std::string VSTree::to_str()
{
//debug
{
stringstream _ss;
_ss << "after build tree, root is:" << endl;
_ss << this->getRoot()->to_str() << endl;
Util::logging(_ss.str());
}
std::stringstream _ss;
std::queue<int> nodeFileLineQueue;
nodeFileLineQueue.push(this->getRoot()->getFileLine());
while(! nodeFileLineQueue.empty())
{
int currentNodeFileLine = nodeFileLineQueue.front();
nodeFileLineQueue.pop();
VNode* currentNodePtr = this->getNode(currentNodeFileLine);
_ss << currentNodePtr->to_str();
int childNum = currentNodePtr->getChildNum();
for(int i = 0; i < childNum; i ++)
{
if(! currentNodePtr->isLeaf())
{
int childNodeFileLine = currentNodePtr->getChildFileLine(i);
nodeFileLineQueue.push(childNodeFileLine);
}
}
}
return _ss.str();
}
void
VSTree::split(VNode* _p_node_being_split, const SigEntry& _insert_entry, VNode* _p_insert_node)
{
#ifdef DEBUG_VSTREE
stringstream _ss;
_ss << "**********************split happen at "
<< _p_node_being_split->getFileLine() << endl;
_ss << _p_node_being_split->to_str() << endl;
Util::logging(_ss.str());
#endif
// first, add the new child node(if not leaf) or child entry(if leaf) to the full node.
bool just_insert_entry = (_p_insert_node == NULL);
if(just_insert_entry)
{
_p_node_being_split->addChildEntry(_insert_entry, true);
}
else
{
_p_node_being_split->addChildNode(_p_insert_node, true);
}
SigEntry entryA, entryB;
//BETTER: use hanming, xor result or the vector included angle to guess the distince.
//And then also use the farest two as seeds.
//
//two seeds to generate two new nodes.
//seedA kernel: the SigEntry with the minimal count of signature.
//seedB kernel: the SigEntry with the maximal count of signature.
int maxCount = 0; // record the minimal signature count.
int entryA_index = 0; // record the seedA kernel index.
for(int i = 0; i < VNode::MAX_CHILD_NUM; i++)
{
int currentCount = (int) _p_node_being_split->getChildEntry(i).getSigCount();
if(maxCount < currentCount)
{
maxCount = currentCount;
entryA_index = i;
}
}
entryA = _p_node_being_split->getChildEntry(entryA_index);
maxCount = 0;
int entryB_index = 0; // record the seedB kernel index.
for(int i = 0; i < VNode::MAX_CHILD_NUM; i++)
{
//NOTICE:I think xOR should be used here to choose the farest two
int currentCount = entryA.xOR(_p_node_being_split->getChildEntry(i));
//int currentCount = entryA.xEpsilen(_p_node_being_split->getChildEntry(i));
if(i != entryA_index && maxCount <= currentCount)
{
maxCount = currentCount;
entryB_index = i;
}
}
entryB = _p_node_being_split->getChildEntry(entryB_index);
// AEntryIndex: the entry index near seedA.
// BEntryIndex: the entry index near seedB.
std::vector<int> entryIndex_nearA, entryIndex_nearB;
entryIndex_nearA.clear();
entryIndex_nearB.clear();
entryIndex_nearA.push_back(entryA_index);
entryIndex_nearB.push_back(entryB_index);
int nearA_max_size, nearB_max_size;
bool nearA_tooSmall, nearB_tooSmall;
for(int i = 0; i < VNode::MAX_CHILD_NUM; i++)
{
if(i == entryA_index || i == entryB_index) continue;
//should guarantee that each new node has at least MIN_CHILD_NUM children.
nearA_max_size = VNode::MAX_CHILD_NUM - entryIndex_nearB.size();
nearA_tooSmall = (nearA_max_size <= VNode::MIN_CHILD_NUM);
if(nearA_tooSmall)
{
for(; i < VNode::MAX_CHILD_NUM; i++)
{
if (i == entryA_index || i == entryB_index) continue;
entryIndex_nearA.push_back(i);
}
break;
}
nearB_max_size = VNode::MAX_CHILD_NUM - entryIndex_nearA.size();
nearB_tooSmall = (nearB_max_size <= VNode::MIN_CHILD_NUM);
if(nearB_tooSmall)
{
for(; i < VNode::MAX_CHILD_NUM; i++)
{
if(i == entryA_index || i == entryB_index) continue;
entryIndex_nearB.push_back(i);
}
break;
}
//calculate the distance from
//the i-th child entry signature to seedA(or seedB).
//NOTICE:we should expect that the candidate can be almost contained!
//However, the precondition there are not too many 1s
int disToSeedA = entryA.xEpsilen(_p_node_being_split->getChildEntry(i));
int disToSeedB = entryB.xEpsilen(_p_node_being_split->getChildEntry(i));
// choose the near one seed to add into
if(disToSeedA <= disToSeedB)
{
entryIndex_nearA.push_back(i);
}
else
{
entryIndex_nearB.push_back(i);
}
}
// then create a new node to act as BEntryIndex's father.
VNode* newNodePtr = this->createNode();
#ifdef DEBUG_VSTREE
stringstream _ss2;
_ss2 << "new Node is :[" << newNodePtr->getFileLine() << "]" << endl;
Util::logging(_ss2.str());
#endif
// the old one acts as AEntryIndex's father.
VNode* oldNodePtr = _p_node_being_split;
// if the old node is leaf, set the new node as a leaf.
if(oldNodePtr->isLeaf())
{
newNodePtr->setAsLeaf(true);
}
//add all the entries in BEntryIndex into the new node child entry array,
//and calculate the new node's entry.
for(unsigned i = 0; i < entryIndex_nearB.size(); i++)
{
if(oldNodePtr->isLeaf())
{
newNodePtr->addChildEntry(oldNodePtr->getChildEntry(entryIndex_nearB[i]), false);
}
else
{
//debug target 2
VNode* childPtr = oldNodePtr->getChild(entryIndex_nearB[i], *(this->node_buffer));
newNodePtr->addChildNode(childPtr);
}
}
newNodePtr->refreshSignature();
//label the child being removed with -1,
//and update the old node's entry.
sort(entryIndex_nearA.begin(), entryIndex_nearA.end(), less<int>());
#ifdef DEBUG_VSTREE
stringstream _ss1;
{
_ss1 << "nearA: ";
for(unsigned i = 0; i < entryIndex_nearA.size(); i++)
{
_ss1 << entryIndex_nearA[i] << " ";
}
_ss1 << endl;
_ss1 << "nearB: ";
for(unsigned i = 0; i < entryIndex_nearB.size(); i++)
{
_ss1 << entryIndex_nearB[i] << " ";
}
_ss1 << endl;
}
Util::logging(_ss1.str());
#endif
for(unsigned i = 0; i < entryIndex_nearA.size(); i++)
{
oldNodePtr->setChildEntry(i, oldNodePtr->getChildEntry(entryIndex_nearA[i]));
oldNodePtr->setChildFileLine(i, oldNodePtr->getChildFileLine(entryIndex_nearA[i]));
}
oldNodePtr->setChildNum(entryIndex_nearA.size());
oldNodePtr->refreshSignature();
int oldNode_index = oldNodePtr->getIndexInFatherNode(*(this->node_buffer));
// full node's father pointer.
VNode* oldNodeFatherPtr = oldNodePtr->getFather(*(this->node_buffer));
if(oldNodePtr->isRoot())
{
//if the old node is root,
//split the root, create a new root,
//and the tree height will be increased.
VNode* RootNewPtr = this->createNode();
//change the old root node to not-root node,
//and set the RootNew to root node.
oldNodePtr->setAsRoot(false);
RootNewPtr->setAsRoot(true);
//set the split two node(old node and new node) as the new root's child,
//and update signatures.
RootNewPtr->addChildNode(oldNodePtr);
RootNewPtr->addChildNode(newNodePtr);
RootNewPtr->refreshSignature();
//debug
// {
// stringstream _ss;
// _ss << "create new root:" << endl;
// _ss << "before swap file line, two sons are: " << oldNodePtr->getFileLine() << " " << newNodePtr->getFileLine() << endl;
// Util::logging(_ss.str());
// }
//should keep the root node always being
//at the first line(line zero) of the tree node file.
this->swapNodeFileLine(RootNewPtr, oldNodePtr);
this->height++;
//debug
// {
// stringstream _ss;
// _ss << "create new root:" << endl;
// _ss << "two sons are: " << oldNodePtr->getFileLine() << " " << newNodePtr->getFileLine() << endl;
// _ss << Signature::BitSet2str(oldNodePtr->getEntry().getEntitySig().entityBitSet) << endl;
// _ss << RootNewPtr->to_str() << endl;
// Util::logging(_ss.str());
// }
}
else
{
//if the (OldNode) is not Root,
//change the old node's signature to A's signature.
oldNodeFatherPtr->setChildEntry(oldNode_index, oldNodePtr->getEntry());
if(oldNodeFatherPtr->isFull())
{
oldNodeFatherPtr->refreshAncestorSignature(*(this->node_buffer));
this->split(oldNodeFatherPtr, newNodePtr->getEntry(), newNodePtr);
}
else
{
oldNodeFatherPtr->addChildNode(newNodePtr);
oldNodeFatherPtr->refreshAncestorSignature(*(this->node_buffer));
}
}
//debug
// if (!oldNodePtr->checkState())
// {
// stringstream _ss;
// _ss << "node " << oldNodePtr->getFileLine() << " childFileLine error. oldNode when split" << endl;
// Util::logging(_ss.str());
// }
// if (!newNodePtr->checkState())
// {
// stringstream _ss;
// _ss << "node " << newNodePtr->getFileLine() << " childFileLine error. newNode when split" << endl;
// Util::logging(_ss.str());
// }
// update the entityID2FileLineMap by these two nodes.
this->updateEntityID2FileLineMap(oldNodePtr);
this->updateEntityID2FileLineMap(newNodePtr);
}
//retrieve the candidate entity ID which signature can cover the _entity_bit_set, and add them to the _p_id_list.
void
VSTree::retrieveEntity(const EntityBitSet& _entity_bit_set, IDList* _p_id_list)
{
Util::logging("IN retrieveEntity");
EntitySig filterSig(_entity_bit_set);
#ifdef DEBUG_VSTREE
cerr << "the filter signature: " << filterSig.to_str() << endl;
#endif
queue<int> nodeQueue; //searching node file line queue.
//debug
{
stringstream _ss;
_ss << "filterSig=" << Signature::BitSet2str(filterSig.entityBitSet) << endl;
Util::logging(_ss.str());
}
const SigEntry& root_entry = (this->getRoot())->getEntry();
Util::logging("Get Root Entry");
if(root_entry.cover(filterSig))
{
nodeQueue.push(this->getRoot()->getFileLine());
Util::logging("root cover the filter_sig");
}
else
{
Util::logging("warning: root is not cover the filter_sig");
}
//debug
// {
// Util::logging(this->getRoot()->to_str());
// Util::logging("Before BFS");
// }
//using BFS algorithm to traverse the VSTree and retrieve the entry.
while (!nodeQueue.empty())
{
int currentNodeFileLine = nodeQueue.front();
nodeQueue.pop();
VNode* currentNodePtr = this->getNode(currentNodeFileLine);
int childNum = currentNodePtr->getChildNum();
//debug
// {
// std::stringstream _ss;
// _ss << "childNum of ["
// << currentNodePtr->getFileLine()
// << "] is " << childNum << endl;
//
// for (int i=0;i<childNum;i++)
// {
// _ss << currentNodePtr->getChildFileLine(i) << " ";
// }
// _ss << endl;
//
// Util::logging(_ss.str());
// }
int valid = 0;
for (int i = 0; i < childNum; i++)
{
const SigEntry& entry = currentNodePtr->getChildEntry(i);
#ifdef DEBUG_VSTREE
//cerr << "current entry: " << entry.to_str() << endl;
#endif
if (entry.cover(filterSig))
{
valid++;
if (currentNodePtr->isLeaf())
{
// if leaf node, add the satisfying entries' entity id to result list.
_p_id_list->addID(entry.getEntityId());
//debug
// {
// stringstream _ss;
// _ss << "child_" << i << " cover filter sig" << endl;
// _ss << Signature::BitSet2str(entry.getEntitySig().entityBitSet)<< endl;
// Util::logging(_ss.str());
// }
}
else
{
// if non-leaf node, add the child node pointer to the searching queue.
//VNode* childPtr = currentNodePtr->getChild(i, *(this->node_buffer));
// if non-leaf node, add the child node file line to the searching queue.
int childNodeFileLine = currentNodePtr->getChildFileLine(i);
nodeQueue.push(childNodeFileLine);
//debug
// {
// stringstream _ss;
// _ss << "child[" << childPtr->getFileLine() << "] cover filter sig" << endl;
// Util::logging(_ss.str());
// }
}
}
}
#ifdef DEBUG_VSTREE
//cerr << "child num: " << childNum << " valid num: " << valid << endl;
#endif
}
Util::logging("OUT retrieveEntity");
}
void VSTree::split(VNode* _p_node_being_split, const SigEntry& _insert_entry, VNode* _p_insert_node)
{
//debug
// {
// stringstream _ss;
// _ss << "**********************split happen at "
// << _p_node_being_split->getFileLine() << endl;
// _ss << _p_node_being_split->to_str() << endl;
// Util::logging(_ss.str());
// }
// first, add the new child node(if not leaf) or child entry(if leaf) to the full node.
bool just_insert_entry = (_p_insert_node == NULL);
if (just_insert_entry)
{
_p_node_being_split->addChildEntry(_insert_entry, true);
}
else
{
_p_node_being_split->addChildNode(_p_insert_node, true);
}
SigEntry entryA, entryB;
/* two seeds to generate two new nodes.
* seedA kernel: the SigEntry with the minimal count of signature.
* seedB kernel: the SigEntry with the second minimal count of signature.
* */
int minCount = 0; // record the minimal signature count.
int entryA_index = 0; // record the seedA kernel index.
for (int i=0;i<VNode::MAX_CHILD_NUM;i++)
{
int currentCount = (int) _p_node_being_split->getChildEntry(i).getSigCount();
if (minCount < currentCount)
{
minCount = currentCount;
entryA_index = i;
}
}
entryA = _p_node_being_split->getChildEntry(entryA_index);
minCount = 0;
int entryB_index = 0; // record the seedB kernel index.
for (int i=0;i<VNode::MAX_CHILD_NUM;i++)
{
int currentCount = entryA.xEpsilen(_p_node_being_split->getChildEntry(i));
if (i != entryA_index && minCount <= currentCount)
{
minCount = currentCount;
entryB_index = i;
}
}
entryB = _p_node_being_split->getChildEntry(entryB_index);
// AEntryIndex: the entry index near seedA.
// BEntryIndex: the entry index near seedB.
std::vector<int> entryIndex_nearA, entryIndex_nearB;
entryIndex_nearA.clear();
entryIndex_nearB.clear();
entryIndex_nearA.push_back(entryA_index);
entryIndex_nearB.push_back(entryB_index);
/* just tmp variables, for more readibility */
int nearA_max_size, nearB_max_size;
bool nearA_tooSmall, nearB_tooSmall;
for (int i=0;i<VNode::MAX_CHILD_NUM;i++)
{
if (i == entryA_index || i == entryB_index) continue;
/* should guarantee that each new node has at least MIN_CHILD_NUM children. */
nearA_max_size = VNode::MAX_CHILD_NUM - entryIndex_nearB.size();
nearA_tooSmall = (nearA_max_size <= VNode::MIN_CHILD_NUM);
if (nearA_tooSmall)
{
for (;i<VNode::MAX_CHILD_NUM;i++)
{
if (i == entryA_index || i == entryB_index) continue;
entryIndex_nearA.push_back(i);
}
break;
}
nearB_max_size = VNode::MAX_CHILD_NUM - entryIndex_nearA.size();
nearB_tooSmall = (nearB_max_size <= VNode::MIN_CHILD_NUM);
if (nearB_tooSmall)
{
for (;i<VNode::MAX_CHILD_NUM;i++)
{
if (i == entryA_index || i == entryB_index) continue;
entryIndex_nearB.push_back(i);
}
break;
}
/* calculate the distance from
* the i-th child entry signature to seedA(or seedB).*/
/*debug target 1*/
int disToSeedA = entryA.xEpsilen(_p_node_being_split->getChildEntry(i));
int disToSeedB = entryB.xEpsilen(_p_node_being_split->getChildEntry(i));
// choose the near one seed to add into
if (disToSeedA <= disToSeedB)
{
entryIndex_nearA.push_back(i);
}
else
{
entryIndex_nearB.push_back(i);
}
}
// then create a new node to act as BEntryIndex's father.
VNode* newNodePtr = this->createNode();
//debug
// {
// stringstream _ss;
// _ss << "new Node is :[" << newNodePtr->getFileLine() << "]" << endl;
// Util::logging(_ss.str());
// }
// the old one acts as AEntryIndex's father.
VNode* oldNodePtr = _p_node_being_split;
// if the old node is leaf, set the new node as a leaf.
if (oldNodePtr->isLeaf())
{
newNodePtr->setAsLeaf(true);
}
/* add all the entries in BEntryIndex into the new node child entry array,
and calculate the new node's entry.*/
for (unsigned i=0;i<entryIndex_nearB.size();i++)
{
if (oldNodePtr->isLeaf())
{
newNodePtr->addChildEntry(oldNodePtr->getChildEntry(entryIndex_nearB[i]), false);
}
else
{
/*debug target 2*/
VNode* childPtr = oldNodePtr->getChild(entryIndex_nearB[i], *(this->node_buffer));
newNodePtr->addChildNode(childPtr);
}
}
newNodePtr->refreshSignature();
/* label the child being removed with -1,
* and update the old node's entry.*/
std::sort(entryIndex_nearA.begin(), entryIndex_nearA.end(), less<int>());
//debug
// {
// stringstream _ss;
// {
// _ss << "nearA: ";
// for(int i = 0; i < entryIndex_nearA.size(); i ++)
// {
// _ss << entryIndex_nearA[i] << " ";
// }
// _ss << endl;
//
// _ss << "nearB: ";
// for(int i = 0; i < entryIndex_nearB.size(); i ++)
// {
// _ss << entryIndex_nearB[i] << " ";
// }
// _ss << endl;
// }
// Util::logging(_ss.str());
// }
for (unsigned i=0;i<entryIndex_nearA.size();i++)
{
oldNodePtr->setChildEntry(i, oldNodePtr->getChildEntry(entryIndex_nearA[i]));
oldNodePtr->setChildFileLine(i, oldNodePtr->getChildFileLine(entryIndex_nearA[i]));
}
oldNodePtr->setChildNum(entryIndex_nearA.size());
oldNodePtr->refreshSignature();
int oldNode_index = oldNodePtr->getIndexInFatherNode(*(this->node_buffer));
// full node's father pointer.
VNode* oldNodeFatherPtr = oldNodePtr->getFather(*(this->node_buffer));
if (oldNodePtr->isRoot())
{
/* if the old node is root,
* split the root, create a new root,
* and the tree height will be increased.*/
VNode* RootNewPtr = this->createNode();
/* change the old root node to not-root node,
* and set the RootNew to root node.*/
oldNodePtr->setAsRoot(false);
RootNewPtr->setAsRoot(true);
/* set the split two node(old node and new node) as the new root's child,
* and update signatures.*/
RootNewPtr->addChildNode(oldNodePtr);
RootNewPtr->addChildNode(newNodePtr);
RootNewPtr->refreshSignature();
//debug
// {
// stringstream _ss;
// _ss << "create new root:" << endl;
// _ss << "before swap file line, two sons are: " << oldNodePtr->getFileLine() << " " << newNodePtr->getFileLine() << endl;
// Util::logging(_ss.str());
// }
/* should keep the root node always being
* at the first line(line zero) of the tree node file.*/
this->swapNodeFileLine(RootNewPtr, oldNodePtr);
this->height ++;
//debug
// {
// stringstream _ss;
// _ss << "create new root:" << endl;
// _ss << "two sons are: " << oldNodePtr->getFileLine() << " " << newNodePtr->getFileLine() << endl;
// _ss << Signature::BitSet2str(oldNodePtr->getEntry().getEntitySig().entityBitSet) << endl;
// _ss << RootNewPtr->to_str() << endl;
// Util::logging(_ss.str());
// }
}
else
{
/* if the (OldNode) is not Root,
* change the old node's signature to A's signature.*/
oldNodeFatherPtr->setChildEntry(oldNode_index, oldNodePtr->getEntry());
if (oldNodeFatherPtr->isFull())
{
oldNodeFatherPtr->refreshAncestorSignature(*(this->node_buffer));
this->split(oldNodeFatherPtr, newNodePtr->getEntry(), newNodePtr);
}
else
{
oldNodeFatherPtr->addChildNode(newNodePtr);
oldNodeFatherPtr->refreshAncestorSignature(*(this->node_buffer));
}
}
//debug
// if (!oldNodePtr->checkState())
// {
// stringstream _ss;
// _ss << "node " << oldNodePtr->getFileLine() << " childFileLine error. oldNode when split" << endl;
// Util::logging(_ss.str());
// }
// if (!newNodePtr->checkState())
// {
// stringstream _ss;
// _ss << "node " << newNodePtr->getFileLine() << " childFileLine error. newNode when split" << endl;
// Util::logging(_ss.str());
// }
// update the entityID2FileLineMap by these two nodes.
this->updateEntityID2FileLineMap(oldNodePtr);
this->updateEntityID2FileLineMap(newNodePtr);
}