kicad/thirdparty/rtree/geometry/rtree.h

2016 lines
60 KiB
C++

//TITLE
//
// R-TREES: A DYNAMIC INDEX STRUCTURE FOR SPATIAL SEARCHING
//
//DESCRIPTION
//
// A C++ templated version of the RTree algorithm.
// For more information please read the comments in RTree.h
//
//AUTHORS
//
// * 1983 Original algorithm and test code by Antonin Guttman and Michael Stonebraker, UC Berkely
// * 1994 ANCI C ported from original test code by Melinda Green - melinda@superliminal.com
// * 1995 Sphere volume fix for degeneracy problem submitted by Paul Brook
// * 2004 Templated C++ port by Greg Douglas
// * 2013 CERN (www.cern.ch)
// * 2020 KiCad Developers - Add std::iterator support for searching
// * 2020 KiCad Developers - Add container nearest neighbor based on Hjaltason & Samet
// * 2022 KiCad Developers - Slight optimizations in RectSphericalVolume
//
/*
* This program source code file is part of KiCad, a free EDA CAD application.
*
* Copyright (C) 2020 KiCad Developers, see AUTHORS.txt for contributors.
* Copyright (C) 2013 CERN
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version 3
* of the License, or (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, you may find one here:
* http://www.gnu.org/licenses/old-licenses/gpl-3.0.html
* or you may search the http://www.gnu.org website for the version 3 license,
* or you may write to the Free Software Foundation, Inc.,
* 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA
*/
#ifndef RTREE_H
#define RTREE_H
// NOTE This file compiles under MSVC 6 SP5 and MSVC .Net 2003 it may not work on other compilers without modification.
// NOTE These next few lines may be win32 specific, you may need to modify them to compile on other platform
#include <cassert>
#include <cmath>
#include <cstdio>
#include <cstdlib>
#include <algorithm>
#include <array>
#include <functional>
#include <iterator>
#include <limits>
#include <queue>
#include <vector>
#ifdef DEBUG
#define ASSERT assert // RTree uses ASSERT( condition )
#else
#define ASSERT( _x )
#endif
//
// RTree.h
//
#define RTREE_TEMPLATE template <class DATATYPE, class ELEMTYPE, int NUMDIMS, \
class ELEMTYPEREAL, int TMAXNODES, int TMINNODES>
#define RTREE_SEARCH_TEMPLATE template <class DATATYPE, class ELEMTYPE, int NUMDIMS, \
class ELEMTYPEREAL, int TMAXNODES, int TMINNODES, class VISITOR>
#define RTREE_QUAL RTree<DATATYPE, ELEMTYPE, NUMDIMS, ELEMTYPEREAL, TMAXNODES, \
TMINNODES>
#define RTREE_SEARCH_QUAL RTree<DATATYPE, ELEMTYPE, NUMDIMS, ELEMTYPEREAL, TMAXNODES, \
TMINNODES, VISITOR>
#define RTREE_DONT_USE_MEMPOOLS // This version does not contain a fixed memory allocator, fill in lines with EXAMPLE to implement one.
#define RTREE_USE_SPHERICAL_VOLUME // Better split classification, may be slower on some systems
// Fwd decl
class RTFileStream; // File I/O helper class, look below for implementation and notes.
/// \class RTree
/// Implementation of RTree, a multidimensional bounding rectangle tree.
/// Example usage: For a 3-dimensional tree use RTree<Object*, float, 3> myTree;
///
/// This modified, templated C++ version by Greg Douglas at Auran (http://www.auran.com)
///
/// DATATYPE Referenced data, should be int, void*, obj* etc. no larger than sizeof<void*> and simple type
/// ELEMTYPE Type of element such as int or float
/// NUMDIMS Number of dimensions such as 2 or 3
/// ELEMTYPEREAL Type of element that allows fractional and large values such as float or double, for use in volume calcs
///
/// NOTES: Inserting and removing data requires the knowledge of its constant Minimal Bounding Rectangle.
/// This version uses new/delete for nodes, I recommend using a fixed size allocator for efficiency.
/// Instead of using a callback function for returned results, I recommend and efficient pre-sized, grow-only memory
/// array similar to MFC CArray or STL Vector for returning search query result.
///
template <class DATATYPE, class ELEMTYPE, int NUMDIMS,
class ELEMTYPEREAL = ELEMTYPE, int TMAXNODES = 8, int TMINNODES = TMAXNODES / 2>
class RTree
{
protected:
struct Node; // Fwd decl. Used by other internal structs and iterator
public:
/// Minimal bounding rectangle (n-dimensional)
struct Rect
{
ELEMTYPE m_min[NUMDIMS]; ///< Min dimensions of bounding box
ELEMTYPE m_max[NUMDIMS]; ///< Max dimensions of bounding box
};
// These constant must be declared after Branch and before Node struct
// Stuck up here for MSVC 6 compiler. NSVC .NET 2003 is much happier.
enum {
MAXNODES = TMAXNODES, ///< Max elements in node
MINNODES = TMINNODES ///< Min elements in node
};
struct Statistics {
int maxDepth;
int avgDepth;
int maxNodeLoad;
int avgNodeLoad;
int totalItems;
};
public:
RTree();
virtual ~RTree();
/// Insert entry
/// \param a_min Min of bounding rect
/// \param a_max Max of bounding rect
/// \param a_dataId Positive Id of data. Maybe zero, but negative numbers not allowed.
void Insert( const ELEMTYPE a_min[NUMDIMS],
const ELEMTYPE a_max[NUMDIMS],
const DATATYPE& a_dataId );
/// Remove entry
/// \param a_min Min of bounding rect
/// \param a_max Max of bounding rect
/// \param a_dataId Positive Id of data. Maybe zero, but negative numbers not allowed.
/// \return 1 if record not found, 0 if success.
bool Remove( const ELEMTYPE a_min[NUMDIMS],
const ELEMTYPE a_max[NUMDIMS],
const DATATYPE& a_dataId );
/// Find all within search rectangle
/// \param a_min Min of search bounding rect
/// \param a_max Max of search bounding rect
/// \param a_callback Callback function to return result. Callback should return 'true' to continue searching
/// \return Returns the number of entries found
int Search( const ELEMTYPE a_min[NUMDIMS],
const ELEMTYPE a_max[NUMDIMS],
std::function<bool (const DATATYPE&)> a_callback ) const;
/// Find all within search rectangle
/// \param a_min Min of search bounding rect
/// \param a_max Max of search bounding rect
/// \param a_callback Callback function to return result. Callback should return 'true' to continue searching
/// \param aFinished This is set to true if the search completed and false if it was interupted
/// \return Returns the number of entries found
int Search( const ELEMTYPE a_min[NUMDIMS], const ELEMTYPE a_max[NUMDIMS],
std::function<bool( const DATATYPE& )> a_callback, bool& aFinished ) const;
template <class VISITOR>
int Search( const ELEMTYPE a_min[NUMDIMS], const ELEMTYPE a_max[NUMDIMS], VISITOR& a_visitor ) const
{
#ifdef _DEBUG
for( int index = 0; index < NUMDIMS; ++index )
{
ASSERT( a_min[index] <= a_max[index] );
}
#endif // _DEBUG
Rect rect;
for( int axis = 0; axis < NUMDIMS; ++axis )
{
rect.m_min[axis] = a_min[axis];
rect.m_max[axis] = a_max[axis];
}
// NOTE: May want to return search result another way, perhaps returning the number of found elements here.
int cnt = 0;
Search( m_root, &rect, a_visitor, cnt );
return cnt;
}
/// Calculate Statistics
Statistics CalcStats();
/// Remove all entries from tree
void RemoveAll();
/// Count the data elements in this container. This is slow as no internal counter is maintained.
int Count() const;
/// Load tree contents from file
bool Load( const char* a_fileName );
/// Load tree contents from stream
bool Load( RTFileStream& a_stream ) const;
/// Save tree contents to file
bool Save( const char* a_fileName );
/// Save tree contents to stream
bool Save( RTFileStream& a_stream ) const;
/**
* Gets an ordered vector of the nearest data elements to a specified point
* @param aPoint coordinate to measure against
* @param aTerminate Callback routine to check when we have gathered sufficient elements
* @param aFilter Callback routine to remove specific elements from the query results
* @param aSquaredDist Callback routine to measure the distance from the point to the data element
* @return vector of matching elements and their distance to the point
*/
std::vector<std::pair<ELEMTYPE, DATATYPE>> NearestNeighbors(
const ELEMTYPE aPoint[NUMDIMS],
std::function<bool( const std::size_t aNumResults, const ELEMTYPE aMinDist )> aTerminate,
std::function<bool( const DATATYPE aElement )> aFilter,
std::function<ELEMTYPE( const ELEMTYPE a_point[NUMDIMS], const DATATYPE a_data )> aSquaredDist ) const;
public:
/// Iterator is not remove safe.
class Iterator
{
private:
enum
{
MAX_STACK = 32
}; // Max stack size. Allows almost n^32 where n is number of branches in node
struct StackElement
{
Node* m_node;
int m_branchIndex;
};
public:
typedef std::forward_iterator_tag iterator_category;
typedef DATATYPE value_type;
typedef ptrdiff_t difference_type;
typedef DATATYPE* pointer;
typedef DATATYPE& reference;
public:
Iterator() : m_stack( {} ), m_tos( 0 )
{
for( int i = 0; i < NUMDIMS; ++i )
{
m_rect.m_min[i] = std::numeric_limits<ELEMTYPE>::min();
m_rect.m_max[i] = std::numeric_limits<ELEMTYPE>::max();
}
}
Iterator( const Rect& aRect ) : m_stack( {} ), m_tos( 0 ), m_rect( aRect )
{
}
~Iterator()
{
}
/// Is iterator pointing to valid data
constexpr bool IsNotNull() const
{
return m_tos > 0;
}
/// Access the current data element. Caller must be sure iterator is not NULL first.
DATATYPE& operator*()
{
ASSERT( IsNotNull() );
StackElement& curTos = m_stack[m_tos - 1];
return curTos.m_node->m_branch[curTos.m_branchIndex].m_data;
}
/// Access the current data element. Caller must be sure iterator is not NULL first.
const DATATYPE& operator*() const
{
ASSERT( IsNotNull() );
StackElement& curTos = m_stack[m_tos - 1];
return curTos.m_node->m_branch[curTos.m_branchIndex].m_data;
}
DATATYPE* operator->()
{
ASSERT( IsNotNull() );
StackElement& curTos = m_stack[m_tos - 1];
return &( curTos.m_node->m_branch[curTos.m_branchIndex].m_data );
}
/// Prefix ++ operator
Iterator& operator++()
{
FindNextData();
return *this;
}
/// Postfix ++ operator
Iterator operator++( int )
{
Iterator retval = *this;
FindNextData();
return retval;
}
bool operator==( const Iterator& rhs ) const
{
return ( ( m_tos <= 0 && rhs.m_tos <= 0 )
|| ( m_tos == rhs.m_tos && m_stack[m_tos].m_node == rhs.m_stack[m_tos].m_node
&& m_stack[m_tos].m_branchIndex
== rhs.m_stack[m_tos].m_branchIndex ) );
}
bool operator!=( const Iterator& rhs ) const
{
return ( ( m_tos > 0 || rhs.m_tos > 0 )
&& ( m_tos != rhs.m_tos || m_stack[m_tos].m_node != rhs.m_stack[m_tos].m_node
|| m_stack[m_tos].m_branchIndex
!= rhs.m_stack[m_tos].m_branchIndex ) );
}
private:
/// Find the next data element in the tree (For internal use only)
void FindNextData()
{
while( m_tos > 0 )
{
StackElement curTos = Pop();
int nextBranch = curTos.m_branchIndex + 1;
if( curTos.m_node->IsLeaf() )
{
// Keep walking through siblings until we find an overlapping leaf
for( int i = nextBranch; i < curTos.m_node->m_count; i++ )
{
if( RTree::Overlap( &m_rect, &curTos.m_node->m_branch[i].m_rect ) )
{
Push( curTos.m_node, i );
return;
}
}
// No more data, so it will fall back to previous level
}
else
{
// Look for an overlapping sibling that we can use as the fall-back node
// when we've iterated down the current branch
for( int i = nextBranch; i < curTos.m_node->m_count; i++ )
{
if( RTree::Overlap( &m_rect, &curTos.m_node->m_branch[i].m_rect ) )
{
Push( curTos.m_node, i );
break;
}
}
Node* nextLevelnode = curTos.m_node->m_branch[curTos.m_branchIndex].m_child;
// Since cur node is not a leaf, push first of next level,
// zero-th branch to get deeper into the tree
Push( nextLevelnode, 0 );
// If the branch is a leaf, and it overlaps, then break with the current data
// Otherwise, we allow it to seed our next iteration as it may have siblings that
// do overlap
if( nextLevelnode->IsLeaf()
&& RTree::Overlap( &m_rect, &nextLevelnode->m_branch[0].m_rect ) )
return;
}
}
}
/// Push node and branch onto iteration stack (For internal use only)
void Push( Node* a_node, int a_branchIndex )
{
m_stack[m_tos].m_node = a_node;
m_stack[m_tos].m_branchIndex = a_branchIndex;
++m_tos;
ASSERT( m_tos <= MAX_STACK );
}
/// Pop element off iteration stack (For internal use only)
StackElement& Pop()
{
ASSERT( m_tos > 0 );
--m_tos;
return m_stack[m_tos];
}
std::array<StackElement, MAX_STACK> m_stack; ///< Stack for iteration
int m_tos; ///< Top Of Stack index
Rect m_rect; ///< Search rectangle
friend class RTree;
// Allow hiding of non-public functions while allowing manipulation by logical owner
};
using iterator = Iterator;
using const_iterator = const Iterator;
iterator begin( const Rect& aRect ) const
{
iterator retval( aRect );
if( !m_root->m_count )
return retval;
retval.Push( m_root, 0 );
// If the first leaf matches, return the root pointer, otherwise,
// increment to the first match or empty if none.
if( m_root->IsLeaf() && Overlap( &aRect, &m_root->m_branch[0].m_rect ) )
return retval;
++retval;
return retval;
}
iterator begin() const
{
Rect full_rect;
std::fill_n( full_rect.m_min, NUMDIMS, INT_MIN );
std::fill_n( full_rect.m_max, NUMDIMS, INT_MAX );
return begin( full_rect );
}
iterator end() const
{
iterator retval;
return retval;
}
iterator end( const Rect& aRect ) const
{
return end();
}
protected:
/// May be data or may be another subtree
/// The parents level determines this.
/// If the parents level is 0, then this is data
struct Branch
{
Rect m_rect; ///< Bounds
union
{
Node* m_child; ///< Child node
DATATYPE m_data; ///< Data Id or Ptr
};
};
/// Node for each branch level
struct Node
{
constexpr bool IsInternalNode() const { return m_level > 0; } // Not a leaf, but a internal node
constexpr bool IsLeaf() const { return m_level == 0; } // A leaf, contains data
int m_count; ///< Count
int m_level; ///< Leaf is zero, others positive
Branch m_branch[MAXNODES]; ///< Branch
};
/// A link list of nodes for reinsertion after a delete operation
struct ListNode
{
ListNode* m_next; ///< Next in list
Node* m_node; ///< Node
};
/// Variables for finding a split partition
struct PartitionVars
{
int m_partition[MAXNODES + 1];
int m_total;
int m_minFill;
bool m_taken[MAXNODES + 1];
int m_count[2];
Rect m_cover[2];
ELEMTYPEREAL m_area[2];
Branch m_branchBuf[MAXNODES + 1];
int m_branchCount;
Rect m_coverSplit;
ELEMTYPEREAL m_coverSplitArea;
};
/// Data structure used for Nearest Neighbor search implementation
struct NNNode
{
Branch m_branch;
ELEMTYPE minDist;
bool isLeaf;
inline bool operator<(const NNNode &other) const
{
/// This is reversed on purpose to use std::priority_queue
return other.minDist < minDist;
}
};
Node* AllocNode() const;
void FreeNode( Node* a_node ) const;
void InitNode( Node* a_node ) const;
void InitRect( Rect* a_rect ) const;
bool InsertRectRec( const Rect* a_rect,
const DATATYPE& a_id,
Node* a_node,
Node** a_newNode,
int a_level ) const;
bool InsertRect( const Rect* a_rect, const DATATYPE& a_id, Node** a_root, int a_level ) const;
Rect NodeCover( Node* a_node ) const;
bool AddBranch( const Branch* a_branch, Node* a_node, Node** a_newNode ) const;
void DisconnectBranch( Node* a_node, int a_index ) const;
int PickBranch( const Rect* a_rect, Node* a_node ) const;
Rect CombineRect( const Rect* a_rectA, const Rect* a_rectB ) const;
void SplitNode( Node* a_node, const Branch* a_branch, Node** a_newNode ) const;
ELEMTYPEREAL RectSphericalVolume( const Rect* a_rect ) const;
ELEMTYPEREAL RectVolume( const Rect* a_rect ) const;
ELEMTYPEREAL CalcRectVolume( const Rect* a_rect ) const;
void GetBranches( Node* a_node, const Branch* a_branch, PartitionVars* a_parVars ) const;
void ChoosePartition( PartitionVars* a_parVars, int a_minFill ) const;
void LoadNodes( Node* a_nodeA, Node* a_nodeB, PartitionVars* a_parVars ) const;
void InitParVars( PartitionVars* a_parVars, int a_maxRects, int a_minFill ) const;
void PickSeeds( PartitionVars* a_parVars ) const;
void Classify( int a_index, int a_group, PartitionVars* a_parVars ) const;
bool RemoveRect( const Rect* a_rect, const DATATYPE& a_id, Node** a_root ) const;
bool RemoveRectRec( const Rect* a_rect,
const DATATYPE& a_id,
Node* a_node,
ListNode** a_listNode ) const;
ListNode* AllocListNode() const;
void FreeListNode( ListNode* a_listNode ) const;
static bool Overlap( const Rect* a_rectA, const Rect* a_rectB );
void ReInsert( Node* a_node, ListNode** a_listNode ) const;
ELEMTYPE MinDist( const ELEMTYPE a_point[NUMDIMS], const Rect& a_rect ) const;
bool Search( const Node* a_node, const Rect* a_rect, int& a_foundCount,
std::function<bool (const DATATYPE&)> a_callback ) const;
template <class VISITOR>
bool Search( const Node* a_node, const Rect* a_rect, VISITOR& a_visitor, int& a_foundCount ) const
{
ASSERT( a_node );
ASSERT( a_node->m_level >= 0 );
ASSERT( a_rect );
if( a_node->IsInternalNode() ) // This is an internal node in the tree
{
for( int index = 0; index < a_node->m_count; ++index )
{
if( Overlap( a_rect, &a_node->m_branch[index].m_rect ) )
{
if( !Search( a_node->m_branch[index].m_child, a_rect, a_visitor, a_foundCount ) )
{
return false; // Don't continue searching
}
}
}
}
else // This is a leaf node
{
for( int index = 0; index < a_node->m_count; ++index )
{
if( Overlap( a_rect, &a_node->m_branch[index].m_rect ) )
{
const DATATYPE& id = a_node->m_branch[index].m_data;
if( !a_visitor( id ) )
return false;
a_foundCount++;
}
}
}
return true; // Continue searching
}
void RemoveAllRec( Node* a_node ) const;
void Reset() const;
void CountRec( const Node* a_node, int& a_count ) const;
bool SaveRec( const Node* a_node, RTFileStream& a_stream ) const;
bool LoadRec( const Node* a_node, RTFileStream& a_stream ) const;
Node* m_root; ///< Root of tree
ELEMTYPEREAL m_unitSphereVolume; ///< Unit sphere constant for required number of dimensions
};
// Because there is not stream support, this is a quick and dirty file I/O helper.
// Users will likely replace its usage with a Stream implementation from their favorite API.
class RTFileStream
{
FILE* m_file;
public:
RTFileStream()
{
m_file = NULL;
}
~RTFileStream()
{
Close();
}
bool OpenRead( const char* a_fileName )
{
m_file = std::fopen( a_fileName, "rb" );
if( !m_file )
{
return false;
}
return true;
}
bool OpenWrite( const char* a_fileName )
{
m_file = std::fopen( a_fileName, "wb" );
if( !m_file )
{
return false;
}
return true;
}
void Close()
{
if( m_file )
{
std::fclose( m_file );
m_file = NULL;
}
}
template <typename TYPE>
size_t Write( const TYPE& a_value )
{
ASSERT( m_file );
return std::fwrite( (void*) &a_value, sizeof(a_value), 1, m_file );
}
template <typename TYPE>
size_t WriteArray( const TYPE* a_array, int a_count )
{
ASSERT( m_file );
return std::fwrite( (void*) a_array, sizeof(TYPE) * a_count, 1, m_file );
}
template <typename TYPE>
size_t Read( TYPE& a_value )
{
ASSERT( m_file );
return std::fread( (void*) &a_value, sizeof(a_value), 1, m_file );
}
template <typename TYPE>
size_t ReadArray( TYPE* a_array, int a_count )
{
ASSERT( m_file );
return std::fread( (void*) a_array, sizeof(TYPE) * a_count, 1, m_file );
}
};
RTREE_TEMPLATE RTREE_QUAL::RTree()
{
ASSERT( MAXNODES > MINNODES );
ASSERT( MINNODES > 0 );
// We only support machine word size simple data type eg. integer index or object pointer.
// Since we are storing as union with non data branch
ASSERT( sizeof(DATATYPE) == sizeof(void*) || sizeof(DATATYPE) == sizeof(int) );
// Precomputed volumes of the unit spheres for the first few dimensions
const float UNIT_SPHERE_VOLUMES[] =
{
0.000000f, 2.000000f, 3.141593f, // Dimension 0,1,2
4.188790f, 4.934802f, 5.263789f, // Dimension 3,4,5
5.167713f, 4.724766f, 4.058712f, // Dimension 6,7,8
3.298509f, 2.550164f, 1.884104f, // Dimension 9,10,11
1.335263f, 0.910629f, 0.599265f, // Dimension 12,13,14
0.381443f, 0.235331f, 0.140981f, // Dimension 15,16,17
0.082146f, 0.046622f, 0.025807f, // Dimension 18,19,20
};
m_root = AllocNode();
m_root->m_level = 0;
m_unitSphereVolume = (ELEMTYPEREAL) UNIT_SPHERE_VOLUMES[NUMDIMS];
}
RTREE_TEMPLATE
RTREE_QUAL::~RTree() {
Reset(); // Free, or reset node memory
}
RTREE_TEMPLATE
void RTREE_QUAL::Insert( const ELEMTYPE a_min[NUMDIMS],
const ELEMTYPE a_max[NUMDIMS],
const DATATYPE& a_dataId )
{
#ifdef _DEBUG
for( int index = 0; index<NUMDIMS; ++index )
{
ASSERT( a_min[index] <= a_max[index] );
}
#endif // _DEBUG
Rect rect;
for( int axis = 0; axis < NUMDIMS; ++axis )
{
rect.m_min[axis] = a_min[axis];
rect.m_max[axis] = a_max[axis];
}
InsertRect( &rect, a_dataId, &m_root, 0 );
}
RTREE_TEMPLATE
bool RTREE_QUAL::Remove( const ELEMTYPE a_min[NUMDIMS],
const ELEMTYPE a_max[NUMDIMS],
const DATATYPE& a_dataId )
{
#ifdef _DEBUG
for( int index = 0; index<NUMDIMS; ++index )
{
ASSERT( a_min[index] <= a_max[index] );
}
#endif // _DEBUG
Rect rect;
for( int axis = 0; axis < NUMDIMS; ++axis )
{
rect.m_min[axis] = a_min[axis];
rect.m_max[axis] = a_max[axis];
}
return RemoveRect( &rect, a_dataId, &m_root );
}
RTREE_TEMPLATE
int RTREE_QUAL::Search( const ELEMTYPE a_min[NUMDIMS],
const ELEMTYPE a_max[NUMDIMS],
std::function<bool (const DATATYPE&)> a_callback ) const
{
#ifdef _DEBUG
for( int index = 0; index < NUMDIMS; ++index )
{
ASSERT( a_min[index] <= a_max[index] );
}
#endif // _DEBUG
Rect rect;
for( int axis = 0; axis < NUMDIMS; ++axis )
{
rect.m_min[axis] = a_min[axis];
rect.m_max[axis] = a_max[axis];
}
// NOTE: May want to return search result another way, perhaps returning the number of found elements here.
int foundCount = 0;
Search( m_root, &rect, foundCount, a_callback );
return foundCount;
}
RTREE_TEMPLATE
int RTREE_QUAL::Search( const ELEMTYPE a_min[NUMDIMS], const ELEMTYPE a_max[NUMDIMS],
std::function<bool( const DATATYPE& )> a_callback, bool& aFinished ) const
{
#ifdef _DEBUG
for( int index = 0; index < NUMDIMS; ++index )
{
ASSERT( a_min[index] <= a_max[index] );
}
#endif // _DEBUG
Rect rect;
for( int axis = 0; axis < NUMDIMS; ++axis )
{
rect.m_min[axis] = a_min[axis];
rect.m_max[axis] = a_max[axis];
}
// NOTE: May want to return search result another way, perhaps returning the number of found elements here.
int foundCount = 0;
aFinished = Search( m_root, &rect, foundCount, a_callback );
return foundCount;
}
RTREE_TEMPLATE
std::vector<std::pair<ELEMTYPE, DATATYPE>> RTREE_QUAL::NearestNeighbors(
const ELEMTYPE a_point[NUMDIMS],
std::function<bool( const std::size_t aNumResults, const ELEMTYPE aMinDist )> aTerminate,
std::function<bool( const DATATYPE aElement )> aFilter,
std::function<ELEMTYPE( const ELEMTYPE a_point[NUMDIMS], const DATATYPE a_data )> aSquaredDist ) const
{
std::vector<std::pair<ELEMTYPE, DATATYPE>> result;
std::priority_queue<NNNode> search_q;
for( int i = 0; i < m_root->m_count; ++i )
{
if( m_root->IsLeaf() )
{
search_q.push( NNNode{ m_root->m_branch[i],
aSquaredDist( a_point, m_root->m_branch[i].m_data ),
m_root->IsLeaf() });
}
else
{
search_q.push( NNNode{ m_root->m_branch[i],
MinDist(a_point, m_root->m_branch[i].m_rect),
m_root->IsLeaf() });
}
}
while( !search_q.empty() )
{
const NNNode curNode = search_q.top();
if( aTerminate( result.size(), curNode.minDist ) )
break;
search_q.pop();
if( curNode.isLeaf )
{
if( aFilter( curNode.m_branch.m_data ) )
result.emplace_back( curNode.minDist, curNode.m_branch.m_data );
}
else
{
Node* node = curNode.m_branch.m_child;
for( int i = 0; i < node->m_count; ++i )
{
NNNode newNode;
newNode.isLeaf = node->IsLeaf();
newNode.m_branch = node->m_branch[i];
if( newNode.isLeaf )
newNode.minDist = aSquaredDist( a_point, newNode.m_branch.m_data );
else
newNode.minDist = this->MinDist( a_point, node->m_branch[i].m_rect );
search_q.push( newNode );
}
}
}
return result;
}
RTREE_TEMPLATE
int RTREE_QUAL::Count() const
{
int count = 0;
CountRec( m_root, count );
return count;
}
RTREE_TEMPLATE
void RTREE_QUAL::CountRec( const Node* a_node, int& a_count ) const
{
if( a_node->IsInternalNode() ) // not a leaf node
{
for( int index = 0; index < a_node->m_count; ++index )
{
CountRec( a_node->m_branch[index].m_child, a_count );
}
}
else // A leaf node
{
a_count += a_node->m_count;
}
}
RTREE_TEMPLATE
bool RTREE_QUAL::Load( const char* a_fileName )
{
RemoveAll(); // Clear existing tree
RTFileStream stream;
if( !stream.OpenRead( a_fileName ) )
{
return false;
}
bool result = Load( stream );
stream.Close();
return result;
}
RTREE_TEMPLATE
bool RTREE_QUAL::Load( RTFileStream& a_stream ) const
{
// Write some kind of header
int _dataFileId = ('R' << 0) | ('T' << 8) | ('R' << 16) | ('E' << 24);
int _dataSize = sizeof(DATATYPE);
int _dataNumDims = NUMDIMS;
int _dataElemSize = sizeof(ELEMTYPE);
int _dataElemRealSize = sizeof(ELEMTYPEREAL);
int _dataMaxNodes = TMAXNODES;
int _dataMinNodes = TMINNODES;
int dataFileId = 0;
int dataSize = 0;
int dataNumDims = 0;
int dataElemSize = 0;
int dataElemRealSize = 0;
int dataMaxNodes = 0;
int dataMinNodes = 0;
a_stream.Read( dataFileId );
a_stream.Read( dataSize );
a_stream.Read( dataNumDims );
a_stream.Read( dataElemSize );
a_stream.Read( dataElemRealSize );
a_stream.Read( dataMaxNodes );
a_stream.Read( dataMinNodes );
bool result = false;
// Test if header was valid and compatible
if( (dataFileId == _dataFileId)
&& (dataSize == _dataSize)
&& (dataNumDims == _dataNumDims)
&& (dataElemSize == _dataElemSize)
&& (dataElemRealSize == _dataElemRealSize)
&& (dataMaxNodes == _dataMaxNodes)
&& (dataMinNodes == _dataMinNodes)
)
{
// Recursively load tree
result = LoadRec( m_root, a_stream );
}
return result;
}
RTREE_TEMPLATE
bool RTREE_QUAL::LoadRec( const Node* a_node, RTFileStream& a_stream ) const
{
a_stream.Read( a_node->m_level );
a_stream.Read( a_node->m_count );
if( a_node->IsInternalNode() ) // not a leaf node
{
for( int index = 0; index < a_node->m_count; ++index )
{
const Branch* curBranch = &a_node->m_branch[index];
a_stream.ReadArray( curBranch->m_rect.m_min, NUMDIMS );
a_stream.ReadArray( curBranch->m_rect.m_max, NUMDIMS );
curBranch->m_child = AllocNode();
LoadRec( curBranch->m_child, a_stream );
}
}
else // A leaf node
{
for( int index = 0; index < a_node->m_count; ++index )
{
const Branch* curBranch = &a_node->m_branch[index];
a_stream.ReadArray( curBranch->m_rect.m_min, NUMDIMS );
a_stream.ReadArray( curBranch->m_rect.m_max, NUMDIMS );
a_stream.Read( curBranch->m_data );
}
}
return true; // Should do more error checking on I/O operations
}
RTREE_TEMPLATE
bool RTREE_QUAL::Save( const char* a_fileName )
{
RTFileStream stream;
if( !stream.OpenWrite( a_fileName ) )
{
return false;
}
bool result = Save( stream );
stream.Close();
return result;
}
RTREE_TEMPLATE
bool RTREE_QUAL::Save( RTFileStream& a_stream ) const
{
// Write some kind of header
int dataFileId = ('R' << 0) | ('T' << 8) | ('R' << 16) | ('E' << 24);
int dataSize = sizeof(DATATYPE);
int dataNumDims = NUMDIMS;
int dataElemSize = sizeof(ELEMTYPE);
int dataElemRealSize = sizeof(ELEMTYPEREAL);
int dataMaxNodes = TMAXNODES;
int dataMinNodes = TMINNODES;
a_stream.Write( dataFileId );
a_stream.Write( dataSize );
a_stream.Write( dataNumDims );
a_stream.Write( dataElemSize );
a_stream.Write( dataElemRealSize );
a_stream.Write( dataMaxNodes );
a_stream.Write( dataMinNodes );
// Recursively save tree
bool result = SaveRec( m_root, a_stream );
return result;
}
RTREE_TEMPLATE
bool RTREE_QUAL::SaveRec( const Node* a_node, RTFileStream& a_stream ) const
{
a_stream.Write( a_node->m_level );
a_stream.Write( a_node->m_count );
if( a_node->IsInternalNode() ) // not a leaf node
{
for( int index = 0; index < a_node->m_count; ++index )
{
const Branch* curBranch = &a_node->m_branch[index];
a_stream.WriteArray( curBranch->m_rect.m_min, NUMDIMS );
a_stream.WriteArray( curBranch->m_rect.m_max, NUMDIMS );
SaveRec( curBranch->m_child, a_stream );
}
}
else // A leaf node
{
for( int index = 0; index < a_node->m_count; ++index )
{
const Branch* curBranch = &a_node->m_branch[index];
a_stream.WriteArray( curBranch->m_rect.m_min, NUMDIMS );
a_stream.WriteArray( curBranch->m_rect.m_max, NUMDIMS );
a_stream.Write( curBranch->m_data );
}
}
return true; // Should do more error checking on I/O operations
}
RTREE_TEMPLATE
void RTREE_QUAL::RemoveAll()
{
// Delete all existing nodes
Reset();
m_root = AllocNode();
m_root->m_level = 0;
}
RTREE_TEMPLATE
void RTREE_QUAL::Reset() const
{
#ifdef RTREE_DONT_USE_MEMPOOLS
// Delete all existing nodes
RemoveAllRec( m_root );
#else // RTREE_DONT_USE_MEMPOOLS
// Just reset memory pools. We are not using complex types
// EXAMPLE
#endif // RTREE_DONT_USE_MEMPOOLS
}
RTREE_TEMPLATE
void RTREE_QUAL::RemoveAllRec( Node* a_node ) const
{
ASSERT( a_node );
ASSERT( a_node->m_level >= 0 );
if( a_node->IsInternalNode() ) // This is an internal node in the tree
{
for( int index = 0; index < a_node->m_count; ++index )
{
RemoveAllRec( a_node->m_branch[index].m_child );
}
}
FreeNode( a_node );
}
RTREE_TEMPLATE
typename RTREE_QUAL::Node* RTREE_QUAL::AllocNode() const
{
Node* newNode;
#ifdef RTREE_DONT_USE_MEMPOOLS
newNode = new Node;
#else // RTREE_DONT_USE_MEMPOOLS
// EXAMPLE
#endif // RTREE_DONT_USE_MEMPOOLS
InitNode( newNode );
return newNode;
}
RTREE_TEMPLATE
void RTREE_QUAL::FreeNode( Node* a_node ) const
{
ASSERT( a_node );
#ifdef RTREE_DONT_USE_MEMPOOLS
delete a_node;
#else // RTREE_DONT_USE_MEMPOOLS
// EXAMPLE
#endif // RTREE_DONT_USE_MEMPOOLS
}
// Allocate space for a node in the list used in DeletRect to
// store Nodes that are too empty.
RTREE_TEMPLATE
typename RTREE_QUAL::ListNode* RTREE_QUAL::AllocListNode() const
{
#ifdef RTREE_DONT_USE_MEMPOOLS
return new ListNode;
#else // RTREE_DONT_USE_MEMPOOLS
// EXAMPLE
#endif // RTREE_DONT_USE_MEMPOOLS
}
RTREE_TEMPLATE
void RTREE_QUAL::FreeListNode( ListNode* a_listNode ) const
{
#ifdef RTREE_DONT_USE_MEMPOOLS
delete a_listNode;
#else // RTREE_DONT_USE_MEMPOOLS
// EXAMPLE
#endif // RTREE_DONT_USE_MEMPOOLS
}
RTREE_TEMPLATE
void RTREE_QUAL::InitNode( Node* a_node ) const
{
a_node->m_count = 0;
a_node->m_level = -1;
}
RTREE_TEMPLATE
void RTREE_QUAL::InitRect( Rect* a_rect ) const
{
for( int index = 0; index < NUMDIMS; ++index )
{
a_rect->m_min[index] = (ELEMTYPE) 0;
a_rect->m_max[index] = (ELEMTYPE) 0;
}
}
// Inserts a new data rectangle into the index structure.
// Recursively descends tree, propagates splits back up.
// Returns 0 if node was not split. Old node updated.
// If node was split, returns 1 and sets the pointer pointed to by
// new_node to point to the new node. Old node updated to become one of two.
// The level argument specifies the number of steps up from the leaf
// level to insert; e.g. a data rectangle goes in at level = 0.
RTREE_TEMPLATE
bool RTREE_QUAL::InsertRectRec( const Rect* a_rect,
const DATATYPE& a_id,
Node* a_node,
Node** a_newNode,
int a_level ) const
{
ASSERT( a_rect && a_node && a_newNode );
ASSERT( a_level >= 0 && a_level <= a_node->m_level );
int index;
Branch branch;
Node* otherNode;
// Still above level for insertion, go down tree recursively
if( a_node->m_level > a_level )
{
index = PickBranch( a_rect, a_node );
if( !InsertRectRec( a_rect, a_id, a_node->m_branch[index].m_child, &otherNode, a_level ) )
{
// Child was not split
a_node->m_branch[index].m_rect =
CombineRect( a_rect, &(a_node->m_branch[index].m_rect) );
return false;
}
else // Child was split
{
a_node->m_branch[index].m_rect = NodeCover( a_node->m_branch[index].m_child );
branch.m_child = otherNode;
branch.m_rect = NodeCover( otherNode );
return AddBranch( &branch, a_node, a_newNode );
}
}
else if( a_node->m_level == a_level ) // Have reached level for insertion. Add rect, split if necessary
{
branch.m_rect = *a_rect;
branch.m_child = (Node*) a_id;
// Child field of leaves contains id of data record
return AddBranch( &branch, a_node, a_newNode );
}
else
{
// Should never occur
ASSERT( 0 );
return false;
}
}
// Insert a data rectangle into an index structure.
// InsertRect provides for splitting the root;
// returns 1 if root was split, 0 if it was not.
// The level argument specifies the number of steps up from the leaf
// level to insert; e.g. a data rectangle goes in at level = 0.
// InsertRect2 does the recursion.
//
RTREE_TEMPLATE
bool RTREE_QUAL::InsertRect( const Rect* a_rect, const DATATYPE& a_id, Node** a_root, int a_level ) const
{
ASSERT( a_rect && a_root );
ASSERT( a_level >= 0 && a_level <= (*a_root)->m_level );
#ifdef _DEBUG
for( int index = 0; index < NUMDIMS; ++index )
{
ASSERT( a_rect->m_min[index] <= a_rect->m_max[index] );
}
#endif // _DEBUG
Node* newRoot;
Node* newNode;
Branch branch;
if( InsertRectRec( a_rect, a_id, *a_root, &newNode, a_level ) ) // Root split
{
newRoot = AllocNode(); // Grow tree taller and new root
newRoot->m_level = (*a_root)->m_level + 1;
branch.m_rect = NodeCover( *a_root );
branch.m_child = *a_root;
AddBranch( &branch, newRoot, NULL );
branch.m_rect = NodeCover( newNode );
branch.m_child = newNode;
AddBranch( &branch, newRoot, NULL );
*a_root = newRoot;
return true;
}
return false;
}
// Find the smallest rectangle that includes all rectangles in branches of a node.
RTREE_TEMPLATE
typename RTREE_QUAL::Rect RTREE_QUAL::NodeCover( Node* a_node ) const
{
ASSERT( a_node );
bool firstTime = true;
Rect rect;
InitRect( &rect );
for( int index = 0; index < a_node->m_count; ++index )
{
if( firstTime )
{
rect = a_node->m_branch[index].m_rect;
firstTime = false;
}
else
{
rect = CombineRect( &rect, &(a_node->m_branch[index].m_rect) );
}
}
return rect;
}
// Add a branch to a node. Split the node if necessary.
// Returns 0 if node not split. Old node updated.
// Returns 1 if node split, sets *new_node to address of new node.
// Old node updated, becomes one of two.
RTREE_TEMPLATE
bool RTREE_QUAL::AddBranch( const Branch* a_branch, Node* a_node, Node** a_newNode ) const
{
ASSERT( a_branch );
ASSERT( a_node );
if( a_node->m_count < MAXNODES ) // Split won't be necessary
{
a_node->m_branch[a_node->m_count] = *a_branch;
++a_node->m_count;
return false;
}
else
{
ASSERT( a_newNode );
SplitNode( a_node, a_branch, a_newNode );
return true;
}
}
// Disconnect a dependent node.
// Caller must return (or stop using iteration index) after this as count has changed
RTREE_TEMPLATE
void RTREE_QUAL::DisconnectBranch( Node* a_node, int a_index ) const
{
ASSERT( a_node && (a_index >= 0) && (a_index < MAXNODES) );
ASSERT( a_node->m_count > 0 );
// Remove element by swapping with the last element to prevent gaps in array
a_node->m_branch[a_index] = a_node->m_branch[a_node->m_count - 1];
--a_node->m_count;
}
// Pick a branch. Pick the one that will need the smallest increase
// in area to accomodate the new rectangle. This will result in the
// least total area for the covering rectangles in the current node.
// In case of a tie, pick the one which was smaller before, to get
// the best resolution when searching.
RTREE_TEMPLATE
int RTREE_QUAL::PickBranch( const Rect* a_rect, Node* a_node ) const
{
ASSERT( a_rect && a_node );
bool firstTime = true;
ELEMTYPEREAL increase;
ELEMTYPEREAL bestIncr = (ELEMTYPEREAL) -1;
ELEMTYPEREAL area;
ELEMTYPEREAL bestArea = 0;
int best = 0;
Rect tempRect;
for( int index = 0; index < a_node->m_count; ++index )
{
Rect* curRect = &a_node->m_branch[index].m_rect;
area = CalcRectVolume( curRect );
tempRect = CombineRect( a_rect, curRect );
increase = CalcRectVolume( &tempRect ) - area;
if( (increase < bestIncr) || firstTime )
{
best = index;
bestArea = area;
bestIncr = increase;
firstTime = false;
}
else if( (increase == bestIncr) && (area < bestArea) )
{
best = index;
bestArea = area;
bestIncr = increase;
}
}
return best;
}
// Combine two rectangles into larger one containing both
RTREE_TEMPLATE
typename RTREE_QUAL::Rect RTREE_QUAL::CombineRect( const Rect* a_rectA, const Rect* a_rectB ) const
{
ASSERT( a_rectA && a_rectB );
Rect newRect;
for( int index = 0; index < NUMDIMS; ++index )
{
newRect.m_min[index] = std::min( a_rectA->m_min[index], a_rectB->m_min[index] );
newRect.m_max[index] = std::max( a_rectA->m_max[index], a_rectB->m_max[index] );
}
return newRect;
}
// Split a node.
// Divides the nodes branches and the extra one between two nodes.
// Old node is one of the new ones, and one really new one is created.
// Tries more than one method for choosing a partition, uses best result.
RTREE_TEMPLATE
void RTREE_QUAL::SplitNode( Node* a_node, const Branch* a_branch, Node** a_newNode ) const
{
ASSERT( a_node );
ASSERT( a_branch );
// Could just use local here, but member or external is faster since it is reused
PartitionVars localVars;
PartitionVars* parVars = &localVars;
int level;
// Load all the branches into a buffer, initialize old node
level = a_node->m_level;
GetBranches( a_node, a_branch, parVars );
// Find partition
ChoosePartition( parVars, MINNODES );
// Put branches from buffer into 2 nodes according to chosen partition
*a_newNode = AllocNode();
(*a_newNode)->m_level = a_node->m_level = level;
LoadNodes( a_node, *a_newNode, parVars );
ASSERT( (a_node->m_count + (*a_newNode)->m_count) == parVars->m_total );
}
// Calculate the n-dimensional volume of a rectangle
RTREE_TEMPLATE
ELEMTYPEREAL RTREE_QUAL::RectVolume( const Rect* a_rect ) const
{
ASSERT( a_rect );
ELEMTYPEREAL volume = (ELEMTYPEREAL) 1;
for( int index = 0; index<NUMDIMS; ++index )
{
volume *= a_rect->m_max[index] - a_rect->m_min[index];
}
ASSERT( volume >= (ELEMTYPEREAL) 0 );
return volume;
}
// The exact volume of the bounding sphere for the given Rect
RTREE_TEMPLATE
ELEMTYPEREAL RTREE_QUAL::RectSphericalVolume( const Rect* a_rect ) const
{
ASSERT( a_rect );
ELEMTYPEREAL sumOfSquares = (ELEMTYPEREAL) 0;
for( int index = 0; index < NUMDIMS; ++index )
{
ELEMTYPEREAL halfExtent =
( (ELEMTYPEREAL) a_rect->m_max[index] - (ELEMTYPEREAL) a_rect->m_min[index] ) * 0.5f;
sumOfSquares += halfExtent * halfExtent;
}
// Pow maybe slow, so test for common dims like 2,3 and just use x*x, x*x*x.
if( NUMDIMS == 2 )
{
return sumOfSquares * m_unitSphereVolume;
}
else if( NUMDIMS == 3 )
{
ELEMTYPEREAL radius = (ELEMTYPEREAL) std::sqrt( sumOfSquares );
return radius * radius * radius * m_unitSphereVolume;
}
else
{
ELEMTYPEREAL radius = (ELEMTYPEREAL) std::sqrt( sumOfSquares );
return (ELEMTYPEREAL) (std::pow( radius, NUMDIMS ) * m_unitSphereVolume);
}
}
// Use one of the methods to calculate retangle volume
RTREE_TEMPLATE
ELEMTYPEREAL RTREE_QUAL::CalcRectVolume( const Rect* a_rect ) const
{
#ifdef RTREE_USE_SPHERICAL_VOLUME
return RectSphericalVolume( a_rect ); // Slower but helps certain merge cases
#else // RTREE_USE_SPHERICAL_VOLUME
return RectVolume( a_rect ); // Faster but can cause poor merges
#endif // RTREE_USE_SPHERICAL_VOLUME
}
// Load branch buffer with branches from full node plus the extra branch.
RTREE_TEMPLATE
void RTREE_QUAL::GetBranches( Node* a_node, const Branch* a_branch, PartitionVars* a_parVars ) const
{
ASSERT( a_node );
ASSERT( a_branch );
ASSERT( a_node->m_count == MAXNODES );
// Load the branch buffer
for( int index = 0; index < MAXNODES; ++index )
{
a_parVars->m_branchBuf[index] = a_node->m_branch[index];
}
a_parVars->m_branchBuf[MAXNODES] = *a_branch;
a_parVars->m_branchCount = MAXNODES + 1;
// Calculate rect containing all in the set
a_parVars->m_coverSplit = a_parVars->m_branchBuf[0].m_rect;
for( int index = 1; index < MAXNODES + 1; ++index )
{
a_parVars->m_coverSplit =
CombineRect( &a_parVars->m_coverSplit, &a_parVars->m_branchBuf[index].m_rect );
}
a_parVars->m_coverSplitArea = CalcRectVolume( &a_parVars->m_coverSplit );
InitNode( a_node );
}
// Method #0 for choosing a partition:
// As the seeds for the two groups, pick the two rects that would waste the
// most area if covered by a single rectangle, i.e. evidently the worst pair
// to have in the same group.
// Of the remaining, one at a time is chosen to be put in one of the two groups.
// The one chosen is the one with the greatest difference in area expansion
// depending on which group - the rect most strongly attracted to one group
// and repelled from the other.
// If one group gets too full (more would force other group to violate min
// fill requirement) then other group gets the rest.
// These last are the ones that can go in either group most easily.
RTREE_TEMPLATE
void RTREE_QUAL::ChoosePartition( PartitionVars* a_parVars, int a_minFill ) const
{
ASSERT( a_parVars );
ELEMTYPEREAL biggestDiff;
int group, chosen = 0, betterGroup = 0;
InitParVars( a_parVars, a_parVars->m_branchCount, a_minFill );
PickSeeds( a_parVars );
while( ( (a_parVars->m_count[0] + a_parVars->m_count[1]) < a_parVars->m_total )
&& ( a_parVars->m_count[0] < (a_parVars->m_total - a_parVars->m_minFill) )
&& ( a_parVars->m_count[1] < (a_parVars->m_total - a_parVars->m_minFill) ) )
{
biggestDiff = (ELEMTYPEREAL) -1;
for( int index = 0; index<a_parVars->m_total; ++index )
{
if( !a_parVars->m_taken[index] )
{
const Rect* curRect = &a_parVars->m_branchBuf[index].m_rect;
const Rect rect0 = CombineRect( curRect, &a_parVars->m_cover[0] );
const Rect rect1 = CombineRect( curRect, &a_parVars->m_cover[1] );
ELEMTYPEREAL growth0 = CalcRectVolume( &rect0 ) - a_parVars->m_area[0];
ELEMTYPEREAL growth1 = CalcRectVolume( &rect1 ) - a_parVars->m_area[1];
ELEMTYPEREAL diff = growth1 - growth0;
if( diff >= 0 )
{
group = 0;
}
else
{
group = 1;
diff = -diff;
}
if( diff > biggestDiff )
{
biggestDiff = diff;
chosen = index;
betterGroup = group;
}
else if( (diff == biggestDiff)
&& (a_parVars->m_count[group] < a_parVars->m_count[betterGroup]) )
{
chosen = index;
betterGroup = group;
}
}
}
Classify( chosen, betterGroup, a_parVars );
}
// If one group too full, put remaining rects in the other
if( (a_parVars->m_count[0] + a_parVars->m_count[1]) < a_parVars->m_total )
{
if( a_parVars->m_count[0] >= a_parVars->m_total - a_parVars->m_minFill )
{
group = 1;
}
else
{
group = 0;
}
for( int index = 0; index<a_parVars->m_total; ++index )
{
if( !a_parVars->m_taken[index] )
{
Classify( index, group, a_parVars );
}
}
}
ASSERT( (a_parVars->m_count[0] + a_parVars->m_count[1]) == a_parVars->m_total );
ASSERT( (a_parVars->m_count[0] >= a_parVars->m_minFill)
&& (a_parVars->m_count[1] >= a_parVars->m_minFill) );
}
// Copy branches from the buffer into two nodes according to the partition.
RTREE_TEMPLATE
void RTREE_QUAL::LoadNodes( Node* a_nodeA, Node* a_nodeB, PartitionVars* a_parVars ) const
{
ASSERT( a_nodeA );
ASSERT( a_nodeB );
ASSERT( a_parVars );
for( int index = 0; index < a_parVars->m_total; ++index )
{
ASSERT( a_parVars->m_partition[index] == 0 || a_parVars->m_partition[index] == 1 );
if( a_parVars->m_partition[index] == 0 )
{
AddBranch( &a_parVars->m_branchBuf[index], a_nodeA, NULL );
}
else if( a_parVars->m_partition[index] == 1 )
{
AddBranch( &a_parVars->m_branchBuf[index], a_nodeB, NULL );
}
}
}
// Initialize a PartitionVars structure.
RTREE_TEMPLATE
void RTREE_QUAL::InitParVars( PartitionVars* a_parVars, int a_maxRects, int a_minFill ) const
{
ASSERT( a_parVars );
a_parVars->m_count[0] = a_parVars->m_count[1] = 0;
a_parVars->m_area[0] = a_parVars->m_area[1] = (ELEMTYPEREAL) 0;
a_parVars->m_total = a_maxRects;
a_parVars->m_minFill = a_minFill;
for( int index = 0; index < a_maxRects; ++index )
{
a_parVars->m_taken[index] = false;
a_parVars->m_partition[index] = -1;
}
}
RTREE_TEMPLATE
void RTREE_QUAL::PickSeeds( PartitionVars* a_parVars ) const
{
int seed0 = 0, seed1 = 0;
ELEMTYPEREAL worst, waste;
ELEMTYPEREAL area[MAXNODES + 1];
for( int index = 0; index<a_parVars->m_total; ++index )
{
area[index] = CalcRectVolume( &a_parVars->m_branchBuf[index].m_rect );
}
worst = -a_parVars->m_coverSplitArea - 1;
for( int indexA = 0; indexA < a_parVars->m_total - 1; ++indexA )
{
for( int indexB = indexA + 1; indexB < a_parVars->m_total; ++indexB )
{
Rect oneRect = CombineRect( &a_parVars->m_branchBuf[indexA].m_rect,
&a_parVars->m_branchBuf[indexB].m_rect );
waste = CalcRectVolume( &oneRect ) - area[indexA] - area[indexB];
if( waste >= worst )
{
worst = waste;
seed0 = indexA;
seed1 = indexB;
}
}
}
Classify( seed0, 0, a_parVars );
Classify( seed1, 1, a_parVars );
}
// Put a branch in one of the groups.
RTREE_TEMPLATE
void RTREE_QUAL::Classify( int a_index, int a_group, PartitionVars* a_parVars ) const
{
ASSERT( a_parVars );
ASSERT( !a_parVars->m_taken[a_index] );
a_parVars->m_partition[a_index] = a_group;
a_parVars->m_taken[a_index] = true;
if( a_parVars->m_count[a_group] == 0 )
{
a_parVars->m_cover[a_group] = a_parVars->m_branchBuf[a_index].m_rect;
}
else
{
a_parVars->m_cover[a_group] = CombineRect( &a_parVars->m_branchBuf[a_index].m_rect,
&a_parVars->m_cover[a_group] );
}
a_parVars->m_area[a_group] = CalcRectVolume( &a_parVars->m_cover[a_group] );
++a_parVars->m_count[a_group];
}
// Delete a data rectangle from an index structure.
// Pass in a pointer to a Rect, the tid of the record, ptr to ptr to root node.
// Returns 1 if record not found, 0 if success.
// RemoveRect provides for eliminating the root.
RTREE_TEMPLATE
bool RTREE_QUAL::RemoveRect( const Rect* a_rect, const DATATYPE& a_id, Node** a_root ) const
{
ASSERT( a_rect && a_root );
ASSERT( *a_root );
Node* tempNode;
ListNode* reInsertList = NULL;
if( !RemoveRectRec( a_rect, a_id, *a_root, &reInsertList ) )
{
// Found and deleted a data item
// Reinsert any branches from eliminated nodes
while( reInsertList )
{
tempNode = reInsertList->m_node;
for( int index = 0; index < tempNode->m_count; ++index )
{
InsertRect( &(tempNode->m_branch[index].m_rect),
tempNode->m_branch[index].m_data,
a_root,
tempNode->m_level );
}
ListNode* remLNode = reInsertList;
reInsertList = reInsertList->m_next;
FreeNode( remLNode->m_node );
FreeListNode( remLNode );
}
// Check for redundant root (not leaf, 1 child) and eliminate
if( (*a_root)->m_count == 1 && (*a_root)->IsInternalNode() )
{
tempNode = (*a_root)->m_branch[0].m_child;
ASSERT( tempNode );
FreeNode( *a_root );
*a_root = tempNode;
}
return false;
}
else
{
return true;
}
}
// Delete a rectangle from non-root part of an index structure.
// Called by RemoveRect. Descends tree recursively,
// merges branches on the way back up.
// Returns 1 if record not found, 0 if success.
RTREE_TEMPLATE
bool RTREE_QUAL::RemoveRectRec( const Rect* a_rect,
const DATATYPE& a_id,
Node* a_node,
ListNode** a_listNode ) const
{
ASSERT( a_rect && a_node && a_listNode );
ASSERT( a_node->m_level >= 0 );
if( a_node->IsInternalNode() ) // not a leaf node
{
for( int index = 0; index < a_node->m_count; ++index )
{
if( Overlap( a_rect, &(a_node->m_branch[index].m_rect) ) )
{
if( !RemoveRectRec( a_rect, a_id, a_node->m_branch[index].m_child, a_listNode ) )
{
if( a_node->m_branch[index].m_child->m_count >= MINNODES )
{
// child removed, just resize parent rect
a_node->m_branch[index].m_rect =
NodeCover( a_node->m_branch[index].m_child );
}
else
{
// child removed, not enough entries in node, eliminate node
ReInsert( a_node->m_branch[index].m_child, a_listNode );
DisconnectBranch( a_node, index ); // Must return after this call as count has changed
}
return false;
}
}
}
return true;
}
else // A leaf node
{
for( int index = 0; index < a_node->m_count; ++index )
{
if( a_node->m_branch[index].m_child == (Node*) a_id )
{
DisconnectBranch( a_node, index ); // Must return after this call as count has changed
return false;
}
}
return true;
}
}
// Decide whether two rectangles overlap.
RTREE_TEMPLATE
bool RTREE_QUAL::Overlap( const Rect* a_rectA, const Rect* a_rectB )
{
ASSERT( a_rectA && a_rectB );
for( int index = 0; index < NUMDIMS; ++index )
{
if( a_rectA->m_min[index] > a_rectB->m_max[index]
|| a_rectB->m_min[index] > a_rectA->m_max[index] )
{
return false;
}
}
return true;
}
// Add a node to the reinsertion list. All its branches will later
// be reinserted into the index structure.
RTREE_TEMPLATE
void RTREE_QUAL::ReInsert( Node* a_node, ListNode** a_listNode ) const
{
ListNode* newListNode;
newListNode = AllocListNode();
newListNode->m_node = a_node;
newListNode->m_next = *a_listNode;
*a_listNode = newListNode;
}
// Search in an index tree or subtree for all data rectangles that overlap the argument rectangle.
RTREE_TEMPLATE
bool RTREE_QUAL::Search( const Node* a_node, const Rect* a_rect, int& a_foundCount,
std::function<bool (const DATATYPE&)> a_callback ) const
{
ASSERT( a_node );
ASSERT( a_node->m_level >= 0 );
ASSERT( a_rect );
if( a_node->IsInternalNode() ) // This is an internal node in the tree
{
for( int index = 0; index < a_node->m_count; ++index )
{
if( Overlap( a_rect, &a_node->m_branch[index].m_rect ) )
{
if( !Search( a_node->m_branch[index].m_child, a_rect, a_foundCount, a_callback ) )
{
return false; // Don't continue searching
}
}
}
}
else // This is a leaf node
{
for( int index = 0; index < a_node->m_count; ++index )
{
if( Overlap( a_rect, &a_node->m_branch[index].m_rect ) )
{
DATATYPE& id = a_node->m_branch[index].m_data;
++a_foundCount;
if( a_callback && !a_callback( id ) )
{
return false; // Don't continue searching
}
}
}
}
return true; // Continue searching
}
//calculate the minimum distance between a point and a rectangle as defined by Manolopoulos et al.
// returns Euclidean norm to ensure value fits in ELEMTYPE
RTREE_TEMPLATE
ELEMTYPE RTREE_QUAL::MinDist( const ELEMTYPE a_point[NUMDIMS], const Rect& a_rect ) const
{
const ELEMTYPE *q, *s, *t;
q = a_point;
s = a_rect.m_min;
t = a_rect.m_max;
double minDist = 0.0;
for( int index = 0; index < NUMDIMS; index++ )
{
int r = q[index];
if( q[index] < s[index] )
{
r = s[index];
}
else if( q[index] > t[index] )
{
r = t[index];
}
double addend = q[index] - r;
minDist += addend * addend;
}
return std::lround( std::sqrt( minDist ) );
}
#undef RTREE_TEMPLATE
#undef RTREE_QUAL
#undef RTREE_SEARCH_TEMPLATE
#undef RTREE_SEARCH_QUAL
#endif // RTREE_H