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 f) :
18 rabund(rav), list(lv), dMatrix(dm), method(f)
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.
55 seqVec = vector<MatVec>(lv->size());
56 for (MatData currentCell = dMatrix->begin(); currentCell != dMatrix->end(); currentCell++) {
57 seqVec[currentCell->row].push_back(currentCell);
58 seqVec[currentCell->column].push_back(currentCell);
61 mapWanted = false; //set to true by mgcluster to speed up overlap merge
63 //save so you can modify as it changes in average neighbor
65 m = MothurOut::getInstance();
68 /***********************************************************************/
70 void Cluster::getRowColCells() {
72 PCell* smallCell = dMatrix->getSmallestCell(); //find the smallest cell - this routine should probably not be in the SpMat class
74 smallRow = smallCell->row; // get its row
75 smallCol = smallCell->column; // get its column
76 smallDist = smallCell->dist; // get the smallest distance
78 rowCells = seqVec[smallRow]; // all distances related to the row index
79 colCells = seqVec[smallCol]; // all distances related to the column index
80 nRowCells = rowCells.size();
81 nColCells = colCells.size();
84 m->errorOut(e, "Cluster", "getRowColCells");
89 /***********************************************************************/
90 // Remove the specified cell from the seqVec and from the sparse
92 void Cluster::removeCell(const MatData& cell, int vrow, int vcol, bool rmMatrix){
96 ull dcol = cell->column;
97 if (((vrow >=0) && (drow != smallRow)) ||
98 ((vcol >=0) && (dcol != smallCol))) {
108 nCells = seqVec[drow].size();
109 for (vrow=0; vrow<nCells;vrow++) {
110 crow = seqVec[drow][vrow]->row;
111 ccol = seqVec[drow][vrow]->column;
112 if (((crow == drow) && (ccol == dcol)) ||
113 ((ccol == drow) && (crow == dcol))) {
119 seqVec[drow].erase(seqVec[drow].begin()+vrow);
121 nCells = seqVec[dcol].size();
122 for (vcol=0; vcol<nCells;vcol++) {
123 crow = seqVec[dcol][vcol]->row;
124 ccol = seqVec[dcol][vcol]->column;
125 if (((crow == drow) && (ccol == dcol)) ||
126 ((ccol == drow) && (crow == dcol))) {
132 seqVec[dcol].erase(seqVec[dcol].begin()+vcol);
135 dMatrix->rmCell(cell);
139 catch(exception& e) {
140 m->errorOut(e, "Cluster", "removeCell");
144 /***********************************************************************/
146 void Cluster::clusterBins(){
148 // cout << smallCol << '\t' << smallRow << '\t' << smallDist << '\t' << rabund->get(smallRow) << '\t' << rabund->get(smallCol);
150 rabund->set(smallCol, rabund->get(smallRow)+rabund->get(smallCol));
151 rabund->set(smallRow, 0);
152 rabund->setLabel(toString(smallDist));
154 // cout << '\t' << rabund->get(smallRow) << '\t' << rabund->get(smallCol) << endl;
156 catch(exception& e) {
157 m->errorOut(e, "Cluster", "clusterBins");
164 /***********************************************************************/
166 void Cluster::clusterNames(){
168 // cout << smallCol << '\t' << smallRow << '\t' << smallDist << '\t' << list->get(smallRow) << '\t' << list->get(smallCol);
169 if (mapWanted) { updateMap(); }
171 list->set(smallCol, list->get(smallRow)+','+list->get(smallCol));
172 list->set(smallRow, "");
173 list->setLabel(toString(smallDist));
175 // cout << '\t' << list->get(smallRow) << '\t' << list->get(smallCol) << endl;
177 catch(exception& e) {
178 m->errorOut(e, "Cluster", "clusterNames");
184 /***********************************************************************/
185 //This function clusters based on the method of the derived class
186 //At the moment only average and complete linkage are covered, because
187 //single linkage uses a different approach.
188 void Cluster::update(double& cutOFF){
192 vector<int> foundCol(nColCells, 0);
197 // The vector has to be traversed in reverse order to preserve the index
198 // for faster removal in removeCell()
199 for (int i=nRowCells-1;i>=0;i--) {
200 if (!((rowCells[i]->row == smallRow) && (rowCells[i]->column == smallCol))) {
201 if (rowCells[i]->row == smallRow) {
202 search = rowCells[i]->column;
204 search = rowCells[i]->row;
208 for (int j=0;j<nColCells;j++) {
209 if (!((colCells[j]->row == smallRow) && (colCells[j]->column == smallCol))) { //if you are not hte smallest distance
210 if (colCells[j]->row == search || colCells[j]->column == search) {
213 changed = updateDistance(colCells[j], rowCells[i]);
214 // If the cell's distance changed and it had the same distance as
215 // the smallest distance, invalidate the mins vector in SparseMatrix
217 if (colCells[j]->vectorMap != NULL) {
218 *(colCells[j]->vectorMap) = NULL;
219 colCells[j]->vectorMap = NULL;
226 //if not merged it you need it for warning
227 if ((!merged) && (method == "average")) {
228 //m->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."); m->mothurOutEndLine();
229 if (cutOFF > rowCells[i]->dist) {
230 cutOFF = rowCells[i]->dist;
231 //m->mothurOut("changing cutoff to " + toString(cutOFF)); m->mothurOutEndLine();
235 removeCell(rowCells[i], i , -1);
242 // Special handling for singlelinkage case, not sure whether this
244 for (int i=nColCells-1;i>=0;i--) {
245 if (foundCol[i] == 0) {
246 if (method == "average") {
247 if (!((colCells[i]->row == smallRow) && (colCells[i]->column == smallCol))) {
248 //m->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."); m->mothurOutEndLine();
249 if (cutOFF > colCells[i]->dist) {
250 cutOFF = colCells[i]->dist;
251 //m->mothurOut("changing cutoff to " + toString(cutOFF)); m->mothurOutEndLine();
255 removeCell(colCells[i], -1, i);
259 catch(exception& e) {
260 m->errorOut(e, "Cluster", "update");
264 /***********************************************************************/
265 void Cluster::setMapWanted(bool f) {
270 for (int i = 0; i < list->getNumBins(); i++) {
273 string names = list->get(i);
274 while (names.find_first_of(',') != -1) {
276 string name = names.substr(0,names.find_first_of(','));
277 //save name and bin number
279 names = names.substr(names.find_first_of(',')+1, names.length());
287 catch(exception& e) {
288 m->errorOut(e, "Cluster", "setMapWanted");
292 /***********************************************************************/
293 void Cluster::updateMap() {
295 //update location of seqs in smallRow since they move to smallCol now
296 string names = list->get(smallRow);
297 while (names.find_first_of(',') != -1) {
299 string name = names.substr(0,names.find_first_of(','));
300 //save name and bin number
301 seq2Bin[name] = smallCol;
302 names = names.substr(names.find_first_of(',')+1, names.length());
306 seq2Bin[names] = smallCol;
309 catch(exception& e) {
310 m->errorOut(e, "Cluster", "updateMap");
314 /***********************************************************************/