preclustercommand.cpp

   1 /*
   2  *  preclustercommand.cpp
   3  *  Mothur
   4  *
   5  *  Created by westcott on 12/21/09.
   6  *  Copyright 2009 Schloss Lab. All rights reserved.
   7  *
   8  */
   9
  10 #include "preclustercommand.h"
  11 #include "deconvolutecommand.h"
  12
  13 //**********************************************************************************************************************
  14 vector<string> PreClusterCommand::setParameters(){
  15         try {
  16                 CommandParameter pfasta("fasta", "InputTypes", "", "", "none", "none", "none",false,true); parameters.push_back(pfasta);
  17                 CommandParameter pname("name", "InputTypes", "", "", "none", "none", "none",false,false); parameters.push_back(pname);
  18                 CommandParameter pgroup("group", "InputTypes", "", "", "none", "none", "none",false,false); parameters.push_back(pgroup);
  19                 CommandParameter pdiffs("diffs", "Number", "", "0", "", "", "",false,false); parameters.push_back(pdiffs);
  20                 CommandParameter pprocessors("processors", "Number", "", "1", "", "", "",false,false); parameters.push_back(pprocessors);
  21                 CommandParameter pinputdir("inputdir", "String", "", "", "", "", "",false,false); parameters.push_back(pinputdir);
  22                 CommandParameter poutputdir("outputdir", "String", "", "", "", "", "",false,false); parameters.push_back(poutputdir);
  23
  24                 vector<string> myArray;
  25                 for (int i = 0; i < parameters.size(); i++) {   myArray.push_back(parameters[i].name);          }
  26                 return myArray;
  27         }
  28         catch(exception& e) {
  29                 m->errorOut(e, "PreClusterCommand", "setParameters");
  30                 exit(1);
  31         }
  32 }
  33 //**********************************************************************************************************************
  34 string PreClusterCommand::getHelpString(){
  35         try {
  36                 string helpString = "";
  37                 helpString += "The pre.cluster command groups sequences that are within a given number of base mismatches.\n";
  38                 helpString += "The pre.cluster command outputs a new fasta and name file.\n";
  39                 helpString += "The pre.cluster command parameters are fasta, names and diffs. The fasta parameter is required. \n";
  40                 helpString += "The names parameter allows you to give a list of seqs that are identical. This file is 2 columns, first column is name or representative sequence, second column is a list of its identical sequences separated by commas.\n";
  41                 helpString += "The group parameter allows you to provide a group file so you can cluster by group. \n";
  42                 helpString += "The diffs parameter allows you to specify maximum number of mismatched bases allowed between sequences in a grouping. The default is 1.\n";
  43                 helpString += "The pre.cluster command should be in the following format: \n";
  44                 helpString += "pre.cluster(fasta=yourFastaFile, names=yourNamesFile, diffs=yourMaxDiffs) \n";
  45                 helpString += "Example pre.cluster(fasta=amazon.fasta, diffs=2).\n";
  46                 helpString += "Note: No spaces between parameter labels (i.e. fasta), '=' and parameters (i.e.yourFasta).\n";
  47                 return helpString;
  48         }
  49         catch(exception& e) {
  50                 m->errorOut(e, "PreClusterCommand", "getHelpString");
  51                 exit(1);
  52         }
  53 }
  54
  55 //**********************************************************************************************************************
  56 PreClusterCommand::PreClusterCommand(){
  57         try {
  58                 abort = true; calledHelp = true;
  59                 setParameters();
  60                 vector<string> tempOutNames;
  61                 outputTypes["fasta"] = tempOutNames;
  62                 outputTypes["name"] = tempOutNames;
  63                 outputTypes["map"] = tempOutNames;
  64         }
  65         catch(exception& e) {
  66                 m->errorOut(e, "PreClusterCommand", "PreClusterCommand");
  67                 exit(1);
  68         }
  69 }
  70 //**********************************************************************************************************************
  71
  72 PreClusterCommand::PreClusterCommand(string option) {
  73         try {
  74                 abort = false; calledHelp = false;
  75
  76                 //allow user to run help
  77                 if(option == "help") { help(); abort = true; calledHelp = true; }
  78                 else if(option == "citation") { citation(); abort = true; calledHelp = true;}
  79
  80                 else {
  81                         vector<string> myArray = setParameters();
  82
  83                         OptionParser parser(option);
  84                         map<string, string> parameters = parser.getParameters();
  85
  86                         ValidParameters validParameter;
  87                         map<string, string>::iterator it;
  88
  89                         //check to make sure all parameters are valid for command
  90                         for (map<string, string>::iterator it2 = parameters.begin(); it2 != parameters.end(); it2++) {
  91                                 if (validParameter.isValidParameter(it2->first, myArray, it2->second) != true) {  abort = true;  }
  92                         }
  93
  94                         //initialize outputTypes
  95                         vector<string> tempOutNames;
  96                         outputTypes["fasta"] = tempOutNames;
  97                         outputTypes["name"] = tempOutNames;
  98                         outputTypes["map"] = tempOutNames;
  99
 100                         //if the user changes the input directory command factory will send this info to us in the output parameter
 101                         string inputDir = validParameter.validFile(parameters, "inputdir", false);
 102                         if (inputDir == "not found"){   inputDir = "";          }
 103                         else {
 104                                 string path;
 105                                 it = parameters.find("fasta");
 106                                 //user has given a template file
 107                                 if(it != parameters.end()){
 108                                         path = m->hasPath(it->second);
 109                                         //if the user has not given a path then, add inputdir. else leave path alone.
 110                                         if (path == "") {       parameters["fasta"] = inputDir + it->second;            }
 111                                 }
 112
 113                                 it = parameters.find("name");
 114                                 //user has given a template file
 115                                 if(it != parameters.end()){
 116                                         path = m->hasPath(it->second);
 117                                         //if the user has not given a path then, add inputdir. else leave path alone.
 118                                         if (path == "") {       parameters["name"] = inputDir + it->second;             }
 119                                 }
 120
 121                                 it = parameters.find("group");
 122                                 //user has given a template file
 123                                 if(it != parameters.end()){
 124                                         path = m->hasPath(it->second);
 125                                         //if the user has not given a path then, add inputdir. else leave path alone.
 126                                         if (path == "") {       parameters["group"] = inputDir + it->second;            }
 127                                 }
 128                         }
 129
 130                         //check for required parameters
 131                         fastafile = validParameter.validFile(parameters, "fasta", true);
 132                         if (fastafile == "not found") {
 133                                 fastafile = m->getFastaFile();
 134                                 if (fastafile != "") { m->mothurOut("Using " + fastafile + " as input file for the fasta parameter."); m->mothurOutEndLine(); }
 135                                 else {  m->mothurOut("You have no current fastafile and the fasta parameter is required."); m->mothurOutEndLine(); abort = true; }
 136                         }
 137                         else if (fastafile == "not open") { abort = true; }
 138                         else { m->setFastaFile(fastafile); }
 139
 140                         //if the user changes the output directory command factory will send this info to us in the output parameter
 141                         outputDir = validParameter.validFile(parameters, "outputdir", false);           if (outputDir == "not found"){
 142                                 outputDir = "";
 143                                 outputDir += m->hasPath(fastafile); //if user entered a file with a path then preserve it
 144                         }
 145
 146                         //check for optional parameter and set defaults
 147                         // ...at some point should added some additional type checking...
 148                         namefile = validParameter.validFile(parameters, "name", true);
 149                         if (namefile == "not found") { namefile =  "";  }
 150                         else if (namefile == "not open") { abort = true; }
 151                         else {  m->setNameFile(namefile); }
 152
 153                         groupfile = validParameter.validFile(parameters, "group", true);
 154                         if (groupfile == "not found") { groupfile =  "";  bygroup = false; }
 155                         else if (groupfile == "not open") { abort = true; groupfile =  ""; }
 156                         else {   m->setGroupFile(groupfile); bygroup = true;  }
 157
 158                         string temp     = validParameter.validFile(parameters, "diffs", false);         if(temp == "not found"){        temp = "1"; }
 159                         convert(temp, diffs);
 160
 161                         temp = validParameter.validFile(parameters, "processors", false);       if (temp == "not found"){       temp = m->getProcessors();      }
 162                         m->setProcessors(temp);
 163                         convert(temp, processors);
 164
 165
 166                 }
 167
 168         }
 169         catch(exception& e) {
 170                 m->errorOut(e, "PreClusterCommand", "PreClusterCommand");
 171                 exit(1);
 172         }
 173 }
 174 //**********************************************************************************************************************
 175
 176 int PreClusterCommand::execute(){
 177         try {
 178
 179                 if (abort == true) { if (calledHelp) { return 0; }  return 2;   }
 180
 181                 int start = time(NULL);
 182
 183                 string fileroot = outputDir + m->getRootName(m->getSimpleName(fastafile));
 184                 string newFastaFile = fileroot + "precluster" + m->getExtension(fastafile);
 185                 string newNamesFile = fileroot + "precluster.names";
 186                 string newMapFile = fileroot + "precluster.map"; //add group name if by group
 187                 outputNames.push_back(newFastaFile); outputTypes["fasta"].push_back(newFastaFile);
 188                 outputNames.push_back(newNamesFile); outputTypes["name"].push_back(newNamesFile);
 189
 190
 191                 if (bygroup) {
 192                         //clear out old files
 193                         ofstream outFasta; m->openOutputFile(newFastaFile, outFasta); outFasta.close();
 194                         ofstream outNames; m->openOutputFile(newNamesFile, outNames);  outNames.close();
 195                         newMapFile = fileroot + "precluster.";
 196
 197                         //parse fasta and name file by group
 198                         SequenceParser* parser;
 199                         if (namefile != "") { parser = new SequenceParser(groupfile, fastafile, namefile);      }
 200                         else                            { parser = new SequenceParser(groupfile, fastafile);                    }
 201
 202                         vector<string> groups = parser->getNamesOfGroups();
 203
 204                         if(processors == 1)     {       driverGroups(parser, newFastaFile, newNamesFile, newMapFile, 0, groups.size(), groups); }
 205                         else                            {       createProcessesGroups(parser, newFastaFile, newNamesFile, newMapFile, groups);                  }
 206
 207                         delete parser;
 208
 209                         if (m->control_pressed) { for (int i = 0; i < outputNames.size(); i++) {        m->mothurRemove(outputNames[i]);        }        return 0; }
 210
 211                         //run unique.seqs for deconvolute results
 212                         string inputString = "fasta=" + newFastaFile;
 213                         if (namefile != "") { inputString += ", name=" + newNamesFile; }
 214                         m->mothurOutEndLine();
 215                         m->mothurOut("/******************************************/"); m->mothurOutEndLine();
 216                         m->mothurOut("Running command: unique.seqs(" + inputString + ")"); m->mothurOutEndLine();
 217
 218                         Command* uniqueCommand = new DeconvoluteCommand(inputString);
 219                         uniqueCommand->execute();
 220
 221                         map<string, vector<string> > filenames = uniqueCommand->getOutputFiles();
 222
 223                         delete uniqueCommand;
 224
 225                         m->mothurOut("/******************************************/"); m->mothurOutEndLine();
 226
 227                         m->renameFile(filenames["fasta"][0], newFastaFile);
 228
 229                         m->mothurOut("It took " + toString(time(NULL) - start) + " secs to run pre.cluster."); m->mothurOutEndLine();
 230
 231                 }else {
 232                         if (namefile != "") { readNameFile(); }
 233
 234                         //reads fasta file and return number of seqs
 235                         int numSeqs = readFASTA(); //fills alignSeqs and makes all seqs active
 236
 237                         if (m->control_pressed) { for (int i = 0; i < outputNames.size(); i++) {        m->mothurRemove(outputNames[i]);        } return 0; }
 238
 239                         if (numSeqs == 0) { m->mothurOut("Error reading fasta file...please correct."); m->mothurOutEndLine(); return 0;  }
 240                         if (diffs > length) { m->mothurOut("Error: diffs is greater than your sequence length."); m->mothurOutEndLine(); return 0;  }
 241
 242                         int count = process(newMapFile);
 243                         outputNames.push_back(newMapFile); outputTypes["map"].push_back(newMapFile);
 244
 245                         if (m->control_pressed) { for (int i = 0; i < outputNames.size(); i++) {        m->mothurRemove(outputNames[i]);        } return 0; }
 246
 247                         m->mothurOut("Total number of sequences before precluster was " + toString(alignSeqs.size()) + "."); m->mothurOutEndLine();
 248                         m->mothurOut("pre.cluster removed " + toString(count) + " sequences."); m->mothurOutEndLine(); m->mothurOutEndLine();
 249                         printData(newFastaFile, newNamesFile);
 250
 251                         m->mothurOut("It took " + toString(time(NULL) - start) + " secs to cluster " + toString(numSeqs) + " sequences."); m->mothurOutEndLine();
 252                 }
 253
 254                 if (m->control_pressed) { for (int i = 0; i < outputNames.size(); i++) {        m->mothurRemove(outputNames[i]);        } return 0; }
 255
 256                 m->mothurOutEndLine();
 257                 m->mothurOut("Output File Names: "); m->mothurOutEndLine();
 258                 for (int i = 0; i < outputNames.size(); i++) {  m->mothurOut(outputNames[i]); m->mothurOutEndLine();    }
 259                 m->mothurOutEndLine();
 260
 261                 //set fasta file as new current fastafile
 262                 string current = "";
 263                 itTypes = outputTypes.find("fasta");
 264                 if (itTypes != outputTypes.end()) {
 265                         if ((itTypes->second).size() != 0) { current = (itTypes->second)[0]; m->setFastaFile(current); }
 266                 }
 267
 268                 itTypes = outputTypes.find("name");
 269                 if (itTypes != outputTypes.end()) {
 270                         if ((itTypes->second).size() != 0) { current = (itTypes->second)[0]; m->setNameFile(current); }
 271                 }
 272
 273                 return 0;
 274
 275         }
 276         catch(exception& e) {
 277                 m->errorOut(e, "PreClusterCommand", "execute");
 278                 exit(1);
 279         }
 280 }
 281 /**************************************************************************************************/
 282 int PreClusterCommand::createProcessesGroups(SequenceParser* parser, string newFName, string newNName, string newMFile, vector<string> groups) {
 283         try {
 284
 285                 vector<int> processIDS;
 286                 int process = 1;
 287                 int num = 0;
 288
 289                 //sanity check
 290                 if (groups.size() < processors) { processors = groups.size(); }
 291
 292                 //divide the groups between the processors
 293                 vector<linePair> lines;
 294                 int numGroupsPerProcessor = groups.size() / processors;
 295                 for (int i = 0; i < processors; i++) {
 296                         int startIndex =  i * numGroupsPerProcessor;
 297                         int endIndex = (i+1) * numGroupsPerProcessor;
 298                         if(i == (processors - 1)){      endIndex = groups.size();       }
 299                         lines.push_back(linePair(startIndex, endIndex));
 300                 }
 301
 302 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
 303
 304                 //loop through and create all the processes you want
 305                 while (process != processors) {
 306                         int pid = fork();
 307
 308                         if (pid > 0) {
 309                                 processIDS.push_back(pid);  //create map from line number to pid so you can append files in correct order later
 310                                 process++;
 311                         }else if (pid == 0){
 312                                 num = driverGroups(parser, newFName + toString(getpid()) + ".temp", newNName + toString(getpid()) + ".temp", newMFile, lines[process].start, lines[process].end, groups);
 313                                 exit(0);
 314                         }else {
 315                                 m->mothurOut("[ERROR]: unable to spawn the necessary processes."); m->mothurOutEndLine();
 316                                 for (int i = 0; i < processIDS.size(); i++) { kill (processIDS[i], SIGINT); }
 317                                 exit(0);
 318                         }
 319                 }
 320
 321                 //do my part
 322                 num = driverGroups(parser, newFName, newNName, newMFile, lines[0].start, lines[0].end, groups);
 323
 324                 //force parent to wait until all the processes are done
 325                 for (int i=0;i<processIDS.size();i++) {
 326                         int temp = processIDS[i];
 327                         wait(&temp);
 328                 }
 329
 330 #else
 331
 332                 //////////////////////////////////////////////////////////////////////////////////////////////////////
 333                 //Windows version shared memory, so be careful when passing variables through the preClusterData struct.
 334                 //Above fork() will clone, so memory is separate, but that's not the case with windows,
 335                 //////////////////////////////////////////////////////////////////////////////////////////////////////
 336
 337                 vector<preClusterData*> pDataArray;
 338                 DWORD   dwThreadIdArray[processors-1];
 339                 HANDLE  hThreadArray[processors-1];
 340
 341                 //Create processor worker threads.
 342                 for( int i=1; i<processors; i++ ){
 343                         // Allocate memory for thread data.
 344                         string extension = toString(i) + ".temp";
 345
 346                         preClusterData* tempPreCluster = new preClusterData(fastafile, namefile, groupfile, (newFName+extension), (newNName+extension), newMFile, groups, m, lines[i].start, lines[i].end, diffs, i);
 347                         pDataArray.push_back(tempPreCluster);
 348                         processIDS.push_back(i);
 349
 350                         //MySeqSumThreadFunction is in header. It must be global or static to work with the threads.
 351                         //default security attributes, thread function name, argument to thread function, use default creation flags, returns the thread identifier
 352                         hThreadArray[i-1] = CreateThread(NULL, 0, MyPreclusterThreadFunction, pDataArray[i-1], 0, &dwThreadIdArray[i-1]);
 353                 }
 354
 355
 356                 //using the main process as a worker saves time and memory
 357                 num = driverGroups(parser, newFName, newNName, newMFile, lines[0].start, lines[0].end, groups);
 358
 359                 //Wait until all threads have terminated.
 360                 WaitForMultipleObjects(processors-1, hThreadArray, TRUE, INFINITE);
 361
 362                 //Close all thread handles and free memory allocations.
 363                 for(int i=0; i < pDataArray.size(); i++){
 364                         for (int j = 0; j < pDataArray[i]->mapFileNames.size(); j++) {
 365                                 outputNames.push_back(pDataArray[i]->mapFileNames[j]); outputTypes["map"].push_back(pDataArray[i]->mapFileNames[j]);
 366                         }
 367                         CloseHandle(hThreadArray[i]);
 368                         delete pDataArray[i];
 369                 }
 370
 371 #endif
 372
 373                 //append output files
 374                 for(int i=0;i<processIDS.size();i++){
 375                         m->appendFiles((newFName + toString(processIDS[i]) + ".temp"), newFName);
 376                         m->mothurRemove((newFName + toString(processIDS[i]) + ".temp"));
 377
 378                         m->appendFiles((newNName + toString(processIDS[i]) + ".temp"), newNName);
 379                         m->mothurRemove((newNName + toString(processIDS[i]) + ".temp"));
 380                 }
 381
 382                 return num;
 383
 384         }
 385         catch(exception& e) {
 386                 m->errorOut(e, "PreClusterCommand", "createProcessesGroups");
 387                 exit(1);
 388         }
 389 }
 390 /**************************************************************************************************/
 391 int PreClusterCommand::driverGroups(SequenceParser* parser, string newFFile, string newNFile, string newMFile, int start, int end, vector<string> groups){
 392         try {
 393
 394                 int numSeqs = 0;
 395
 396                 //precluster each group
 397                 for (int i = start; i < end; i++) {
 398
 399                         start = time(NULL);
 400
 401                         if (m->control_pressed) {  return 0; }
 402
 403                         m->mothurOutEndLine(); m->mothurOut("Processing group " + groups[i] + ":"); m->mothurOutEndLine();
 404
 405                         map<string, string> thisNameMap;
 406                         if (namefile != "") { thisNameMap = parser->getNameMap(groups[i]); }
 407                         vector<Sequence> thisSeqs = parser->getSeqs(groups[i]);
 408
 409                         //fill alignSeqs with this groups info.
 410                         numSeqs = loadSeqs(thisNameMap, thisSeqs);
 411
 412                         if (m->control_pressed) {   return 0; }
 413
 414                         if (diffs > length) { m->mothurOut("Error: diffs is greater than your sequence length."); m->mothurOutEndLine(); m->control_pressed = true; return 0;  }
 415
 416                         int count = process(newMFile+groups[i]+".map");
 417                         outputNames.push_back(newMFile+groups[i]+".map"); outputTypes["map"].push_back(newMFile+groups[i]+".map");
 418
 419                         if (m->control_pressed) {  return 0; }
 420
 421                         m->mothurOut("Total number of sequences before pre.cluster was " + toString(alignSeqs.size()) + "."); m->mothurOutEndLine();
 422                         m->mothurOut("pre.cluster removed " + toString(count) + " sequences."); m->mothurOutEndLine(); m->mothurOutEndLine();
 423                         printData(newFFile, newNFile);
 424
 425                         m->mothurOut("It took " + toString(time(NULL) - start) + " secs to cluster " + toString(numSeqs) + " sequences."); m->mothurOutEndLine();
 426
 427                 }
 428
 429                 return numSeqs;
 430         }
 431         catch(exception& e) {
 432                 m->errorOut(e, "PreClusterCommand", "driverGroups");
 433                 exit(1);
 434         }
 435 }
 436 /**************************************************************************************************/
 437 int PreClusterCommand::process(string newMapFile){
 438         try {
 439                 ofstream out;
 440                 m->openOutputFile(newMapFile, out);
 441
 442                 //sort seqs by number of identical seqs
 443                 sort(alignSeqs.begin(), alignSeqs.end(), comparePriority);
 444
 445                 int count = 0;
 446                 int numSeqs = alignSeqs.size();
 447
 448                 //think about running through twice...
 449                 for (int i = 0; i < numSeqs; i++) {
 450
 451                         //are you active
 452                         //                      itActive = active.find(alignSeqs[i].seq.getName());
 453
 454                         if (alignSeqs[i].active) {  //this sequence has not been merged yet
 455
 456                                 string chunk = alignSeqs[i].seq.getName() + "\t" + toString(alignSeqs[i].numIdentical) + "\t" + toString(0) + "\t" + alignSeqs[i].seq.getAligned() + "\n";
 457
 458                                 //try to merge it with all smaller seqs
 459                                 for (int j = i+1; j < numSeqs; j++) {
 460
 461                                         if (m->control_pressed) { out.close(); return 0; }
 462
 463                                         if (alignSeqs[j].active) {  //this sequence has not been merged yet
 464                                                 //are you within "diff" bases
 465                                                 int mismatch = calcMisMatches(alignSeqs[i].seq.getAligned(), alignSeqs[j].seq.getAligned());
 466
 467                                                 if (mismatch <= diffs) {
 468                                                         //merge
 469                                                         alignSeqs[i].names += ',' + alignSeqs[j].names;
 470                                                         alignSeqs[i].numIdentical += alignSeqs[j].numIdentical;
 471
 472                                                         chunk += alignSeqs[j].seq.getName() + "\t" + toString(alignSeqs[j].numIdentical) + "\t" + toString(mismatch) + "\t" + alignSeqs[j].seq.getAligned() + "\n";
 473
 474                                                         alignSeqs[j].active = 0;
 475                                                         alignSeqs[j].numIdentical = 0;
 476                                                         count++;
 477                                                 }
 478                                         }//end if j active
 479                                 }//end for loop j
 480
 481                                 //remove from active list
 482                                 alignSeqs[i].active = 0;
 483
 484                                 out << "ideal_seq_" << (i+1) << '\t' << alignSeqs[i].numIdentical << endl << chunk << endl;;
 485
 486                         }//end if active i
 487                         if(i % 100 == 0)        { m->mothurOut(toString(i) + "\t" + toString(numSeqs - count) + "\t" + toString(count)); m->mothurOutEndLine(); }
 488                 }
 489                 out.close();
 490
 491                 if(numSeqs % 100 != 0)  { m->mothurOut(toString(numSeqs) + "\t" + toString(numSeqs - count) + "\t" + toString(count)); m->mothurOutEndLine();   }
 492
 493                 return count;
 494
 495         }
 496         catch(exception& e) {
 497                 m->errorOut(e, "PreClusterCommand", "process");
 498                 exit(1);
 499         }
 500 }
 501 /**************************************************************************************************/
 502 int PreClusterCommand::readFASTA(){
 503         try {
 504                 //ifstream inNames;
 505                 ifstream inFasta;
 506
 507                 //m->openInputFile(namefile, inNames);
 508                 m->openInputFile(fastafile, inFasta);
 509
 510                 //string firstCol, secondCol, nameString;
 511                 length = 0;
 512
 513                 while (!inFasta.eof()) {
 514
 515                         if (m->control_pressed) { inFasta.close(); return 0; }
 516
 517                         //inNames >> firstCol >> secondCol;
 518                         //nameString = secondCol;
 519
 520                         //m->gobble(inNames);
 521                         //int size = 1;
 522                         //while (secondCol.find_first_of(',') != -1) {
 523                         //      size++;
 524                         //      secondCol = secondCol.substr(secondCol.find_first_of(',')+1, secondCol.length());
 525                         //}
 526
 527                         Sequence seq(inFasta);  m->gobble(inFasta);
 528
 529                         if (seq.getName() != "") {  //can get "" if commented line is at end of fasta file
 530                                 if (namefile != "") {
 531                                         itSize = sizes.find(seq.getName());
 532
 533                                         if (itSize == sizes.end()) { m->mothurOut(seq.getName() + " is not in your names file, please correct."); m->mothurOutEndLine(); exit(1); }
 534                                         else{
 535                                                 seqPNode tempNode(itSize->second, seq, names[seq.getName()]);
 536                                                 alignSeqs.push_back(tempNode);
 537                                                 if (seq.getAligned().length() > length) {  length = seq.getAligned().length();  }
 538                                         }
 539                                 }else { //no names file, you are identical to yourself
 540                                         seqPNode tempNode(1, seq, seq.getName());
 541                                         alignSeqs.push_back(tempNode);
 542                                         if (seq.getAligned().length() > length) {  length = seq.getAligned().length();  }
 543                                 }
 544                         }
 545                 }
 546                 inFasta.close();
 547                 //inNames.close();
 548                 return alignSeqs.size();
 549         }
 550
 551         catch(exception& e) {
 552                 m->errorOut(e, "PreClusterCommand", "readFASTA");
 553                 exit(1);
 554         }
 555 }
 556 /**************************************************************************************************/
 557 int PreClusterCommand::loadSeqs(map<string, string>& thisName, vector<Sequence>& thisSeqs){
 558         try {
 559                 length = 0;
 560                 alignSeqs.clear();
 561                 map<string, string>::iterator it;
 562                 bool error = false;
 563
 564                 for (int i = 0; i < thisSeqs.size(); i++) {
 565
 566                         if (m->control_pressed) { return 0; }
 567
 568                         if (namefile != "") {
 569                                 it = thisName.find(thisSeqs[i].getName());
 570
 571                                 //should never be true since parser checks for this
 572                                 if (it == thisName.end()) { m->mothurOut(thisSeqs[i].getName() + " is not in your names file, please correct."); m->mothurOutEndLine(); error = true; }
 573                                 else{
 574                                         //get number of reps
 575                                         int numReps = 1;
 576                                         for(int j=0;j<(it->second).length();j++){
 577                                                 if((it->second)[j] == ','){     numReps++;      }
 578                                         }
 579
 580                                         seqPNode tempNode(numReps, thisSeqs[i], it->second);
 581                                         alignSeqs.push_back(tempNode);
 582                                         if (thisSeqs[i].getAligned().length() > length) {  length = thisSeqs[i].getAligned().length();  }
 583                                 }
 584                         }else { //no names file, you are identical to yourself
 585                                 seqPNode tempNode(1, thisSeqs[i], thisSeqs[i].getName());
 586                                 alignSeqs.push_back(tempNode);
 587                                 if (thisSeqs[i].getAligned().length() > length) {  length = thisSeqs[i].getAligned().length();  }
 588                         }
 589                 }
 590
 591                 //sanity check
 592                 if (error) { m->control_pressed = true; }
 593
 594                 thisSeqs.clear();
 595
 596                 return alignSeqs.size();
 597         }
 598
 599         catch(exception& e) {
 600                 m->errorOut(e, "PreClusterCommand", "loadSeqs");
 601                 exit(1);
 602         }
 603 }
 604
 605 /**************************************************************************************************/
 606
 607 int PreClusterCommand::calcMisMatches(string seq1, string seq2){
 608         try {
 609                 int numBad = 0;
 610
 611                 for (int i = 0; i < seq1.length(); i++) {
 612                         //do they match
 613                         if (seq1[i] != seq2[i]) { numBad++; }
 614                         if (numBad > diffs) { return length;  } //to far to cluster
 615                 }
 616
 617                 return numBad;
 618         }
 619         catch(exception& e) {
 620                 m->errorOut(e, "PreClusterCommand", "calcMisMatches");
 621                 exit(1);
 622         }
 623 }
 624
 625 /**************************************************************************************************/
 626
 627 void PreClusterCommand::printData(string newfasta, string newname){
 628         try {
 629                 ofstream outFasta;
 630                 ofstream outNames;
 631
 632                 if (bygroup) {
 633                         m->openOutputFileAppend(newfasta, outFasta);
 634                         m->openOutputFileAppend(newname, outNames);
 635                 }else {
 636                         m->openOutputFile(newfasta, outFasta);
 637                         m->openOutputFile(newname, outNames);
 638                 }
 639
 640                 for (int i = 0; i < alignSeqs.size(); i++) {
 641                         if (alignSeqs[i].numIdentical != 0) {
 642                                 alignSeqs[i].seq.printSequence(outFasta);
 643                                 outNames << alignSeqs[i].seq.getName() << '\t' << alignSeqs[i].names << endl;
 644                         }
 645                 }
 646
 647                 outFasta.close();
 648                 outNames.close();
 649
 650         }
 651         catch(exception& e) {
 652                 m->errorOut(e, "PreClusterCommand", "printData");
 653                 exit(1);
 654         }
 655 }
 656 /**************************************************************************************************/
 657
 658 void PreClusterCommand::readNameFile(){
 659         try {
 660                 ifstream in;
 661                 m->openInputFile(namefile, in);
 662                 string firstCol, secondCol;
 663
 664                 while (!in.eof()) {
 665                         in >> firstCol >> secondCol; m->gobble(in);
 666                         names[firstCol] = secondCol;
 667                         int size = 1;
 668
 669                         for(int i=0;i<secondCol.size();i++){
 670                                 if(secondCol[i] == ','){        size++; }
 671                         }
 672                         sizes[firstCol] = size;
 673                 }
 674                 in.close();
 675         }
 676         catch(exception& e) {
 677                 m->errorOut(e, "PreClusterCommand", "readNameFile");
 678                 exit(1);
 679         }
 680 }
 681
 682 /**************************************************************************************************/
 683
 684