| 1 | /* |
| 2 | * Copyright 2001-2005 Daniel F. Savarese |
| 3 | * Copyright 2006-2009 Savarese Software Research Corporation |
| 4 | * |
| 5 | * Licensed under the Apache License, Version 2.0 (the "License"); |
| 6 | * you may not use this file except in compliance with the License. |
| 7 | * You may obtain a copy of the License at |
| 8 | * |
| 9 | * https://www.savarese.com/software/ApacheLicense-2.0 |
| 10 | * |
| 11 | * Unless required by applicable law or agreed to in writing, software |
| 12 | * distributed under the License is distributed on an "AS IS" BASIS, |
| 13 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 14 | * See the License for the specific language governing permissions and |
| 15 | * limitations under the License. |
| 16 | */ |
| 17 | |
| 18 | package com.savarese.spatial; |
| 19 | |
| 20 | import java.lang.reflect.Array; |
| 21 | import java.util.*; |
| 22 | |
| 23 | // All the view classes are inefficient for anything other than iteration. |
| 24 | /** |
| 25 | * <p>A k-d tree divides a k-dimensional space relative to the points it |
| 26 | * contains by storing them in a binary tree, discriminating by a |
| 27 | * different dimension at each level of the tree. It allows efficient |
| 28 | * point data retrieval (<em>O(lg(n))</em>) and range searching.</p> |
| 29 | * |
| 30 | * <p>KDTree conforms to the java.util.Map interface except that |
| 31 | * Iterator.remove is not supported by the returned views.</p> |
| 32 | */ |
| 33 | public class KDTree<Coord extends Comparable<? super Coord>, |
| 34 | P extends Point<Coord>, V> |
| 35 | implements RangeSearchTree<Coord, P, V> |
| 36 | { |
| 37 | final class KDNode implements Map.Entry<P,V>{ |
| 38 | int _discriminator; |
| 39 | P _point; |
| 40 | V _value; |
| 41 | KDNode _low, _high; |
| 42 | |
| 43 | KDNode(int discriminator, P point, V value) { |
| 44 | _point = point; |
| 45 | _value = value; |
| 46 | _low = _high = null; |
| 47 | _discriminator = discriminator; |
| 48 | } |
| 49 | |
| 50 | public boolean equals(Object o) { |
| 51 | KDNode node = (KDNode)o; |
| 52 | |
| 53 | if(node == this) |
| 54 | return true; |
| 55 | |
| 56 | return |
| 57 | ((getKey() == null ? |
| 58 | node.getKey() == null : getKey().equals(node.getKey())) && |
| 59 | (getValue() == null ? |
| 60 | node.getValue() == null : getValue().equals(node.getValue()))); |
| 61 | } |
| 62 | |
| 63 | public P getKey() { |
| 64 | return _point; |
| 65 | } |
| 66 | |
| 67 | public V getValue() { |
| 68 | return _value; |
| 69 | } |
| 70 | |
| 71 | // Only call if the node is in the tree. |
| 72 | public V setValue(V value) { |
| 73 | V old = _value; |
| 74 | _hashCode-=hashCode(); |
| 75 | _value = value; |
| 76 | _hashCode+=hashCode(); |
| 77 | return old; |
| 78 | } |
| 79 | |
| 80 | public int hashCode() { |
| 81 | return |
| 82 | ((getKey() == null ? 0 : getKey().hashCode()) ^ |
| 83 | (getValue() == null ? 0 : getValue().hashCode())); |
| 84 | } |
| 85 | } |
| 86 | |
| 87 | final class MapEntryIterator implements Iterator<Map.Entry<P,V>> { |
| 88 | LinkedList<KDNode> _stack; |
| 89 | KDNode _next; |
| 90 | P _lower, _upper; |
| 91 | |
| 92 | MapEntryIterator(P lower, P upper) { |
| 93 | _stack = new LinkedList<KDNode>(); |
| 94 | _lower = lower; |
| 95 | _upper = upper; |
| 96 | _next = null; |
| 97 | |
| 98 | if(_root != null) |
| 99 | _stack.addLast(_root); |
| 100 | next(); |
| 101 | } |
| 102 | |
| 103 | MapEntryIterator() { |
| 104 | this(null, null); |
| 105 | } |
| 106 | |
| 107 | public boolean hasNext() { |
| 108 | return (_next != null); |
| 109 | } |
| 110 | |
| 111 | public Map.Entry<P,V> next() { |
| 112 | KDNode old = _next; |
| 113 | |
| 114 | while(!_stack.isEmpty()) { |
| 115 | KDNode node = _stack.removeLast(); |
| 116 | int discriminator = node._discriminator; |
| 117 | |
| 118 | if((_upper == null || |
| 119 | node._point.getCoord(discriminator).compareTo( |
| 120 | _upper.getCoord(discriminator)) <= 0) && node._high != null) |
| 121 | _stack.addLast(node._high); |
| 122 | |
| 123 | if((_lower == null || |
| 124 | node._point.getCoord(discriminator).compareTo( |
| 125 | _lower.getCoord(discriminator)) > 0) && node._low != null) |
| 126 | _stack.addLast(node._low); |
| 127 | |
| 128 | if(isInRange(node._point, _lower, _upper)) { |
| 129 | _next = node; |
| 130 | return old; |
| 131 | } |
| 132 | } |
| 133 | |
| 134 | _next = null; |
| 135 | |
| 136 | return old; |
| 137 | } |
| 138 | |
| 139 | // This violates the contract for entrySet, but we can't support |
| 140 | // in a reasonable fashion the removal of mappings through the iterator. |
| 141 | // Java iterators require a hasNext() function, which forces the stack |
| 142 | // to reflect a future search state, making impossible to adjust the current |
| 143 | // stack after a removal. Implementation alternatives are all too |
| 144 | // expensive. Yet another reason to favor the C++ implementation... |
| 145 | public void remove() |
| 146 | throws UnsupportedOperationException |
| 147 | { |
| 148 | throw new UnsupportedOperationException(); |
| 149 | } |
| 150 | } |
| 151 | |
| 152 | final class KeyIterator implements Iterator<P> { |
| 153 | MapEntryIterator iterator; |
| 154 | |
| 155 | KeyIterator(MapEntryIterator it) { |
| 156 | iterator = it; |
| 157 | } |
| 158 | |
| 159 | public boolean hasNext() { |
| 160 | return iterator.hasNext(); |
| 161 | } |
| 162 | |
| 163 | public P next() { |
| 164 | Map.Entry<P,V> next = iterator.next(); |
| 165 | return (next == null ? null : next.getKey()); |
| 166 | } |
| 167 | |
| 168 | public void remove() |
| 169 | throws UnsupportedOperationException |
| 170 | { |
| 171 | iterator.remove(); |
| 172 | } |
| 173 | } |
| 174 | |
| 175 | final class ValueIterator implements Iterator<V> { |
| 176 | MapEntryIterator iterator; |
| 177 | |
| 178 | ValueIterator(MapEntryIterator it) { |
| 179 | iterator = it; |
| 180 | } |
| 181 | |
| 182 | public boolean hasNext() { |
| 183 | return iterator.hasNext(); |
| 184 | } |
| 185 | |
| 186 | public V next() { |
| 187 | Map.Entry<P,V> next = iterator.next(); |
| 188 | return (next == null ? null : next.getValue()); |
| 189 | } |
| 190 | |
| 191 | public void remove() |
| 192 | throws UnsupportedOperationException |
| 193 | { |
| 194 | iterator.remove(); |
| 195 | } |
| 196 | } |
| 197 | |
| 198 | abstract class CollectionView<E> implements Collection<E> { |
| 199 | |
| 200 | public boolean add(E o) |
| 201 | throws UnsupportedOperationException |
| 202 | { |
| 203 | throw new UnsupportedOperationException(); |
| 204 | } |
| 205 | |
| 206 | public boolean addAll(Collection<? extends E> c) |
| 207 | throws UnsupportedOperationException |
| 208 | { |
| 209 | throw new UnsupportedOperationException(); |
| 210 | } |
| 211 | |
| 212 | public void clear() { |
| 213 | KDTree.this.clear(); |
| 214 | } |
| 215 | |
| 216 | public boolean containsAll(Collection<?> c) { |
| 217 | for(Object o : c) { |
| 218 | if(!contains(o)) |
| 219 | return false; |
| 220 | } |
| 221 | return true; |
| 222 | } |
| 223 | |
| 224 | public int hashCode() { |
| 225 | return KDTree.this.hashCode(); |
| 226 | } |
| 227 | |
| 228 | public boolean isEmpty() { |
| 229 | return KDTree.this.isEmpty(); |
| 230 | } |
| 231 | |
| 232 | public int size() { |
| 233 | return KDTree.this.size(); |
| 234 | } |
| 235 | |
| 236 | public Object[] toArray() { |
| 237 | Object[] obja = new Object[size()]; |
| 238 | int i=0; |
| 239 | |
| 240 | for(E e : this) { |
| 241 | obja[i] = e; |
| 242 | ++i; |
| 243 | } |
| 244 | |
| 245 | return obja; |
| 246 | } |
| 247 | |
| 248 | public <T> T[] toArray(T[] a) { |
| 249 | Object[] array = a; |
| 250 | |
| 251 | if(array.length < size()) |
| 252 | array = a = |
| 253 | (T[])Array.newInstance(a.getClass().getComponentType(), size()); |
| 254 | |
| 255 | if(array.length > size()) |
| 256 | array[size()] = null; |
| 257 | |
| 258 | int i = 0; |
| 259 | for(E e : this) { |
| 260 | array[i] = e; |
| 261 | ++i; |
| 262 | } |
| 263 | |
| 264 | return a; |
| 265 | } |
| 266 | } |
| 267 | |
| 268 | abstract class SetView<E> extends CollectionView<E> implements Set<E> { |
| 269 | public boolean equals(Object o) { |
| 270 | if(!(o instanceof Set)) |
| 271 | return false; |
| 272 | |
| 273 | if(o == this) |
| 274 | return true; |
| 275 | |
| 276 | Set<?> set = (Set<?>)o; |
| 277 | |
| 278 | if(set.size() != size()) |
| 279 | return false; |
| 280 | |
| 281 | try { |
| 282 | return containsAll(set); |
| 283 | } catch(ClassCastException cce) { |
| 284 | return false; |
| 285 | } |
| 286 | } |
| 287 | } |
| 288 | |
| 289 | final class MapEntrySet extends SetView<Map.Entry<P,V>> { |
| 290 | public boolean contains(Object o) |
| 291 | throws ClassCastException, NullPointerException |
| 292 | { |
| 293 | Map.Entry<P,V> e = (Map.Entry<P,V>)o; |
| 294 | KDNode node = getNode(e.getKey()); |
| 295 | |
| 296 | if(node == null) |
| 297 | return false; |
| 298 | |
| 299 | return e.getValue().equals(node.getValue()); |
| 300 | } |
| 301 | |
| 302 | public Iterator<Map.Entry<P,V>> iterator() { |
| 303 | return new MapEntryIterator(); |
| 304 | } |
| 305 | |
| 306 | public boolean remove(Object o) |
| 307 | throws ClassCastException |
| 308 | { |
| 309 | int size = size(); |
| 310 | Map.Entry<P,V> e = (Map.Entry<P,V>)o; |
| 311 | |
| 312 | KDTree.this.remove(e.getKey()); |
| 313 | |
| 314 | return (size != size()); |
| 315 | } |
| 316 | |
| 317 | public boolean removeAll(Collection<?> c) |
| 318 | throws ClassCastException |
| 319 | { |
| 320 | int size = size(); |
| 321 | |
| 322 | for(Object o : c) { |
| 323 | Map.Entry<P,V> e = (Map.Entry<P,V>)o; |
| 324 | KDTree.this.remove(e.getKey()); |
| 325 | } |
| 326 | |
| 327 | return (size != size()); |
| 328 | } |
| 329 | |
| 330 | public boolean retainAll(Collection<?> c) |
| 331 | throws ClassCastException |
| 332 | { |
| 333 | for(Object o : c) { |
| 334 | if(contains(o)) { |
| 335 | Collection<Map.Entry<P,V>> col = (Collection<Map.Entry<P,V>>)c; |
| 336 | clear(); |
| 337 | for(Map.Entry<P,V> e : col) |
| 338 | put(e.getKey(), e.getValue()); |
| 339 | return true; |
| 340 | } |
| 341 | } |
| 342 | return false; |
| 343 | } |
| 344 | } |
| 345 | |
| 346 | final class KeySet extends SetView<P> { |
| 347 | |
| 348 | public boolean contains(Object o) |
| 349 | throws ClassCastException, NullPointerException |
| 350 | { |
| 351 | return KDTree.this.containsKey(o); |
| 352 | } |
| 353 | |
| 354 | public Iterator<P> iterator() { |
| 355 | return new KeyIterator(new MapEntryIterator()); |
| 356 | } |
| 357 | |
| 358 | public boolean remove(Object o) |
| 359 | throws ClassCastException |
| 360 | { |
| 361 | int size = size(); |
| 362 | KDTree.this.remove(o); |
| 363 | return (size != size()); |
| 364 | } |
| 365 | |
| 366 | public boolean removeAll(Collection<?> c) |
| 367 | throws ClassCastException |
| 368 | { |
| 369 | int size = size(); |
| 370 | |
| 371 | for(Object o : c) |
| 372 | KDTree.this.remove(o); |
| 373 | |
| 374 | return (size != size()); |
| 375 | } |
| 376 | |
| 377 | public boolean retainAll(Collection<?> c) |
| 378 | throws ClassCastException |
| 379 | { |
| 380 | HashMap<P,V> map = new HashMap<P,V>(); |
| 381 | int size = size(); |
| 382 | |
| 383 | for(Object o : c) { |
| 384 | V val = get(o); |
| 385 | |
| 386 | if(val != null || contains(o)) |
| 387 | map.put((P)o, val); |
| 388 | } |
| 389 | |
| 390 | clear(); |
| 391 | putAll(map); |
| 392 | |
| 393 | return (size != size()); |
| 394 | } |
| 395 | } |
| 396 | |
| 397 | final class ValueCollection extends CollectionView<V> { |
| 398 | |
| 399 | public boolean contains(Object o) |
| 400 | throws ClassCastException, NullPointerException |
| 401 | { |
| 402 | return KDTree.this.containsValue(o); |
| 403 | } |
| 404 | |
| 405 | public Iterator<V> iterator() { |
| 406 | return new ValueIterator(new MapEntryIterator()); |
| 407 | } |
| 408 | |
| 409 | public boolean remove(Object o) |
| 410 | throws ClassCastException |
| 411 | { |
| 412 | KDNode node = findValue(_root, o); |
| 413 | |
| 414 | if(node != null) { |
| 415 | KDTree.this.remove(node.getKey()); |
| 416 | return true; |
| 417 | } |
| 418 | |
| 419 | return false; |
| 420 | } |
| 421 | |
| 422 | public boolean removeAll(Collection<?> c) |
| 423 | throws ClassCastException |
| 424 | { |
| 425 | int size = size(); |
| 426 | |
| 427 | for(Object o : c) { |
| 428 | KDNode node = findValue(_root, o); |
| 429 | |
| 430 | while(node != null) { |
| 431 | KDTree.this.remove(o); |
| 432 | node = findValue(_root, o); |
| 433 | } |
| 434 | } |
| 435 | |
| 436 | return (size != size()); |
| 437 | } |
| 438 | |
| 439 | public boolean retainAll(Collection<?> c) |
| 440 | throws ClassCastException |
| 441 | { |
| 442 | HashMap<P,V> map = new HashMap<P,V>(); |
| 443 | int size = size(); |
| 444 | |
| 445 | for(Object o : c) { |
| 446 | KDNode node = findValue(_root, o); |
| 447 | |
| 448 | while(node != null) { |
| 449 | map.put(node.getKey(), node.getValue()); |
| 450 | node = findValue(_root, o); |
| 451 | } |
| 452 | } |
| 453 | |
| 454 | clear(); |
| 455 | putAll(map); |
| 456 | |
| 457 | return (size != size()); |
| 458 | } |
| 459 | } |
| 460 | |
| 461 | int _size, _hashCode, _dimensions; |
| 462 | KDNode _root; |
| 463 | |
| 464 | KDNode getNode(P point, KDNode[] parent) { |
| 465 | int discriminator; |
| 466 | KDNode node = _root, current, last = null; |
| 467 | Coord c1, c2; |
| 468 | |
| 469 | while(node != null) { |
| 470 | discriminator = node._discriminator; |
| 471 | c1 = point.getCoord(discriminator); |
| 472 | c2 = node._point.getCoord(discriminator); |
| 473 | current = node; |
| 474 | |
| 475 | if(c1.compareTo(c2) > 0) |
| 476 | node = node._high; |
| 477 | else if(c1.compareTo(c2) < 0) |
| 478 | node = node._low; |
| 479 | else if(node._point.equals(point)) { |
| 480 | if(parent != null) |
| 481 | parent[0] = last; |
| 482 | return node; |
| 483 | } else |
| 484 | node = node._high; |
| 485 | |
| 486 | last = current; |
| 487 | } |
| 488 | |
| 489 | if(parent != null) |
| 490 | parent[0] = last; |
| 491 | |
| 492 | return null; |
| 493 | } |
| 494 | |
| 495 | KDNode getNode(P point) { |
| 496 | return getNode(point, null); |
| 497 | } |
| 498 | |
| 499 | KDNode getMinimumNode(KDNode node, KDNode p, int discriminator, |
| 500 | KDNode[] parent) |
| 501 | { |
| 502 | KDNode result; |
| 503 | |
| 504 | if(discriminator == node._discriminator) { |
| 505 | if(node._low != null) |
| 506 | return getMinimumNode(node._low, node, discriminator, parent); |
| 507 | else |
| 508 | result = node; |
| 509 | } else { |
| 510 | KDNode nlow = null, nhigh = null; |
| 511 | KDNode[] plow = new KDTree.KDNode[1], phigh = new KDTree.KDNode[1]; |
| 512 | |
| 513 | if(node._low != null) |
| 514 | nlow = getMinimumNode(node._low, node, discriminator, plow); |
| 515 | |
| 516 | if(node._high != null) |
| 517 | nhigh = getMinimumNode(node._high, node, discriminator, phigh); |
| 518 | |
| 519 | if(nlow != null && nhigh != null) { |
| 520 | if(nlow._point.getCoord(discriminator).compareTo(nhigh._point.getCoord(discriminator)) < 0) { |
| 521 | result = nlow; |
| 522 | parent[0] = plow[0]; |
| 523 | } else { |
| 524 | result = nhigh; |
| 525 | parent[0] = phigh[0]; |
| 526 | } |
| 527 | } else if(nlow != null) { |
| 528 | result = nlow; |
| 529 | parent[0] = plow[0]; |
| 530 | } else if(nhigh != null) { |
| 531 | result = nhigh; |
| 532 | parent[0] = phigh[0]; |
| 533 | } else |
| 534 | result = node; |
| 535 | } |
| 536 | |
| 537 | if(result == node) |
| 538 | parent[0] = p; |
| 539 | else if(node._point.getCoord(discriminator).compareTo(result._point.getCoord(discriminator)) < 0) { |
| 540 | result = node; |
| 541 | parent[0] = p; |
| 542 | } |
| 543 | |
| 544 | return result; |
| 545 | } |
| 546 | |
| 547 | KDNode recursiveRemoveNode(KDNode node) { |
| 548 | int discriminator; |
| 549 | |
| 550 | if(node._low == null && node._high == null) |
| 551 | return null; |
| 552 | else |
| 553 | discriminator = node._discriminator; |
| 554 | |
| 555 | if(node._high == null) { |
| 556 | node._high = node._low; |
| 557 | node._low = null; |
| 558 | } |
| 559 | |
| 560 | KDNode[] parent = new KDTree.KDNode[1]; |
| 561 | KDNode newRoot = |
| 562 | getMinimumNode(node._high, node, discriminator, parent); |
| 563 | KDNode child = recursiveRemoveNode(newRoot); |
| 564 | |
| 565 | if(parent[0]._low == newRoot) |
| 566 | parent[0]._low = child; |
| 567 | else |
| 568 | parent[0]._high = child; |
| 569 | |
| 570 | newRoot._low = node._low; |
| 571 | newRoot._high = node._high; |
| 572 | newRoot._discriminator = node._discriminator; |
| 573 | |
| 574 | return newRoot; |
| 575 | } |
| 576 | |
| 577 | KDNode findValue(KDNode node, Object value) { |
| 578 | if(node == null || (value == null ? node.getValue() == null : |
| 579 | value.equals(node.getValue()))) |
| 580 | return node; |
| 581 | |
| 582 | KDNode result; |
| 583 | |
| 584 | if((result = findValue(node._low, value)) == null) |
| 585 | result = findValue(node._high, value); |
| 586 | |
| 587 | return result; |
| 588 | } |
| 589 | |
| 590 | boolean isInRange(P point, P lower, P upper) { |
| 591 | Coord coordinate1, coordinate2 = null, coordinate3 = null; |
| 592 | |
| 593 | if(lower != null || upper != null) { |
| 594 | int dimensions; |
| 595 | dimensions = point.getDimensions(); |
| 596 | |
| 597 | for(int i = 0; i < dimensions; ++i) { |
| 598 | coordinate1 = point.getCoord(i); |
| 599 | if(lower != null) |
| 600 | coordinate2 = lower.getCoord(i); |
| 601 | if(upper != null) |
| 602 | coordinate3 = upper.getCoord(i); |
| 603 | if((coordinate2 != null && coordinate1.compareTo(coordinate2) < 0) || |
| 604 | (coordinate3 != null && coordinate1.compareTo(coordinate3) > 0)) |
| 605 | return false; |
| 606 | } |
| 607 | } |
| 608 | |
| 609 | return true; |
| 610 | } |
| 611 | |
| 612 | /** |
| 613 | * Creates a two-dimensional KDTree. |
| 614 | */ |
| 615 | public KDTree() { |
| 616 | this(2); |
| 617 | } |
| 618 | |
| 619 | /** |
| 620 | * Creates a KDTree of the specified number of dimensions. |
| 621 | * |
| 622 | * @param dimensions The number of dimensions. Must be greater than 0. |
| 623 | */ |
| 624 | public KDTree(int dimensions) { |
| 625 | assert(dimensions > 0); |
| 626 | _dimensions = dimensions; |
| 627 | clear(); |
| 628 | } |
| 629 | |
| 630 | // Begin Map interface methods |
| 631 | |
| 632 | /** |
| 633 | * Removes all elements from the container, leaving it empty. |
| 634 | */ |
| 635 | public void clear() { |
| 636 | _root = null; |
| 637 | _size = _hashCode = 0; |
| 638 | } |
| 639 | |
| 640 | /** |
| 641 | * Returns true if the container contains a mapping for the specified key. |
| 642 | * |
| 643 | * @param key The point key to search for. |
| 644 | * @return true if the container contains a mapping for the specified key. |
| 645 | * @exception ClassCastException if the key is not an instance of P. |
| 646 | */ |
| 647 | public boolean containsKey(Object key) |
| 648 | throws ClassCastException |
| 649 | { |
| 650 | return (getNode((P)key) != null); |
| 651 | } |
| 652 | |
| 653 | /** |
| 654 | * Returns true if the container contains a mapping with the specified value. |
| 655 | * Note: this is very inefficient for KDTrees because it requires searching |
| 656 | * the entire tree. |
| 657 | * |
| 658 | * @param value The value to search for. |
| 659 | * @return true If the container contains a mapping with the specified value. |
| 660 | */ |
| 661 | public boolean containsValue(Object value) { |
| 662 | return (findValue(_root, value) != null); |
| 663 | } |
| 664 | |
| 665 | /** |
| 666 | * Returns a Set view of the point to value mappings in the KDTree. |
| 667 | * Modifications to the resulting set will be reflected in the KDTree |
| 668 | * and vice versa, except that {@code Iterator.remove} is not supported. |
| 669 | * |
| 670 | * @return A Set view of the point to value mappings in the KDTree. |
| 671 | */ |
| 672 | public Set<Map.Entry<P,V>> entrySet() { |
| 673 | return new MapEntrySet(); |
| 674 | } |
| 675 | |
| 676 | /** |
| 677 | * Returns true if the object contains the same mappings, false if not. |
| 678 | * |
| 679 | * @param o The object to test for equality. |
| 680 | * @return true if the object contains the same mappings, false if not. |
| 681 | */ |
| 682 | public boolean equals(Object o) |
| 683 | throws ClassCastException |
| 684 | { |
| 685 | if(!(o instanceof Map)) |
| 686 | return false; |
| 687 | |
| 688 | if(o == this) |
| 689 | return true; |
| 690 | |
| 691 | Map map = (Map)o; |
| 692 | |
| 693 | return (entrySet().equals(map.entrySet())); |
| 694 | } |
| 695 | |
| 696 | /** |
| 697 | * Retrieves the value at the given location. |
| 698 | * |
| 699 | * @param point The location from which to retrieve the value. |
| 700 | * @return The value at the given location, or null if no value is present. |
| 701 | * @exception ClassCastException If the given point is not of the |
| 702 | * expected type. |
| 703 | */ |
| 704 | public V get(Object point) throws ClassCastException { |
| 705 | KDNode node = getNode((P)point); |
| 706 | |
| 707 | return (node == null ? null : node.getValue()); |
| 708 | } |
| 709 | |
| 710 | /** |
| 711 | * Returns the hash code value for this map. |
| 712 | * |
| 713 | * @return The sum of the hash codes of all of the map entries. |
| 714 | */ |
| 715 | public int hashCode() { |
| 716 | return _hashCode; |
| 717 | } |
| 718 | |
| 719 | /** |
| 720 | * Returns true if the container has no elements, false if it |
| 721 | * contains one or more elements. |
| 722 | * |
| 723 | * @return true if the container has no elements, false if it |
| 724 | * contains one or more elements. |
| 725 | */ |
| 726 | public boolean isEmpty() { |
| 727 | return (_root == null); |
| 728 | } |
| 729 | |
| 730 | /** |
| 731 | * Returns a Set view of the point keys for the mappings in the |
| 732 | * KDTree. Changes to the Set are reflected in the KDTree and vice |
| 733 | * versa, except that {@code Iterator.remove} is not supported. |
| 734 | * |
| 735 | * @return A Set view of the point keys for the mappings in the KDTree. |
| 736 | */ |
| 737 | public Set<P> keySet() { |
| 738 | return new KeySet(); |
| 739 | } |
| 740 | |
| 741 | /** |
| 742 | * Inserts a point value pair into the tree, preserving the |
| 743 | * spatial ordering. |
| 744 | * |
| 745 | * @param point The point serving as a key. |
| 746 | * @param value The value to insert at the point. |
| 747 | * @return The old value if an existing value is replaced by the |
| 748 | * inserted value. |
| 749 | */ |
| 750 | public V put(P point, V value) { |
| 751 | KDNode[] parent = new KDTree.KDNode[1]; |
| 752 | KDNode node = getNode(point, parent); |
| 753 | V old = null; |
| 754 | |
| 755 | if(node != null) { |
| 756 | old = node.getValue(); |
| 757 | _hashCode-=node.hashCode(); |
| 758 | node._value = value; |
| 759 | } else { |
| 760 | if(parent[0] == null) |
| 761 | node = _root = new KDNode(0, point, value); |
| 762 | else { |
| 763 | int discriminator = parent[0]._discriminator; |
| 764 | |
| 765 | if(point.getCoord(discriminator).compareTo( |
| 766 | parent[0]._point.getCoord(discriminator)) >= 0) |
| 767 | node = parent[0]._high = |
| 768 | new KDNode((discriminator + 1) % _dimensions, point, value); |
| 769 | else |
| 770 | node = parent[0]._low = |
| 771 | new KDNode((discriminator + 1) % _dimensions, point, value); |
| 772 | } |
| 773 | |
| 774 | ++_size; |
| 775 | } |
| 776 | |
| 777 | _hashCode+=node.hashCode(); |
| 778 | |
| 779 | return old; |
| 780 | } |
| 781 | |
| 782 | /** |
| 783 | * Copies all of the point-value mappings from the given Map into the KDTree. |
| 784 | * |
| 785 | * @param map The Map from which to copy the mappings. |
| 786 | */ |
| 787 | public void putAll(Map<? extends P, ? extends V> map) { |
| 788 | for(Map.Entry<? extends P, ? extends V> pair : map.entrySet()) |
| 789 | put(pair.getKey(), pair.getValue()); |
| 790 | } |
| 791 | |
| 792 | /** |
| 793 | * Removes the point-value mapping corresponding to the given point key. |
| 794 | * |
| 795 | * @param key The point key of the mapping to remove. |
| 796 | * @return The value part of the mapping, if a mapping existed and |
| 797 | * was removed. Null if not. |
| 798 | * @exception ClassCastException If the key is not an instance of P. |
| 799 | */ |
| 800 | public V remove(Object key) |
| 801 | throws ClassCastException |
| 802 | { |
| 803 | KDNode[] parent = new KDTree.KDNode[1]; |
| 804 | KDNode node = getNode((P)key, parent); |
| 805 | V old = null; |
| 806 | |
| 807 | if(node != null) { |
| 808 | KDNode child = node; |
| 809 | |
| 810 | node = recursiveRemoveNode(child); |
| 811 | |
| 812 | if(parent[0] == null) |
| 813 | _root = node; |
| 814 | else if(child == parent[0]._low) |
| 815 | parent[0]._low = node; |
| 816 | else if(child == parent[0]._high) |
| 817 | parent[0]._high = node; |
| 818 | |
| 819 | --_size; |
| 820 | _hashCode-=child.hashCode(); |
| 821 | old = child.getValue(); |
| 822 | } |
| 823 | |
| 824 | return old; |
| 825 | } |
| 826 | |
| 827 | /** |
| 828 | * Returns the number of point-value mappings in the KDTree. |
| 829 | * |
| 830 | * @return The number of point-value mappings in the KDTree. |
| 831 | */ |
| 832 | public int size() { |
| 833 | return _size; |
| 834 | } |
| 835 | |
| 836 | /** |
| 837 | * Returns a Collection view of the values contained in the KDTree. |
| 838 | * Changes to the Collection are reflected in the KDTree and vice versa. |
| 839 | * Note: the resulting Collection is very inefficient. |
| 840 | * |
| 841 | * @return A Collection view of the values contained in the KDTree. |
| 842 | */ |
| 843 | public Collection<V> values() { |
| 844 | return new ValueCollection(); |
| 845 | } |
| 846 | |
| 847 | // End Map interface methods |
| 848 | |
| 849 | public Iterator<Map.Entry<P,V>> iterator(P lower, P upper) { |
| 850 | return new MapEntryIterator(lower, upper); |
| 851 | } |
| 852 | |
| 853 | int fillArray(KDNode[] a, int index, KDNode node) { |
| 854 | if(node == null) |
| 855 | return index; |
| 856 | a[index] = node; |
| 857 | index = fillArray(a, index + 1, node._low); |
| 858 | return fillArray(a, index, node._high); |
| 859 | } |
| 860 | |
| 861 | final class NodeComparator implements Comparator<KDNode> { |
| 862 | int _discriminator = 0; |
| 863 | |
| 864 | void setDiscriminator(int val) { |
| 865 | _discriminator = val; |
| 866 | } |
| 867 | |
| 868 | int getDiscriminator() { |
| 869 | return _discriminator; |
| 870 | } |
| 871 | |
| 872 | public int compare(KDNode n1, KDNode n2) { |
| 873 | return |
| 874 | n1._point.getCoord(_discriminator).compareTo(n2._point.getCoord(_discriminator)); |
| 875 | } |
| 876 | } |
| 877 | |
| 878 | KDNode optimize(KDNode[] nodes, int begin, int end, NodeComparator comp) { |
| 879 | KDNode midpoint= null; |
| 880 | int size = end - begin; |
| 881 | |
| 882 | if(size > 1) { |
| 883 | int nth = begin + (size >> 1); |
| 884 | int nthprev = nth - 1; |
| 885 | int d = comp.getDiscriminator(); |
| 886 | |
| 887 | Arrays.sort(nodes, begin, end, comp); |
| 888 | |
| 889 | while(nth > begin && |
| 890 | nodes[nth]._point.getCoord(d).compareTo( |
| 891 | nodes[nthprev]._point.getCoord(d)) == 0) |
| 892 | { |
| 893 | --nth; |
| 894 | --nthprev; |
| 895 | } |
| 896 | |
| 897 | midpoint = nodes[nth]; |
| 898 | midpoint._discriminator = d; |
| 899 | |
| 900 | if(++d >= _dimensions) |
| 901 | d = 0; |
| 902 | |
| 903 | comp.setDiscriminator(d); |
| 904 | |
| 905 | midpoint._low = optimize(nodes, begin, nth, comp); |
| 906 | |
| 907 | comp.setDiscriminator(d); |
| 908 | |
| 909 | midpoint._high = optimize(nodes, nth + 1, end, comp); |
| 910 | } else if(size == 1) { |
| 911 | midpoint = nodes[begin]; |
| 912 | midpoint._discriminator = comp.getDiscriminator(); |
| 913 | midpoint._low = midpoint._high = null; |
| 914 | } |
| 915 | |
| 916 | return midpoint; |
| 917 | } |
| 918 | |
| 919 | /** |
| 920 | * Optimizes the performance of future search operations by balancing the |
| 921 | * KDTree. The balancing operation is relatively expensive, but can |
| 922 | * significantly improve the performance of searches. Usually, you |
| 923 | * don't have to optimize a tree which contains random key values |
| 924 | * inserted in a random order. |
| 925 | */ |
| 926 | public void optimize() { |
| 927 | if(isEmpty()) |
| 928 | return; |
| 929 | |
| 930 | KDNode[] nodes = |
| 931 | (KDNode[])Array.newInstance(KDNode.class, size()); |
| 932 | fillArray(nodes, 0, _root); |
| 933 | |
| 934 | _root = optimize(nodes, 0, nodes.length, new NodeComparator()); |
| 935 | } |
| 936 | } |
