) -> Double): KDTreeNode<T> {
fun <T> buildTreeTask(
scope: CoroutineScope,
node: KDTreeNode<T>,
items: MutableList<T>,
dimensions: Int,
levels: Int,
mapper: (T, Int) -> Double
): KDTreeNode<T> {
if (items.size > 0) {
val dimension = levels % dimensions
@@ -119,7 +144,7 @@ fun <T> buildKDTree(items: MutableList<T>, dimensions: Int, mapper: (T, Int) ->
}
if (leftItems.size > 0) {
node.children[0] = KDTreeNode()
node.children[0] = KDTreeNode(dimensions, mapper)
node.children[0]?.let {
it.parent = node
@@ -129,7 +154,7 @@ fun <T> buildKDTree(items: MutableList<T>, dimensions: Int, mapper: (T, Int) ->
}
}
if (rightItems.size > 0) {
node.children[1] = KDTreeNode()
node.children[1] = KDTreeNode(dimensions, mapper)
node.children[1]?.let {
it.parent = node
scope.launch {
@@ -147,7 +172,7 @@ fun <T> buildKDTree(items: MutableList<T>, dimensions: Int, mapper: (T, Int) ->
runBlocking {
job.join()
}
println("building took ${System.currentTimeMillis()-start}ms")
println("building took ${System.currentTimeMillis() - start}ms")
return root
}
@@ -185,45 +210,45 @@ fun <T> findAllNodes(root: KDTreeNode<T>): List<KDTreeNode<T>> {
fun <T> findKNearest(
root: KDTreeNode<T>,
item: T,
query: T,
k: Int,
dimensions: Int,
mapper: (T, Int) -> Double
includeQuery: Boolean = false
): List<T> {
// max-heap with size k
val queue = PriorityQueue<Pair<KDTreeNode<T>, Double>>(k + 1) {
nodeA, nodeB -> compareValues(nodeB.second, nodeA.second)
val queue = PriorityQueue<Pair<KDTreeNode<T>, Double>>(k + 1) { nodeA, nodeB ->
compareValues(nodeB.second, nodeA.second)
}
fun nearest(node: KDTreeNode<T>?, item: T) {
fun nearest(node: KDTreeNode<T>?) {
if (node != null) {
val dimensionValue = mapper(item, node.dimension)
val dimensionValue = node.mapper(query, node.dimension)
val route: Int = if (dimensionValue < node.median) {
nearest(node.children[0], item)
nearest(node.children[0])
0
} else {
nearest(node.children[1], item)
nearest(node.children[1])
1
}
val distance = sqrDistance(item, node.item
?: throw IllegalStateException("item is null"), dimensions, mapper)
val distance = sqrDistance(query, node.item ?: error("item is null"), node.dimensions, node.mapper)
if (queue.size < k || distance < queue.peek().second) {
queue.add(Pair(node, distance))
if (queue.size > k) {
queue.poll()
if (includeQuery || node.item !== query) {
if (queue.size < k || distance < queue.peek().second) {
queue.add(Pair(node, distance))
if (queue.size > k) {
queue.poll()
}
}
}
val d = abs(node.median - dimensionValue)
if (d * d < queue.peek().second || queue.size < k) {
nearest(node.children[1 - route], item)
nearest(node.children[1 - route])
}
}
}
nearest(root, item)
nearest(root)
return generateSequence { queue.poll() }
.map { it.first.item }
@@ -231,77 +256,79 @@ fun <T> findKNearest(
.toList().reversed()
}
fun <T> findNearest(root: KDTreeNode<T>, item: T, dimensions: Int, mapper: (T, Int) -> Double): T? {
private fun <T> findNearest(root: KDTreeNode<T>, query: T, includeQuery: Boolean = false): T? {
var nearest = java.lang.Double.POSITIVE_INFINITY
var nearestArg: KDTreeNode<T>? = null
fun nearest(node: KDTreeNode<T>?, item: T) {
fun nearest(node: KDTreeNode<T>?) {
if (node != null) {
if (node.item == null) {
println(node)
}
val route: Int = if (mapper(item, node.dimension) < node.median) {
nearest(node.children[0], item)
val route: Int = if (root.mapper(query, node.dimension) < node.median) {
nearest(node.children[0])
0
} else {
nearest(node.children[1], item)
nearest(node.children[1])
1
}
val distance = sqrDistance(item, node.item
?: throw IllegalStateException("item is null"), dimensions, mapper)
if (distance < nearest) {
val distance = sqrDistance(
query, node.item
?: error("item is null"), root.dimensions, root.mapper
)
if (distance < nearest && (includeQuery || node.item !== query)) {
nearest = distance
nearestArg = node
}
val d = abs(node.median - mapper(item, node.dimension))
val d = abs(node.median - root.mapper(query, node.dimension))
if (d * d < nearest) {
nearest(node.children[1 - route], item)
nearest(node.children[1 - route])
}
}
}
nearest(root, item)
nearest(root)
return nearestArg?.item
}
fun <T> findAllInRange(
private fun <T> findAllInRadius(
root: KDTreeNode<T>,
item: T,
maxDistance: Double,
dimensions: Int,
mapper: (T, Int) -> Double
) : List<T> {
query: T,
radius: Double,
includeQuery: Boolean = false
): List<T> {
val sqrMaxDist = maxDistance * maxDistance
val queue = kotlin.collections.ArrayDeque<KDTreeNode<T>?>()
val sqrMaxDist = radius * radius
val queue = ArrayDeque<KDTreeNode<T>>()
queue.add(root)
val results = mutableListOf<T?>()
while (queue.isNotEmpty()) {
val node = queue.removeFirst()
if (node != null) {
val dimensionValue = mapper(item, node.dimension)
val distance = sqrDistance(item, node.item
?: throw IllegalStateException("item is null"), dimensions, mapper)
if (distance <= sqrMaxDist) {
results.add(node.item)
}
val dimensionValue = node.mapper(query, node.dimension)
val distance = sqrDistance(
query, node.item
?: error("item is null"), node.dimensions, node.mapper
)
if (distance <= sqrMaxDist && (includeQuery || node.item != query)) {
results.add(node.item)
}
val route: Int = if (dimensionValue < node.median) {
queue.add(node.children[0])
0
} else {
queue.add(node.children[1])
1
}
val route: Int = if (dimensionValue < node.median && node.children[0] != null) {
queue.add(node.children[0])
0
} else if (node.children[1] != null) {
queue.add(node.children[1])
1
} else {
-1
}
if (route != -1) {
val d = abs(node.median - dimensionValue)
if (d * d <= sqrMaxDist) {
queue.add(node.children[1 - route])
val c = node.children[1 - route]
if (c != null) {
queue.add(c)
}
}
}
}
@@ -309,7 +336,7 @@ fun <T> findAllInRange(
return results.filterNotNull()
}
fun <T> insert(root: KDTreeNode<T>, item: T, dimensions: Int, mapper: (T, Int) -> Double): KDTreeNode<T> {
private fun <T> insert(root: KDTreeNode<T>, item: T, dimensions: Int, mapper: (T, Int) -> Double): KDTreeNode<T> {
val stack = Stack<KDTreeNode<T>>()
stack.push(root)
@@ -324,7 +351,7 @@ fun <T> insert(root: KDTreeNode<T>, item: T, dimensions: Int, mapper: (T, Int) -
stack.push(node.children[0])
} else {
// sit here
node.children[0] = KDTreeNode()
node.children[0] = KDTreeNode(dimensions, mapper)
node.children[0]?.item = item
node.children[0]?.dimension = (node.dimension + 1) % dimensions
node.children[0]?.median = mapper(item, (node.dimension + 1) % dimensions)
@@ -336,7 +363,7 @@ fun <T> insert(root: KDTreeNode<T>, item: T, dimensions: Int, mapper: (T, Int) -
stack.push(node.children[1])
} else {
// sit here
node.children[1] = KDTreeNode()
node.children[1] = KDTreeNode(dimensions, mapper)
node.children[1]?.item = item
node.children[1]?.dimension = (node.dimension + 1) % dimensions
node.children[1]?.median = mapper(item, (node.dimension + 1) % dimensions)
@@ -348,7 +375,7 @@ fun <T> insert(root: KDTreeNode<T>, item: T, dimensions: Int, mapper: (T, Int) -
}
}
fun <T> remove(toRemove: KDTreeNode<T>, mapper: (T, Int) -> Double): KDTreeNode<T>? {
private fun <T> remove(toRemove: KDTreeNode<T>, mapper: (T, Int) -> Double): KDTreeNode<T>? {
// trivial case
if (toRemove.isLeaf) {
val p = toRemove.parent
@@ -439,3 +466,21 @@ fun <T> remove(toRemove: KDTreeNode<T>, mapper: (T, Int) -> Double): KDTreeNode<
}
return null
}
@JvmName("kdTreeVector2")
fun Iterable<Vector2>.kdTree(): KDTreeNode<Vector2> {
val items = this.toMutableList()
return buildKDTree(items, 2, ::vector2Mapper)
}
@JvmName("kdTreeVector3")
fun Iterable<Vector3>.kdTree(): KDTreeNode<Vector3> {
val items = this.toMutableList()
return buildKDTree(items, 3, ::vector3Mapper)
}
@JvmName("kdTreeVector4")
fun Iterable<Vector4>.kdTree(): KDTreeNode<Vector4> {
val items = this.toMutableList()
return buildKDTree(items, 4, ::vector4Mapper)
}