classifyseqscommand.cpp

   1 /*
   2  *  classifyseqscommand.cpp
   3  *  Mothur
   4  *
   5  *  Created by westcott on 11/2/09.
   6  *  Copyright 2009 Schloss Lab. All rights reserved.
   7  *
   8  */
   9
  10 #include "classifyseqscommand.h"
  11 #include "sequence.hpp"
  12 #include "bayesian.h"
  13 #include "phylotree.h"
  14 #include "phylosummary.h"
  15 #include "knn.h"
  16
  17 //**********************************************************************************************************************
  18
  19 ClassifySeqsCommand::ClassifySeqsCommand(string option)  {
  20         try {
  21                 abort = false;
  22
  23                 //allow user to run help
  24                 if(option == "help") { help(); abort = true; }
  25
  26                 else {
  27
  28                         //valid paramters for this command
  29                         string AlignArray[] =  {"template","fasta","name","group","search","ksize","method","processors","taxonomy","match","mismatch","gapopen","gapextend","numwanted","cutoff","probs","iters", "outputdir","inputdir"};
  30                         vector<string> myArray (AlignArray, AlignArray+(sizeof(AlignArray)/sizeof(string)));
  31
  32                         OptionParser parser(option);
  33                         map<string, string> parameters = parser.getParameters();
  34
  35                         ValidParameters validParameter("classify.seqs");
  36                         map<string, string>::iterator it;
  37
  38                         //check to make sure all parameters are valid for command
  39                         for (it = parameters.begin(); it != parameters.end(); it++) {
  40                                 if (validParameter.isValidParameter(it->first, myArray, it->second) != true) {  abort = true;  }
  41                         }
  42
  43                         //if the user changes the output directory command factory will send this info to us in the output parameter
  44                         outputDir = validParameter.validFile(parameters, "outputdir", false);           if (outputDir == "not found"){  outputDir = "";         }
  45
  46                         //if the user changes the input directory command factory will send this info to us in the output parameter
  47                         string inputDir = validParameter.validFile(parameters, "inputdir", false);
  48                         if (inputDir == "not found"){   inputDir = "";          }
  49                         else {
  50                                 string path;
  51                                 it = parameters.find("template");
  52                                 //user has given a template file
  53                                 if(it != parameters.end()){
  54                                         path = m->hasPath(it->second);
  55                                         //if the user has not given a path then, add inputdir. else leave path alone.
  56                                         if (path == "") {       parameters["template"] = inputDir + it->second;         }
  57                                 }
  58
  59                                 it = parameters.find("taxonomy");
  60                                 //user has given a template file
  61                                 if(it != parameters.end()){
  62                                         path = m->hasPath(it->second);
  63                                         //if the user has not given a path then, add inputdir. else leave path alone.
  64                                         if (path == "") {       parameters["taxonomy"] = inputDir + it->second;         }
  65                                 }
  66
  67                                 it = parameters.find("group");
  68                                 //user has given a template file
  69                                 if(it != parameters.end()){
  70                                         path = m->hasPath(it->second);
  71                                         //if the user has not given a path then, add inputdir. else leave path alone.
  72                                         if (path == "") {       parameters["group"] = inputDir + it->second;            }
  73                                 }
  74                         }
  75
  76                         //check for required parameters
  77                         templateFileName = validParameter.validFile(parameters, "template", true);
  78                         if (templateFileName == "not found") {
  79                                 m->mothurOut("template is a required parameter for the classify.seqs command.");
  80                                 m->mothurOutEndLine();
  81                                 abort = true;
  82                         }
  83                         else if (templateFileName == "not open") { abort = true; }
  84
  85
  86                         fastaFileName = validParameter.validFile(parameters, "fasta", false);
  87                         if (fastaFileName == "not found") { m->mothurOut("fasta is a required parameter for the classify.seqs command."); m->mothurOutEndLine(); abort = true;  }
  88                         else {
  89                                 m->splitAtDash(fastaFileName, fastaFileNames);
  90
  91                                 //go through files and make sure they are good, if not, then disregard them
  92                                 for (int i = 0; i < fastaFileNames.size(); i++) {
  93                                         if (inputDir != "") {
  94                                                 string path = m->hasPath(fastaFileNames[i]);
  95                                                 //if the user has not given a path then, add inputdir. else leave path alone.
  96                                                 if (path == "") {       fastaFileNames[i] = inputDir + fastaFileNames[i];               }
  97                                         }
  98
  99                                         int ableToOpen;
 100
 101                                         ifstream in;
 102                                         ableToOpen = m->openInputFile(fastaFileNames[i], in, "noerror");
 103
 104                                         //if you can't open it, try default location
 105                                         if (ableToOpen == 1) {
 106                                                 if (m->getDefaultPath() != "") { //default path is set
 107                                                         string tryPath = m->getDefaultPath() + m->getSimpleName(fastaFileNames[i]);
 108                                                         m->mothurOut("Unable to open " + fastaFileNames[i] + ". Trying default " + tryPath); m->mothurOutEndLine();
 109                                                         ableToOpen = m->openInputFile(tryPath, in, "noerror");
 110                                                         fastaFileNames[i] = tryPath;
 111                                                 }
 112                                         }
 113                                         in.close();
 114
 115                                         if (ableToOpen == 1) {
 116                                                 m->mothurOut("Unable to open " + fastaFileNames[i] + ". It will be disregarded."); m->mothurOutEndLine();
 117                                                 //erase from file list
 118                                                 fastaFileNames.erase(fastaFileNames.begin()+i);
 119                                                 i--;
 120                                         }
 121
 122                                 }
 123
 124                                 //make sure there is at least one valid file left
 125                                 if (fastaFileNames.size() == 0) { m->mothurOut("no valid files."); m->mothurOutEndLine(); abort = true; }
 126                         }
 127
 128
 129                         taxonomyFileName = validParameter.validFile(parameters, "taxonomy", true);
 130                         if (taxonomyFileName == "not found") {
 131                                 m->mothurOut("taxonomy is a required parameter for the classify.seqs command.");
 132                                 m->mothurOutEndLine();
 133                                 abort = true;
 134                         }
 135                         else if (taxonomyFileName == "not open") { abort = true; }
 136
 137
 138                         namefile = validParameter.validFile(parameters, "name", false);
 139                         if (namefile == "not found") { namefile = "";  }
 140
 141                         else {
 142                                 m->splitAtDash(namefile, namefileNames);
 143
 144                                 //go through files and make sure they are good, if not, then disregard them
 145                                 for (int i = 0; i < namefileNames.size(); i++) {
 146                                         if (inputDir != "") {
 147                                                 string path = m->hasPath(namefileNames[i]);
 148                                                 //if the user has not given a path then, add inputdir. else leave path alone.
 149                                                 if (path == "") {       namefileNames[i] = inputDir + namefileNames[i];         }
 150                                         }
 151                                         int ableToOpen;
 152
 153                                         ifstream in;
 154                                         ableToOpen = m->openInputFile(namefileNames[i], in, "noerror");
 155
 156                                         //if you can't open it, try default location
 157                                         if (ableToOpen == 1) {
 158                                                 if (m->getDefaultPath() != "") { //default path is set
 159                                                         string tryPath = m->getDefaultPath() + m->getSimpleName(namefileNames[i]);
 160                                                         m->mothurOut("Unable to open " + namefileNames[i] + ". Trying default " + tryPath); m->mothurOutEndLine();
 161                                                         ableToOpen = m->openInputFile(tryPath, in, "noerror");
 162                                                         namefileNames[i] = tryPath;
 163                                                 }
 164                                         }
 165                                         in.close();
 166
 167                                         if (ableToOpen == 1) {
 168                                                 m->mothurOut("Unable to open " + namefileNames[i] + ". It will be disregarded."); m->mothurOutEndLine();  abort = true;
 169                                                 //erase from file list
 170                                                 namefileNames.erase(namefileNames.begin()+i);
 171                                                 i--;
 172                                         }
 173
 174                                 }
 175                         }
 176
 177                         if (namefile != "") {
 178                                 if (namefileNames.size() != fastaFileNames.size()) { abort = true; m->mothurOut("If you provide a name file, you must have one for each fasta file."); m->mothurOutEndLine(); }
 179                         }
 180
 181                         groupfile = validParameter.validFile(parameters, "group", false);
 182                         if (groupfile == "not found") { groupfile = "";  }
 183                         else {
 184                                 m->splitAtDash(groupfile, groupfileNames);
 185
 186                                 //go through files and make sure they are good, if not, then disregard them
 187                                 for (int i = 0; i < groupfileNames.size(); i++) {
 188                                         if (inputDir != "") {
 189                                                 string path = m->hasPath(groupfileNames[i]);
 190                                                 //if the user has not given a path then, add inputdir. else leave path alone.
 191                                                 if (path == "") {       groupfileNames[i] = inputDir + groupfileNames[i];               }
 192                                         }
 193                                         int ableToOpen;
 194
 195                                         ifstream in;
 196                                         ableToOpen = m->openInputFile(groupfileNames[i], in, "noerror");
 197
 198                                         //if you can't open it, try default location
 199                                         if (ableToOpen == 1) {
 200                                                 if (m->getDefaultPath() != "") { //default path is set
 201                                                         string tryPath = m->getDefaultPath() + m->getSimpleName(groupfileNames[i]);
 202                                                         m->mothurOut("Unable to open " + groupfileNames[i] + ". Trying default " + tryPath); m->mothurOutEndLine();
 203                                                         ableToOpen = m->openInputFile(tryPath, in, "noerror");
 204                                                         groupfileNames[i] = tryPath;
 205                                                 }
 206                                         }
 207                                         in.close();
 208
 209                                         if (ableToOpen == 1) {
 210                                                 m->mothurOut("Unable to open " + groupfileNames[i] + ". It will be disregarded."); m->mothurOutEndLine(); groupfileNames[i] = "";
 211                                                 //erase from file list
 212                                                 groupfileNames.erase(groupfileNames.begin()+i);
 213                                                 i--;
 214                                         }
 215                                 }
 216                         }
 217
 218                         if (groupfile != "") {
 219                                 if (groupfileNames.size() != fastaFileNames.size()) { abort = true; m->mothurOut("If you provide a group file, you must have one for each fasta file."); m->mothurOutEndLine(); }
 220                         }else {
 221                                 for (int i = 0; i < fastaFileNames.size(); i++) {  groupfileNames.push_back("");  }
 222                         }
 223
 224                         //check for optional parameter and set defaults
 225                         // ...at some point should added some additional type checking...
 226                         string temp;
 227                         temp = validParameter.validFile(parameters, "ksize", false);            if (temp == "not found"){       temp = "8";                             }
 228                         convert(temp, kmerSize);
 229
 230                         temp = validParameter.validFile(parameters, "processors", false);       if (temp == "not found"){       temp = "1";                             }
 231                         convert(temp, processors);
 232
 233                         search = validParameter.validFile(parameters, "search", false);         if (search == "not found"){     search = "kmer";                }
 234
 235                         method = validParameter.validFile(parameters, "method", false);         if (method == "not found"){     method = "bayesian";    }
 236
 237                         temp = validParameter.validFile(parameters, "match", false);            if (temp == "not found"){       temp = "1.0";                   }
 238                         convert(temp, match);
 239
 240                         temp = validParameter.validFile(parameters, "mismatch", false);         if (temp == "not found"){       temp = "-1.0";                  }
 241                         convert(temp, misMatch);
 242
 243                         temp = validParameter.validFile(parameters, "gapopen", false);          if (temp == "not found"){       temp = "-2.0";                  }
 244                         convert(temp, gapOpen);
 245
 246                         temp = validParameter.validFile(parameters, "gapextend", false);        if (temp == "not found"){       temp = "-1.0";                  }
 247                         convert(temp, gapExtend);
 248
 249                         temp = validParameter.validFile(parameters, "numwanted", false);        if (temp == "not found"){       temp = "10";                    }
 250                         convert(temp, numWanted);
 251
 252                         temp = validParameter.validFile(parameters, "cutoff", false);           if (temp == "not found"){       temp = "0";                             }
 253                         convert(temp, cutoff);
 254
 255                         temp = validParameter.validFile(parameters, "probs", false);            if (temp == "not found"){       temp = "true";                  }
 256                         probs = m->isTrue(temp);
 257
 258                         temp = validParameter.validFile(parameters, "iters", false);            if (temp == "not found") { temp = "100";                        }
 259                         convert(temp, iters);
 260
 261
 262
 263                         if ((method == "bayesian") && (search != "kmer"))  {
 264                                 m->mothurOut("The bayesian method requires the kmer search." + search + "will be disregarded." ); m->mothurOutEndLine();
 265                                 search = "kmer";
 266                         }
 267                 }
 268
 269         }
 270         catch(exception& e) {
 271                 m->errorOut(e, "ClassifySeqsCommand", "ClassifySeqsCommand");
 272                 exit(1);
 273         }
 274 }
 275
 276 //**********************************************************************************************************************
 277
 278 ClassifySeqsCommand::~ClassifySeqsCommand(){
 279
 280         if (abort == false) {
 281                 for (int i = 0; i < lines.size(); i++) {  delete lines[i];  }  lines.clear();
 282         }
 283 }
 284
 285 //**********************************************************************************************************************
 286
 287 void ClassifySeqsCommand::help(){
 288         try {
 289                 m->mothurOut("The classify.seqs command reads a fasta file containing sequences and creates a .taxonomy file and a .tax.summary file.\n");
 290                 m->mothurOut("The classify.seqs command parameters are template, fasta, name, search, ksize, method, taxonomy, processors, match, mismatch, gapopen, gapextend, numwanted and probs.\n");
 291                 m->mothurOut("The template, fasta and taxonomy parameters are required. You may enter multiple fasta files by separating their names with dashes. ie. fasta=abrecovery.fasta-amzon.fasta \n");
 292                 m->mothurOut("The search parameter allows you to specify the method to find most similar template.  Your options are: suffix, kmer, blast and distance. The default is kmer.\n");
 293                 m->mothurOut("The name parameter allows you add a names file with your fasta file, if you enter multiple fasta files, you must enter matching names files for them.\n");
 294                 m->mothurOut("The group parameter allows you add a group file so you can have the summary totals broken up by group.\n");
 295                 m->mothurOut("The method parameter allows you to specify classification method to use.  Your options are: bayesian and knn. The default is bayesian.\n");
 296                 m->mothurOut("The ksize parameter allows you to specify the kmer size for finding most similar template to candidate.  The default is 8.\n");
 297                 m->mothurOut("The processors parameter allows you to specify the number of processors to use. The default is 1.\n");
 298                 #ifdef USE_MPI
 299                 m->mothurOut("When using MPI, the processors parameter is set to the number of MPI processes running. \n");
 300                 #endif
 301                 m->mothurOut("The match parameter allows you to specify the bonus for having the same base. The default is 1.0.\n");
 302                 m->mothurOut("The mistmatch parameter allows you to specify the penalty for having different bases.  The default is -1.0.\n");
 303                 m->mothurOut("The gapopen parameter allows you to specify the penalty for opening a gap in an alignment. The default is -2.0.\n");
 304                 m->mothurOut("The gapextend parameter allows you to specify the penalty for extending a gap in an alignment.  The default is -1.0.\n");
 305                 m->mothurOut("The numwanted parameter allows you to specify the number of sequence matches you want with the knn method.  The default is 10.\n");
 306                 m->mothurOut("The cutoff parameter allows you to specify a bootstrap confidence threshold for your taxonomy.  The default is 0.\n");
 307                 m->mothurOut("The probs parameter shuts off the bootstrapping results for the bayesian method. The default is true, meaning you want the bootstrapping to be shown.\n");
 308                 m->mothurOut("The iters parameter allows you to specify how many iterations to do when calculating the bootstrap confidence score for your taxonomy with the bayesian method.  The default is 100.\n");
 309                 m->mothurOut("The classify.seqs command should be in the following format: \n");
 310                 m->mothurOut("classify.seqs(template=yourTemplateFile, fasta=yourFastaFile, method=yourClassificationMethod, search=yourSearchmethod, ksize=yourKmerSize, taxonomy=yourTaxonomyFile, processors=yourProcessors) \n");
 311                 m->mothurOut("Example classify.seqs(fasta=amazon.fasta, template=core.filtered, method=knn, search=gotoh, ksize=8, processors=2)\n");
 312                 m->mothurOut("The .taxonomy file consists of 2 columns: 1 = your sequence name, 2 = the taxonomy for your sequence. \n");
 313                 m->mothurOut("The .tax.summary is a summary of the different taxonomies represented in your fasta file. \n");
 314                 m->mothurOut("Note: No spaces between parameter labels (i.e. fasta), '=' and parameters (i.e.yourFastaFile).\n\n");
 315         }
 316         catch(exception& e) {
 317                 m->errorOut(e, "ClassifySeqsCommand", "help");
 318                 exit(1);
 319         }
 320 }
 321
 322
 323 //**********************************************************************************************************************
 324
 325 int ClassifySeqsCommand::execute(){
 326         try {
 327                 if (abort == true) {    return 0;       }
 328
 329                 if(method == "bayesian"){       classify = new Bayesian(taxonomyFileName, templateFileName, search, kmerSize, cutoff, iters);           }
 330                 else if(method == "knn"){       classify = new Knn(taxonomyFileName, templateFileName, search, kmerSize, gapOpen, gapExtend, match, misMatch, numWanted);                               }
 331                 else {
 332                         m->mothurOut(search + " is not a valid method option. I will run the command using bayesian.");
 333                         m->mothurOutEndLine();
 334                         classify = new Bayesian(taxonomyFileName, templateFileName, search, kmerSize, cutoff, iters);
 335                 }
 336
 337                 if (m->control_pressed) { delete classify; return 0; }
 338
 339                 vector<string> outputNames;
 340
 341                 for (int s = 0; s < fastaFileNames.size(); s++) {
 342
 343                         m->mothurOut("Classifying sequences from " + fastaFileNames[s] + " ..." ); m->mothurOutEndLine();
 344
 345                         string RippedTaxName = m->getRootName(m->getSimpleName(taxonomyFileName));
 346                         RippedTaxName = m->getExtension(RippedTaxName.substr(0, RippedTaxName.length()-1));
 347                         if (RippedTaxName[0] == '.') { RippedTaxName = RippedTaxName.substr(1, RippedTaxName.length()); }
 348                         RippedTaxName +=  ".";
 349
 350                         if (outputDir == "") { outputDir += m->hasPath(fastaFileNames[s]); }
 351                         string newTaxonomyFile = outputDir + m->getRootName(m->getSimpleName(fastaFileNames[s])) + RippedTaxName + "taxonomy";
 352                         string tempTaxonomyFile = outputDir + m->getRootName(m->getSimpleName(fastaFileNames[s])) + "taxonomy.temp";
 353                         string taxSummary = outputDir + m->getRootName(m->getSimpleName(fastaFileNames[s])) + RippedTaxName + "tax.summary";
 354
 355                         if ((method == "knn") && (search == "distance")) {
 356                                 string DistName = outputDir + m->getRootName(m->getSimpleName(fastaFileNames[s])) + "match.dist";
 357                                 classify->setDistName(DistName);  outputNames.push_back(DistName);
 358                         }
 359
 360                         outputNames.push_back(newTaxonomyFile);
 361                         outputNames.push_back(taxSummary);
 362
 363                         int start = time(NULL);
 364                         int numFastaSeqs = 0;
 365                         for (int i = 0; i < lines.size(); i++) {  delete lines[i];  }  lines.clear();
 366
 367 #ifdef USE_MPI
 368                                 int pid, end, numSeqsPerProcessor;
 369                                 int tag = 2001;
 370                                 vector<unsigned long int> MPIPos;
 371
 372                                 MPI_Status status;
 373                                 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 374                                 MPI_Comm_size(MPI_COMM_WORLD, &processors);
 375
 376                                 MPI_File inMPI;
 377                                 MPI_File outMPINewTax;
 378                                 MPI_File outMPITempTax;
 379
 380                                 int outMode=MPI_MODE_CREATE|MPI_MODE_WRONLY;
 381                                 int inMode=MPI_MODE_RDONLY;
 382
 383                                 //char* outNewTax = new char[newTaxonomyFile.length()];
 384                                 //memcpy(outNewTax, newTaxonomyFile.c_str(), newTaxonomyFile.length());
 385
 386                                 char outNewTax[1024];
 387                                 strcpy(outNewTax, newTaxonomyFile.c_str());
 388
 389                                 //char* outTempTax = new char[tempTaxonomyFile.length()];
 390                                 //memcpy(outTempTax, tempTaxonomyFile.c_str(), tempTaxonomyFile.length());
 391
 392                                 char outTempTax[1024];
 393                                 strcpy(outTempTax, tempTaxonomyFile.c_str());
 394
 395                                 //char* inFileName = new char[fastaFileNames[s].length()];
 396                                 //memcpy(inFileName, fastaFileNames[s].c_str(), fastaFileNames[s].length());
 397
 398                                 char inFileName[1024];
 399                                 strcpy(inFileName, fastaFileNames[s].c_str());
 400
 401                                 MPI_File_open(MPI_COMM_WORLD, inFileName, inMode, MPI_INFO_NULL, &inMPI);  //comm, filename, mode, info, filepointer
 402                                 MPI_File_open(MPI_COMM_WORLD, outNewTax, outMode, MPI_INFO_NULL, &outMPINewTax);
 403                                 MPI_File_open(MPI_COMM_WORLD, outTempTax, outMode, MPI_INFO_NULL, &outMPITempTax);
 404
 405                                 //delete outNewTax;
 406                                 //delete outTempTax;
 407                                 //delete inFileName;
 408
 409                                 if (m->control_pressed) {  MPI_File_close(&inMPI);  MPI_File_close(&outMPINewTax);   MPI_File_close(&outMPITempTax);  delete classify; return 0;  }
 410
 411                                 if (pid == 0) { //you are the root process
 412
 413                                         MPIPos = m->setFilePosFasta(fastaFileNames[s], numFastaSeqs); //fills MPIPos, returns numSeqs
 414
 415                                         //send file positions to all processes
 416                                         for(int i = 1; i < processors; i++) {
 417                                                 MPI_Send(&numFastaSeqs, 1, MPI_INT, i, tag, MPI_COMM_WORLD);
 418                                                 MPI_Send(&MPIPos[0], (numFastaSeqs+1), MPI_LONG, i, tag, MPI_COMM_WORLD);
 419                                         }
 420
 421                                         //figure out how many sequences you have to align
 422                                         numSeqsPerProcessor = numFastaSeqs / processors;
 423                                         int startIndex =  pid * numSeqsPerProcessor;
 424                                         if(pid == (processors - 1)){    numSeqsPerProcessor = numFastaSeqs - pid * numSeqsPerProcessor;         }
 425
 426
 427                                         //align your part
 428                                         driverMPI(startIndex, numSeqsPerProcessor, inMPI, outMPINewTax, outMPITempTax, MPIPos);
 429
 430                                         if (m->control_pressed) {  MPI_File_close(&inMPI);  MPI_File_close(&outMPINewTax);   MPI_File_close(&outMPITempTax);  for (int i = 0; i < outputNames.size(); i++) {    remove(outputNames[i].c_str()); } delete classify; return 0;  }
 431
 432                                         for (int i = 1; i < processors; i++) {
 433                                                 int done;
 434                                                 MPI_Recv(&done, 1, MPI_INT, i, tag, MPI_COMM_WORLD, &status);
 435                                         }
 436                                 }else{ //you are a child process
 437                                         MPI_Recv(&numFastaSeqs, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
 438                                         MPIPos.resize(numFastaSeqs+1);
 439                                         MPI_Recv(&MPIPos[0], (numFastaSeqs+1), MPI_LONG, 0, tag, MPI_COMM_WORLD, &status);
 440
 441                                         //figure out how many sequences you have to align
 442                                         numSeqsPerProcessor = numFastaSeqs / processors;
 443                                         int startIndex =  pid * numSeqsPerProcessor;
 444                                         if(pid == (processors - 1)){    numSeqsPerProcessor = numFastaSeqs - pid * numSeqsPerProcessor;         }
 445
 446
 447                                         //align your part
 448                                         driverMPI(startIndex, numSeqsPerProcessor, inMPI, outMPINewTax, outMPITempTax, MPIPos);
 449
 450                                         if (m->control_pressed) {  MPI_File_close(&inMPI);  MPI_File_close(&outMPINewTax);   MPI_File_close(&outMPITempTax);  delete classify; return 0;  }
 451
 452                                         int done = 0;
 453                                         MPI_Send(&done, 1, MPI_INT, 0, tag, MPI_COMM_WORLD);
 454                                 }
 455
 456                                 //close files
 457                                 MPI_File_close(&inMPI);
 458                                 MPI_File_close(&outMPINewTax);
 459                                 MPI_File_close(&outMPITempTax);
 460                                 MPI_Barrier(MPI_COMM_WORLD); //make everyone wait - just in case
 461
 462 #else
 463
 464                         vector<unsigned long int> positions = m->divideFile(fastaFileNames[s], processors);
 465
 466                         for (int i = 0; i < (positions.size()-1); i++) {
 467                                 lines.push_back(new linePair(positions[i], positions[(i+1)]));
 468                         }
 469
 470                 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
 471                         if(processors == 1){
 472                                 numFastaSeqs = driver(lines[0], newTaxonomyFile, tempTaxonomyFile, fastaFileNames[s]);
 473                         }
 474                         else{
 475                                 processIDS.resize(0);
 476
 477                                 numFastaSeqs = createProcesses(newTaxonomyFile, tempTaxonomyFile, fastaFileNames[s]);
 478
 479                                 rename((newTaxonomyFile + toString(processIDS[0]) + ".temp").c_str(), newTaxonomyFile.c_str());
 480                                 rename((tempTaxonomyFile + toString(processIDS[0]) + ".temp").c_str(), tempTaxonomyFile.c_str());
 481
 482                                 for(int i=1;i<processors;i++){
 483                                         appendTaxFiles((newTaxonomyFile + toString(processIDS[i]) + ".temp"), newTaxonomyFile);
 484                                         appendTaxFiles((tempTaxonomyFile + toString(processIDS[i]) + ".temp"), tempTaxonomyFile);
 485                                         remove((newTaxonomyFile + toString(processIDS[i]) + ".temp").c_str());
 486                                         remove((tempTaxonomyFile + toString(processIDS[i]) + ".temp").c_str());
 487                                 }
 488
 489                         }
 490         #else
 491                         numFastaSeqs = driver(lines[0], newTaxonomyFile, tempTaxonomyFile, fastaFileNames[s]);
 492         #endif
 493 #endif
 494
 495                 m->mothurOutEndLine();
 496                 m->mothurOut("It took " + toString(time(NULL) - start) + " secs to classify " + toString(numFastaSeqs) + " sequences."); m->mothurOutEndLine(); m->mothurOutEndLine();
 497                 start = time(NULL);
 498
 499
 500                 #ifdef USE_MPI
 501                         if (pid == 0) {  //this part does not need to be paralellized
 502
 503                                 if(namefile != "") { m->mothurOut("Reading " + namefileNames[s] + "..."); cout.flush();  MPIReadNamesFile(namefileNames[s]);  m->mothurOut("  Done."); m->mothurOutEndLine(); }
 504                 #else
 505                         //read namefile
 506                         if(namefile != "") {
 507
 508                             m->mothurOut("Reading " + namefileNames[s] + "..."); cout.flush();
 509
 510                                 nameMap.clear(); //remove old names
 511
 512                                 ifstream inNames;
 513                                 m->openInputFile(namefileNames[s], inNames);
 514
 515                                 string firstCol, secondCol;
 516                                 while(!inNames.eof()) {
 517                                         inNames >> firstCol >> secondCol; m->gobble(inNames);
 518
 519                                         vector<string> temp;
 520                                         m->splitAtComma(secondCol, temp);
 521
 522                                         nameMap[firstCol] = temp;
 523                                 }
 524                                 inNames.close();
 525
 526                                 m->mothurOut("  Done."); m->mothurOutEndLine();
 527                         }
 528                 #endif
 529
 530                         string group = "";
 531                         if (groupfile != "") {  group = groupfileNames[s]; }
 532
 533                         PhyloSummary taxaSum(taxonomyFileName, group);
 534
 535                         if (m->control_pressed) {  for (int i = 0; i < outputNames.size(); i++) {       remove(outputNames[i].c_str()); } delete classify; return 0; }
 536
 537                         if (namefile == "") {  taxaSum.summarize(tempTaxonomyFile);  }
 538                         else {
 539                                 ifstream in;
 540                                 m->openInputFile(tempTaxonomyFile, in);
 541
 542                                 //read in users taxonomy file and add sequences to tree
 543                                 string name, taxon;
 544
 545                                 while(!in.eof()){
 546                                         in >> name >> taxon; m->gobble(in);
 547
 548                                         itNames = nameMap.find(name);
 549
 550                                         if (itNames == nameMap.end()) {
 551                                                 m->mothurOut(name + " is not in your name file please correct."); m->mothurOutEndLine(); exit(1);
 552                                         }else{
 553                                                 for (int i = 0; i < itNames->second.size(); i++) {
 554                                                         taxaSum.addSeqToTree(itNames->second[i], taxon);  //add it as many times as there are identical seqs
 555                                                 }
 556                                                 itNames->second.clear();
 557                                                 nameMap.erase(itNames->first);
 558                                         }
 559                                 }
 560                                 in.close();
 561                         }
 562                         remove(tempTaxonomyFile.c_str());
 563
 564                         if (m->control_pressed) {  for (int i = 0; i < outputNames.size(); i++) {       remove(outputNames[i].c_str()); } delete classify; return 0; }
 565
 566                         //print summary file
 567                         ofstream outTaxTree;
 568                         m->openOutputFile(taxSummary, outTaxTree);
 569                         taxaSum.print(outTaxTree);
 570                         outTaxTree.close();
 571
 572                         //output taxonomy with the unclassified bins added
 573                         ifstream inTax;
 574                         m->openInputFile(newTaxonomyFile, inTax);
 575
 576                         ofstream outTax;
 577                         string unclass = newTaxonomyFile + ".unclass.temp";
 578                         m->openOutputFile(unclass, outTax);
 579
 580                         //get maxLevel from phylotree so you know how many 'unclassified's to add
 581                         int maxLevel = taxaSum.getMaxLevel();
 582
 583                         //read taxfile - this reading and rewriting is done to preserve the confidence scores.
 584                         string name, taxon;
 585                         while (!inTax.eof()) {
 586                                 if (m->control_pressed) {  for (int i = 0; i < outputNames.size(); i++) {       remove(outputNames[i].c_str()); } remove(unclass.c_str()); delete classify; return 0; }
 587
 588                                 inTax >> name >> taxon; m->gobble(inTax);
 589
 590                                 string newTax = addUnclassifieds(taxon, maxLevel);
 591
 592                                 outTax << name << '\t' << newTax << endl;
 593                         }
 594                         inTax.close();
 595                         outTax.close();
 596
 597                         remove(newTaxonomyFile.c_str());
 598                         rename(unclass.c_str(), newTaxonomyFile.c_str());
 599
 600                         m->mothurOutEndLine();
 601                         m->mothurOut("It took " + toString(time(NULL) - start) + " secs to create the summary file for " + toString(numFastaSeqs) + " sequences."); m->mothurOutEndLine(); m->mothurOutEndLine();
 602
 603                         #ifdef USE_MPI
 604                                 }
 605                         #endif
 606
 607                         m->mothurOutEndLine();
 608                         m->mothurOut("Output File Names: "); m->mothurOutEndLine();
 609                         for (int i = 0; i < outputNames.size(); i++) {  m->mothurOut(outputNames[i]); m->mothurOutEndLine();    }
 610                         m->mothurOutEndLine();
 611                 }
 612
 613                 delete classify;
 614                 return 0;
 615         }
 616         catch(exception& e) {
 617                 m->errorOut(e, "ClassifySeqsCommand", "execute");
 618                 exit(1);
 619         }
 620 }
 621
 622 /**************************************************************************************************/
 623 string ClassifySeqsCommand::addUnclassifieds(string tax, int maxlevel) {
 624         try{
 625                 string newTax, taxon;
 626                 int level = 0;
 627
 628                 //keep what you have counting the levels
 629                 while (tax.find_first_of(';') != -1) {
 630                         //get taxon
 631                         taxon = tax.substr(0,tax.find_first_of(';'))+';';
 632                         tax = tax.substr(tax.find_first_of(';')+1, tax.length());
 633                         newTax += taxon;
 634                         level++;
 635                 }
 636
 637                 //add "unclassified" until you reach maxLevel
 638                 while (level < maxlevel) {
 639                         newTax += "unclassified;";
 640                         level++;
 641                 }
 642
 643                 return newTax;
 644         }
 645         catch(exception& e) {
 646                 m->errorOut(e, "ClassifySeqsCommand", "addUnclassifieds");
 647                 exit(1);
 648         }
 649 }
 650
 651 /**************************************************************************************************/
 652
 653 int ClassifySeqsCommand::createProcesses(string taxFileName, string tempTaxFile, string filename) {
 654         try {
 655 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
 656                 int process = 0;
 657                 int num = 0;
 658
 659                 //loop through and create all the processes you want
 660                 while (process != processors) {
 661                         int pid = fork();
 662
 663                         if (pid > 0) {
 664                                 processIDS.push_back(pid);  //create map from line number to pid so you can append files in correct order later
 665                                 process++;
 666                         }else if (pid == 0){
 667                                 num = driver(lines[process], taxFileName + toString(getpid()) + ".temp", tempTaxFile + toString(getpid()) + ".temp", filename);
 668
 669                                 //pass numSeqs to parent
 670                                 ofstream out;
 671                                 string tempFile = filename + toString(getpid()) + ".num.temp";
 672                                 m->openOutputFile(tempFile, out);
 673                                 out << num << endl;
 674                                 out.close();
 675
 676                                 exit(0);
 677                         }else { m->mothurOut("unable to spawn the necessary processes."); m->mothurOutEndLine(); exit(0); }
 678                 }
 679
 680                 //force parent to wait until all the processes are done
 681                 for (int i=0;i<processors;i++) {
 682                         int temp = processIDS[i];
 683                         wait(&temp);
 684                 }
 685
 686                 for (int i = 0; i < processIDS.size(); i++) {
 687                         ifstream in;
 688                         string tempFile =  filename + toString(processIDS[i]) + ".num.temp";
 689                         m->openInputFile(tempFile, in);
 690                         if (!in.eof()) { int tempNum = 0; in >> tempNum; num += tempNum; }
 691                         in.close(); remove(tempFile.c_str());
 692                 }
 693
 694                 return num;
 695 #endif
 696         }
 697         catch(exception& e) {
 698                 m->errorOut(e, "ClassifySeqsCommand", "createProcesses");
 699                 exit(1);
 700         }
 701 }
 702 /**************************************************************************************************/
 703
 704 void ClassifySeqsCommand::appendTaxFiles(string temp, string filename) {
 705         try{
 706
 707                 ofstream output;
 708                 ifstream input;
 709                 m->openOutputFileAppend(filename, output);
 710                 m->openInputFile(temp, input);
 711
 712                 while(char c = input.get()){
 713                         if(input.eof())         {       break;                  }
 714                         else                            {       output << c;    }
 715                 }
 716
 717                 input.close();
 718                 output.close();
 719         }
 720         catch(exception& e) {
 721                 m->errorOut(e, "ClassifySeqsCommand", "appendTaxFiles");
 722                 exit(1);
 723         }
 724 }
 725
 726 //**********************************************************************************************************************
 727
 728 int ClassifySeqsCommand::driver(linePair* filePos, string taxFName, string tempTFName, string filename){
 729         try {
 730                 ofstream outTax;
 731                 m->openOutputFile(taxFName, outTax);
 732
 733                 ofstream outTaxSimple;
 734                 m->openOutputFile(tempTFName, outTaxSimple);
 735
 736                 ifstream inFASTA;
 737                 m->openInputFile(filename, inFASTA);
 738
 739                 string taxonomy;
 740
 741                 inFASTA.seekg(filePos->start);
 742
 743                 bool done = false;
 744                 int count = 0;
 745
 746                 while (!done) {
 747                         if (m->control_pressed) { return 0; }
 748
 749                         Sequence* candidateSeq = new Sequence(inFASTA); m->gobble(inFASTA);
 750
 751                         if (candidateSeq->getName() != "") {
 752                                 taxonomy = classify->getTaxonomy(candidateSeq);
 753
 754                                 if (m->control_pressed) { delete candidateSeq; return 0; }
 755
 756                                 if (taxonomy != "bad seq") {
 757                                         //output confidence scores or not
 758                                         if (probs) {
 759                                                 outTax << candidateSeq->getName() << '\t' << taxonomy << endl;
 760                                         }else{
 761                                                 outTax << candidateSeq->getName() << '\t' << classify->getSimpleTax() << endl;
 762                                         }
 763
 764                                         outTaxSimple << candidateSeq->getName() << '\t' << classify->getSimpleTax() << endl;
 765                                 }
 766                                 count++;
 767                         }
 768                         delete candidateSeq;
 769
 770                         #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
 771                                 unsigned long int pos = inFASTA.tellg();
 772                                 if ((pos == -1) || (pos >= filePos->end)) { break; }
 773                         #else
 774                                 if (inFASTA.eof()) { break; }
 775                         #endif
 776
 777                         //report progress
 778                         if((count) % 100 == 0){ m->mothurOut("Processing sequence: " + toString(count)); m->mothurOutEndLine();         }
 779                 }
 780                 //report progress
 781                 if((count) % 100 != 0){ m->mothurOut("Processing sequence: " + toString(count)); m->mothurOutEndLine();         }
 782
 783                 inFASTA.close();
 784                 outTax.close();
 785                 outTaxSimple.close();
 786
 787                 return count;
 788         }
 789         catch(exception& e) {
 790                 m->errorOut(e, "ClassifySeqsCommand", "driver");
 791                 exit(1);
 792         }
 793 }
 794 //**********************************************************************************************************************
 795 #ifdef USE_MPI
 796 int ClassifySeqsCommand::driverMPI(int start, int num, MPI_File& inMPI, MPI_File& newFile, MPI_File& tempFile, vector<unsigned long int>& MPIPos){
 797         try {
 798                 MPI_Status statusNew;
 799                 MPI_Status statusTemp;
 800                 MPI_Status status;
 801
 802                 int pid;
 803                 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 804
 805                 string taxonomy;
 806                 string outputString;
 807
 808                 for(int i=0;i<num;i++){
 809
 810                         if (m->control_pressed) { return 0; }
 811
 812                         //read next sequence
 813                         int length = MPIPos[start+i+1] - MPIPos[start+i];
 814                         char* buf4 = new char[length];
 815                         MPI_File_read_at(inMPI, MPIPos[start+i], buf4, length, MPI_CHAR, &status);
 816
 817                         string tempBuf = buf4;
 818                         if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length);  }
 819                         istringstream iss (tempBuf,istringstream::in);
 820                         delete buf4;
 821
 822                         Sequence* candidateSeq = new Sequence(iss);
 823
 824                         if (candidateSeq->getName() != "") {
 825                                 taxonomy = classify->getTaxonomy(candidateSeq);
 826
 827                                 if (taxonomy != "bad seq") {
 828                                         //output confidence scores or not
 829                                         if (probs) {
 830                                                 outputString =  candidateSeq->getName() + "\t" + taxonomy + "\n";
 831                                         }else{
 832                                                 outputString =  candidateSeq->getName() + "\t" + classify->getSimpleTax() + "\n";
 833                                         }
 834
 835                                         int length = outputString.length();
 836                                         char* buf2 = new char[length];
 837                                         memcpy(buf2, outputString.c_str(), length);
 838
 839                                         MPI_File_write_shared(newFile, buf2, length, MPI_CHAR, &statusNew);
 840                                         delete buf2;
 841
 842                                         outputString =  candidateSeq->getName() + "\t" + classify->getSimpleTax() + "\n";
 843                                         length = outputString.length();
 844                                         char* buf = new char[length];
 845                                         memcpy(buf, outputString.c_str(), length);
 846
 847                                         MPI_File_write_shared(tempFile, buf, length, MPI_CHAR, &statusTemp);
 848                                         delete buf;
 849                                 }
 850                         }
 851                         delete candidateSeq;
 852
 853                         if((i+1) % 100 == 0){   cout << "Classifying sequence " << (i+1) << endl;       }
 854                 }
 855
 856                 if(num % 100 != 0){     cout << "Classifying sequence " << (num) << endl;       }
 857
 858
 859                 return 1;
 860         }
 861         catch(exception& e) {
 862                 m->errorOut(e, "ClassifySeqsCommand", "driverMPI");
 863                 exit(1);
 864         }
 865 }
 866
 867 //**********************************************************************************************************************
 868 int ClassifySeqsCommand::MPIReadNamesFile(string nameFilename){
 869         try {
 870
 871                 nameMap.clear(); //remove old names
 872
 873                 MPI_File inMPI;
 874                 MPI_Offset size;
 875                 MPI_Status status;
 876
 877                 //char* inFileName = new char[nameFilename.length()];
 878                 //memcpy(inFileName, nameFilename.c_str(), nameFilename.length());
 879
 880                 char inFileName[1024];
 881                 strcpy(inFileName, nameFilename.c_str());
 882
 883                 MPI_File_open(MPI_COMM_WORLD, inFileName, MPI_MODE_RDONLY, MPI_INFO_NULL, &inMPI);
 884                 MPI_File_get_size(inMPI, &size);
 885                 //delete inFileName;
 886
 887                 char* buffer = new char[size];
 888                 MPI_File_read(inMPI, buffer, size, MPI_CHAR, &status);
 889
 890                 string tempBuf = buffer;
 891                 if (tempBuf.length() > size) { tempBuf = tempBuf.substr(0, size);  }
 892                 istringstream iss (tempBuf,istringstream::in);
 893                 delete buffer;
 894
 895                 string firstCol, secondCol;
 896                 while(!iss.eof()) {
 897                         iss >> firstCol >> secondCol; m->gobble(iss);
 898
 899                         vector<string> temp;
 900                         m->splitAtComma(secondCol, temp);
 901
 902                         nameMap[firstCol] = temp;
 903                 }
 904
 905                 MPI_File_close(&inMPI);
 906
 907                 return 1;
 908         }
 909         catch(exception& e) {
 910                 m->errorOut(e, "ClassifySeqsCommand", "MPIReadNamesFile");
 911                 exit(1);
 912         }
 913 }
 914 #endif
 915 /**************************************************************************************************/