source/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") { namefile = ""; 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                         m->mothurConvert(temp, diffs);
 160
 161                         temp = validParameter.validFile(parameters, "processors", false);       if (temp == "not found"){       temp = m->getProcessors();      }
 162                         m->setProcessors(temp);
 163                         m->mothurConvert(temp, processors);
 164
 165                         if (namefile == "") {
 166                                 vector<string> files; files.push_back(fastafile);
 167                                 parser.getNameFile(files);
 168                         }
 169                 }
 170
 171         }
 172         catch(exception& e) {
 173                 m->errorOut(e, "PreClusterCommand", "PreClusterCommand");
 174                 exit(1);
 175         }
 176 }
 177 //**********************************************************************************************************************
 178
 179 int PreClusterCommand::execute(){
 180         try {
 181
 182                 if (abort == true) { if (calledHelp) { return 0; }  return 2;   }
 183
 184                 int start = time(NULL);
 185
 186                 string fileroot = outputDir + m->getRootName(m->getSimpleName(fastafile));
 187                 string newFastaFile = fileroot + "precluster" + m->getExtension(fastafile);
 188                 string newNamesFile = fileroot + "precluster.names";
 189                 string newMapFile = fileroot + "precluster.map"; //add group name if by group
 190                 outputNames.push_back(newFastaFile); outputTypes["fasta"].push_back(newFastaFile);
 191                 outputNames.push_back(newNamesFile); outputTypes["name"].push_back(newNamesFile);
 192
 193
 194                 if (bygroup) {
 195                         //clear out old files
 196                         ofstream outFasta; m->openOutputFile(newFastaFile, outFasta); outFasta.close();
 197                         ofstream outNames; m->openOutputFile(newNamesFile, outNames);  outNames.close();
 198                         newMapFile = fileroot + "precluster.";
 199
 200                         //parse fasta and name file by group
 201                         SequenceParser* parser;
 202                         if (namefile != "") { parser = new SequenceParser(groupfile, fastafile, namefile);      }
 203                         else                            { parser = new SequenceParser(groupfile, fastafile);                    }
 204
 205                         vector<string> groups = parser->getNamesOfGroups();
 206
 207                         if(processors == 1)     {       driverGroups(parser, newFastaFile, newNamesFile, newMapFile, 0, groups.size(), groups); }
 208                         else                            {       createProcessesGroups(parser, newFastaFile, newNamesFile, newMapFile, groups);                  }
 209
 210                         delete parser;
 211
 212                         if (m->control_pressed) { for (int i = 0; i < outputNames.size(); i++) {        m->mothurRemove(outputNames[i]);        }        return 0; }
 213
 214                         //run unique.seqs for deconvolute results
 215                         string inputString = "fasta=" + newFastaFile;
 216                         if (namefile != "") { inputString += ", name=" + newNamesFile; }
 217                         m->mothurOutEndLine();
 218                         m->mothurOut("/******************************************/"); m->mothurOutEndLine();
 219                         m->mothurOut("Running command: unique.seqs(" + inputString + ")"); m->mothurOutEndLine();
 220                         m->mothurCalling = true;
 221
 222                         Command* uniqueCommand = new DeconvoluteCommand(inputString);
 223                         uniqueCommand->execute();
 224
 225                         map<string, vector<string> > filenames = uniqueCommand->getOutputFiles();
 226
 227                         delete uniqueCommand;
 228                         m->mothurCalling = false;
 229                         m->mothurOut("/******************************************/"); m->mothurOutEndLine();
 230
 231                         m->renameFile(filenames["fasta"][0], newFastaFile);
 232             m->renameFile(filenames["name"][0], newNamesFile);
 233
 234                         m->mothurOut("It took " + toString(time(NULL) - start) + " secs to run pre.cluster."); m->mothurOutEndLine();
 235
 236                 }else {
 237                         if (namefile != "") { readNameFile(); }
 238
 239                         //reads fasta file and return number of seqs
 240                         int numSeqs = readFASTA(); //fills alignSeqs and makes all seqs active
 241
 242                         if (m->control_pressed) { for (int i = 0; i < outputNames.size(); i++) {        m->mothurRemove(outputNames[i]);        } return 0; }
 243
 244                         if (numSeqs == 0) { m->mothurOut("Error reading fasta file...please correct."); m->mothurOutEndLine(); return 0;  }
 245                         if (diffs > length) { m->mothurOut("Error: diffs is greater than your sequence length."); m->mothurOutEndLine(); return 0;  }
 246
 247                         int count = process(newMapFile);
 248                         outputNames.push_back(newMapFile); outputTypes["map"].push_back(newMapFile);
 249
 250                         if (m->control_pressed) { for (int i = 0; i < outputNames.size(); i++) {        m->mothurRemove(outputNames[i]);        } return 0; }
 251
 252                         m->mothurOut("Total number of sequences before precluster was " + toString(alignSeqs.size()) + "."); m->mothurOutEndLine();
 253                         m->mothurOut("pre.cluster removed " + toString(count) + " sequences."); m->mothurOutEndLine(); m->mothurOutEndLine();
 254                         printData(newFastaFile, newNamesFile);
 255
 256                         m->mothurOut("It took " + toString(time(NULL) - start) + " secs to cluster " + toString(numSeqs) + " sequences."); m->mothurOutEndLine();
 257                 }
 258
 259                 if (m->control_pressed) { for (int i = 0; i < outputNames.size(); i++) {        m->mothurRemove(outputNames[i]);        } return 0; }
 260
 261                 m->mothurOutEndLine();
 262                 m->mothurOut("Output File Names: "); m->mothurOutEndLine();
 263                 for (int i = 0; i < outputNames.size(); i++) {  m->mothurOut(outputNames[i]); m->mothurOutEndLine();    }
 264                 m->mothurOutEndLine();
 265
 266                 //set fasta file as new current fastafile
 267                 string current = "";
 268                 itTypes = outputTypes.find("fasta");
 269                 if (itTypes != outputTypes.end()) {
 270                         if ((itTypes->second).size() != 0) { current = (itTypes->second)[0]; m->setFastaFile(current); }
 271                 }
 272
 273                 itTypes = outputTypes.find("name");
 274                 if (itTypes != outputTypes.end()) {
 275                         if ((itTypes->second).size() != 0) { current = (itTypes->second)[0]; m->setNameFile(current); }
 276                 }
 277
 278                 return 0;
 279
 280         }
 281         catch(exception& e) {
 282                 m->errorOut(e, "PreClusterCommand", "execute");
 283                 exit(1);
 284         }
 285 }
 286 /**************************************************************************************************/
 287 int PreClusterCommand::createProcessesGroups(SequenceParser* parser, string newFName, string newNName, string newMFile, vector<string> groups) {
 288         try {
 289
 290                 vector<int> processIDS;
 291                 int process = 1;
 292                 int num = 0;
 293
 294                 //sanity check
 295                 if (groups.size() < processors) { processors = groups.size(); }
 296
 297                 //divide the groups between the processors
 298                 vector<linePair> lines;
 299                 int numGroupsPerProcessor = groups.size() / processors;
 300                 for (int i = 0; i < processors; i++) {
 301                         int startIndex =  i * numGroupsPerProcessor;
 302                         int endIndex = (i+1) * numGroupsPerProcessor;
 303                         if(i == (processors - 1)){      endIndex = groups.size();       }
 304                         lines.push_back(linePair(startIndex, endIndex));
 305                 }
 306
 307 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) || (__linux__) || (__unix__) || (__unix)
 308
 309                 //loop through and create all the processes you want
 310                 while (process != processors) {
 311                         int pid = fork();
 312
 313                         if (pid > 0) {
 314                                 processIDS.push_back(pid);  //create map from line number to pid so you can append files in correct order later
 315                                 process++;
 316                         }else if (pid == 0){
 317                 outputNames.clear();
 318                                 num = driverGroups(parser, newFName + toString(getpid()) + ".temp", newNName + toString(getpid()) + ".temp", newMFile, lines[process].start, lines[process].end, groups);
 319
 320                 string tempFile = toString(getpid()) + ".outputNames.temp";
 321                 ofstream outTemp;
 322                 m->openOutputFile(tempFile, outTemp);
 323
 324                 outTemp << outputNames.size();
 325                 for (int i = 0; i < outputNames.size(); i++) { outTemp << outputNames[i] << endl; }
 326                 outTemp.close();
 327
 328                                 exit(0);
 329                         }else {
 330                                 m->mothurOut("[ERROR]: unable to spawn the necessary processes."); m->mothurOutEndLine();
 331                                 for (int i = 0; i < processIDS.size(); i++) { kill (processIDS[i], SIGINT); }
 332                                 exit(0);
 333                         }
 334                 }
 335
 336                 //do my part
 337                 num = driverGroups(parser, newFName, newNName, newMFile, lines[0].start, lines[0].end, groups);
 338
 339                 //force parent to wait until all the processes are done
 340                 for (int i=0;i<processIDS.size();i++) {
 341                         int temp = processIDS[i];
 342                         wait(&temp);
 343                 }
 344
 345         for (int i = 0; i < processIDS.size(); i++) {
 346             string tempFile = toString(processIDS[i]) +  ".outputNames.temp";
 347             ifstream intemp;
 348             m->openInputFile(tempFile, intemp);
 349
 350             int num;
 351             intemp >> num;
 352             for (int k = 0; k < num; k++) {
 353                 string name = "";
 354                 intemp >> name; m->gobble(intemp);
 355
 356                 outputNames.push_back(name); outputTypes["map"].push_back(name);
 357             }
 358             intemp.close();
 359             m->mothurRemove(tempFile);
 360         }
 361 #else
 362
 363                 //////////////////////////////////////////////////////////////////////////////////////////////////////
 364                 //Windows version shared memory, so be careful when passing variables through the preClusterData struct.
 365                 //Above fork() will clone, so memory is separate, but that's not the case with windows,
 366                 //////////////////////////////////////////////////////////////////////////////////////////////////////
 367
 368                 vector<preClusterData*> pDataArray;
 369                 DWORD   dwThreadIdArray[processors-1];
 370                 HANDLE  hThreadArray[processors-1];
 371
 372                 //Create processor worker threads.
 373                 for( int i=1; i<processors; i++ ){
 374                         // Allocate memory for thread data.
 375                         string extension = toString(i) + ".temp";
 376
 377                         preClusterData* tempPreCluster = new preClusterData(fastafile, namefile, groupfile, (newFName+extension), (newNName+extension), newMFile, groups, m, lines[i].start, lines[i].end, diffs, i);
 378                         pDataArray.push_back(tempPreCluster);
 379                         processIDS.push_back(i);
 380
 381                         //MySeqSumThreadFunction is in header. It must be global or static to work with the threads.
 382                         //default security attributes, thread function name, argument to thread function, use default creation flags, returns the thread identifier
 383                         hThreadArray[i-1] = CreateThread(NULL, 0, MyPreclusterThreadFunction, pDataArray[i-1], 0, &dwThreadIdArray[i-1]);
 384                 }
 385
 386
 387                 //using the main process as a worker saves time and memory
 388                 num = driverGroups(parser, newFName, newNName, newMFile, lines[0].start, lines[0].end, groups);
 389
 390                 //Wait until all threads have terminated.
 391                 WaitForMultipleObjects(processors-1, hThreadArray, TRUE, INFINITE);
 392
 393                 //Close all thread handles and free memory allocations.
 394                 for(int i=0; i < pDataArray.size(); i++){
 395                         for (int j = 0; j < pDataArray[i]->mapFileNames.size(); j++) {
 396                                 outputNames.push_back(pDataArray[i]->mapFileNames[j]); outputTypes["map"].push_back(pDataArray[i]->mapFileNames[j]);
 397                         }
 398                         CloseHandle(hThreadArray[i]);
 399                         delete pDataArray[i];
 400                 }
 401
 402 #endif
 403
 404                 //append output files
 405                 for(int i=0;i<processIDS.size();i++){
 406                         //newFName = m->getFullPathName(".\\" + newFName);
 407                         //newNName = m->getFullPathName(".\\" + newNName);
 408
 409                         m->appendFiles((newFName + toString(processIDS[i]) + ".temp"), newFName);
 410                         m->mothurRemove((newFName + toString(processIDS[i]) + ".temp"));
 411
 412                         m->appendFiles((newNName + toString(processIDS[i]) + ".temp"), newNName);
 413                         m->mothurRemove((newNName + toString(processIDS[i]) + ".temp"));
 414                 }
 415
 416                 return num;
 417
 418         }
 419         catch(exception& e) {
 420                 m->errorOut(e, "PreClusterCommand", "createProcessesGroups");
 421                 exit(1);
 422         }
 423 }
 424 /**************************************************************************************************/
 425 int PreClusterCommand::driverGroups(SequenceParser* parser, string newFFile, string newNFile, string newMFile, int start, int end, vector<string> groups){
 426         try {
 427
 428                 int numSeqs = 0;
 429
 430                 //precluster each group
 431                 for (int i = start; i < end; i++) {
 432
 433                         start = time(NULL);
 434
 435                         if (m->control_pressed) {  return 0; }
 436
 437                         m->mothurOutEndLine(); m->mothurOut("Processing group " + groups[i] + ":"); m->mothurOutEndLine();
 438
 439                         map<string, string> thisNameMap;
 440                         if (namefile != "") { thisNameMap = parser->getNameMap(groups[i]); }
 441                         vector<Sequence> thisSeqs = parser->getSeqs(groups[i]);
 442
 443                         //fill alignSeqs with this groups info.
 444                         numSeqs = loadSeqs(thisNameMap, thisSeqs);
 445
 446                         if (m->control_pressed) {   return 0; }
 447
 448                         if (diffs > length) { m->mothurOut("Error: diffs is greater than your sequence length."); m->mothurOutEndLine(); m->control_pressed = true; return 0;  }
 449
 450                         int count = process(newMFile+groups[i]+".map");
 451                         outputNames.push_back(newMFile+groups[i]+".map"); outputTypes["map"].push_back(newMFile+groups[i]+".map");
 452
 453                         if (m->control_pressed) {  return 0; }
 454
 455                         m->mothurOut("Total number of sequences before pre.cluster was " + toString(alignSeqs.size()) + "."); m->mothurOutEndLine();
 456                         m->mothurOut("pre.cluster removed " + toString(count) + " sequences."); m->mothurOutEndLine(); m->mothurOutEndLine();
 457                         printData(newFFile, newNFile);
 458
 459                         m->mothurOut("It took " + toString(time(NULL) - start) + " secs to cluster " + toString(numSeqs) + " sequences."); m->mothurOutEndLine();
 460
 461                 }
 462
 463                 return numSeqs;
 464         }
 465         catch(exception& e) {
 466                 m->errorOut(e, "PreClusterCommand", "driverGroups");
 467                 exit(1);
 468         }
 469 }
 470 /**************************************************************************************************/
 471 int PreClusterCommand::process(string newMapFile){
 472         try {
 473                 ofstream out;
 474                 m->openOutputFile(newMapFile, out);
 475
 476                 //sort seqs by number of identical seqs
 477                 sort(alignSeqs.begin(), alignSeqs.end(), comparePriority);
 478
 479                 int count = 0;
 480                 int numSeqs = alignSeqs.size();
 481
 482                 //think about running through twice...
 483                 for (int i = 0; i < numSeqs; i++) {
 484
 485                         //are you active
 486                         //                      itActive = active.find(alignSeqs[i].seq.getName());
 487
 488                         if (alignSeqs[i].active) {  //this sequence has not been merged yet
 489
 490                                 string chunk = alignSeqs[i].seq.getName() + "\t" + toString(alignSeqs[i].numIdentical) + "\t" + toString(0) + "\t" + alignSeqs[i].seq.getAligned() + "\n";
 491
 492                                 //try to merge it with all smaller seqs
 493                                 for (int j = i+1; j < numSeqs; j++) {
 494
 495                                         if (m->control_pressed) { out.close(); return 0; }
 496
 497                                         if (alignSeqs[j].active) {  //this sequence has not been merged yet
 498                                                 //are you within "diff" bases
 499                                                 int mismatch = calcMisMatches(alignSeqs[i].seq.getAligned(), alignSeqs[j].seq.getAligned());
 500
 501                                                 if (mismatch <= diffs) {
 502                                                         //merge
 503                                                         alignSeqs[i].names += ',' + alignSeqs[j].names;
 504                                                         alignSeqs[i].numIdentical += alignSeqs[j].numIdentical;
 505
 506                                                         chunk += alignSeqs[j].seq.getName() + "\t" + toString(alignSeqs[j].numIdentical) + "\t" + toString(mismatch) + "\t" + alignSeqs[j].seq.getAligned() + "\n";
 507
 508                                                         alignSeqs[j].active = 0;
 509                                                         alignSeqs[j].numIdentical = 0;
 510                                                         count++;
 511                                                 }
 512                                         }//end if j active
 513                                 }//end for loop j
 514
 515                                 //remove from active list
 516                                 alignSeqs[i].active = 0;
 517
 518                                 out << "ideal_seq_" << (i+1) << '\t' << alignSeqs[i].numIdentical << endl << chunk << endl;;
 519
 520                         }//end if active i
 521                         if(i % 100 == 0)        { m->mothurOut(toString(i) + "\t" + toString(numSeqs - count) + "\t" + toString(count)); m->mothurOutEndLine(); }
 522                 }
 523                 out.close();
 524
 525                 if(numSeqs % 100 != 0)  { m->mothurOut(toString(numSeqs) + "\t" + toString(numSeqs - count) + "\t" + toString(count)); m->mothurOutEndLine();   }
 526
 527                 return count;
 528
 529         }
 530         catch(exception& e) {
 531                 m->errorOut(e, "PreClusterCommand", "process");
 532                 exit(1);
 533         }
 534 }
 535 /**************************************************************************************************/
 536 int PreClusterCommand::readFASTA(){
 537         try {
 538                 //ifstream inNames;
 539                 ifstream inFasta;
 540
 541                 //m->openInputFile(namefile, inNames);
 542                 m->openInputFile(fastafile, inFasta);
 543
 544                 //string firstCol, secondCol, nameString;
 545                 set<int> lengths;
 546
 547                 while (!inFasta.eof()) {
 548
 549                         if (m->control_pressed) { inFasta.close(); return 0; }
 550
 551                         //inNames >> firstCol >> secondCol;
 552                         //nameString = secondCol;
 553
 554                         //m->gobble(inNames);
 555                         //int size = 1;
 556                         //while (secondCol.find_first_of(',') != -1) {
 557                         //      size++;
 558                         //      secondCol = secondCol.substr(secondCol.find_first_of(',')+1, secondCol.length());
 559                         //}
 560
 561                         Sequence seq(inFasta);  m->gobble(inFasta);
 562
 563                         if (seq.getName() != "") {  //can get "" if commented line is at end of fasta file
 564                                 if (namefile != "") {
 565                                         itSize = sizes.find(seq.getName());
 566
 567                                         if (itSize == sizes.end()) { m->mothurOut(seq.getName() + " is not in your names file, please correct."); m->mothurOutEndLine(); exit(1); }
 568                                         else{
 569                                                 seqPNode tempNode(itSize->second, seq, names[seq.getName()]);
 570                                                 alignSeqs.push_back(tempNode);
 571                                                 lengths.insert(seq.getAligned().length());
 572                                         }
 573                                 }else { //no names file, you are identical to yourself
 574                                         seqPNode tempNode(1, seq, seq.getName());
 575                                         alignSeqs.push_back(tempNode);
 576                                         lengths.insert(seq.getAligned().length());
 577                                 }
 578                         }
 579                 }
 580                 inFasta.close();
 581                 //inNames.close();
 582
 583         if (lengths.size() > 1) { m->control_pressed = true; m->mothurOut("[ERROR]: your sequences are not all the same length. pre.cluster requires sequences to be aligned."); m->mothurOutEndLine(); }
 584         else if (lengths.size() == 1) { length = *(lengths.begin()); }
 585
 586                 return alignSeqs.size();
 587         }
 588
 589         catch(exception& e) {
 590                 m->errorOut(e, "PreClusterCommand", "readFASTA");
 591                 exit(1);
 592         }
 593 }
 594 /**************************************************************************************************/
 595 int PreClusterCommand::loadSeqs(map<string, string>& thisName, vector<Sequence>& thisSeqs){
 596         try {
 597                 set<int> lengths;
 598                 alignSeqs.clear();
 599                 map<string, string>::iterator it;
 600                 bool error = false;
 601
 602                 for (int i = 0; i < thisSeqs.size(); i++) {
 603
 604                         if (m->control_pressed) { return 0; }
 605
 606                         if (namefile != "") {
 607                                 it = thisName.find(thisSeqs[i].getName());
 608
 609                                 //should never be true since parser checks for this
 610                                 if (it == thisName.end()) { m->mothurOut(thisSeqs[i].getName() + " is not in your names file, please correct."); m->mothurOutEndLine(); error = true; }
 611                                 else{
 612                                         //get number of reps
 613                                         int numReps = 1;
 614                                         for(int j=0;j<(it->second).length();j++){
 615                                                 if((it->second)[j] == ','){     numReps++;      }
 616                                         }
 617
 618                                         seqPNode tempNode(numReps, thisSeqs[i], it->second);
 619                                         alignSeqs.push_back(tempNode);
 620                     lengths.insert(thisSeqs[i].getAligned().length());
 621                                 }
 622                         }else { //no names file, you are identical to yourself
 623                                 seqPNode tempNode(1, thisSeqs[i], thisSeqs[i].getName());
 624                                 alignSeqs.push_back(tempNode);
 625                                 lengths.insert(thisSeqs[i].getAligned().length());
 626                         }
 627                 }
 628
 629         if (lengths.size() > 1) { error = true; m->mothurOut("[ERROR]: your sequences are not all the same length. pre.cluster requires sequences to be aligned."); m->mothurOutEndLine(); }
 630         else if (lengths.size() == 1) { length = *(lengths.begin()); }
 631
 632                 //sanity check
 633                 if (error) { m->control_pressed = true; }
 634
 635                 thisSeqs.clear();
 636
 637                 return alignSeqs.size();
 638         }
 639
 640         catch(exception& e) {
 641                 m->errorOut(e, "PreClusterCommand", "loadSeqs");
 642                 exit(1);
 643         }
 644 }
 645
 646 /**************************************************************************************************/
 647
 648 int PreClusterCommand::calcMisMatches(string seq1, string seq2){
 649         try {
 650                 int numBad = 0;
 651
 652                 for (int i = 0; i < seq1.length(); i++) {
 653                         //do they match
 654                         if (seq1[i] != seq2[i]) { numBad++; }
 655                         if (numBad > diffs) { return length;  } //to far to cluster
 656                 }
 657
 658                 return numBad;
 659         }
 660         catch(exception& e) {
 661                 m->errorOut(e, "PreClusterCommand", "calcMisMatches");
 662                 exit(1);
 663         }
 664 }
 665
 666 /**************************************************************************************************/
 667
 668 void PreClusterCommand::printData(string newfasta, string newname){
 669         try {
 670                 ofstream outFasta;
 671                 ofstream outNames;
 672
 673                 if (bygroup) {
 674                         m->openOutputFileAppend(newfasta, outFasta);
 675                         m->openOutputFileAppend(newname, outNames);
 676                 }else {
 677                         m->openOutputFile(newfasta, outFasta);
 678                         m->openOutputFile(newname, outNames);
 679                 }
 680
 681                 for (int i = 0; i < alignSeqs.size(); i++) {
 682                         if (alignSeqs[i].numIdentical != 0) {
 683                                 alignSeqs[i].seq.printSequence(outFasta);
 684                                 outNames << alignSeqs[i].seq.getName() << '\t' << alignSeqs[i].names << endl;
 685                         }
 686                 }
 687
 688                 outFasta.close();
 689                 outNames.close();
 690
 691         }
 692         catch(exception& e) {
 693                 m->errorOut(e, "PreClusterCommand", "printData");
 694                 exit(1);
 695         }
 696 }
 697 /**************************************************************************************************/
 698
 699 void PreClusterCommand::readNameFile(){
 700         try {
 701                 ifstream in;
 702                 m->openInputFile(namefile, in);
 703                 string firstCol, secondCol;
 704
 705                 while (!in.eof()) {
 706                         in >> firstCol >> secondCol; m->gobble(in);
 707                         names[firstCol] = secondCol;
 708                         int size = 1;
 709
 710                         for(int i=0;i<secondCol.size();i++){
 711                                 if(secondCol[i] == ','){        size++; }
 712                         }
 713                         sizes[firstCol] = size;
 714                 }
 715                 in.close();
 716         }
 717         catch(exception& e) {
 718                 m->errorOut(e, "PreClusterCommand", "readNameFile");
 719                 exit(1);
 720         }
 721 }
 722
 723 /**************************************************************************************************/
 724
 725