hclustercommand.cpp

   1 /*
   2  *  hclustercommand.cpp
   3  *  Mothur
   4  *
   5  *  Created by westcott on 10/13/09.
   6  *  Copyright 2009 Schloss Lab. All rights reserved.
   7  *
   8  */
   9
  10 #include "hclustercommand.h"
  11
  12 //**********************************************************************************************************************
  13 //This function checks to make sure the cluster command has no errors and then clusters based on the method chosen.
  14 HClusterCommand::HClusterCommand(string option){
  15         try{
  16                 globaldata = GlobalData::getInstance();
  17                 abort = false;
  18
  19                 //allow user to run help
  20                 if(option == "help") { help(); abort = true; }
  21
  22                 else {
  23                         //valid paramters for this command
  24                         string Array[] =  {"cutoff","precision","method","showabund","timing","phylip","column","name","sorted"};
  25                         vector<string> myArray (Array, Array+(sizeof(Array)/sizeof(string)));
  26
  27                         OptionParser parser(option);
  28                         map<string,string> parameters = parser.getParameters();
  29
  30                         ValidParameters validParameter;
  31
  32                         //check to make sure all parameters are valid for command
  33                         for (map<string,string>::iterator it = parameters.begin(); it != parameters.end(); it++) {
  34                                 if (validParameter.isValidParameter(it->first, myArray, it->second) != true) {
  35                                         abort = true;
  36                                 }
  37                         }
  38
  39                         globaldata->newRead();
  40
  41                         //check for required parameters
  42                         phylipfile = validParameter.validFile(parameters, "phylip", true);
  43                         if (phylipfile == "not open") { abort = true; }
  44                         else if (phylipfile == "not found") { phylipfile = ""; }
  45                         else {  distfile = phylipfile;  format = "phylip";      }
  46
  47                         columnfile = validParameter.validFile(parameters, "column", true);
  48                         if (columnfile == "not open") { abort = true; }
  49                         else if (columnfile == "not found") { columnfile = ""; }
  50                         else {  distfile = columnfile; format = "column";       }
  51
  52                         namefile = validParameter.validFile(parameters, "name", true);
  53                         if (namefile == "not open") { abort = true; }
  54                         else if (namefile == "not found") { namefile = ""; }
  55
  56                         if ((phylipfile == "") && (columnfile == "")) { mothurOut("When executing a cluster command you must enter a phylip or a column."); mothurOutEndLine(); abort = true; }
  57                         else if ((phylipfile != "") && (columnfile != "")) { mothurOut("When executing a cluster command you must enter ONLY ONE of the following: phylip or column."); mothurOutEndLine(); abort = true; }
  58
  59                         if (columnfile != "") {
  60                                 if (namefile == "") {  cout << "You need to provide a namefile if you are going to use the column format." << endl; abort = true; }
  61                         }
  62
  63                         //check for optional parameter and set defaults
  64                         // ...at some point should added some additional type checking...
  65                         //get user cutoff and precision or use defaults
  66                         string temp;
  67                         temp = validParameter.validFile(parameters, "precision", false);
  68                         if (temp == "not found") { temp = "100"; }
  69                         //saves precision legnth for formatting below
  70                         length = temp.length();
  71                         convert(temp, precision);
  72
  73                         temp = validParameter.validFile(parameters, "cutoff", false);
  74                         if (temp == "not found") { temp = "10"; }
  75                         convert(temp, cutoff);
  76                         cutoff += (5 / (precision * 10.0));
  77
  78                         method = validParameter.validFile(parameters, "method", false);
  79                         if (method == "not found") { method = "nearest"; }
  80
  81                         if ((method == "furthest") || (method == "nearest") || (method == "average")) { }
  82                         else { mothurOut("Not a valid clustering method.  Valid clustering algorithms are furthest, nearest or average."); mothurOutEndLine(); abort = true; }
  83
  84                         showabund = validParameter.validFile(parameters, "showabund", false);
  85                         if (showabund == "not found") { showabund = "T"; }
  86
  87                         sort = validParameter.validFile(parameters, "sorted", false);
  88                         if (sort == "not found") { sort = "F"; }
  89                         sorted = isTrue(sort);
  90
  91                         timing = validParameter.validFile(parameters, "timing", false);
  92                         if (timing == "not found") { timing = "F"; }
  93
  94
  95                         if (abort == false) {
  96
  97                                 fileroot = getRootName(distfile);
  98
  99                                 tag = "fn";  //until we figure out average and nearest methods
 100
 101                                 openOutputFile(fileroot+ tag + ".sabund",       sabundFile);
 102                                 openOutputFile(fileroot+ tag + ".rabund",       rabundFile);
 103                                 openOutputFile(fileroot+ tag + ".list",         listFile);
 104                         }
 105                 }
 106         }
 107         catch(exception& e) {
 108                 errorOut(e, "HClusterCommand", "HClusterCommand");
 109                 exit(1);
 110         }
 111 }
 112
 113 //**********************************************************************************************************************
 114
 115 void HClusterCommand::help(){
 116         try {
 117                 mothurOut("The cluster command can only be executed after a successful read.dist command.\n");
 118                 mothurOut("The cluster command parameter options are method, cuttoff, precision, showabund and timing. No parameters are required.\n");
 119                 mothurOut("The cluster command should be in the following format: \n");
 120                 mothurOut("cluster(method=yourMethod, cutoff=yourCutoff, precision=yourPrecision) \n");
 121                 mothurOut("The acceptable cluster methods are furthest, nearest and average.  If no method is provided then furthest is assumed.\n\n");
 122         }
 123         catch(exception& e) {
 124                 errorOut(e, "HClusterCommand", "help");
 125                 exit(1);
 126         }
 127 }
 128
 129 //**********************************************************************************************************************
 130
 131 HClusterCommand::~HClusterCommand(){}
 132
 133 //**********************************************************************************************************************
 134
 135 int HClusterCommand::execute(){
 136         try {
 137
 138                 if (abort == true) {    return 0;       }
 139
 140                 if(namefile != ""){
 141                         globaldata->nameMap = new NameAssignment(namefile);
 142                         globaldata->nameMap->readMap();
 143                 }else{
 144                         globaldata->nameMap = NULL;
 145                 }
 146
 147                 time_t estart = time(NULL);
 148
 149                 if (!sorted) {
 150                         read = new ReadCluster(distfile, cutoff);
 151                         read->setFormat(format);
 152                         read->read(globaldata->nameMap);
 153                         distfile = read->getOutputFile();
 154
 155                         list = read->getListVector();
 156                         delete read;
 157                 }else {
 158                         list = new ListVector(globaldata->nameMap->getListVector());
 159                 }
 160
 161                 mothurOut("It took " + toString(time(NULL) - estart) + " seconds to sort. "); mothurOutEndLine();
 162                 estart = time(NULL);
 163
 164                 //list vector made by read contains all sequence names
 165                 if(list != NULL){
 166                         rabund = new RAbundVector(list->getRAbundVector());
 167                 }else{
 168                         mothurOut("Error: no list vector!"); mothurOutEndLine(); return 0;
 169                 }
 170
 171                 float previousDist = 0.00000;
 172                 float rndPreviousDist = 0.00000;
 173                 oldRAbund = *rabund;
 174                 oldList = *list;
 175
 176                 print_start = true;
 177                 start = time(NULL);
 178
 179 //cout << "here" << endl;
 180                 ifstream in;
 181                 openInputFile(distfile, in);
 182                 string firstName, secondName;
 183                 float distance;
 184
 185                 cluster = new HCluster(rabund, list);
 186                 bool clusteredSomething;
 187                 vector<seqDist> seqs; seqs.resize(1); // to start loop
 188                 exitedBreak = false;  //lets you know if there is a distance stored in next
 189
 190                 while (seqs.size() != 0){
 191
 192                         seqs = getSeqs(in);
 193                         random_shuffle(seqs.begin(), seqs.end());
 194
 195                         if (seqs.size() == 0) { break; } //there are no more distances
 196
 197                         for (int i = 0; i < seqs.size(); i++) {  //-1 means skip me
 198
 199                                 if (print_start && isTrue(timing)) {
 200                                         mothurOut("Clustering (" + tag + ") dist " + toString(distance) + "/"
 201                                                           + toString(roundDist(distance, precision))
 202                                                           + "\t(precision: " + toString(precision) + ")");
 203                                         cout.flush();
 204                                         print_start = false;
 205                                 }
 206
 207         ///cout << "before cluster update" << endl;
 208                                 if (seqs[i].seq1 != seqs[i].seq2) {
 209                                         clusteredSomething = cluster->update(seqs[i].seq1, seqs[i].seq2, seqs[i].dist);
 210
 211                                         float rndDist = roundDist(seqs[i].dist, precision);
 212                                         //cout << "after cluster update clusterSomething = " << clusteredSomething << " rndDist = " << rndDist << " rndPreviousDist = " << rndPreviousDist << endl;
 213
 214
 215                                         if((previousDist <= 0.0000) && (seqs[i].dist != previousDist)){
 216                                                 printData("unique");
 217                                         }
 218                                         else if((rndDist != rndPreviousDist)){
 219                                                 printData(toString(rndPreviousDist,  length-1));
 220                                         }
 221
 222                                         previousDist = seqs[i].dist;
 223                                         rndPreviousDist = rndDist;
 224                                         oldRAbund = *rabund;
 225                                         oldList = *list;
 226                                 }
 227                         }
 228                 }
 229
 230                 in.close();
 231
 232                 if (print_start && isTrue(timing)) {
 233                         //mothurOut("Clustering (" + tag + ") for distance " + toString(previousDist) + "/" + toString(rndPreviousDist)
 234                                          //+ "\t(precision: " + toString(precision) + ", Nodes: " + toString(matrix->getNNodes()) + ")");
 235                         cout.flush();
 236                         print_start = false;
 237                 }
 238
 239                 if(previousDist <= 0.0000){
 240                         printData("unique");
 241                 }
 242                 else if(rndPreviousDist<cutoff){
 243                         printData(toString(rndPreviousDist, length-1));
 244                 }
 245
 246                 //delete globaldata's copy of the sparsematrix and listvector to free up memory
 247                 delete globaldata->gListVector;  globaldata->gListVector = NULL;
 248
 249                 //saves .list file so you can do the collect, rarefaction and summary commands without doing a read.list
 250                 if (globaldata->getFormat() == "phylip") { globaldata->setPhylipFile(""); }
 251                 else if (globaldata->getFormat() == "column") { globaldata->setColumnFile(""); }
 252
 253                 globaldata->setListFile(fileroot+ tag + ".list");
 254                 globaldata->setNameFile("");
 255                 globaldata->setFormat("list");
 256
 257                 sabundFile.close();
 258                 rabundFile.close();
 259                 listFile.close();
 260
 261                 delete cluster;
 262                 //if (isTrue(timing)) {
 263                         mothurOut("It took " + toString(time(NULL) - estart) + " seconds to cluster. "); mothurOutEndLine();
 264                 //}
 265                 return 0;
 266         }
 267         catch(exception& e) {
 268                 errorOut(e, "HClusterCommand", "execute");
 269                 exit(1);
 270         }
 271 }
 272
 273 //**********************************************************************************************************************
 274
 275 void HClusterCommand::printData(string label){
 276         try {
 277                 if (isTrue(timing)) {
 278                         mothurOut("\tTime: " + toString(time(NULL) - start) + "\tsecs for " + toString(oldRAbund.getNumBins())
 279                      + "\tclusters. Updates: " + toString(loops)); mothurOutEndLine();
 280                 }
 281                 print_start = true;
 282                 loops = 0;
 283                 start = time(NULL);
 284
 285                 oldRAbund.setLabel(label);
 286                 if (isTrue(showabund)) {
 287                         oldRAbund.getSAbundVector().print(cout);
 288                 }
 289                 oldRAbund.print(rabundFile);
 290                 oldRAbund.getSAbundVector().print(sabundFile);
 291
 292                 oldList.setLabel(label);
 293                 oldList.print(listFile);
 294         }
 295         catch(exception& e) {
 296                 errorOut(e, "HClusterCommand", "printData");
 297                 exit(1);
 298         }
 299
 300
 301 }
 302 //**********************************************************************************************************************
 303
 304 vector<seqDist> HClusterCommand::getSeqs(ifstream& filehandle){
 305         try {
 306                 string firstName, secondName;
 307                 float distance, prevDistance;
 308                 vector<seqDist> sameSeqs;
 309                 prevDistance = -1;
 310
 311                 //if you are not at the beginning of the file
 312                 if (exitedBreak) {
 313                         sameSeqs.push_back(next);
 314                         prevDistance = next.dist;
 315                         exitedBreak = false;
 316                 }
 317
 318                 //get entry
 319                 while (filehandle) {
 320
 321                         filehandle >> firstName >> secondName >> distance;
 322 //cout << firstName << '\t' << secondName << '\t' << distance << endl;
 323                         gobble(filehandle);
 324
 325                         //save first one
 326                         if (prevDistance == -1) { prevDistance = distance; }
 327
 328                         map<string,int>::iterator itA = globaldata->nameMap->find(firstName);
 329                         map<string,int>::iterator itB = globaldata->nameMap->find(secondName);
 330
 331                         if(itA == globaldata->nameMap->end()){
 332                                 cerr << "AAError: Sequence '" << firstName << "' was not found in the names file, please correct\n";
 333                         }
 334                         if(itB == globaldata->nameMap->end()){
 335                                 cerr << "ABError: Sequence '" << secondName << "' was not found in the names file, please correct\n";
 336                         }
 337
 338                         //using cutoff
 339                         if (distance > cutoff) { break; }
 340
 341                         if (distance != -1) { //-1 means skip me
 342
 343                                 //are the distances the same
 344                                 if (distance == prevDistance) { //save in vector
 345                                         seqDist temp;
 346                                         temp.seq1 = itA->second;
 347                                         temp.seq2 = itB->second;
 348                                         temp.dist = distance;
 349                                         sameSeqs.push_back(temp);
 350                                         exitedBreak = false;
 351                                         //what about precision??
 352
 353                                 }else{
 354                                         next.seq1 = itA->second;
 355                                         next.seq2 = itB->second;
 356                                         next.dist = distance;
 357                                         exitedBreak = true;
 358                                         break;
 359                                 }
 360
 361                         }
 362                 }
 363
 364                 return sameSeqs;
 365         }
 366         catch(exception& e) {
 367                 errorOut(e, "HClusterCommand", "getSeqs");
 368                 exit(1);
 369         }
 370
 371
 372 }
 373
 374 //**********************************************************************************************************************
 375