60#include <unordered_set>
64#define NANOFLANN_VERSION 0x170
67#if !defined(NOMINMAX) && \
68 (defined(_WIN32) || defined(_WIN32_) || defined(WIN32) || defined(_WIN64))
89 return static_cast<T
>(3.14159265358979323846);
96template <
typename T,
typename =
int>
107template <
typename T,
typename =
int>
121template <
typename Container>
122inline typename std::enable_if<has_resize<Container>::value,
void>::type
resize(
123 Container& c,
const size_t nElements)
132template <
typename Container>
133inline typename std::enable_if<!has_resize<Container>::value,
void>::type
134 resize(Container& c,
const size_t nElements)
136 if (nElements != c.size())
137 throw std::logic_error(
"Try to change the size of a std::array.");
143template <
typename Container,
typename T>
144inline typename std::enable_if<has_assign<Container>::value,
void>::type
assign(
145 Container& c,
const size_t nElements,
const T& value)
147 c.assign(nElements, value);
153template <
typename Container,
typename T>
154inline typename std::enable_if<!has_assign<Container>::value,
void>::type
155 assign(Container& c,
const size_t nElements,
const T& value)
157 for (
size_t i = 0; i < nElements; i++) c[i] = value;
164 template <
typename PairType>
165 bool operator()(
const PairType& p1,
const PairType& p2)
const
167 return p1.second < p2.second;
179template <
typename IndexType =
size_t,
typename DistanceType =
double>
183 ResultItem(
const IndexType index,
const DistanceType distance)
197 typename _DistanceType,
typename _IndexType = size_t,
198 typename _CountType =
size_t>
202 using DistanceType = _DistanceType;
203 using IndexType = _IndexType;
204 using CountType = _CountType;
214 : indices(
nullptr), dists(
nullptr), capacity(capacity_), count(0)
218 void init(IndexType* indices_, DistanceType* dists_)
225 CountType size()
const {
return count; }
226 bool empty()
const {
return count == 0; }
227 bool full()
const {
return count == capacity; }
237 for (i = count; i > 0; --i)
241#ifdef NANOFLANN_FIRST_MATCH
242 if ((dists[i - 1] > dist) ||
243 ((dist == dists[i - 1]) && (indices[i - 1] > index)))
246 if (dists[i - 1] > dist)
251 dists[i] = dists[i - 1];
252 indices[i] = indices[i - 1];
263 if (count < capacity) count++;
273 return count < capacity ? std::numeric_limits<DistanceType>::max()
285 typename _DistanceType,
typename _IndexType = size_t,
286 typename _CountType =
size_t>
290 using DistanceType = _DistanceType;
291 using IndexType = _IndexType;
292 using CountType = _CountType;
299 DistanceType maximumSearchDistanceSquared;
303 CountType capacity_, DistanceType maximumSearchDistanceSquared_)
308 maximumSearchDistanceSquared(maximumSearchDistanceSquared_)
312 void init(IndexType* indices_, DistanceType* dists_)
317 if (capacity) dists[capacity - 1] = maximumSearchDistanceSquared;
320 CountType size()
const {
return count; }
321 bool empty()
const {
return count == 0; }
322 bool full()
const {
return count == capacity; }
332 for (i = count; i > 0; --i)
336#ifdef NANOFLANN_FIRST_MATCH
337 if ((dists[i - 1] > dist) ||
338 ((dist == dists[i - 1]) && (indices[i - 1] > index)))
341 if (dists[i - 1] > dist)
346 dists[i] = dists[i - 1];
347 indices[i] = indices[i - 1];
358 if (count < capacity) count++;
368 return count < capacity ? maximumSearchDistanceSquared
381template <
typename _DistanceType,
typename _IndexType =
size_t>
385 using DistanceType = _DistanceType;
386 using IndexType = _IndexType;
389 const DistanceType radius;
391 std::vector<ResultItem<IndexType, DistanceType>>& m_indices_dists;
394 DistanceType radius_,
396 : radius(radius_), m_indices_dists(indices_dists)
401 void init() { clear(); }
402 void clear() { m_indices_dists.clear(); }
404 size_t size()
const {
return m_indices_dists.size(); }
405 size_t empty()
const {
return m_indices_dists.empty(); }
407 bool full()
const {
return true; }
416 if (dist < radius) m_indices_dists.emplace_back(index, dist);
420 DistanceType worstDist()
const {
return radius; }
428 if (m_indices_dists.empty())
429 throw std::runtime_error(
430 "Cannot invoke RadiusResultSet::worst_item() on "
431 "an empty list of results.");
432 auto it = std::max_element(
449void save_value(std::ostream& stream,
const T& value)
451 stream.write(
reinterpret_cast<const char*
>(&value),
sizeof(T));
455void save_value(std::ostream& stream,
const std::vector<T>& value)
457 size_t size = value.size();
458 stream.write(
reinterpret_cast<const char*
>(&size),
sizeof(
size_t));
459 stream.write(
reinterpret_cast<const char*
>(value.data()),
sizeof(T) * size);
463void load_value(std::istream& stream, T& value)
465 stream.read(
reinterpret_cast<char*
>(&value),
sizeof(T));
469void load_value(std::istream& stream, std::vector<T>& value)
472 stream.read(
reinterpret_cast<char*
>(&size),
sizeof(
size_t));
474 stream.read(
reinterpret_cast<char*
>(value.data()),
sizeof(T) * size);
496 class T,
class DataSource,
typename _DistanceType = T,
497 typename IndexType = uint32_t>
500 using ElementType = T;
501 using DistanceType = _DistanceType;
503 const DataSource& data_source;
505 L1_Adaptor(
const DataSource& _data_source) : data_source(_data_source) {}
507 DistanceType evalMetric(
508 const T* a,
const IndexType b_idx,
size_t size,
509 DistanceType worst_dist = -1)
const
511 DistanceType result = DistanceType();
512 const T* last = a + size;
513 const T* lastgroup = last - 3;
517 while (a < lastgroup)
519 const DistanceType diff0 =
520 std::abs(a[0] - data_source.kdtree_get_pt(b_idx, d++));
521 const DistanceType diff1 =
522 std::abs(a[1] - data_source.kdtree_get_pt(b_idx, d++));
523 const DistanceType diff2 =
524 std::abs(a[2] - data_source.kdtree_get_pt(b_idx, d++));
525 const DistanceType diff3 =
526 std::abs(a[3] - data_source.kdtree_get_pt(b_idx, d++));
527 result += diff0 + diff1 + diff2 + diff3;
529 if ((worst_dist > 0) && (result > worst_dist)) {
return result; }
535 result += std::abs(*a++ - data_source.kdtree_get_pt(b_idx, d++));
540 template <
typename U,
typename V>
541 DistanceType accum_dist(
const U a,
const V b,
const size_t)
const
543 return std::abs(a - b);
558 class T,
class DataSource,
typename _DistanceType = T,
559 typename IndexType = uint32_t>
562 using ElementType = T;
563 using DistanceType = _DistanceType;
565 const DataSource& data_source;
567 L2_Adaptor(
const DataSource& _data_source) : data_source(_data_source) {}
569 DistanceType evalMetric(
570 const T* a,
const IndexType b_idx,
size_t size,
571 DistanceType worst_dist = -1)
const
573 DistanceType result = DistanceType();
574 const T* last = a + size;
575 const T* lastgroup = last - 3;
579 while (a < lastgroup)
581 const DistanceType diff0 =
582 a[0] - data_source.kdtree_get_pt(b_idx, d++);
583 const DistanceType diff1 =
584 a[1] - data_source.kdtree_get_pt(b_idx, d++);
585 const DistanceType diff2 =
586 a[2] - data_source.kdtree_get_pt(b_idx, d++);
587 const DistanceType diff3 =
588 a[3] - data_source.kdtree_get_pt(b_idx, d++);
590 diff0 * diff0 + diff1 * diff1 + diff2 * diff2 + diff3 * diff3;
592 if ((worst_dist > 0) && (result > worst_dist)) {
return result; }
598 const DistanceType diff0 =
599 *a++ - data_source.kdtree_get_pt(b_idx, d++);
600 result += diff0 * diff0;
605 template <
typename U,
typename V>
606 DistanceType accum_dist(
const U a,
const V b,
const size_t)
const
608 return (a - b) * (a - b);
623 class T,
class DataSource,
typename _DistanceType = T,
624 typename IndexType = uint32_t>
627 using ElementType = T;
628 using DistanceType = _DistanceType;
630 const DataSource& data_source;
633 : data_source(_data_source)
637 DistanceType evalMetric(
638 const T* a,
const IndexType b_idx,
size_t size)
const
640 DistanceType result = DistanceType();
641 for (
size_t i = 0; i < size; ++i)
643 const DistanceType diff =
644 a[i] - data_source.kdtree_get_pt(b_idx, i);
645 result += diff * diff;
650 template <
typename U,
typename V>
651 DistanceType accum_dist(
const U a,
const V b,
const size_t)
const
653 return (a - b) * (a - b);
668 class T,
class DataSource,
typename _DistanceType = T,
669 typename IndexType = uint32_t>
672 using ElementType = T;
673 using DistanceType = _DistanceType;
675 const DataSource& data_source;
677 SO2_Adaptor(
const DataSource& _data_source) : data_source(_data_source) {}
679 DistanceType evalMetric(
680 const T* a,
const IndexType b_idx,
size_t size)
const
683 a[size - 1], data_source.kdtree_get_pt(b_idx, size - 1), size - 1);
688 template <
typename U,
typename V>
689 DistanceType
accum_dist(
const U a,
const V b,
const size_t)
const
691 DistanceType result = DistanceType();
692 DistanceType PI = pi_const<DistanceType>();
696 else if (result < -PI)
713 class T,
class DataSource,
typename _DistanceType = T,
714 typename IndexType = uint32_t>
717 using ElementType = T;
718 using DistanceType = _DistanceType;
724 : distance_L2_Simple(_data_source)
728 DistanceType evalMetric(
729 const T* a,
const IndexType b_idx,
size_t size)
const
731 return distance_L2_Simple.evalMetric(a, b_idx, size);
734 template <
typename U,
typename V>
735 DistanceType accum_dist(
const U a,
const V b,
const size_t idx)
const
737 return distance_L2_Simple.accum_dist(a, b, idx);
744 template <
class T,
class DataSource,
typename IndexType = u
int32_t>
754 template <
class T,
class DataSource,
typename IndexType = u
int32_t>
764 template <
class T,
class DataSource,
typename IndexType = u
int32_t>
773 template <
class T,
class DataSource,
typename IndexType = u
int32_t>
782 template <
class T,
class DataSource,
typename IndexType = u
int32_t>
794enum class KDTreeSingleIndexAdaptorFlags
797 SkipInitialBuildIndex = 1
800inline std::underlying_type<KDTreeSingleIndexAdaptorFlags>::type operator&(
801 KDTreeSingleIndexAdaptorFlags lhs, KDTreeSingleIndexAdaptorFlags rhs)
804 typename std::underlying_type<KDTreeSingleIndexAdaptorFlags>::type;
805 return static_cast<underlying
>(lhs) &
static_cast<underlying
>(rhs);
812 size_t _leaf_max_size = 10,
813 KDTreeSingleIndexAdaptorFlags _flags =
814 KDTreeSingleIndexAdaptorFlags::None,
815 unsigned int _n_thread_build = 1)
816 : leaf_max_size(_leaf_max_size),
818 n_thread_build(_n_thread_build)
822 size_t leaf_max_size;
823 KDTreeSingleIndexAdaptorFlags flags;
824 unsigned int n_thread_build;
831 : eps(eps_), sorted(sorted_)
860 static constexpr size_t WORDSIZE = 16;
861 static constexpr size_t BLOCKSIZE = 8192;
872 void* base_ =
nullptr;
873 void* loc_ =
nullptr;
885 Size wastedMemory = 0;
900 while (base_ !=
nullptr)
903 void* prev = *(
static_cast<void**
>(base_));
920 const Size size = (req_size + (WORDSIZE - 1)) & ~(WORDSIZE - 1);
925 if (size > remaining_)
927 wastedMemory += remaining_;
930 const Size blocksize =
931 size > BLOCKSIZE ? size + WORDSIZE : BLOCKSIZE + WORDSIZE;
934 void* m = ::malloc(blocksize);
937 fprintf(stderr,
"Failed to allocate memory.\n");
938 throw std::bad_alloc();
942 static_cast<void**
>(m)[0] = base_;
945 remaining_ = blocksize - WORDSIZE;
946 loc_ =
static_cast<char*
>(m) + WORDSIZE;
949 loc_ =
static_cast<char*
>(loc_) + size;
964 template <
typename T>
967 T* mem =
static_cast<T*
>(this->malloc(
sizeof(T) * count));
979template <
int32_t DIM,
typename T>
982 using type = std::array<T, DIM>;
988 using type = std::vector<T>;
1008 class Derived,
typename Distance,
class DatasetAdaptor, int32_t DIM = -1,
1009 typename index_t = uint32_t>
1017 obj.pool_.free_all();
1018 obj.root_node_ =
nullptr;
1019 obj.size_at_index_build_ = 0;
1022 using ElementType =
typename Distance::ElementType;
1023 using DistanceType =
typename Distance::DistanceType;
1024 using IndexType = index_t;
1031 using Offset =
typename decltype(vAcc_)::size_type;
1032 using Size =
typename decltype(vAcc_)::size_type;
1033 using Dimension = int32_t;
1057 Node *child1 =
nullptr, *child2 =
nullptr;
1065 ElementType low, high;
1070 Size leaf_max_size_ = 0;
1073 Size n_thread_build_ = 1;
1077 Size size_at_index_build_ = 0;
1101 Size
size(
const Derived& obj)
const {
return obj.size_; }
1104 Size
veclen(
const Derived& obj) {
return DIM > 0 ? DIM : obj.dim; }
1108 const Derived& obj, IndexType element, Dimension component)
const
1110 return obj.dataset_.kdtree_get_pt(element, component);
1119 return obj.pool_.usedMemory + obj.pool_.wastedMemory +
1120 obj.dataset_.kdtree_get_point_count() *
1125 const Derived& obj, Offset ind, Size count, Dimension element,
1126 ElementType& min_elem, ElementType& max_elem)
1128 min_elem = dataset_get(obj, vAcc_[ind], element);
1129 max_elem = min_elem;
1130 for (Offset i = 1; i < count; ++i)
1132 ElementType val = dataset_get(obj, vAcc_[ind + i], element);
1133 if (val < min_elem) min_elem = val;
1134 if (val > max_elem) max_elem = val;
1146 Derived& obj,
const Offset left,
const Offset right,
BoundingBox& bbox)
1148 assert(left < obj.dataset_.kdtree_get_point_count());
1150 NodePtr node = obj.pool_.template allocate<Node>();
1151 const auto dims = (DIM > 0 ? DIM : obj.dim_);
1154 if ((right - left) <=
static_cast<Offset
>(obj.leaf_max_size_))
1156 node->
child1 = node->child2 =
nullptr;
1161 for (Dimension i = 0; i < dims; ++i)
1163 bbox[i].low = dataset_get(obj, obj.vAcc_[left], i);
1164 bbox[i].high = dataset_get(obj, obj.vAcc_[left], i);
1166 for (Offset k = left + 1; k < right; ++k)
1168 for (Dimension i = 0; i < dims; ++i)
1170 const auto val = dataset_get(obj, obj.vAcc_[k], i);
1171 if (bbox[i].low > val) bbox[i].low = val;
1172 if (bbox[i].high < val) bbox[i].high = val;
1180 DistanceType cutval;
1181 middleSplit_(obj, left, right - left, idx, cutfeat, cutval, bbox);
1186 left_bbox[cutfeat].high = cutval;
1187 node->
child1 = this->divideTree(obj, left, left + idx, left_bbox);
1190 right_bbox[cutfeat].low = cutval;
1191 node->child2 = this->divideTree(obj, left + idx, right, right_bbox);
1193 node->
node_type.sub.divlow = left_bbox[cutfeat].high;
1194 node->
node_type.sub.divhigh = right_bbox[cutfeat].low;
1196 for (Dimension i = 0; i < dims; ++i)
1198 bbox[i].low = std::min(left_bbox[i].low, right_bbox[i].low);
1199 bbox[i].high = std::max(left_bbox[i].high, right_bbox[i].high);
1217 Derived& obj,
const Offset left,
const Offset right,
BoundingBox& bbox,
1218 std::atomic<unsigned int>& thread_count, std::mutex& mutex)
1220 std::unique_lock<std::mutex> lock(mutex);
1221 NodePtr node = obj.pool_.template allocate<Node>();
1224 const auto dims = (DIM > 0 ? DIM : obj.dim_);
1227 if ((right - left) <=
static_cast<Offset
>(obj.leaf_max_size_))
1229 node->
child1 = node->child2 =
nullptr;
1234 for (Dimension i = 0; i < dims; ++i)
1236 bbox[i].low = dataset_get(obj, obj.vAcc_[left], i);
1237 bbox[i].high = dataset_get(obj, obj.vAcc_[left], i);
1239 for (Offset k = left + 1; k < right; ++k)
1241 for (Dimension i = 0; i < dims; ++i)
1243 const auto val = dataset_get(obj, obj.vAcc_[k], i);
1244 if (bbox[i].low > val) bbox[i].low = val;
1245 if (bbox[i].high < val) bbox[i].high = val;
1253 DistanceType cutval;
1254 middleSplit_(obj, left, right - left, idx, cutfeat, cutval, bbox);
1258 std::future<NodePtr> right_future;
1261 right_bbox[cutfeat].low = cutval;
1262 if (++thread_count < n_thread_build_)
1265 right_future = std::async(
1266 std::launch::async, &KDTreeBaseClass::divideTreeConcurrent,
1267 this, std::ref(obj), left + idx, right,
1268 std::ref(right_bbox), std::ref(thread_count),
1271 else { --thread_count; }
1274 left_bbox[cutfeat].high = cutval;
1275 node->
child1 = this->divideTreeConcurrent(
1276 obj, left, left + idx, left_bbox, thread_count, mutex);
1278 if (right_future.valid())
1281 node->child2 = right_future.get();
1286 node->child2 = this->divideTreeConcurrent(
1287 obj, left + idx, right, right_bbox, thread_count, mutex);
1290 node->
node_type.sub.divlow = left_bbox[cutfeat].high;
1291 node->
node_type.sub.divhigh = right_bbox[cutfeat].low;
1293 for (Dimension i = 0; i < dims; ++i)
1295 bbox[i].low = std::min(left_bbox[i].low, right_bbox[i].low);
1296 bbox[i].high = std::max(left_bbox[i].high, right_bbox[i].high);
1304 const Derived& obj,
const Offset ind,
const Size count, Offset& index,
1305 Dimension& cutfeat, DistanceType& cutval,
const BoundingBox& bbox)
1307 const auto dims = (DIM > 0 ? DIM : obj.dim_);
1308 const auto EPS =
static_cast<DistanceType
>(0.00001);
1309 ElementType max_span = bbox[0].high - bbox[0].low;
1310 for (Dimension i = 1; i < dims; ++i)
1312 ElementType span = bbox[i].high - bbox[i].low;
1313 if (span > max_span) { max_span = span; }
1315 ElementType max_spread = -1;
1317 ElementType min_elem = 0, max_elem = 0;
1318 for (Dimension i = 0; i < dims; ++i)
1320 ElementType span = bbox[i].high - bbox[i].low;
1321 if (span >= (1 - EPS) * max_span)
1323 ElementType min_elem_, max_elem_;
1324 computeMinMax(obj, ind, count, i, min_elem_, max_elem_);
1325 ElementType spread = max_elem_ - min_elem_;
1326 if (spread > max_spread)
1329 max_spread = spread;
1330 min_elem = min_elem_;
1331 max_elem = max_elem_;
1336 DistanceType split_val = (bbox[cutfeat].low + bbox[cutfeat].high) / 2;
1338 if (split_val < min_elem)
1340 else if (split_val > max_elem)
1346 planeSplit(obj, ind, count, cutfeat, cutval, lim1, lim2);
1348 if (lim1 > count / 2)
1350 else if (lim2 < count / 2)
1366 const Derived& obj,
const Offset ind,
const Size count,
1367 const Dimension cutfeat,
const DistanceType& cutval, Offset& lim1,
1372 Offset right = count - 1;
1375 while (left <= right &&
1376 dataset_get(obj, vAcc_[ind + left], cutfeat) < cutval)
1378 while (right && left <= right &&
1379 dataset_get(obj, vAcc_[ind + right], cutfeat) >= cutval)
1381 if (left > right || !right)
1383 std::swap(vAcc_[ind + left], vAcc_[ind + right]);
1394 while (left <= right &&
1395 dataset_get(obj, vAcc_[ind + left], cutfeat) <= cutval)
1397 while (right && left <= right &&
1398 dataset_get(obj, vAcc_[ind + right], cutfeat) > cutval)
1400 if (left > right || !right)
1402 std::swap(vAcc_[ind + left], vAcc_[ind + right]);
1409 DistanceType computeInitialDistances(
1410 const Derived& obj,
const ElementType* vec,
1411 distance_vector_t& dists)
const
1414 DistanceType dist = DistanceType();
1416 for (Dimension i = 0; i < (DIM > 0 ? DIM : obj.dim_); ++i)
1418 if (vec[i] < obj.root_bbox_[i].low)
1421 obj.distance_.accum_dist(vec[i], obj.root_bbox_[i].low, i);
1424 if (vec[i] > obj.root_bbox_[i].high)
1427 obj.distance_.accum_dist(vec[i], obj.root_bbox_[i].high, i);
1434 static void save_tree(
1435 const Derived& obj, std::ostream& stream,
const NodeConstPtr tree)
1437 save_value(stream, *tree);
1438 if (tree->child1 !=
nullptr) { save_tree(obj, stream, tree->child1); }
1439 if (tree->child2 !=
nullptr) { save_tree(obj, stream, tree->child2); }
1442 static void load_tree(Derived& obj, std::istream& stream, NodePtr& tree)
1444 tree = obj.pool_.template allocate<Node>();
1445 load_value(stream, *tree);
1446 if (tree->child1 !=
nullptr) { load_tree(obj, stream, tree->child1); }
1447 if (tree->child2 !=
nullptr) { load_tree(obj, stream, tree->child2); }
1455 void saveIndex(
const Derived& obj, std::ostream& stream)
const
1457 save_value(stream, obj.size_);
1458 save_value(stream, obj.dim_);
1459 save_value(stream, obj.root_bbox_);
1460 save_value(stream, obj.leaf_max_size_);
1461 save_value(stream, obj.vAcc_);
1462 if (obj.root_node_) save_tree(obj, stream, obj.root_node_);
1472 load_value(stream, obj.size_);
1473 load_value(stream, obj.dim_);
1474 load_value(stream, obj.root_bbox_);
1475 load_value(stream, obj.leaf_max_size_);
1476 load_value(stream, obj.vAcc_);
1477 load_tree(obj, stream, obj.root_node_);
1523 typename Distance,
class DatasetAdaptor, int32_t DIM = -1,
1524 typename index_t = uint32_t>
1527 KDTreeSingleIndexAdaptor<Distance, DatasetAdaptor, DIM, index_t>,
1528 Distance, DatasetAdaptor, DIM, index_t>
1534 Distance, DatasetAdaptor, DIM, index_t>&) =
delete;
1545 Distance, DatasetAdaptor, DIM, index_t>,
1546 Distance, DatasetAdaptor, DIM, index_t>;
1548 using Offset =
typename Base::Offset;
1549 using Size =
typename Base::Size;
1550 using Dimension =
typename Base::Dimension;
1552 using ElementType =
typename Base::ElementType;
1553 using DistanceType =
typename Base::DistanceType;
1554 using IndexType =
typename Base::IndexType;
1556 using Node =
typename Base::Node;
1557 using NodePtr = Node*;
1559 using Interval =
typename Base::Interval;
1589 template <
class... Args>
1591 const Dimension dimensionality,
const DatasetAdaptor& inputData,
1593 : dataset_(inputData),
1594 indexParams(params),
1595 distance_(inputData, std::forward<Args>(args)...)
1597 init(dimensionality, params);
1601 const Dimension dimensionality,
const DatasetAdaptor& inputData,
1603 : dataset_(inputData), indexParams(params), distance_(inputData)
1605 init(dimensionality, params);
1610 const Dimension dimensionality,
1611 const KDTreeSingleIndexAdaptorParams& params)
1613 Base::size_ = dataset_.kdtree_get_point_count();
1614 Base::size_at_index_build_ = Base::size_;
1615 Base::dim_ = dimensionality;
1616 if (DIM > 0) Base::dim_ = DIM;
1617 Base::leaf_max_size_ = params.leaf_max_size;
1618 if (params.n_thread_build > 0)
1620 Base::n_thread_build_ = params.n_thread_build;
1624 Base::n_thread_build_ =
1625 std::max(std::thread::hardware_concurrency(), 1u);
1628 if (!(params.flags &
1629 KDTreeSingleIndexAdaptorFlags::SkipInitialBuildIndex))
1642 Base::size_ = dataset_.kdtree_get_point_count();
1643 Base::size_at_index_build_ = Base::size_;
1645 this->freeIndex(*
this);
1646 Base::size_at_index_build_ = Base::size_;
1647 if (Base::size_ == 0)
return;
1648 computeBoundingBox(Base::root_bbox_);
1650 if (Base::n_thread_build_ == 1)
1653 this->divideTree(*
this, 0, Base::size_, Base::root_bbox_);
1657#ifndef NANOFLANN_NO_THREADS
1658 std::atomic<unsigned int> thread_count(0u);
1660 Base::root_node_ = this->divideTreeConcurrent(
1661 *
this, 0, Base::size_, Base::root_bbox_, thread_count, mutex);
1663 throw std::runtime_error(
"Multithreading is disabled");
1687 template <
typename RESULTSET>
1689 RESULTSET& result,
const ElementType* vec,
1693 if (this->size(*
this) == 0)
return false;
1694 if (!Base::root_node_)
1695 throw std::runtime_error(
1696 "[nanoflann] findNeighbors() called before building the "
1698 float epsError = 1 + searchParams.eps;
1701 distance_vector_t dists;
1703 auto zero =
static_cast<typename RESULTSET::DistanceType
>(0);
1704 assign(dists, (DIM > 0 ? DIM : Base::dim_), zero);
1705 DistanceType dist = this->computeInitialDistances(*
this, vec, dists);
1706 searchLevel(result, vec, Base::root_node_, dist, dists, epsError);
1708 if (searchParams.sorted) result.sort();
1710 return result.full();
1729 const ElementType* query_point,
const Size num_closest,
1730 IndexType* out_indices, DistanceType* out_distances)
const
1733 resultSet.init(out_indices, out_distances);
1734 findNeighbors(resultSet, query_point);
1735 return resultSet.size();
1758 const ElementType* query_point,
const DistanceType& radius,
1763 radius, IndicesDists);
1765 radiusSearchCustomCallback(query_point, resultSet, searchParams);
1774 template <
class SEARCH_CALLBACK>
1776 const ElementType* query_point, SEARCH_CALLBACK& resultSet,
1779 findNeighbors(resultSet, query_point, searchParams);
1780 return resultSet.size();
1800 const ElementType* query_point,
const Size num_closest,
1801 IndexType* out_indices, DistanceType* out_distances,
1802 const DistanceType& radius)
const
1805 num_closest, radius);
1806 resultSet.init(out_indices, out_distances);
1807 findNeighbors(resultSet, query_point);
1808 return resultSet.size();
1819 Base::size_ = dataset_.kdtree_get_point_count();
1820 if (Base::vAcc_.size() != Base::size_) Base::vAcc_.resize(Base::size_);
1821 for (IndexType i = 0; i < static_cast<IndexType>(Base::size_); i++)
1825 void computeBoundingBox(BoundingBox& bbox)
1827 const auto dims = (DIM > 0 ? DIM : Base::dim_);
1829 if (dataset_.kdtree_get_bbox(bbox))
1835 const Size N = dataset_.kdtree_get_point_count();
1837 throw std::runtime_error(
1838 "[nanoflann] computeBoundingBox() called but "
1839 "no data points found.");
1840 for (Dimension i = 0; i < dims; ++i)
1842 bbox[i].low = bbox[i].high =
1843 this->dataset_get(*
this, Base::vAcc_[0], i);
1845 for (Offset k = 1; k < N; ++k)
1847 for (Dimension i = 0; i < dims; ++i)
1850 this->dataset_get(*
this, Base::vAcc_[k], i);
1851 if (val < bbox[i].low) bbox[i].low = val;
1852 if (val > bbox[i].high) bbox[i].high = val;
1864 template <
class RESULTSET>
1866 RESULTSET& result_set,
const ElementType* vec,
const NodePtr node,
1868 const float epsError)
const
1871 if ((node->child1 ==
nullptr) && (node->child2 ==
nullptr))
1873 DistanceType worst_dist = result_set.worstDist();
1874 for (Offset i = node->node_type.lr.left;
1875 i < node->node_type.lr.right; ++i)
1877 const IndexType accessor = Base::vAcc_[i];
1878 DistanceType dist = distance_.evalMetric(
1879 vec, accessor, (DIM > 0 ? DIM : Base::dim_));
1880 if (dist < worst_dist)
1882 if (!result_set.addPoint(dist, Base::vAcc_[i]))
1894 Dimension idx = node->node_type.sub.divfeat;
1895 ElementType val = vec[idx];
1896 DistanceType diff1 = val - node->node_type.sub.divlow;
1897 DistanceType diff2 = val - node->node_type.sub.divhigh;
1901 DistanceType cut_dist;
1902 if ((diff1 + diff2) < 0)
1904 bestChild = node->child1;
1905 otherChild = node->child2;
1907 distance_.accum_dist(val, node->node_type.sub.divhigh, idx);
1911 bestChild = node->child2;
1912 otherChild = node->child1;
1914 distance_.accum_dist(val, node->node_type.sub.divlow, idx);
1918 if (!searchLevel(result_set, vec, bestChild, mindist, dists, epsError))
1925 DistanceType dst = dists[idx];
1926 mindist = mindist + cut_dist - dst;
1927 dists[idx] = cut_dist;
1928 if (mindist * epsError <= result_set.worstDist())
1931 result_set, vec, otherChild, mindist, dists, epsError))
1950 Base::saveIndex(*
this, stream);
1958 void loadIndex(std::istream& stream) { Base::loadIndex(*
this, stream); }
2000 typename Distance,
class DatasetAdaptor, int32_t DIM = -1,
2001 typename IndexType = uint32_t>
2004 KDTreeSingleIndexDynamicAdaptor_<
2005 Distance, DatasetAdaptor, DIM, IndexType>,
2006 Distance, DatasetAdaptor, DIM, IndexType>
2016 std::vector<int>& treeIndex_;
2022 Distance, DatasetAdaptor, DIM, IndexType>,
2023 Distance, DatasetAdaptor, DIM, IndexType>;
2025 using ElementType =
typename Base::ElementType;
2026 using DistanceType =
typename Base::DistanceType;
2028 using Offset =
typename Base::Offset;
2029 using Size =
typename Base::Size;
2030 using Dimension =
typename Base::Dimension;
2032 using Node =
typename Base::Node;
2033 using NodePtr = Node*;
2035 using Interval =
typename Base::Interval;
2060 const Dimension dimensionality,
const DatasetAdaptor& inputData,
2061 std::vector<int>& treeIndex,
2064 : dataset_(inputData),
2065 index_params_(params),
2066 treeIndex_(treeIndex),
2067 distance_(inputData)
2070 Base::size_at_index_build_ = 0;
2071 for (
auto& v : Base::root_bbox_) v = {};
2072 Base::dim_ = dimensionality;
2073 if (DIM > 0) Base::dim_ = DIM;
2074 Base::leaf_max_size_ = params.leaf_max_size;
2075 if (params.n_thread_build > 0)
2077 Base::n_thread_build_ = params.n_thread_build;
2081 Base::n_thread_build_ =
2082 std::max(std::thread::hardware_concurrency(), 1u);
2095 std::swap(Base::vAcc_, tmp.Base::vAcc_);
2096 std::swap(Base::leaf_max_size_, tmp.Base::leaf_max_size_);
2097 std::swap(index_params_, tmp.index_params_);
2098 std::swap(treeIndex_, tmp.treeIndex_);
2099 std::swap(Base::size_, tmp.Base::size_);
2100 std::swap(Base::size_at_index_build_, tmp.Base::size_at_index_build_);
2101 std::swap(Base::root_node_, tmp.Base::root_node_);
2102 std::swap(Base::root_bbox_, tmp.Base::root_bbox_);
2103 std::swap(Base::pool_, tmp.Base::pool_);
2112 Base::size_ = Base::vAcc_.size();
2113 this->freeIndex(*
this);
2114 Base::size_at_index_build_ = Base::size_;
2115 if (Base::size_ == 0)
return;
2116 computeBoundingBox(Base::root_bbox_);
2118 if (Base::n_thread_build_ == 1)
2121 this->divideTree(*
this, 0, Base::size_, Base::root_bbox_);
2125#ifndef NANOFLANN_NO_THREADS
2126 std::atomic<unsigned int> thread_count(0u);
2128 Base::root_node_ = this->divideTreeConcurrent(
2129 *
this, 0, Base::size_, Base::root_bbox_, thread_count, mutex);
2131 throw std::runtime_error(
"Multithreading is disabled");
2159 template <
typename RESULTSET>
2161 RESULTSET& result,
const ElementType* vec,
2165 if (this->size(*
this) == 0)
return false;
2166 if (!Base::root_node_)
return false;
2167 float epsError = 1 + searchParams.eps;
2170 distance_vector_t dists;
2173 dists, (DIM > 0 ? DIM : Base::dim_),
2174 static_cast<typename distance_vector_t::value_type>(0));
2175 DistanceType dist = this->computeInitialDistances(*
this, vec, dists);
2176 searchLevel(result, vec, Base::root_node_, dist, dists, epsError);
2177 return result.full();
2195 const ElementType* query_point,
const Size num_closest,
2196 IndexType* out_indices, DistanceType* out_distances,
2200 resultSet.init(out_indices, out_distances);
2201 findNeighbors(resultSet, query_point, searchParams);
2202 return resultSet.size();
2225 const ElementType* query_point,
const DistanceType& radius,
2230 radius, IndicesDists);
2231 const size_t nFound =
2232 radiusSearchCustomCallback(query_point, resultSet, searchParams);
2241 template <
class SEARCH_CALLBACK>
2243 const ElementType* query_point, SEARCH_CALLBACK& resultSet,
2246 findNeighbors(resultSet, query_point, searchParams);
2247 return resultSet.size();
2253 void computeBoundingBox(BoundingBox& bbox)
2255 const auto dims = (DIM > 0 ? DIM : Base::dim_);
2258 if (dataset_.kdtree_get_bbox(bbox))
2264 const Size N = Base::size_;
2266 throw std::runtime_error(
2267 "[nanoflann] computeBoundingBox() called but "
2268 "no data points found.");
2269 for (Dimension i = 0; i < dims; ++i)
2271 bbox[i].low = bbox[i].high =
2272 this->dataset_get(*
this, Base::vAcc_[0], i);
2274 for (Offset k = 1; k < N; ++k)
2276 for (Dimension i = 0; i < dims; ++i)
2279 this->dataset_get(*
this, Base::vAcc_[k], i);
2280 if (val < bbox[i].low) bbox[i].low = val;
2281 if (val > bbox[i].high) bbox[i].high = val;
2291 template <
class RESULTSET>
2293 RESULTSET& result_set,
const ElementType* vec,
const NodePtr node,
2295 const float epsError)
const
2298 if ((node->child1 ==
nullptr) && (node->child2 ==
nullptr))
2300 DistanceType worst_dist = result_set.worstDist();
2301 for (Offset i = node->node_type.lr.left;
2302 i < node->node_type.lr.right; ++i)
2304 const IndexType index = Base::vAcc_[i];
2305 if (treeIndex_[index] == -1)
continue;
2306 DistanceType dist = distance_.evalMetric(
2307 vec, index, (DIM > 0 ? DIM : Base::dim_));
2308 if (dist < worst_dist)
2310 if (!result_set.addPoint(
2311 static_cast<typename RESULTSET::DistanceType
>(dist),
2312 static_cast<typename RESULTSET::IndexType
>(
2325 Dimension idx = node->node_type.sub.divfeat;
2326 ElementType val = vec[idx];
2327 DistanceType diff1 = val - node->node_type.sub.divlow;
2328 DistanceType diff2 = val - node->node_type.sub.divhigh;
2332 DistanceType cut_dist;
2333 if ((diff1 + diff2) < 0)
2335 bestChild = node->child1;
2336 otherChild = node->child2;
2338 distance_.accum_dist(val, node->node_type.sub.divhigh, idx);
2342 bestChild = node->child2;
2343 otherChild = node->child1;
2345 distance_.accum_dist(val, node->node_type.sub.divlow, idx);
2349 searchLevel(result_set, vec, bestChild, mindist, dists, epsError);
2351 DistanceType dst = dists[idx];
2352 mindist = mindist + cut_dist - dst;
2353 dists[idx] = cut_dist;
2354 if (mindist * epsError <= result_set.worstDist())
2356 searchLevel(result_set, vec, otherChild, mindist, dists, epsError);
2392 typename Distance,
class DatasetAdaptor, int32_t DIM = -1,
2393 typename IndexType = uint32_t>
2397 using ElementType =
typename Distance::ElementType;
2398 using DistanceType =
typename Distance::DistanceType;
2401 Distance, DatasetAdaptor, DIM>::Offset;
2403 Distance, DatasetAdaptor, DIM>::Size;
2405 Distance, DatasetAdaptor, DIM>::Dimension;
2408 Size leaf_max_size_;
2420 std::unordered_set<int> removedPoints_;
2427 Distance, DatasetAdaptor, DIM, IndexType>;
2428 std::vector<index_container_t> index_;
2440 int First0Bit(IndexType num)
2454 using my_kd_tree_t = KDTreeSingleIndexDynamicAdaptor_<
2455 Distance, DatasetAdaptor, DIM, IndexType>;
2456 std::vector<my_kd_tree_t> index(
2458 my_kd_tree_t(dim_ , dataset_, treeIndex_, index_params_));
2481 const int dimensionality,
const DatasetAdaptor& inputData,
2484 const size_t maximumPointCount = 1000000000U)
2485 : dataset_(inputData), index_params_(params), distance_(inputData)
2487 treeCount_ =
static_cast<size_t>(std::log2(maximumPointCount)) + 1;
2489 dim_ = dimensionality;
2491 if (DIM > 0) dim_ = DIM;
2492 leaf_max_size_ = params.leaf_max_size;
2494 const size_t num_initial_points = dataset_.kdtree_get_point_count();
2495 if (num_initial_points > 0) { addPoints(0, num_initial_points - 1); }
2501 Distance, DatasetAdaptor, DIM, IndexType>&) =
delete;
2506 const Size count = end - start + 1;
2508 treeIndex_.resize(treeIndex_.size() + count);
2509 for (IndexType idx = start; idx <= end; idx++)
2511 const int pos = First0Bit(pointCount_);
2512 maxIndex = std::max(pos, maxIndex);
2513 treeIndex_[pointCount_] = pos;
2515 const auto it = removedPoints_.find(idx);
2516 if (it != removedPoints_.end())
2518 removedPoints_.erase(it);
2519 treeIndex_[idx] = pos;
2522 for (
int i = 0; i < pos; i++)
2524 for (
int j = 0; j < static_cast<int>(index_[i].vAcc_.size());
2527 index_[pos].vAcc_.push_back(index_[i].vAcc_[j]);
2528 if (treeIndex_[index_[i].vAcc_[j]] != -1)
2529 treeIndex_[index_[i].vAcc_[j]] = pos;
2531 index_[i].vAcc_.clear();
2533 index_[pos].vAcc_.push_back(idx);
2537 for (
int i = 0; i <= maxIndex; ++i)
2539 index_[i].freeIndex(index_[i]);
2540 if (!index_[i].vAcc_.empty()) index_[i].buildIndex();
2547 if (idx >= pointCount_)
return;
2548 removedPoints_.insert(idx);
2549 treeIndex_[idx] = -1;
2568 template <
typename RESULTSET>
2570 RESULTSET& result,
const ElementType* vec,
2573 for (
size_t i = 0; i < treeCount_; i++)
2575 index_[i].findNeighbors(result, &vec[0], searchParams);
2577 return result.full();
2608 bool row_major =
true>
2613 using num_t =
typename MatrixType::Scalar;
2614 using IndexType =
typename MatrixType::Index;
2615 using metric_t =
typename Distance::template traits<
2616 num_t,
self_t, IndexType>::distance_t;
2620 row_major ? MatrixType::ColsAtCompileTime
2621 : MatrixType::RowsAtCompileTime,
2628 using Size =
typename index_t::Size;
2629 using Dimension =
typename index_t::Dimension;
2633 const Dimension dimensionality,
2634 const std::reference_wrapper<const MatrixType>& mat,
2635 const int leaf_max_size = 10,
const unsigned int n_thread_build = 1)
2636 : m_data_matrix(mat)
2638 const auto dims = row_major ? mat.get().cols() : mat.get().rows();
2639 if (
static_cast<Dimension
>(dims) != dimensionality)
2640 throw std::runtime_error(
2641 "Error: 'dimensionality' must match column count in data "
2643 if (DIM > 0 &&
static_cast<int32_t
>(dims) != DIM)
2644 throw std::runtime_error(
2645 "Data set dimensionality does not match the 'DIM' template "
2650 leaf_max_size, nanoflann::KDTreeSingleIndexAdaptorFlags::None,
2660 const std::reference_wrapper<const MatrixType> m_data_matrix;
2671 const num_t* query_point,
const Size num_closest,
2672 IndexType* out_indices, num_t* out_distances)
const
2675 resultSet.init(out_indices, out_distances);
2682 const self_t& derived()
const {
return *
this; }
2683 self_t& derived() {
return *
this; }
2686 Size kdtree_get_point_count()
const
2689 return m_data_matrix.get().rows();
2691 return m_data_matrix.get().cols();
2695 num_t kdtree_get_pt(
const IndexType idx,
size_t dim)
const
2698 return m_data_matrix.get().coeff(idx, IndexType(dim));
2700 return m_data_matrix.get().coeff(IndexType(dim), idx);
2708 template <
class BBOX>
2709 bool kdtree_get_bbox(BBOX& )
const
// end of grouping
Definition nanoflann.hpp:1011
void freeIndex(Derived &obj)
Definition nanoflann.hpp:1015
BoundingBox root_bbox_
Definition nanoflann.hpp:1089
Size veclen(const Derived &obj)
Definition nanoflann.hpp:1104
void saveIndex(const Derived &obj, std::ostream &stream) const
Definition nanoflann.hpp:1455
Size usedMemory(Derived &obj)
Definition nanoflann.hpp:1117
typename array_or_vector< DIM, DistanceType >::type distance_vector_t
Definition nanoflann.hpp:1086
void planeSplit(const Derived &obj, const Offset ind, const Size count, const Dimension cutfeat, const DistanceType &cutval, Offset &lim1, Offset &lim2)
Definition nanoflann.hpp:1365
NodePtr divideTree(Derived &obj, const Offset left, const Offset right, BoundingBox &bbox)
Definition nanoflann.hpp:1145
std::vector< IndexType > vAcc_
Definition nanoflann.hpp:1029
Size size(const Derived &obj) const
Definition nanoflann.hpp:1101
NodePtr divideTreeConcurrent(Derived &obj, const Offset left, const Offset right, BoundingBox &bbox, std::atomic< unsigned int > &thread_count, std::mutex &mutex)
Definition nanoflann.hpp:1216
void loadIndex(Derived &obj, std::istream &stream)
Definition nanoflann.hpp:1470
PooledAllocator pool_
Definition nanoflann.hpp:1098
ElementType dataset_get(const Derived &obj, IndexType element, Dimension component) const
Helper accessor to the dataset points:
Definition nanoflann.hpp:1107
typename array_or_vector< DIM, Interval >::type BoundingBox
Definition nanoflann.hpp:1082
Definition nanoflann.hpp:1529
bool searchLevel(RESULTSET &result_set, const ElementType *vec, const NodePtr node, DistanceType mindist, distance_vector_t &dists, const float epsError) const
Definition nanoflann.hpp:1865
void saveIndex(std::ostream &stream) const
Definition nanoflann.hpp:1948
Size radiusSearch(const ElementType *query_point, const DistanceType &radius, std::vector< ResultItem< IndexType, DistanceType > > &IndicesDists, const SearchParameters &searchParams={}) const
Definition nanoflann.hpp:1757
void init_vind()
Definition nanoflann.hpp:1816
void buildIndex()
Definition nanoflann.hpp:1640
const DatasetAdaptor & dataset_
Definition nanoflann.hpp:1537
KDTreeSingleIndexAdaptor(const KDTreeSingleIndexAdaptor< Distance, DatasetAdaptor, DIM, index_t > &)=delete
bool findNeighbors(RESULTSET &result, const ElementType *vec, const SearchParameters &searchParams={}) const
Definition nanoflann.hpp:1688
Size rknnSearch(const ElementType *query_point, const Size num_closest, IndexType *out_indices, DistanceType *out_distances, const DistanceType &radius) const
Definition nanoflann.hpp:1799
Size radiusSearchCustomCallback(const ElementType *query_point, SEARCH_CALLBACK &resultSet, const SearchParameters &searchParams={}) const
Definition nanoflann.hpp:1775
typename Base::distance_vector_t distance_vector_t
Definition nanoflann.hpp:1567
void loadIndex(std::istream &stream)
Definition nanoflann.hpp:1958
typename Base::BoundingBox BoundingBox
Definition nanoflann.hpp:1563
KDTreeSingleIndexAdaptor(const Dimension dimensionality, const DatasetAdaptor &inputData, const KDTreeSingleIndexAdaptorParams ¶ms, Args &&... args)
Definition nanoflann.hpp:1590
Size knnSearch(const ElementType *query_point, const Size num_closest, IndexType *out_indices, DistanceType *out_distances) const
Definition nanoflann.hpp:1728
Definition nanoflann.hpp:2007
Size radiusSearch(const ElementType *query_point, const DistanceType &radius, std::vector< ResultItem< IndexType, DistanceType > > &IndicesDists, const SearchParameters &searchParams={}) const
Definition nanoflann.hpp:2224
KDTreeSingleIndexDynamicAdaptor_(const Dimension dimensionality, const DatasetAdaptor &inputData, std::vector< int > &treeIndex, const KDTreeSingleIndexAdaptorParams ¶ms=KDTreeSingleIndexAdaptorParams())
Definition nanoflann.hpp:2059
typename Base::BoundingBox BoundingBox
Definition nanoflann.hpp:2038
const DatasetAdaptor & dataset_
The source of our data.
Definition nanoflann.hpp:2012
Size radiusSearchCustomCallback(const ElementType *query_point, SEARCH_CALLBACK &resultSet, const SearchParameters &searchParams={}) const
Definition nanoflann.hpp:2242
KDTreeSingleIndexDynamicAdaptor_(const KDTreeSingleIndexDynamicAdaptor_ &rhs)=default
void buildIndex()
Definition nanoflann.hpp:2110
void saveIndex(std::ostream &stream)
Definition nanoflann.hpp:2367
typename Base::distance_vector_t distance_vector_t
Definition nanoflann.hpp:2042
void loadIndex(std::istream &stream)
Definition nanoflann.hpp:2374
Size knnSearch(const ElementType *query_point, const Size num_closest, IndexType *out_indices, DistanceType *out_distances, const SearchParameters &searchParams={}) const
Definition nanoflann.hpp:2194
void searchLevel(RESULTSET &result_set, const ElementType *vec, const NodePtr node, DistanceType mindist, distance_vector_t &dists, const float epsError) const
Definition nanoflann.hpp:2292
bool findNeighbors(RESULTSET &result, const ElementType *vec, const SearchParameters &searchParams={}) const
Definition nanoflann.hpp:2160
KDTreeSingleIndexDynamicAdaptor_ operator=(const KDTreeSingleIndexDynamicAdaptor_ &rhs)
Definition nanoflann.hpp:2091
Definition nanoflann.hpp:2395
bool findNeighbors(RESULTSET &result, const ElementType *vec, const SearchParameters &searchParams={}) const
Definition nanoflann.hpp:2569
const DatasetAdaptor & dataset_
The source of our data.
Definition nanoflann.hpp:2415
void removePoint(size_t idx)
Definition nanoflann.hpp:2545
void addPoints(IndexType start, IndexType end)
Definition nanoflann.hpp:2504
KDTreeSingleIndexDynamicAdaptor(const int dimensionality, const DatasetAdaptor &inputData, const KDTreeSingleIndexAdaptorParams ¶ms=KDTreeSingleIndexAdaptorParams(), const size_t maximumPointCount=1000000000U)
Definition nanoflann.hpp:2480
std::vector< int > treeIndex_
Definition nanoflann.hpp:2419
const std::vector< index_container_t > & getAllIndices() const
Definition nanoflann.hpp:2433
Dimension dim_
Dimensionality of each data point.
Definition nanoflann.hpp:2424
KDTreeSingleIndexDynamicAdaptor(const KDTreeSingleIndexDynamicAdaptor< Distance, DatasetAdaptor, DIM, IndexType > &)=delete
Definition nanoflann.hpp:200
bool addPoint(DistanceType dist, IndexType index)
Definition nanoflann.hpp:234
DistanceType worstDist() const
Definition nanoflann.hpp:271
Definition nanoflann.hpp:859
~PooledAllocator()
Definition nanoflann.hpp:895
void free_all()
Definition nanoflann.hpp:898
void * malloc(const size_t req_size)
Definition nanoflann.hpp:914
T * allocate(const size_t count=1)
Definition nanoflann.hpp:965
PooledAllocator()
Definition nanoflann.hpp:890
Definition nanoflann.hpp:288
bool addPoint(DistanceType dist, IndexType index)
Definition nanoflann.hpp:329
DistanceType worstDist() const
Definition nanoflann.hpp:366
Definition nanoflann.hpp:383
ResultItem< IndexType, DistanceType > worst_item() const
Definition nanoflann.hpp:426
bool addPoint(DistanceType dist, IndexType index)
Definition nanoflann.hpp:414
std::enable_if< has_assign< Container >::value, void >::type assign(Container &c, const size_t nElements, const T &value)
Definition nanoflann.hpp:144
T pi_const()
Definition nanoflann.hpp:87
std::enable_if< has_resize< Container >::value, void >::type resize(Container &c, const size_t nElements)
Definition nanoflann.hpp:122
Definition nanoflann.hpp:162
bool operator()(const PairType &p1, const PairType &p2) const
Definition nanoflann.hpp:165
Definition nanoflann.hpp:1064
Definition nanoflann.hpp:1039
DistanceType divlow
The values used for subdivision.
Definition nanoflann.hpp:1052
Offset right
Indices of points in leaf node.
Definition nanoflann.hpp:1046
union nanoflann::KDTreeBaseClass::Node::@0 node_type
Dimension divfeat
Definition nanoflann.hpp:1050
Node * child1
Definition nanoflann.hpp:1057
Definition nanoflann.hpp:2610
void query(const num_t *query_point, const Size num_closest, IndexType *out_indices, num_t *out_distances) const
Definition nanoflann.hpp:2670
KDTreeEigenMatrixAdaptor(const self_t &)=delete
typename index_t::Offset Offset
Definition nanoflann.hpp:2627
KDTreeEigenMatrixAdaptor(const Dimension dimensionality, const std::reference_wrapper< const MatrixType > &mat, const int leaf_max_size=10, const unsigned int n_thread_build=1)
Constructor: takes a const ref to the matrix object with the data points.
Definition nanoflann.hpp:2632
Definition nanoflann.hpp:810
Definition nanoflann.hpp:499
Definition nanoflann.hpp:561
Definition nanoflann.hpp:626
Definition nanoflann.hpp:482
Definition nanoflann.hpp:181
DistanceType second
Distance from sample to query point.
Definition nanoflann.hpp:189
IndexType first
Index of the sample in the dataset.
Definition nanoflann.hpp:188
Definition nanoflann.hpp:671
DistanceType accum_dist(const U a, const V b, const size_t) const
Definition nanoflann.hpp:689
Definition nanoflann.hpp:716
Definition nanoflann.hpp:829
bool sorted
distance (default: true)
Definition nanoflann.hpp:836
float eps
search for eps-approximate neighbours (default: 0)
Definition nanoflann.hpp:835
Definition nanoflann.hpp:981
Definition nanoflann.hpp:109
Definition nanoflann.hpp:98
Definition nanoflann.hpp:746
Definition nanoflann.hpp:743
Definition nanoflann.hpp:756
Definition nanoflann.hpp:766
Definition nanoflann.hpp:763
Definition nanoflann.hpp:753
Definition nanoflann.hpp:775
Definition nanoflann.hpp:772
Definition nanoflann.hpp:784
Definition nanoflann.hpp:781