suffixtree.cpp

   1 /*
   2  *  suffixtree.cpp
   3  *
   4  *
   5  *  Created by Pat Schloss on 12/15/08.
   6  *  Copyright 2008 Patrick D. Schloss. All rights reserved.
   7  *
   8  *      This is my half-assed attempt to implement a suffix tree.  This is a cobbled together algorithm using materials that
   9  *      I found at http://marknelson.us/1996/08/01/suffix-trees/ and:
  10  *
  11  *              Ukkonen E. (1995). On-line construction of suffix trees. Algorithmica 14 (3): 249--260
  12  *              Gusfield, Dan (1999). Algorithms on Strings, Trees and Sequences: Computer Science and Computational Biology.
  13  *                      USA: Cambridge University Press
  14  *
  15  *      The Ukkonen paper is the seminal paper describing the on-line method of constructing a suffix tree.
  16  *
  17  *      I have chosen to store the nodes of the tree as a vector of pointers to SuffixNode objects.  The root is stored at
  18  *      nodeVector[0].  Each tree also stores the sequence name and the string that corresponds to the actual sequence.
  19  *      Finally, this class provides a way of counting the number of suffixes that are needed in one tree to generate a new
  20  *      sequence (countSuffixes).  This method is used to determine similarity between sequences and was inspired by the
  21  *      article and Perl source code provided at http://www.ddj.com/web-development/184416093.
  22  *
  23  */
  24
  25 #include "sequence.hpp"
  26 #include "suffixnodes.hpp"
  27 #include "suffixtree.hpp"
  28
  29
  30 //********************************************************************************************************************
  31
  32 inline bool compareParents(SuffixNode* left, SuffixNode* right){//      this is necessary to print the tree and to sort the
  33         return (left->getParentNode() < right->getParentNode());        //      nodes in order of their parent
  34 }
  35
  36 //********************************************************************************************************************
  37
  38 SuffixTree::SuffixTree(){ m = MothurOut::getInstance(); }
  39
  40 //********************************************************************************************************************
  41
  42 SuffixTree::~SuffixTree(){
  43         for(int i=0;i<nodeVector.size();i++){   delete nodeVector[i];   }
  44         nodeVector.clear();
  45 }
  46
  47 //********************************************************************************************************************
  48
  49 void SuffixTree::loadSequence(Sequence seq){
  50         nodeCounter = 0;                                                        //      initially there are 0 nodes in the tree
  51         activeStartPosition = 0;
  52         activeEndPosition = -1;
  53         seqName = seq.getName();
  54         sequence = seq.convert2ints();
  55         sequence += '5';                                                        //      this essentially concatenates a '$' to the end of the sequence to
  56         int seqLength = sequence.length();                      //      make it a cononical suffix tree
  57
  58         nodeVector.push_back(new SuffixBranch(-1, 0, -1));      //      enter the root of the suffix tree
  59
  60         activeNode = root = 0;
  61         string hold;
  62         for(int i=0;i<seqLength;i++){
  63                 addPrefix(i);                                                   //      step through the sequence adding each prefix
  64         }
  65 }
  66
  67 //********************************************************************************************************************
  68
  69 string SuffixTree::getSeqName() {
  70         return seqName;
  71 }
  72
  73 //********************************************************************************************************************
  74
  75 void SuffixTree::print(){
  76         vector<SuffixNode*> hold = nodeVector;
  77         sort(hold.begin(), hold.end(), compareParents);
  78         m->mothurOut("Address\t\tParent\tNode\tSuffix\tStartC\tEndC\tSuffix"); m->mothurOutEndLine();
  79         for(int i=1;i<=nodeCounter;i++){
  80                 hold[i]->print(sequence, i);
  81         }
  82 }
  83
  84 //********************************************************************************************************************
  85
  86 int SuffixTree::countSuffixes(string compareSequence, int& minValue){   //      here we count the number of suffix parts
  87                                                                                                                         //      we need to rewrite a user supplied sequence.  if the
  88         int numSuffixes = 0;                                                                    //      count exceeds the supplied minValue, bail out.  The
  89         int seqLength = compareSequence.length();                               //      time complexity should be O(L)
  90         int position = 0;
  91
  92         int presentNode = 0;
  93
  94         while(position < seqLength){            //      while the position in the query sequence isn't at the end...
  95
  96                 if(numSuffixes > minValue)      {       return 1000000;         }       //      bail if the count gets too high
  97
  98                 int newNode = nodeVector[presentNode]->getChild(compareSequence[position]);     //      see if the current node has a
  99                                                                                                                                 //      child that matches the next character in the query
 100                 if(newNode == -1){
 101                         if(presentNode == 0){   position++;             }                       //      if not, go back to the root and increase the count
 102                         numSuffixes++;                                                                          //      by one.
 103                         presentNode = 0;
 104                 }
 105                 else{                                                                                                   //      if there is, move to that node and see how far down
 106                         presentNode = newNode;                                                          //      it we can get
 107
 108                         for(int i=nodeVector[newNode]->getStartCharPos(); i<=nodeVector[newNode]->getEndCharPos(); i++){
 109                                 if(compareSequence[position] == sequence[i]){
 110                                         position++;                                                                     //      as long as the query and branch agree, keep going
 111                                 }
 112                                 else{
 113                                         numSuffixes++;                                                          //      if there is a mismatch, increase the number of
 114                                         presentNode = 0;                                                        //      suffixes and go back to the root
 115                                         break;
 116                                 }
 117                         }
 118                 }
 119                 //      if we get all the way through the node we'll go to the top of the while loop and find the child node
 120                 //      that corresponds to what we are interested in
 121         }
 122         numSuffixes--;                                                                                          //      the method puts an extra count on numSuffixes
 123
 124         if(numSuffixes < minValue)      {       minValue = numSuffixes; }       //      if the count is less than the previous minValue,
 125         return numSuffixes;                                                                                     //      change the value and return the number of suffixes
 126
 127 }
 128 //********************************************************************************************************************
 129
 130 int SuffixTree::countSuffixes(string compareSequence){  //      here we count the number of suffix parts
 131                                                                                                                         //      we need to rewrite a user supplied sequence.  if the
 132         int numSuffixes = 0;                                                                    //      count exceeds the supplied minValue, bail out.  The
 133         int seqLength = compareSequence.length();                               //      time complexity should be O(L)
 134         int position = 0;
 135
 136         int presentNode = 0;
 137
 138         while(position < seqLength){            //      while the position in the query sequence isn't at the end...
 139
 140                 int newNode = nodeVector[presentNode]->getChild(compareSequence[position]);     //      see if the current node has a
 141                                                                                                                                 //      child that matches the next character in the query
 142                 if(newNode == -1){
 143                         if(presentNode == 0){   position++;             }                       //      if not, go back to the root and increase the count
 144                         numSuffixes++;                                                                          //      by one.
 145                         presentNode = 0;
 146                 }
 147                 else{                                                                                                   //      if there is, move to that node and see how far down
 148                         presentNode = newNode;                                                          //      it we can get
 149
 150                         for(int i=nodeVector[newNode]->getStartCharPos(); i<=nodeVector[newNode]->getEndCharPos(); i++){
 151                                 if(compareSequence[position] == sequence[i]){
 152                                         position++;                                                                     //      as long as the query and branch agree, keep going
 153                                 }
 154                                 else{
 155                                         numSuffixes++;                                                          //      if there is a mismatch, increase the number of
 156                                         presentNode = 0;                                                        //      suffixes and go back to the root
 157                                         break;
 158                                 }
 159                         }
 160                 }
 161                 //      if we get all the way through the node we'll go to the top of the while loop and find the child node
 162                 //      that corresponds to what we are interested in
 163         }
 164         numSuffixes--;                                                                                          //      the method puts an extra count on numSuffixes
 165
 166         return numSuffixes;                                                                                     //      change the value and return the number of suffixes
 167 }
 168 //********************************************************************************************************************
 169
 170 void SuffixTree::canonize(){    //      if you have to ask how this works, you don't really want to know and this really
 171                                                                 //      isn't the place to ask.
 172         if ( isExplicit() == 0 ) {      //      if the node has no children...
 173
 174                 int tempNodeIndex = nodeVector[activeNode]->getChild(sequence[activeStartPosition]);
 175                 SuffixNode* tempNode = nodeVector[tempNodeIndex];
 176
 177                 int span = tempNode->getEndCharPos() - tempNode->getStartCharPos();
 178
 179                 while ( span <= ( activeEndPosition - activeStartPosition ) ) {
 180
 181             activeStartPosition = activeStartPosition + span + 1;
 182
 183                         activeNode = tempNodeIndex;
 184
 185             if ( activeStartPosition <= activeEndPosition ) {
 186                                 tempNodeIndex = nodeVector[tempNodeIndex]->getChild(sequence[activeStartPosition]);
 187                                 tempNode = nodeVector[tempNodeIndex];
 188                                 span = tempNode->getEndCharPos() - tempNode->getStartCharPos();
 189             }
 190
 191         }
 192     }
 193 }
 194
 195 //********************************************************************************************************************
 196
 197 int SuffixTree::split(int nodeIndex, int position){     //      leaves stay leaves, etc, to split a leaf we make a new interior
 198                                                                                                         //      node and reconnect everything
 199         SuffixNode* node = nodeVector[nodeIndex];                                       //      get the node that needs to be split
 200         SuffixNode* parentNode = nodeVector[node->getParentNode()];     //      get it's parent node
 201
 202         parentNode->eraseChild(sequence[node->getStartCharPos()]);      //      erase the present node from the registry of its parent
 203
 204         nodeCounter++;
 205         SuffixNode* newNode = new SuffixBranch(node->getParentNode(), node->getStartCharPos(), node->getStartCharPos() + activeEndPosition - activeStartPosition);      //      create a new node that will link the parent with the old child
 206         parentNode->setChildren(sequence[newNode->getStartCharPos()], nodeCounter);//   give the parent the new child
 207         nodeVector.push_back(newNode);
 208
 209         node->setParentNode(nodeCounter);       //      give the original node the new node as its parent
 210         newNode->setChildren(sequence[node->getStartCharPos() + activeEndPosition - activeStartPosition + 1], nodeIndex);
 211         //      put the original node in the registry of the new node's children
 212         newNode->setSuffixNode(activeNode);//link the new node with the old active node
 213
 214         //      recalculate the startCharPosition of the outermost node
 215         node->setStartCharPos(node->getStartCharPos() + activeEndPosition - activeStartPosition + 1 );
 216
 217         return node->getParentNode();
 218 }
 219
 220 //********************************************************************************************************************
 221
 222 void SuffixTree::makeSuffixLink(int& previous, int present){
 223
 224 //      here we link the nodes that are suffixes of one another to rapidly speed through the tree
 225         if ( previous > 0 ) {   nodeVector[previous]->setSuffixNode(present);   }
 226         else                            {       /*      do nothing                                                              */      }
 227
 228     previous = present;
 229 }
 230
 231 //********************************************************************************************************************
 232
 233 void SuffixTree::addPrefix(int prefixPosition){
 234
 235         int lastParentNode = -1;        //      we need to place a new prefix in the suffix tree
 236         int parentNode = 0;
 237
 238         while(1){
 239
 240                 parentNode = activeNode;
 241
 242                 if(isExplicit() == 1){  //      if the node is explicit (has kids), try to follow it down the branch if its there...
 243                         if(nodeVector[activeNode]->getChild(sequence[prefixPosition]) != -1){   //      break out and get next prefix...
 244                                 break;
 245                         }
 246                         else{                           //      ...otherwise continue, we'll need to make a new node later on...
 247                         }
 248                 }
 249                 else{                                   //      if it's not explicit (no kids), read through and see if all of the chars agree...
 250                         int tempNode = nodeVector[activeNode]->getChild(sequence[activeStartPosition]);
 251                         int span = activeEndPosition - activeStartPosition;
 252
 253                         if(sequence[nodeVector[tempNode]->getStartCharPos() + span + 1] == sequence[prefixPosition] ){
 254                                 break;                  //      if the existing suffix agrees with the new one, grab a new prefix...
 255                         }
 256                         else{
 257                                 parentNode = split(tempNode, prefixPosition);   //      ... otherwise we need to split the node
 258                         }
 259
 260                 }
 261
 262                 nodeCounter++;  //      we need to generate a new node here if the kid didn't exist, or we split a node
 263                 SuffixNode* newSuffixLeaf = new SuffixLeaf(parentNode, prefixPosition, sequence.length()-1);
 264                 nodeVector[parentNode]->setChildren(sequence[prefixPosition], nodeCounter);
 265                 nodeVector.push_back(newSuffixLeaf);
 266
 267                 makeSuffixLink( lastParentNode, parentNode );           //      make a suffix link for the parent node
 268
 269                 if(nodeVector[activeNode]->getParentNode() == -1){      //      move along the start position for the tree
 270             activeStartPosition++;
 271         }
 272                 else {
 273             activeNode = nodeVector[activeNode]->getSuffixNode();
 274                 }
 275                 canonize();                                                                                     //      frankly, i'm not entirely clear on what canonize does.
 276         }
 277
 278         makeSuffixLink( lastParentNode, parentNode );
 279         activeEndPosition++;                                                                    //      move along the end position for the tree
 280
 281         canonize();                                                                                             //      frankly, i'm not entirely clear on what canonize does.
 282
 283 }
 284
 285 //********************************************************************************************************************
 286