Libraryless. Click here for Pure Java version (4623L/25K).
// TODO: find splitPoint faster (just access the median element!) // TODO: really slow making the initial tree (5s for ~250 points) sclass PtTree extends AbstractSet<Pt> { int size; Node root = new Leaf(null); // config int maxPointsPerNode = 4; abstract class Node { Node parent; abstract bool add(Pt p); abstract void collectPointsIn(Rect r, L<Pt> out); abstract void replaceChild(Node from, Node to); void addAll(Iterable<Pt> l) { fOr (Pt p : l) add(p); } } class Leaf extends Node { Cl<Pt> points; *(Node *parent) {} bool add(Pt p) { if (main contains(points, p)) false; if (points == null) points = new L; points.add(p); ++size; possiblySplit(); true; } void possiblySplit { if (l(points) > maxPointsPerNode) split(); } void collectPointsIn(Rect r, L<Pt> out) { fOr (p : points) if (containsPt(r, p)) out.add(p); } record noeq ProposedSplit(int dimension, int splitPoint, int count) { int error() { ret abs(count-half(l(points))); } } Split split() { int n = l(points), half = n/2; ProposedSplit xSplit = checkSplit(0); ProposedSplit ySplit = checkSplit(1); var best = xSplit.error() < ySplit.error() ? xSplit : ySplit; new Split split; split.dimension = (byte) best.dimension; split.splitPoint = best.splitPoint; Leaf a = new Leaf(split); IPred<Pt> pred = p -> ptCoord(split.dimension, p) >= split.splitPoint; a.addAll(antiFilter(pred, points)); split.a = a; Leaf b = new Leaf(split); b.addAll(filter(pred, points)); split.b = b; a.possiblySplit(); b.possiblySplit(); ret replaceMeWith(split); } <A extends Node> A replaceMeWith(A node) { node.parent = parent; if (parent != null) parent.replaceChild(this, node); else if (root == this) root = node; ret node; } ProposedSplit checkSplit(int dimension) { L<Pt> lx = sortedBy(points, p -> ptCoord(dimension, p)); new ProposedSplit ps; ps.dimension = dimension; int n = l(points), half = n/2; int i = 0; int splitPoint = Int.MIN_VALUE; while true { int lastSplitPoint = splitPoint; splitPoint = ptCoord(lx.get(i++), dimension)+1; int lastI = i; while (i < n && ptCoord(lx.get(i), dimension) < splitPoint) ++i; if (i >= half) { if (abs(lastI-half) < abs(i-half)) { ps.splitPoint = lastSplitPoint; ps.count = lastI; } else { ps.splitPoint = splitPoint; ps.count = i; } ret ps; } } } void replaceChild(Node from, Node to) { unimplemented(); } // doesn't need } static final int minInf = Int.MIN_VALUE/4; static final int inf = Int.MAX_VALUE/4; class Split extends Node { byte dimension; // 0 for X, 1 for Y int splitPoint; Node a, b; bool add(Pt p) { ret (ptCoord(p, dimension) >= splitPoint ? b : a).add(p); } Rect aRect() { ret rectFromPoints(minInf, minInf, dimension == 0 ? splitPoint : inf, dimension == 1 ? splitPoint : inf); } Rect bRect() { ret rectFromPoints( dimension == 0 ? splitPoint : minInf, dimension == 1 ? splitPoint : minInf, inf, inf); } void collectPointsIn(Rect r, L<Pt> out) { Rect rA = intersectRect(r, aRect()); if (nempty(rA)) a.collectPointsIn(rA, out); Rect rB = intersectRect(r, bRect()); if (nempty(rB)) b.collectPointsIn(rB, out); } void replaceChild(Node from, Node to) { if (a == from) a = to; if (b == from) b = to; } } public bool add(Pt p) { ret root.add(p); } L<Pt> pointsIn aka pointsInRect(Rect r) { new L<Pt> out; root.collectPointsIn(r, out); ret out; } public bool contains(O p) { ret p instanceof Pt && contains(p/Pt); } bool contains(Pt p) { ret nempty(pointsIn(rect(p.x, p.y, 1, 1))); } // use this to make a PtTree! // It's actually makes a balanced tree. // It is assumed that points contains no duplicates static PtTree fromPointSet(Iterable<Pt> points) { new PtTree tree; Leaf root = cast tree.root; root.points = cloneList(points); tree.size = l(root.points); root.possiblySplit(); ret tree; } public int size() { ret size; } public Iterator<Pt> iterator() { /* What we actually want is a coroutine yielding the values. // Like this, if we had this syntax: ret coroutine { void iterate(Node node) { if (node cast Leaf) returnAll(node.points); else { cast node to Split; iterate(node.a); iterate(node.b); } run { iterate(root); } } }; // So how do we turn this into an iterator? // Maybe through VStack.Computable. */ new Var<Iterator<Pt>> out; new VStack stack; stack.push(new co_iterate(root, out)); ret iff_endOnNull(new IF0<Pt> { bool done; public Pt get() { while ping (true) { // something in the pipe? return it if (out.has()) if (out->hasNext()) ret out->next(); else out.clear(); // and clean the pipe afterwards, boy! if (done) null; // end of the line // nothing in the pipe, gotta walk the tree some more if (!step(stack)) set done; // prepare to get off the train } } }); } record noeq co_iterate(Node node, Var<Iterator<Pt>> out) extends VStackComputableWithStep { void step(VStack stack) { if (step == 0) { if (node cast Leaf) { out.set(main iterator(node.points)); stack.return(); } else { cast node to Split; stack.call(new co_iterate(node.a, out)); ret with step = 1; } } else { cast node to Split; stack.replace(new co_iterate(node.b, out)); } } } }
download show line numbers debug dex old transpilations
Travelled to 3 computer(s): bhatertpkbcr, mqqgnosmbjvj, pyentgdyhuwx
No comments. add comment
Snippet ID: | #1032229 |
Snippet name: | PtTree - tree structure for quick finding of points in any rectangle [probably OK but super slow to make] |
Eternal ID of this version: | #1032229/29 |
Text MD5: | fbd207613c282e9244d96da4ca3cd560 |
Transpilation MD5: | d89e72e91ab65f5db75e707df0555b61 |
Author: | stefan |
Category: | javax / 2d space |
Type: | JavaX fragment (include) |
Public (visible to everyone): | Yes |
Archived (hidden from active list): | No |
Created/modified: | 2021-08-22 19:11:42 |
Source code size: | 6597 bytes / 253 lines |
Pitched / IR pitched: | No / No |
Views / Downloads: | 220 / 412 |
Version history: | 28 change(s) |
Referenced in: | #1032244 - PtTree - tree structure for quick finding of points in any rectangle [backup with the cool co_iterate] #1034167 - Standard Classes + Interfaces (LIVE, continuation of #1003674) |