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","fasta-name",false,true,true); parameters.push_back(pfasta);
  17                 CommandParameter pname("name", "InputTypes", "", "", "NameCount", "none", "none","name",false,false,true); parameters.push_back(pname);
  18         CommandParameter pcount("count", "InputTypes", "", "", "NameCount-CountGroup", "none", "none","count",false,false,true); parameters.push_back(pcount);
  19                 CommandParameter pgroup("group", "InputTypes", "", "", "CountGroup", "none", "none","",false,false,true); parameters.push_back(pgroup);
  20                 CommandParameter pdiffs("diffs", "Number", "", "1", "", "", "","",false,false,true); parameters.push_back(pdiffs);
  21                 CommandParameter pprocessors("processors", "Number", "", "1", "", "", "","",false,false,true); parameters.push_back(pprocessors);
  22         CommandParameter ptopdown("topdown", "Boolean", "", "T", "", "", "","",false,false); parameters.push_back(ptopdown);
  23
  24                 CommandParameter pinputdir("inputdir", "String", "", "", "", "", "","",false,false); parameters.push_back(pinputdir);
  25                 CommandParameter poutputdir("outputdir", "String", "", "", "", "", "","",false,false); parameters.push_back(poutputdir);
  26
  27                 vector<string> myArray;
  28                 for (int i = 0; i < parameters.size(); i++) {   myArray.push_back(parameters[i].name);          }
  29                 return myArray;
  30         }
  31         catch(exception& e) {
  32                 m->errorOut(e, "PreClusterCommand", "setParameters");
  33                 exit(1);
  34         }
  35 }
  36 //**********************************************************************************************************************
  37 string PreClusterCommand::getHelpString(){
  38         try {
  39                 string helpString = "";
  40                 helpString += "The pre.cluster command groups sequences that are within a given number of base mismatches.\n";
  41                 helpString += "The pre.cluster command outputs a new fasta and name file.\n";
  42                 helpString += "The pre.cluster command parameters are fasta, name, group, count, topdown, processors and diffs. The fasta parameter is required. \n";
  43                 helpString += "The name 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";
  44                 helpString += "The group parameter allows you to provide a group file so you can cluster by group. \n";
  45         helpString += "The count parameter allows you to provide a count file so you can cluster by group. \n";
  46                 helpString += "The diffs parameter allows you to specify maximum number of mismatched bases allowed between sequences in a grouping. The default is 1.\n";
  47         helpString += "The topdown parameter allows you to specify whether to cluster from largest abundance to smallest or smallest to largest.  Default=T, meaning largest to smallest.\n";
  48                 helpString += "The pre.cluster command should be in the following format: \n";
  49                 helpString += "pre.cluster(fasta=yourFastaFile, names=yourNamesFile, diffs=yourMaxDiffs) \n";
  50                 helpString += "Example pre.cluster(fasta=amazon.fasta, diffs=2).\n";
  51                 helpString += "Note: No spaces between parameter labels (i.e. fasta), '=' and parameters (i.e.yourFasta).\n";
  52                 return helpString;
  53         }
  54         catch(exception& e) {
  55                 m->errorOut(e, "PreClusterCommand", "getHelpString");
  56                 exit(1);
  57         }
  58 }
  59 //**********************************************************************************************************************
  60 string PreClusterCommand::getOutputPattern(string type) {
  61     try {
  62         string pattern = "";
  63
  64         if (type == "fasta") {  pattern = "[filename],precluster,[extension]"; }
  65         else if (type == "name") {  pattern = "[filename],precluster.names"; }
  66         else if (type == "count") {  pattern = "[filename],precluster.count_table"; }
  67         else if (type == "map") {  pattern =  "[filename],precluster.map"; }
  68         else { m->mothurOut("[ERROR]: No definition for type " + type + " output pattern.\n"); m->control_pressed = true;  }
  69
  70         return pattern;
  71     }
  72     catch(exception& e) {
  73         m->errorOut(e, "PreClusterCommand", "getOutputPattern");
  74         exit(1);
  75     }
  76 }
  77 //**********************************************************************************************************************
  78 PreClusterCommand::PreClusterCommand(){
  79         try {
  80                 abort = true; calledHelp = true;
  81                 setParameters();
  82                 vector<string> tempOutNames;
  83                 outputTypes["fasta"] = tempOutNames;
  84                 outputTypes["name"] = tempOutNames;
  85         outputTypes["count"] = tempOutNames;
  86                 outputTypes["map"] = tempOutNames;
  87         }
  88         catch(exception& e) {
  89                 m->errorOut(e, "PreClusterCommand", "PreClusterCommand");
  90                 exit(1);
  91         }
  92 }
  93 //**********************************************************************************************************************
  94
  95 PreClusterCommand::PreClusterCommand(string option) {
  96         try {
  97                 abort = false; calledHelp = false;
  98
  99                 //allow user to run help
 100                 if(option == "help") { help(); abort = true; calledHelp = true; }
 101                 else if(option == "citation") { citation(); abort = true; calledHelp = true;}
 102
 103                 else {
 104                         vector<string> myArray = setParameters();
 105
 106                         OptionParser parser(option);
 107                         map<string, string> parameters = parser.getParameters();
 108
 109                         ValidParameters validParameter;
 110                         map<string, string>::iterator it;
 111
 112                         //check to make sure all parameters are valid for command
 113                         for (map<string, string>::iterator it2 = parameters.begin(); it2 != parameters.end(); it2++) {
 114                                 if (validParameter.isValidParameter(it2->first, myArray, it2->second) != true) {  abort = true;  }
 115                         }
 116
 117                         //initialize outputTypes
 118                         vector<string> tempOutNames;
 119                         outputTypes["fasta"] = tempOutNames;
 120                         outputTypes["name"] = tempOutNames;
 121                         outputTypes["map"] = tempOutNames;
 122             outputTypes["count"] = tempOutNames;
 123
 124                         //if the user changes the input directory command factory will send this info to us in the output parameter
 125                         string inputDir = validParameter.validFile(parameters, "inputdir", false);
 126                         if (inputDir == "not found"){   inputDir = "";          }
 127                         else {
 128                                 string path;
 129                                 it = parameters.find("fasta");
 130                                 //user has given a template file
 131                                 if(it != parameters.end()){
 132                                         path = m->hasPath(it->second);
 133                                         //if the user has not given a path then, add inputdir. else leave path alone.
 134                                         if (path == "") {       parameters["fasta"] = inputDir + it->second;            }
 135                                 }
 136
 137                                 it = parameters.find("name");
 138                                 //user has given a template file
 139                                 if(it != parameters.end()){
 140                                         path = m->hasPath(it->second);
 141                                         //if the user has not given a path then, add inputdir. else leave path alone.
 142                                         if (path == "") {       parameters["name"] = inputDir + it->second;             }
 143                                 }
 144
 145                                 it = parameters.find("group");
 146                                 //user has given a template file
 147                                 if(it != parameters.end()){
 148                                         path = m->hasPath(it->second);
 149                                         //if the user has not given a path then, add inputdir. else leave path alone.
 150                                         if (path == "") {       parameters["group"] = inputDir + it->second;            }
 151                                 }
 152
 153                 it = parameters.find("count");
 154                                 //user has given a template file
 155                                 if(it != parameters.end()){
 156                                         path = m->hasPath(it->second);
 157                                         //if the user has not given a path then, add inputdir. else leave path alone.
 158                                         if (path == "") {       parameters["count"] = inputDir + it->second;            }
 159                                 }
 160                         }
 161
 162                         //check for required parameters
 163                         fastafile = validParameter.validFile(parameters, "fasta", true);
 164                         if (fastafile == "not found") {
 165                                 fastafile = m->getFastaFile();
 166                                 if (fastafile != "") { m->mothurOut("Using " + fastafile + " as input file for the fasta parameter."); m->mothurOutEndLine(); }
 167                                 else {  m->mothurOut("You have no current fastafile and the fasta parameter is required."); m->mothurOutEndLine(); abort = true; }
 168                         }
 169                         else if (fastafile == "not open") { abort = true; }
 170                         else { m->setFastaFile(fastafile); }
 171
 172                         //if the user changes the output directory command factory will send this info to us in the output parameter
 173                         outputDir = validParameter.validFile(parameters, "outputdir", false);           if (outputDir == "not found"){
 174                                 outputDir = "";
 175                                 outputDir += m->hasPath(fastafile); //if user entered a file with a path then preserve it
 176                         }
 177
 178                         //check for optional parameter and set defaults
 179                         // ...at some point should added some additional type checking...
 180                         namefile = validParameter.validFile(parameters, "name", true);
 181                         if (namefile == "not found") { namefile =  "";  }
 182                         else if (namefile == "not open") { namefile = ""; abort = true; }
 183                         else {  m->setNameFile(namefile); }
 184
 185                         groupfile = validParameter.validFile(parameters, "group", true);
 186                         if (groupfile == "not found") { groupfile =  "";  bygroup = false; }
 187                         else if (groupfile == "not open") { abort = true; groupfile =  ""; }
 188                         else {   m->setGroupFile(groupfile); bygroup = true;  }
 189
 190             countfile = validParameter.validFile(parameters, "count", true);
 191                         if (countfile == "not found") { countfile =  "";   }
 192                         else if (countfile == "not open") { abort = true; countfile =  ""; }
 193                         else {
 194                 m->setCountTableFile(countfile);
 195                 ct.readTable(countfile, true);
 196                 if (ct.hasGroupInfo()) { bygroup = true; }
 197                 else { bygroup = false;  }
 198             }
 199
 200             if ((namefile != "") && (countfile != "")) {
 201                 m->mothurOut("[ERROR]: you may only use one of the following: name or count."); m->mothurOutEndLine(); abort = true;
 202             }
 203
 204             if ((groupfile != "") && (countfile != "")) {
 205                 m->mothurOut("[ERROR]: you may only use one of the following: group or count."); m->mothurOutEndLine(); abort=true;
 206             }
 207
 208
 209                         string temp     = validParameter.validFile(parameters, "diffs", false);         if(temp == "not found"){        temp = "1"; }
 210                         m->mothurConvert(temp, diffs);
 211
 212                         temp = validParameter.validFile(parameters, "processors", false);       if (temp == "not found"){       temp = m->getProcessors();      }
 213                         m->setProcessors(temp);
 214                         m->mothurConvert(temp, processors);
 215
 216             temp = validParameter.validFile(parameters, "topdown", false);              if(temp == "not found"){  temp = "T"; }
 217                         topdown = m->isTrue(temp);
 218
 219             if (countfile == "") {
 220                 if (namefile == "") {
 221                     vector<string> files; files.push_back(fastafile);
 222                     parser.getNameFile(files);
 223                 }
 224             }
 225                 }
 226
 227         }
 228         catch(exception& e) {
 229                 m->errorOut(e, "PreClusterCommand", "PreClusterCommand");
 230                 exit(1);
 231         }
 232 }
 233 //**********************************************************************************************************************
 234
 235 int PreClusterCommand::execute(){
 236         try {
 237
 238                 if (abort == true) { if (calledHelp) { return 0; }  return 2;   }
 239
 240                 int start = time(NULL);
 241
 242                 string fileroot = outputDir + m->getRootName(m->getSimpleName(fastafile));
 243         map<string, string> variables;
 244         variables["[filename]"] = fileroot;
 245                 string newNamesFile = getOutputFileName("name",variables);
 246         string newCountFile = getOutputFileName("count",variables);
 247                 string newMapFile = getOutputFileName("map",variables); //add group name if by group
 248         variables["[extension]"] = m->getExtension(fastafile);
 249                 string newFastaFile = getOutputFileName("fasta", variables);
 250                 outputNames.push_back(newFastaFile); outputTypes["fasta"].push_back(newFastaFile);
 251                 if (countfile == "") { outputNames.push_back(newNamesFile); outputTypes["name"].push_back(newNamesFile); }
 252                 else { outputNames.push_back(newCountFile); outputTypes["count"].push_back(newCountFile); }
 253
 254                 if (bygroup) {
 255                         //clear out old files
 256                         ofstream outFasta; m->openOutputFile(newFastaFile, outFasta); outFasta.close();
 257                         ofstream outNames; m->openOutputFile(newNamesFile, outNames);  outNames.close();
 258                         newMapFile = fileroot + "precluster.";
 259
 260                         //parse fasta and name file by group
 261             vector<string> groups;
 262                         if (countfile != "") {
 263                 cparser = new SequenceCountParser(countfile, fastafile);
 264                 groups = cparser->getNamesOfGroups();
 265             }else {
 266                 if (namefile != "") { parser = new SequenceParser(groupfile, fastafile, namefile);      }
 267                 else                            { parser = new SequenceParser(groupfile, fastafile);                    }
 268                 groups = parser->getNamesOfGroups();
 269                         }
 270
 271                         if(processors == 1)     {       driverGroups(newFastaFile, newNamesFile, newMapFile, 0, groups.size(), groups); }
 272                         else                            {       createProcessesGroups(newFastaFile, newNamesFile, newMapFile, groups);                  }
 273
 274                         if (countfile != "") {
 275                 mergeGroupCounts(newCountFile, newNamesFile, newFastaFile);
 276                 delete cparser;
 277             }else {
 278                 delete parser;
 279                 //run unique.seqs for deconvolute results
 280                 string inputString = "fasta=" + newFastaFile;
 281                 if (namefile != "") { inputString += ", name=" + newNamesFile; }
 282                 m->mothurOutEndLine();
 283                 m->mothurOut("/******************************************/"); m->mothurOutEndLine();
 284                 m->mothurOut("Running command: unique.seqs(" + inputString + ")"); m->mothurOutEndLine();
 285                 m->mothurCalling = true;
 286
 287                 Command* uniqueCommand = new DeconvoluteCommand(inputString);
 288                 uniqueCommand->execute();
 289
 290                 map<string, vector<string> > filenames = uniqueCommand->getOutputFiles();
 291
 292                 delete uniqueCommand;
 293                 m->mothurCalling = false;
 294                 m->mothurOut("/******************************************/"); m->mothurOutEndLine();
 295
 296                 m->renameFile(filenames["fasta"][0], newFastaFile);
 297                 m->renameFile(filenames["name"][0], newNamesFile);
 298                         }
 299             if (m->control_pressed) { for (int i = 0; i < outputNames.size(); i++) {    m->mothurRemove(outputNames[i]);        }        return 0; }
 300                         m->mothurOut("It took " + toString(time(NULL) - start) + " secs to run pre.cluster."); m->mothurOutEndLine();
 301
 302                 }else {
 303                         if (namefile != "") { readNameFile(); }
 304
 305                         //reads fasta file and return number of seqs
 306                         int numSeqs = readFASTA(); //fills alignSeqs and makes all seqs active
 307
 308                         if (m->control_pressed) { for (int i = 0; i < outputNames.size(); i++) {        m->mothurRemove(outputNames[i]);        } return 0; }
 309
 310                         if (numSeqs == 0) { m->mothurOut("Error reading fasta file...please correct."); m->mothurOutEndLine(); return 0;  }
 311                         if (diffs > length) { m->mothurOut("Error: diffs is greater than your sequence length."); m->mothurOutEndLine(); return 0;  }
 312
 313                         int count = process(newMapFile);
 314                         outputNames.push_back(newMapFile); outputTypes["map"].push_back(newMapFile);
 315
 316                         if (m->control_pressed) { for (int i = 0; i < outputNames.size(); i++) {        m->mothurRemove(outputNames[i]);        } return 0; }
 317
 318                         m->mothurOut("Total number of sequences before precluster was " + toString(alignSeqs.size()) + "."); m->mothurOutEndLine();
 319                         m->mothurOut("pre.cluster removed " + toString(count) + " sequences."); m->mothurOutEndLine(); m->mothurOutEndLine();
 320                         if (countfile != "") { newNamesFile = newCountFile; }
 321             printData(newFastaFile, newNamesFile, "");
 322
 323                         m->mothurOut("It took " + toString(time(NULL) - start) + " secs to cluster " + toString(numSeqs) + " sequences."); m->mothurOutEndLine();
 324                 }
 325
 326                 if (m->control_pressed) { for (int i = 0; i < outputNames.size(); i++) {        m->mothurRemove(outputNames[i]);        } return 0; }
 327
 328                 m->mothurOutEndLine();
 329                 m->mothurOut("Output File Names: "); m->mothurOutEndLine();
 330                 for (int i = 0; i < outputNames.size(); i++) {  m->mothurOut(outputNames[i]); m->mothurOutEndLine();    }
 331                 m->mothurOutEndLine();
 332
 333                 //set fasta file as new current fastafile
 334                 string current = "";
 335                 itTypes = outputTypes.find("fasta");
 336                 if (itTypes != outputTypes.end()) {
 337                         if ((itTypes->second).size() != 0) { current = (itTypes->second)[0]; m->setFastaFile(current); }
 338                 }
 339
 340                 itTypes = outputTypes.find("name");
 341                 if (itTypes != outputTypes.end()) {
 342                         if ((itTypes->second).size() != 0) { current = (itTypes->second)[0]; m->setNameFile(current); }
 343                 }
 344
 345         itTypes = outputTypes.find("count");
 346                 if (itTypes != outputTypes.end()) {
 347                         if ((itTypes->second).size() != 0) { current = (itTypes->second)[0]; m->setCountTableFile(current); }
 348                 }
 349
 350                 return 0;
 351
 352         }
 353         catch(exception& e) {
 354                 m->errorOut(e, "PreClusterCommand", "execute");
 355                 exit(1);
 356         }
 357 }
 358 /**************************************************************************************************/
 359 int PreClusterCommand::createProcessesGroups(string newFName, string newNName, string newMFile, vector<string> groups) {
 360         try {
 361
 362                 vector<int> processIDS;
 363                 int process = 1;
 364                 int num = 0;
 365
 366                 //sanity check
 367                 if (groups.size() < processors) { processors = groups.size(); }
 368
 369                 //divide the groups between the processors
 370                 vector<linePair> lines;
 371                 int numGroupsPerProcessor = groups.size() / processors;
 372                 for (int i = 0; i < processors; i++) {
 373                         int startIndex =  i * numGroupsPerProcessor;
 374                         int endIndex = (i+1) * numGroupsPerProcessor;
 375                         if(i == (processors - 1)){      endIndex = groups.size();       }
 376                         lines.push_back(linePair(startIndex, endIndex));
 377                 }
 378
 379 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) || (__linux__) || (__unix__) || (__unix)
 380
 381                 //loop through and create all the processes you want
 382                 while (process != processors) {
 383                         int pid = fork();
 384
 385                         if (pid > 0) {
 386                                 processIDS.push_back(pid);  //create map from line number to pid so you can append files in correct order later
 387                                 process++;
 388                         }else if (pid == 0){
 389                 outputNames.clear();
 390                                 num = driverGroups(newFName + toString(getpid()) + ".temp", newNName + toString(getpid()) + ".temp", newMFile, lines[process].start, lines[process].end, groups);
 391
 392                 string tempFile = toString(getpid()) + ".outputNames.temp";
 393                 ofstream outTemp;
 394                 m->openOutputFile(tempFile, outTemp);
 395
 396                 outTemp << outputNames.size();
 397                 for (int i = 0; i < outputNames.size(); i++) { outTemp << outputNames[i] << endl; }
 398                 outTemp.close();
 399
 400                                 exit(0);
 401                         }else {
 402                                 m->mothurOut("[ERROR]: unable to spawn the necessary processes."); m->mothurOutEndLine();
 403                                 for (int i = 0; i < processIDS.size(); i++) { kill (processIDS[i], SIGINT); }
 404                                 exit(0);
 405                         }
 406                 }
 407
 408                 //do my part
 409                 num = driverGroups(newFName, newNName, newMFile, lines[0].start, lines[0].end, groups);
 410
 411                 //force parent to wait until all the processes are done
 412                 for (int i=0;i<processIDS.size();i++) {
 413                         int temp = processIDS[i];
 414                         wait(&temp);
 415                 }
 416
 417         for (int i = 0; i < processIDS.size(); i++) {
 418             string tempFile = toString(processIDS[i]) +  ".outputNames.temp";
 419             ifstream intemp;
 420             m->openInputFile(tempFile, intemp);
 421
 422             int num;
 423             intemp >> num;
 424             for (int k = 0; k < num; k++) {
 425                 string name = "";
 426                 intemp >> name; m->gobble(intemp);
 427
 428                 outputNames.push_back(name); outputTypes["map"].push_back(name);
 429             }
 430             intemp.close();
 431             m->mothurRemove(tempFile);
 432         }
 433 #else
 434
 435                 //////////////////////////////////////////////////////////////////////////////////////////////////////
 436                 //Windows version shared memory, so be careful when passing variables through the preClusterData struct.
 437                 //Above fork() will clone, so memory is separate, but that's not the case with windows,
 438                 //////////////////////////////////////////////////////////////////////////////////////////////////////
 439
 440                 vector<preClusterData*> pDataArray;
 441                 DWORD   dwThreadIdArray[processors-1];
 442                 HANDLE  hThreadArray[processors-1];
 443
 444                 //Create processor worker threads.
 445                 for( int i=1; i<processors; i++ ){
 446                         // Allocate memory for thread data.
 447                         string extension = toString(i) + ".temp";
 448
 449                         preClusterData* tempPreCluster = new preClusterData(fastafile, namefile, groupfile, countfile, (newFName+extension), (newNName+extension), newMFile, groups, m, lines[i].start, lines[i].end, diffs, topdown, i);
 450                         pDataArray.push_back(tempPreCluster);
 451                         processIDS.push_back(i);
 452
 453                         //MySeqSumThreadFunction is in header. It must be global or static to work with the threads.
 454                         //default security attributes, thread function name, argument to thread function, use default creation flags, returns the thread identifier
 455                         hThreadArray[i-1] = CreateThread(NULL, 0, MyPreclusterThreadFunction, pDataArray[i-1], 0, &dwThreadIdArray[i-1]);
 456                 }
 457
 458
 459                 //using the main process as a worker saves time and memory
 460                 num = driverGroups(newFName, newNName, newMFile, lines[0].start, lines[0].end, groups);
 461
 462                 //Wait until all threads have terminated.
 463                 WaitForMultipleObjects(processors-1, hThreadArray, TRUE, INFINITE);
 464
 465                 //Close all thread handles and free memory allocations.
 466                 for(int i=0; i < pDataArray.size(); i++){
 467             if (pDataArray[i]->count != (pDataArray[i]->end-pDataArray[i]->start)) {
 468                 m->mothurOut("[ERROR]: process " + toString(i) + " only processed " + toString(pDataArray[i]->count) + " of " + toString(pDataArray[i]->end-pDataArray[i]->start) + " groups assigned to it, quitting. \n"); m->control_pressed = true;
 469             }
 470                         for (int j = 0; j < pDataArray[i]->mapFileNames.size(); j++) {
 471                                 outputNames.push_back(pDataArray[i]->mapFileNames[j]); outputTypes["map"].push_back(pDataArray[i]->mapFileNames[j]);
 472                         }
 473                         CloseHandle(hThreadArray[i]);
 474                         delete pDataArray[i];
 475                 }
 476
 477 #endif
 478
 479                 //append output files
 480                 for(int i=0;i<processIDS.size();i++){
 481                         //newFName = m->getFullPathName(".\\" + newFName);
 482                         //newNName = m->getFullPathName(".\\" + newNName);
 483
 484                         m->appendFiles((newFName + toString(processIDS[i]) + ".temp"), newFName);
 485                         m->mothurRemove((newFName + toString(processIDS[i]) + ".temp"));
 486
 487                         m->appendFiles((newNName + toString(processIDS[i]) + ".temp"), newNName);
 488                         m->mothurRemove((newNName + toString(processIDS[i]) + ".temp"));
 489                 }
 490
 491                 return num;
 492
 493         }
 494         catch(exception& e) {
 495                 m->errorOut(e, "PreClusterCommand", "createProcessesGroups");
 496                 exit(1);
 497         }
 498 }
 499 /**************************************************************************************************/
 500 int PreClusterCommand::driverGroups(string newFFile, string newNFile, string newMFile, int start, int end, vector<string> groups){
 501         try {
 502
 503                 int numSeqs = 0;
 504
 505                 //precluster each group
 506                 for (int i = start; i < end; i++) {
 507
 508                         start = time(NULL);
 509
 510                         if (m->control_pressed) {  return 0; }
 511
 512                         m->mothurOutEndLine(); m->mothurOut("Processing group " + groups[i] + ":"); m->mothurOutEndLine();
 513
 514                         map<string, string> thisNameMap;
 515             vector<Sequence> thisSeqs;
 516                         if (groupfile != "") {
 517                 thisSeqs = parser->getSeqs(groups[i]);
 518             }else if (countfile != "") {
 519                 thisSeqs = cparser->getSeqs(groups[i]);
 520             }
 521                         if (namefile != "") {  thisNameMap = parser->getNameMap(groups[i]); }
 522
 523                         //fill alignSeqs with this groups info.
 524                         numSeqs = loadSeqs(thisNameMap, thisSeqs, groups[i]);
 525
 526                         if (m->control_pressed) {   return 0; }
 527
 528                         if (diffs > length) { m->mothurOut("Error: diffs is greater than your sequence length."); m->mothurOutEndLine(); m->control_pressed = true; return 0;  }
 529
 530                         int count= process(newMFile+groups[i]+".map");
 531                         outputNames.push_back(newMFile+groups[i]+".map"); outputTypes["map"].push_back(newMFile+groups[i]+".map");
 532
 533                         if (m->control_pressed) {  return 0; }
 534
 535                         m->mothurOut("Total number of sequences before pre.cluster was " + toString(alignSeqs.size()) + "."); m->mothurOutEndLine();
 536                         m->mothurOut("pre.cluster removed " + toString(count) + " sequences."); m->mothurOutEndLine(); m->mothurOutEndLine();
 537                         printData(newFFile, newNFile, groups[i]);
 538
 539                         m->mothurOut("It took " + toString(time(NULL) - start) + " secs to cluster " + toString(numSeqs) + " sequences."); m->mothurOutEndLine();
 540
 541                 }
 542
 543                 return numSeqs;
 544         }
 545         catch(exception& e) {
 546                 m->errorOut(e, "PreClusterCommand", "driverGroups");
 547                 exit(1);
 548         }
 549 }
 550 /**************************************************************************************************/
 551 int PreClusterCommand::process(string newMapFile){
 552         try {
 553                 ofstream out;
 554                 m->openOutputFile(newMapFile, out);
 555
 556                 //sort seqs by number of identical seqs
 557         if (topdown) { sort(alignSeqs.begin(), alignSeqs.end(), comparePriorityTopDown);  }
 558         else {  sort(alignSeqs.begin(), alignSeqs.end(), comparePriorityDownTop);  }
 559
 560                 int count = 0;
 561                 int numSeqs = alignSeqs.size();
 562
 563         if (topdown) {
 564             //think about running through twice...
 565             for (int i = 0; i < numSeqs; i++) {
 566
 567                 if (alignSeqs[i].active) {  //this sequence has not been merged yet
 568
 569                     string chunk = alignSeqs[i].seq.getName() + "\t" + toString(alignSeqs[i].numIdentical) + "\t" + toString(0) + "\t" + alignSeqs[i].seq.getAligned() + "\n";
 570
 571                     //try to merge it with all smaller seqs
 572                     for (int j = i+1; j < numSeqs; j++) {
 573
 574                         if (m->control_pressed) { out.close(); return 0; }
 575
 576                         if (alignSeqs[j].active) {  //this sequence has not been merged yet
 577                             //are you within "diff" bases
 578                             int mismatch = calcMisMatches(alignSeqs[i].seq.getAligned(), alignSeqs[j].seq.getAligned());
 579
 580                             if (mismatch <= diffs) {
 581                                 //merge
 582                                 alignSeqs[i].names += ',' + alignSeqs[j].names;
 583                                 alignSeqs[i].numIdentical += alignSeqs[j].numIdentical;
 584
 585                                 chunk += alignSeqs[j].seq.getName() + "\t" + toString(alignSeqs[j].numIdentical) + "\t" + toString(mismatch) + "\t" + alignSeqs[j].seq.getAligned() + "\n";
 586
 587                                 alignSeqs[j].active = 0;
 588                                 alignSeqs[j].numIdentical = 0;
 589                                 count++;
 590                             }
 591                         }//end if j active
 592                     }//end for loop j
 593
 594                     //remove from active list
 595                     alignSeqs[i].active = 0;
 596
 597                     out << "ideal_seq_" << (i+1) << '\t' << alignSeqs[i].numIdentical << endl << chunk << endl;;
 598
 599                 }//end if active i
 600                 if(i % 100 == 0)        { m->mothurOutJustToScreen(toString(i) + "\t" + toString(numSeqs - count) + "\t" + toString(count)+"\n");       }
 601             }
 602         }else {
 603             map<int, string> mapFile;
 604             map<int, int> originalCount;
 605             map<int, int>::iterator itCount;
 606             for (int i = 0; i < numSeqs; i++) { mapFile[i] = ""; originalCount[i] = alignSeqs[i].numIdentical; }
 607
 608             //think about running through twice...
 609             for (int i = 0; i < numSeqs; i++) {
 610
 611                 //try to merge it into larger seqs
 612                 for (int j = i+1; j < numSeqs; j++) {
 613
 614                     if (m->control_pressed) { out.close(); return 0; }
 615
 616                     if (originalCount[j] > originalCount[i]) {  //this sequence is more abundant than I am
 617                         //are you within "diff" bases
 618                         int mismatch = calcMisMatches(alignSeqs[i].seq.getAligned(), alignSeqs[j].seq.getAligned());
 619
 620                         if (mismatch <= diffs) {
 621                             //merge
 622                             alignSeqs[j].names += ',' + alignSeqs[i].names;
 623                             alignSeqs[j].numIdentical += alignSeqs[i].numIdentical;
 624
 625                             mapFile[j] = alignSeqs[i].seq.getName() + "\t" + toString(alignSeqs[i].numIdentical) + "\t" + toString(mismatch) + "\t" + alignSeqs[i].seq.getAligned() + "\n" + mapFile[i];
 626                             alignSeqs[i].numIdentical = 0;
 627                             originalCount.erase(i);
 628                             mapFile[i] = "";
 629                             count++;
 630                             j+=numSeqs; //exit search, we merged this one in.
 631                         }
 632                     }//end abundance check
 633                 }//end for loop j
 634
 635                 if(i % 100 == 0)        { m->mothurOutJustToScreen(toString(i) + "\t" + toString(numSeqs - count) + "\t" + toString(count)+"\n");       }
 636             }
 637
 638             for (int i = 0; i < numSeqs; i++) {
 639                 if (alignSeqs[i].numIdentical != 0) {
 640                     out << "ideal_seq_" << (i+1) << '\t' << alignSeqs[i].numIdentical << endl  << alignSeqs[i].seq.getName() + "\t" + toString(alignSeqs[i].numIdentical) + "\t" + toString(0) + "\t" + alignSeqs[i].seq.getAligned() + "\n" << mapFile[i] << endl;
 641                 }
 642             }
 643
 644         }
 645                 out.close();
 646
 647                 if(numSeqs % 100 != 0)  { m->mothurOut(toString(numSeqs) + "\t" + toString(numSeqs - count) + "\t" + toString(count)); m->mothurOutEndLine();   }
 648
 649                 return count;
 650
 651         }
 652         catch(exception& e) {
 653                 m->errorOut(e, "PreClusterCommand", "process");
 654                 exit(1);
 655         }
 656 }
 657 /**************************************************************************************************/
 658 int PreClusterCommand::readFASTA(){
 659         try {
 660                 //ifstream inNames;
 661                 ifstream inFasta;
 662
 663                 m->openInputFile(fastafile, inFasta);
 664                 set<int> lengths;
 665
 666                 while (!inFasta.eof()) {
 667
 668                         if (m->control_pressed) { inFasta.close(); return 0; }
 669
 670                         Sequence seq(inFasta);  m->gobble(inFasta);
 671
 672                         if (seq.getName() != "") {  //can get "" if commented line is at end of fasta file
 673                                 if (namefile != "") {
 674                                         itSize = sizes.find(seq.getName());
 675
 676                                         if (itSize == sizes.end()) { m->mothurOut(seq.getName() + " is not in your names file, please correct."); m->mothurOutEndLine(); exit(1); }
 677                                         else{
 678                                                 seqPNode tempNode(itSize->second, seq, names[seq.getName()]);
 679                                                 alignSeqs.push_back(tempNode);
 680                                                 lengths.insert(seq.getAligned().length());
 681                                         }
 682                                 }else { //no names file, you are identical to yourself
 683                     int numRep = 1;
 684                     if (countfile != "") { numRep = ct.getNumSeqs(seq.getName()); }
 685                                         seqPNode tempNode(numRep, seq, seq.getName());
 686                                         alignSeqs.push_back(tempNode);
 687                                         lengths.insert(seq.getAligned().length());
 688                                 }
 689                         }
 690                 }
 691                 inFasta.close();
 692
 693         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(); }
 694         else if (lengths.size() == 1) { length = *(lengths.begin()); }
 695
 696                 return alignSeqs.size();
 697         }
 698
 699         catch(exception& e) {
 700                 m->errorOut(e, "PreClusterCommand", "readFASTA");
 701                 exit(1);
 702         }
 703 }
 704 /**************************************************************************************************/
 705 int PreClusterCommand::loadSeqs(map<string, string>& thisName, vector<Sequence>& thisSeqs, string group){
 706         try {
 707                 set<int> lengths;
 708                 alignSeqs.clear();
 709                 map<string, string>::iterator it;
 710                 bool error = false;
 711         map<string, int> thisCount;
 712         if (countfile != "") { thisCount = cparser->getCountTable(group);  }
 713
 714                 for (int i = 0; i < thisSeqs.size(); i++) {
 715
 716                         if (m->control_pressed) { return 0; }
 717
 718                         if (namefile != "") {
 719                                 it = thisName.find(thisSeqs[i].getName());
 720
 721                                 //should never be true since parser checks for this
 722                                 if (it == thisName.end()) { m->mothurOut(thisSeqs[i].getName() + " is not in your names file, please correct."); m->mothurOutEndLine(); error = true; }
 723                                 else{
 724                                         //get number of reps
 725                                         int numReps = 1;
 726                                         for(int j=0;j<(it->second).length();j++){
 727                                                 if((it->second)[j] == ','){     numReps++;      }
 728                                         }
 729
 730                                         seqPNode tempNode(numReps, thisSeqs[i], it->second);
 731                                         alignSeqs.push_back(tempNode);
 732                     lengths.insert(thisSeqs[i].getAligned().length());
 733                                 }
 734                         }else { //no names file, you are identical to yourself
 735                 int numRep = 1;
 736                 if (countfile != "") {
 737                     map<string, int>::iterator it2 = thisCount.find(thisSeqs[i].getName());
 738
 739                     //should never be true since parser checks for this
 740                     if (it2 == thisCount.end()) { m->mothurOut(thisSeqs[i].getName() + " is not in your count file, please correct."); m->mothurOutEndLine(); error = true; }
 741                     else { numRep = it2->second;  }
 742                 }
 743                                 seqPNode tempNode(numRep, thisSeqs[i], thisSeqs[i].getName());
 744                                 alignSeqs.push_back(tempNode);
 745                                 lengths.insert(thisSeqs[i].getAligned().length());
 746                         }
 747                 }
 748
 749         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(); }
 750         else if (lengths.size() == 1) { length = *(lengths.begin()); }
 751
 752                 //sanity check
 753                 if (error) { m->control_pressed = true; }
 754
 755                 thisSeqs.clear();
 756
 757                 return alignSeqs.size();
 758         }
 759
 760         catch(exception& e) {
 761                 m->errorOut(e, "PreClusterCommand", "loadSeqs");
 762                 exit(1);
 763         }
 764 }
 765
 766 /**************************************************************************************************/
 767
 768 int PreClusterCommand::calcMisMatches(string seq1, string seq2){
 769         try {
 770                 int numBad = 0;
 771
 772                 for (int i = 0; i < seq1.length(); i++) {
 773                         //do they match
 774                         if (seq1[i] != seq2[i]) { numBad++; }
 775                         if (numBad > diffs) { return length;  } //to far to cluster
 776                 }
 777
 778                 return numBad;
 779         }
 780         catch(exception& e) {
 781                 m->errorOut(e, "PreClusterCommand", "calcMisMatches");
 782                 exit(1);
 783         }
 784 }
 785 /**************************************************************************************************/
 786
 787 int PreClusterCommand::mergeGroupCounts(string newcount, string newname, string newfasta){
 788         try {
 789                 ifstream inNames;
 790         m->openInputFile(newname, inNames);
 791
 792         string group, first, second;
 793         set<string> uniqueNames;
 794         while (!inNames.eof()) {
 795             if (m->control_pressed) { break; }
 796             inNames >> group; m->gobble(inNames);
 797             inNames >> first; m->gobble(inNames);
 798             inNames >> second; m->gobble(inNames);
 799
 800             vector<string> names;
 801             m->splitAtComma(second, names);
 802
 803             uniqueNames.insert(first);
 804
 805             int total = ct.getGroupCount(first, group);
 806             for (int i = 1; i < names.size(); i++) {
 807                 total += ct.getGroupCount(names[i], group);
 808                 ct.setAbund(names[i], group, 0);
 809             }
 810             ct.setAbund(first, group, total);
 811         }
 812         inNames.close();
 813
 814         vector<string> namesOfSeqs = ct.getNamesOfSeqs();
 815         for (int i = 0; i < namesOfSeqs.size(); i++) {
 816             if (ct.getNumSeqs(namesOfSeqs[i]) == 0) {
 817                 ct.remove(namesOfSeqs[i]);
 818             }
 819         }
 820
 821         ct.printTable(newcount);
 822         m->mothurRemove(newname);
 823
 824         if (bygroup) { //if by group, must remove the duplicate seqs that are named the same
 825             ifstream in;
 826             m->openInputFile(newfasta, in);
 827
 828             ofstream out;
 829             m->openOutputFile(newfasta+"temp", out);
 830
 831             int count = 0;
 832             set<string> already;
 833             while(!in.eof()) {
 834                 if (m->control_pressed) { break; }
 835
 836                 Sequence seq(in); m->gobble(in);
 837
 838                 if (seq.getName() != "") {
 839                     count++;
 840                     if (already.count(seq.getName()) == 0) {
 841                         seq.printSequence(out);
 842                         already.insert(seq.getName());
 843                     }
 844                 }
 845             }
 846             in.close();
 847             out.close();
 848             m->mothurRemove(newfasta);
 849             m->renameFile(newfasta+"temp", newfasta);
 850         }
 851                         return 0;
 852
 853         }
 854         catch(exception& e) {
 855                 m->errorOut(e, "PreClusterCommand", "mergeGroupCounts");
 856                 exit(1);
 857         }
 858 }
 859
 860 /**************************************************************************************************/
 861
 862 void PreClusterCommand::printData(string newfasta, string newname, string group){
 863         try {
 864                 ofstream outFasta;
 865                 ofstream outNames;
 866
 867                 if (bygroup) {
 868                         m->openOutputFileAppend(newfasta, outFasta);
 869                         m->openOutputFileAppend(newname, outNames);
 870                 }else {
 871                         m->openOutputFile(newfasta, outFasta);
 872                         m->openOutputFile(newname, outNames);
 873                 }
 874
 875         if ((countfile != "") && (group == ""))  { outNames << "Representative_Sequence\ttotal\n";  }
 876                 for (int i = 0; i < alignSeqs.size(); i++) {
 877                         if (alignSeqs[i].numIdentical != 0) {
 878                                 alignSeqs[i].seq.printSequence(outFasta);
 879                                 if (countfile != "") {
 880                     if (group != "") {  outNames << group << '\t' << alignSeqs[i].seq.getName() << '\t' << alignSeqs[i].names << endl; }
 881                     else {  outNames << alignSeqs[i].seq.getName() << '\t' << alignSeqs[i].numIdentical << endl;  }
 882                 }else {  outNames << alignSeqs[i].seq.getName() << '\t' << alignSeqs[i].names << endl;  }
 883                         }
 884                 }
 885
 886                 outFasta.close();
 887                 outNames.close();
 888
 889         }
 890         catch(exception& e) {
 891                 m->errorOut(e, "PreClusterCommand", "printData");
 892                 exit(1);
 893         }
 894 }
 895 /**************************************************************************************************/
 896
 897 void PreClusterCommand::readNameFile(){
 898         try {
 899                 ifstream in;
 900                 m->openInputFile(namefile, in);
 901                 string firstCol, secondCol;
 902
 903                 while (!in.eof()) {
 904                         in >> firstCol >> secondCol; m->gobble(in);
 905
 906             m->checkName(firstCol);
 907             m->checkName(secondCol);
 908             int size = m->getNumNames(secondCol);
 909
 910                         names[firstCol] = secondCol;
 911             sizes[firstCol] = size;
 912                 }
 913                 in.close();
 914         }
 915         catch(exception& e) {
 916                 m->errorOut(e, "PreClusterCommand", "readNameFile");
 917                 exit(1);
 918         }
 919 }
 920
 921 /**************************************************************************************************/
 922
 923