5 * Created by Pat Schloss on 8/14/08.
6 * Copyright 2008 Patrick D. Schloss. All rights reserved.
10 #include "cluster.hpp"
11 #include "rabundvector.hpp"
12 #include "listvector.hpp"
13 #include "sparsematrix.hpp"
15 /***********************************************************************/
17 Cluster::Cluster(RAbundVector* rav, ListVector* lv, SparseMatrix* dm, float c, string m) :
18 rabund(rav), list(lv), dMatrix(dm), method(m)
21 cout << "sizeof(MatData): " << sizeof(MatData) << endl;
22 cout << "sizeof(PCell*): " << sizeof(PCell*) << endl;
24 int nCells = dMatrix->getNNodes();
25 time_t start = time(NULL);
27 MatVec matvec = MatVec(nCells);
29 for (MatData currentCell = dMatrix->begin(); currentCell != dMatrix->end(); currentCell++) {
30 matvec[i++] = currentCell;
32 for (i= matvec.size();i>0;i--) {
33 dMatrix->rmCell(matvec[i-1]);
35 MatData it = dMatrix->begin();
36 while (it != dMatrix->end()) {
37 it = dMatrix->rmCell(it);
39 cout << "Time to remove " << nCells << " cells: " << time(NULL) - start << " seconds" << endl;
41 MatData it = dMatrix->begin();
42 cout << it->row << "/" << it->column << "/" << it->dist << endl;
43 dMatrix->rmCell(dMatrix->begin());
44 cout << it->row << "/" << it->column << "/" << it->dist << endl;
48 // Create a data structure to quickly access the PCell information
49 // for a certain sequence. It consists of a vector of lists, where
50 // a list contains pointers (iterators) to the all distances related
51 // to a certain sequence. The Vector is accessed via the index of a
52 // sequence in the distance matrix.
53 seqVec = vector<MatVec>(lv->size());
54 for (MatData currentCell = dMatrix->begin(); currentCell != dMatrix->end(); currentCell++) {
55 seqVec[currentCell->row].push_back(currentCell);
56 seqVec[currentCell->column].push_back(currentCell);
58 mapWanted = false; //set to true by mgcluster to speed up overlap merge
60 //save so you can modify as it changes in average neighbor
64 /***********************************************************************/
66 void Cluster::getRowColCells() {
68 PCell* smallCell = dMatrix->getSmallestCell(); //find the smallest cell - this routine should probably not be in the SpMat class
70 smallRow = smallCell->row; // get its row
71 smallCol = smallCell->column; // get its column
72 smallDist = smallCell->dist; // get the smallest distance
74 rowCells = seqVec[smallRow]; // all distances related to the row index
75 colCells = seqVec[smallCol]; // all distances related to the column index
76 nRowCells = rowCells.size();
77 nColCells = colCells.size();
80 errorOut(e, "Cluster", "getRowColCells");
85 /***********************************************************************/
86 // Remove the specified cell from the seqVec and from the sparse
88 void Cluster::removeCell(const MatData& cell, int vrow, int vcol, bool rmMatrix)
91 ull dcol = cell->column;
92 if (((vrow >=0) && (drow != smallRow)) ||
93 ((vcol >=0) && (dcol != smallCol))) {
103 nCells = seqVec[drow].size();
104 for (vrow=0; vrow<nCells;vrow++) {
105 crow = seqVec[drow][vrow]->row;
106 ccol = seqVec[drow][vrow]->column;
107 if (((crow == drow) && (ccol == dcol)) ||
108 ((ccol == drow) && (crow == dcol))) {
113 seqVec[drow].erase(seqVec[drow].begin()+vrow);
115 nCells = seqVec[dcol].size();
116 for (vcol=0; vcol<nCells;vcol++) {
117 crow = seqVec[dcol][vcol]->row;
118 ccol = seqVec[dcol][vcol]->column;
119 if (((crow == drow) && (ccol == dcol)) ||
120 ((ccol == drow) && (crow == dcol))) {
125 seqVec[dcol].erase(seqVec[dcol].begin()+vcol);
127 dMatrix->rmCell(cell);
132 /***********************************************************************/
134 void Cluster::clusterBins(){
136 // cout << smallCol << '\t' << smallRow << '\t' << smallDist << '\t' << rabund->get(smallRow) << '\t' << rabund->get(smallCol);
138 rabund->set(smallCol, rabund->get(smallRow)+rabund->get(smallCol));
139 rabund->set(smallRow, 0);
140 rabund->setLabel(toString(smallDist));
142 // cout << '\t' << rabund->get(smallRow) << '\t' << rabund->get(smallCol) << endl;
144 catch(exception& e) {
145 errorOut(e, "Cluster", "clusterBins");
152 /***********************************************************************/
154 void Cluster::clusterNames(){
156 // cout << smallCol << '\t' << smallRow << '\t' << smallDist << '\t' << list->get(smallRow) << '\t' << list->get(smallCol);
157 if (mapWanted) { updateMap(); }
159 list->set(smallCol, list->get(smallRow)+','+list->get(smallCol));
160 list->set(smallRow, "");
161 list->setLabel(toString(smallDist));
163 // cout << '\t' << list->get(smallRow) << '\t' << list->get(smallCol) << endl;
165 catch(exception& e) {
166 errorOut(e, "Cluster", "clusterNames");
172 /***********************************************************************/
173 //This function clusters based on the method of the derived class
174 //At the moment only average and complete linkage are covered, because
175 //single linkage uses a different approach.
176 void Cluster::update(double& cutOFF){
180 vector<int> foundCol(nColCells, 0);
185 // The vector has to be traversed in reverse order to preserve the index
186 // for faster removal in removeCell()
187 for (int i=nRowCells-1;i>=0;i--) {
188 if (!((rowCells[i]->row == smallRow) && (rowCells[i]->column == smallCol))) {
189 if (rowCells[i]->row == smallRow) {
190 search = rowCells[i]->column;
192 search = rowCells[i]->row;
196 for (int j=0;j<nColCells;j++) {
197 if (!((colCells[j]->row == smallRow) && (colCells[j]->column == smallCol))) { //if you are not hte smallest distance
198 if (colCells[j]->row == search || colCells[j]->column == search) {
201 changed = updateDistance(colCells[j], rowCells[i]);
202 // If the cell's distance changed and it had the same distance as
203 // the smallest distance, invalidate the mins vector in SparseMatrix
205 if (colCells[j]->vectorMap != NULL) {
206 *(colCells[j]->vectorMap) = NULL;
207 colCells[j]->vectorMap = NULL;
214 //if not merged it you need it for warning
215 if ((!merged) && (method == "average")) {
216 //mothurOut("Warning: trying to merge cell " + toString(rowCells[i]->row+1) + " " + toString(rowCells[i]->column+1) + " distance " + toString(rowCells[i]->dist) + " with value above cutoff. Results may vary from using cutoff at cluster command instead of read.dist."); mothurOutEndLine();
217 if (cutOFF > rowCells[i]->dist) {
218 cutOFF = rowCells[i]->dist;
219 //mothurOut("changing cutoff to " + toString(cutOFF)); mothurOutEndLine();
223 removeCell(rowCells[i], i , -1);
230 // Special handling for singlelinkage case, not sure whether this
232 for (int i=nColCells-1;i>=0;i--) {
233 if (foundCol[i] == 0) {
234 if (method == "average") {
235 if (!((colCells[i]->row == smallRow) && (colCells[i]->column == smallCol))) {
236 //mothurOut("Warning: merging cell " + toString(colCells[i]->row+1) + " " + toString(colCells[i]->column+1) + " distance " + toString(colCells[i]->dist) + " value above cutoff. Results may vary from using cutoff at cluster command instead of read.dist."); mothurOutEndLine();
237 if (cutOFF > colCells[i]->dist) {
238 cutOFF = colCells[i]->dist;
239 //mothurOut("changing cutoff to " + toString(cutOFF)); mothurOutEndLine();
243 removeCell(colCells[i], -1, i);
247 catch(exception& e) {
248 errorOut(e, "Cluster", "update");
252 /***********************************************************************/
253 void Cluster::setMapWanted(bool m) {
258 for (int i = 0; i < list->getNumBins(); i++) {
261 string names = list->get(i);
262 while (names.find_first_of(',') != -1) {
264 string name = names.substr(0,names.find_first_of(','));
265 //save name and bin number
267 names = names.substr(names.find_first_of(',')+1, names.length());
275 catch(exception& e) {
276 errorOut(e, "Cluster", "setMapWanted");
280 /***********************************************************************/
281 void Cluster::updateMap() {
283 //update location of seqs in smallRow since they move to smallCol now
284 string names = list->get(smallRow);
285 while (names.find_first_of(',') != -1) {
287 string name = names.substr(0,names.find_first_of(','));
288 //save name and bin number
289 seq2Bin[name] = smallCol;
290 names = names.substr(names.find_first_of(',')+1, names.length());
294 seq2Bin[names] = smallCol;
297 catch(exception& e) {
298 errorOut(e, "Cluster", "updateMap");
302 /***********************************************************************/