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 "knn.h"
  15
  16 //**********************************************************************************************************************
  17
  18 ClassifySeqsCommand::ClassifySeqsCommand(string option)  {
  19         try {
  20                 abort = false;
  21
  22                 //allow user to run help
  23                 if(option == "help") { help(); abort = true; }
  24
  25                 else {
  26
  27                         //valid paramters for this command
  28                         string AlignArray[] =  {"template","fasta","name","search","ksize","method","processors","taxonomy","match","mismatch","gapopen","gapextend","numwanted","cutoff","probs","iters", "outputdir","inputdir"};
  29                         vector<string> myArray (AlignArray, AlignArray+(sizeof(AlignArray)/sizeof(string)));
  30
  31                         OptionParser parser(option);
  32                         map<string, string> parameters = parser.getParameters();
  33
  34                         ValidParameters validParameter;
  35                         map<string, string>::iterator it;
  36
  37                         //check to make sure all parameters are valid for command
  38                         for (it = parameters.begin(); it != parameters.end(); it++) {
  39                                 if (validParameter.isValidParameter(it->first, myArray, it->second) != true) {  abort = true;  }
  40                         }
  41
  42                         //if the user changes the output directory command factory will send this info to us in the output parameter
  43                         outputDir = validParameter.validFile(parameters, "outputdir", false);           if (outputDir == "not found"){  outputDir = "";         }
  44
  45                         //if the user changes the input directory command factory will send this info to us in the output parameter
  46                         string inputDir = validParameter.validFile(parameters, "inputdir", false);
  47                         if (inputDir == "not found"){   inputDir = "";          }
  48                         else {
  49                                 string path;
  50                                 it = parameters.find("template");
  51                                 //user has given a template file
  52                                 if(it != parameters.end()){
  53                                         path = hasPath(it->second);
  54                                         //if the user has not given a path then, add inputdir. else leave path alone.
  55                                         if (path == "") {       parameters["template"] = inputDir + it->second;         }
  56                                 }
  57
  58                                 it = parameters.find("taxonomy");
  59                                 //user has given a template file
  60                                 if(it != parameters.end()){
  61                                         path = hasPath(it->second);
  62                                         //if the user has not given a path then, add inputdir. else leave path alone.
  63                                         if (path == "") {       parameters["taxonomy"] = inputDir + it->second;         }
  64                                 }
  65                         }
  66
  67                         //check for required parameters
  68                         templateFileName = validParameter.validFile(parameters, "template", true);
  69                         if (templateFileName == "not found") {
  70                                 m->mothurOut("template is a required parameter for the classify.seqs command.");
  71                                 m->mothurOutEndLine();
  72                                 abort = true;
  73                         }
  74                         else if (templateFileName == "not open") { abort = true; }
  75
  76                         fastaFileName = validParameter.validFile(parameters, "fasta", false);
  77                         if (fastaFileName == "not found") { m->mothurOut("fasta is a required parameter for the classify.seqs command."); m->mothurOutEndLine(); abort = true;  }
  78                         else {
  79                                 splitAtDash(fastaFileName, fastaFileNames);
  80
  81                                 //go through files and make sure they are good, if not, then disregard them
  82                                 for (int i = 0; i < fastaFileNames.size(); i++) {
  83                                         if (inputDir != "") {
  84                                                 string path = hasPath(fastaFileNames[i]);
  85                                                 //if the user has not given a path then, add inputdir. else leave path alone.
  86                                                 if (path == "") {       fastaFileNames[i] = inputDir + fastaFileNames[i];               }
  87                                         }
  88
  89                                         int ableToOpen;
  90
  91                                         #ifdef USE_MPI
  92                                                 int pid;
  93                                                 MPI_Comm_size(MPI_COMM_WORLD, &processors); //set processors to the number of mpi processes running
  94                                                 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
  95
  96                                                 if (pid == 0) {
  97                                         #endif
  98
  99                                         ifstream in;
 100                                         ableToOpen = openInputFile(fastaFileNames[i], in);
 101                                         in.close();
 102
 103                                         #ifdef USE_MPI
 104                                                         for (int j = 1; j < processors; j++) {
 105                                                                 MPI_Send(&ableToOpen, 1, MPI_INT, j, 2001, MPI_COMM_WORLD);
 106                                                         }
 107                                                 }else{
 108                                                         MPI_Status status;
 109                                                         MPI_Recv(&ableToOpen, 1, MPI_INT, 0, 2001, MPI_COMM_WORLD, &status);
 110                                                 }
 111
 112                                         #endif
 113
 114                                         if (ableToOpen == 1) {
 115                                                 m->mothurOut(fastaFileNames[i] + " will be disregarded."); m->mothurOutEndLine();
 116                                                 //erase from file list
 117                                                 fastaFileNames.erase(fastaFileNames.begin()+i);
 118                                                 i--;
 119                                         }
 120
 121                                 }
 122
 123                                 //make sure there is at least one valid file left
 124                                 if (fastaFileNames.size() == 0) { m->mothurOut("no valid files."); m->mothurOutEndLine(); abort = true; }
 125                         }
 126
 127
 128                         taxonomyFileName = validParameter.validFile(parameters, "taxonomy", true);
 129                         if (taxonomyFileName == "not found") {
 130                                 m->mothurOut("taxonomy is a required parameter for the classify.seqs command.");
 131                                 m->mothurOutEndLine();
 132                                 abort = true;
 133                         }
 134                         else if (taxonomyFileName == "not open") { abort = true; }
 135
 136
 137                         namefile = validParameter.validFile(parameters, "name", false);
 138                         if (namefile == "not found") { namefile = "";  }
 139
 140                         else {
 141                                 splitAtDash(namefile, namefileNames);
 142
 143                                 //go through files and make sure they are good, if not, then disregard them
 144                                 for (int i = 0; i < namefileNames.size(); i++) {
 145                                         if (inputDir != "") {
 146                                                 string path = hasPath(namefileNames[i]);
 147                                                 //if the user has not given a path then, add inputdir. else leave path alone.
 148                                                 if (path == "") {       namefileNames[i] = inputDir + namefileNames[i];         }
 149                                         }
 150                                         int ableToOpen;
 151
 152                                         #ifdef USE_MPI
 153                                                 int pid;
 154                                                 MPI_Comm_size(MPI_COMM_WORLD, &processors); //set processors to the number of mpi processes running
 155                                                 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 156
 157                                                 if (pid == 0) {
 158                                         #endif
 159
 160                                         ifstream in;
 161                                         ableToOpen = openInputFile(namefileNames[i], in);
 162                                         in.close();
 163
 164                                         #ifdef USE_MPI
 165                                                         for (int j = 1; j < processors; j++) {
 166                                                                 MPI_Send(&ableToOpen, 1, MPI_INT, j, 2001, MPI_COMM_WORLD);
 167                                                         }
 168                                                 }else{
 169                                                         MPI_Status status;
 170                                                         MPI_Recv(&ableToOpen, 1, MPI_INT, 0, 2001, MPI_COMM_WORLD, &status);
 171                                                 }
 172
 173                                         #endif
 174                                         if (ableToOpen == 1) {  m->mothurOut("Unable to match name file with fasta file."); m->mothurOutEndLine(); abort = true;        }
 175
 176                                 }
 177                         }
 178
 179                         if (namefile != "") {
 180                                 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(); }
 181                         }
 182
 183                         //check for optional parameter and set defaults
 184                         // ...at some point should added some additional type checking...
 185                         string temp;
 186                         temp = validParameter.validFile(parameters, "ksize", false);            if (temp == "not found"){       temp = "8";                             }
 187                         convert(temp, kmerSize);
 188
 189                         temp = validParameter.validFile(parameters, "processors", false);       if (temp == "not found"){       temp = "1";                             }
 190                         convert(temp, processors);
 191
 192                         search = validParameter.validFile(parameters, "search", false);         if (search == "not found"){     search = "kmer";                }
 193
 194                         method = validParameter.validFile(parameters, "method", false);         if (method == "not found"){     method = "bayesian";    }
 195
 196                         temp = validParameter.validFile(parameters, "match", false);            if (temp == "not found"){       temp = "1.0";                   }
 197                         convert(temp, match);
 198
 199                         temp = validParameter.validFile(parameters, "mismatch", false);         if (temp == "not found"){       temp = "-1.0";                  }
 200                         convert(temp, misMatch);
 201
 202                         temp = validParameter.validFile(parameters, "gapopen", false);          if (temp == "not found"){       temp = "-2.0";                  }
 203                         convert(temp, gapOpen);
 204
 205                         temp = validParameter.validFile(parameters, "gapextend", false);        if (temp == "not found"){       temp = "-1.0";                  }
 206                         convert(temp, gapExtend);
 207
 208                         temp = validParameter.validFile(parameters, "numwanted", false);        if (temp == "not found"){       temp = "10";                    }
 209                         convert(temp, numWanted);
 210
 211                         temp = validParameter.validFile(parameters, "cutoff", false);           if (temp == "not found"){       temp = "0";                             }
 212                         convert(temp, cutoff);
 213
 214                         temp = validParameter.validFile(parameters, "probs", false);            if (temp == "not found"){       temp = "true";                  }
 215                         probs = isTrue(temp);
 216
 217                         temp = validParameter.validFile(parameters, "iters", false);            if (temp == "not found") { temp = "100";                        }
 218                         convert(temp, iters);
 219
 220
 221
 222                         if ((method == "bayesian") && (search != "kmer"))  {
 223                                 m->mothurOut("The bayesian method requires the kmer search." + search + "will be disregarded." ); m->mothurOutEndLine();
 224                                 search = "kmer";
 225                         }
 226                 }
 227
 228         }
 229         catch(exception& e) {
 230                 m->errorOut(e, "ClassifySeqsCommand", "ClassifySeqsCommand");
 231                 exit(1);
 232         }
 233 }
 234
 235 //**********************************************************************************************************************
 236
 237 ClassifySeqsCommand::~ClassifySeqsCommand(){
 238
 239         if (abort == false) {
 240                 for (int i = 0; i < lines.size(); i++) {  delete lines[i];  }  lines.clear();
 241         }
 242 }
 243
 244 //**********************************************************************************************************************
 245
 246 void ClassifySeqsCommand::help(){
 247         try {
 248                 m->mothurOut("The classify.seqs command reads a fasta file containing sequences and creates a .taxonomy file and a .tax.summary file.\n");
 249                 m->mothurOut("The classify.seqs command parameters are template, fasta, name, search, ksize, method, taxonomy, processors, match, mismatch, gapopen, gapextend, numwanted and probs.\n");
 250                 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");
 251                 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");
 252                 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");
 253                 m->mothurOut("The method parameter allows you to specify classification method to use.  Your options are: bayesian and knn. The default is bayesian.\n");
 254                 m->mothurOut("The ksize parameter allows you to specify the kmer size for finding most similar template to candidate.  The default is 8.\n");
 255                 m->mothurOut("The processors parameter allows you to specify the number of processors to use. The default is 1.\n");
 256                 #ifdef USE_MPI
 257                 m->mothurOut("When using MPI, the processors parameter is set to the number of MPI processes running. \n");
 258                 #endif
 259                 m->mothurOut("The match parameter allows you to specify the bonus for having the same base. The default is 1.0.\n");
 260                 m->mothurOut("The mistmatch parameter allows you to specify the penalty for having different bases.  The default is -1.0.\n");
 261                 m->mothurOut("The gapopen parameter allows you to specify the penalty for opening a gap in an alignment. The default is -2.0.\n");
 262                 m->mothurOut("The gapextend parameter allows you to specify the penalty for extending a gap in an alignment.  The default is -1.0.\n");
 263                 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");
 264                 m->mothurOut("The cutoff parameter allows you to specify a bootstrap confidence threshold for your taxonomy.  The default is 0.\n");
 265                 m->mothurOut("The probs parameter shut off the bootstrapping results for the bayesian method. The default is true, meaning you want the bootstrapping to be run.\n");
 266                 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");
 267                 m->mothurOut("The classify.seqs command should be in the following format: \n");
 268                 m->mothurOut("classify.seqs(template=yourTemplateFile, fasta=yourFastaFile, method=yourClassificationMethod, search=yourSearchmethod, ksize=yourKmerSize, taxonomy=yourTaxonomyFile, processors=yourProcessors) \n");
 269                 m->mothurOut("Example classify.seqs(fasta=amazon.fasta, template=core.filtered, method=knn, search=gotoh, ksize=8, processors=2)\n");
 270                 m->mothurOut("The .taxonomy file consists of 2 columns: 1 = your sequence name, 2 = the taxonomy for your sequence. \n");
 271                 m->mothurOut("The .tax.summary is a summary of the different taxonomies represented in your fasta file. \n");
 272                 m->mothurOut("Note: No spaces between parameter labels (i.e. fasta), '=' and parameters (i.e.yourFastaFile).\n\n");
 273         }
 274         catch(exception& e) {
 275                 m->errorOut(e, "ClassifySeqsCommand", "help");
 276                 exit(1);
 277         }
 278 }
 279
 280
 281 //**********************************************************************************************************************
 282
 283 int ClassifySeqsCommand::execute(){
 284         try {
 285                 if (abort == true) {    return 0;       }
 286
 287                 if(method == "bayesian"){       classify = new Bayesian(taxonomyFileName, templateFileName, search, kmerSize, cutoff, iters);           }
 288                 else if(method == "knn"){       classify = new Knn(taxonomyFileName, templateFileName, search, kmerSize, gapOpen, gapExtend, match, misMatch, numWanted);                               }
 289                 else {
 290                         m->mothurOut(search + " is not a valid method option. I will run the command using bayesian.");
 291                         m->mothurOutEndLine();
 292                         classify = new Bayesian(taxonomyFileName, templateFileName, search, kmerSize, cutoff, iters);
 293                 }
 294
 295                 if (m->control_pressed) { delete classify; return 0; }
 296
 297                 vector<string> outputNames;
 298
 299                 for (int s = 0; s < fastaFileNames.size(); s++) {
 300
 301                         m->mothurOut("Classifying sequences from " + fastaFileNames[s] + " ..." ); m->mothurOutEndLine();
 302
 303                         if (outputDir == "") { outputDir += hasPath(fastaFileNames[s]); }
 304                         string newTaxonomyFile = outputDir + getRootName(getSimpleName(fastaFileNames[s])) + getRootName(getSimpleName(taxonomyFileName)) + "taxonomy";
 305                         string tempTaxonomyFile = outputDir + getRootName(getSimpleName(fastaFileNames[s])) + "taxonomy.temp";
 306                         string taxSummary = outputDir + getRootName(getSimpleName(fastaFileNames[s])) + getRootName(getSimpleName(taxonomyFileName)) + "tax.summary";
 307
 308                         outputNames.push_back(newTaxonomyFile);
 309                         outputNames.push_back(taxSummary);
 310
 311                         int start = time(NULL);
 312                         int numFastaSeqs = 0;
 313                         for (int i = 0; i < lines.size(); i++) {  delete lines[i];  }  lines.clear();
 314
 315 #ifdef USE_MPI
 316                                 int pid, end, numSeqsPerProcessor;
 317                                 int tag = 2001;
 318                                 vector<long> MPIPos;
 319
 320                                 MPI_Status status;
 321                                 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 322                                 MPI_Comm_size(MPI_COMM_WORLD, &processors);
 323
 324                                 MPI_File inMPI;
 325                                 MPI_File outMPINewTax;
 326                                 MPI_File outMPITempTax;
 327
 328                                 int outMode=MPI_MODE_CREATE|MPI_MODE_WRONLY;
 329                                 int inMode=MPI_MODE_RDONLY;
 330
 331                                 char outNewTax[newTaxonomyFile.length()];
 332                                 strcpy(outNewTax, newTaxonomyFile.c_str());
 333
 334                                 char outTempTax[tempTaxonomyFile.length()];
 335                                 strcpy(outTempTax, tempTaxonomyFile.c_str());
 336
 337                                 char inFileName[fastaFileNames[s].length()];
 338                                 strcpy(inFileName, fastaFileNames[s].c_str());
 339
 340                                 MPI_File_open(MPI_COMM_WORLD, inFileName, inMode, MPI_INFO_NULL, &inMPI);  //comm, filename, mode, info, filepointer
 341                                 MPI_File_open(MPI_COMM_WORLD, outNewTax, outMode, MPI_INFO_NULL, &outMPINewTax);
 342                                 MPI_File_open(MPI_COMM_WORLD, outTempTax, outMode, MPI_INFO_NULL, &outMPITempTax);
 343
 344                                 if (m->control_pressed) {  MPI_File_close(&inMPI);  MPI_File_close(&outMPINewTax);   MPI_File_close(&outMPITempTax);  delete classify; return 0;  }
 345
 346                                 if(namefile != "") {  MPIReadNamesFile(namefileNames[s]);  }
 347
 348                                 if (pid == 0) { //you are the root process
 349
 350                                         MPIPos = setFilePosFasta(fastaFileNames[s], numFastaSeqs); //fills MPIPos, returns numSeqs
 351
 352                                         //send file positions to all processes
 353                                         MPI_Bcast(&numFastaSeqs, 1, MPI_INT, 0, MPI_COMM_WORLD);  //send numSeqs
 354                                         MPI_Bcast(&MPIPos[0], (numFastaSeqs+1), MPI_LONG, 0, MPI_COMM_WORLD); //send file pos
 355
 356                                         //figure out how many sequences you have to align
 357                                         numSeqsPerProcessor = numFastaSeqs / processors;
 358                                         if(pid == (processors - 1)){    numSeqsPerProcessor = numFastaSeqs - pid * numSeqsPerProcessor;         }
 359                                         int startIndex =  pid * numSeqsPerProcessor;
 360
 361                                         //align your part
 362                                         driverMPI(startIndex, numSeqsPerProcessor, inMPI, outMPINewTax, outMPITempTax, MPIPos);
 363
 364                                         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;  }
 365
 366                                         for (int i = 1; i < processors; i++) {
 367                                                 int done;
 368                                                 MPI_Recv(&done, 1, MPI_INT, i, tag, MPI_COMM_WORLD, &status);
 369                                         }
 370                                 }else{ //you are a child process
 371                                         MPI_Bcast(&numFastaSeqs, 1, MPI_INT, 0, MPI_COMM_WORLD); //get numSeqs
 372                                         MPIPos.resize(numFastaSeqs+1);
 373                                         MPI_Bcast(&MPIPos[0], (numFastaSeqs+1), MPI_LONG, 0, MPI_COMM_WORLD); //get file positions
 374
 375                                         //figure out how many sequences you have to align
 376                                         numSeqsPerProcessor = numFastaSeqs / processors;
 377                                         if(pid == (processors - 1)){    numSeqsPerProcessor = numFastaSeqs - pid * numSeqsPerProcessor;         }
 378                                         int startIndex =  pid * numSeqsPerProcessor;
 379
 380                                         //align your part
 381                                         driverMPI(startIndex, numSeqsPerProcessor, inMPI, outMPINewTax, outMPITempTax, MPIPos);
 382
 383                                         if (m->control_pressed) {  MPI_File_close(&inMPI);  MPI_File_close(&outMPINewTax);   MPI_File_close(&outMPITempTax);  delete classify; return 0;  }
 384
 385                                         int done = 0;
 386                                         MPI_Send(&done, 1, MPI_INT, 0, tag, MPI_COMM_WORLD);
 387                                 }
 388
 389                                 //close files
 390                                 MPI_File_close(&inMPI);
 391                                 MPI_File_close(&outMPINewTax);
 392                                 MPI_File_close(&outMPITempTax);
 393
 394 #else
 395                         //read namefile
 396                         if(namefile != "") {
 397                                 nameMap.clear(); //remove old names
 398
 399                                 ifstream inNames;
 400                                 openInputFile(namefileNames[s], inNames);
 401
 402                                 string firstCol, secondCol;
 403                                 while(!inNames.eof()) {
 404                                         inNames >> firstCol >> secondCol; gobble(inNames);
 405                                         nameMap[firstCol] = getNumNames(secondCol);  //ex. seq1 seq1,seq3,seq5 -> seq1 = 3.
 406                                 }
 407                                 inNames.close();
 408                         }
 409
 410         #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
 411                         if(processors == 1){
 412                                 ifstream inFASTA;
 413                                 openInputFile(fastaFileNames[s], inFASTA);
 414                                 numFastaSeqs=count(istreambuf_iterator<char>(inFASTA),istreambuf_iterator<char>(), '>');
 415                                 inFASTA.close();
 416
 417                                 lines.push_back(new linePair(0, numFastaSeqs));
 418
 419                                 driver(lines[0], newTaxonomyFile, tempTaxonomyFile, fastaFileNames[s]);
 420                         }
 421                         else{
 422                                 vector<int> positions;
 423                                 processIDS.resize(0);
 424
 425                                 ifstream inFASTA;
 426                                 openInputFile(fastaFileNames[s], inFASTA);
 427
 428                                 string input;
 429                                 while(!inFASTA.eof()){
 430                                         input = getline(inFASTA);
 431                                         if (input.length() != 0) {
 432                                                 if(input[0] == '>'){    int pos = inFASTA.tellg(); positions.push_back(pos - input.length() - 1);       }
 433                                         }
 434                                 }
 435                                 inFASTA.close();
 436
 437                                 numFastaSeqs = positions.size();
 438
 439                                 int numSeqsPerProcessor = numFastaSeqs / processors;
 440
 441                                 for (int i = 0; i < processors; i++) {
 442                                         int startPos = positions[ i * numSeqsPerProcessor ];
 443                                         if(i == processors - 1){
 444                                                 numSeqsPerProcessor = numFastaSeqs - i * numSeqsPerProcessor;
 445                                         }
 446                                         lines.push_back(new linePair(startPos, numSeqsPerProcessor));
 447                                 }
 448                                 createProcesses(newTaxonomyFile, tempTaxonomyFile, fastaFileNames[s]);
 449
 450                                 rename((newTaxonomyFile + toString(processIDS[0]) + ".temp").c_str(), newTaxonomyFile.c_str());
 451                                 rename((tempTaxonomyFile + toString(processIDS[0]) + ".temp").c_str(), tempTaxonomyFile.c_str());
 452
 453                                 for(int i=1;i<processors;i++){
 454                                         appendTaxFiles((newTaxonomyFile + toString(processIDS[i]) + ".temp"), newTaxonomyFile);
 455                                         appendTaxFiles((tempTaxonomyFile + toString(processIDS[i]) + ".temp"), tempTaxonomyFile);
 456                                         remove((newTaxonomyFile + toString(processIDS[i]) + ".temp").c_str());
 457                                         remove((tempTaxonomyFile + toString(processIDS[i]) + ".temp").c_str());
 458                                 }
 459
 460                         }
 461         #else
 462                         ifstream inFASTA;
 463                         openInputFile(fastaFileNames[s], inFASTA);
 464                         numFastaSeqs=count(istreambuf_iterator<char>(inFASTA),istreambuf_iterator<char>(), '>');
 465                         inFASTA.close();
 466
 467                         lines.push_back(new linePair(0, numFastaSeqs));
 468
 469                         driver(lines[0], newTaxonomyFile, tempTaxonomyFile, fastaFileNames[s]);
 470         #endif
 471 #endif
 472
 473                 #ifdef USE_MPI
 474                         if (pid == 0) {  //this part does not need to be paralellized
 475                 #endif
 476
 477                         //make taxonomy tree from new taxonomy file
 478                         PhyloTree taxaBrowser;
 479
 480                         if (m->control_pressed) {  for (int i = 0; i < outputNames.size(); i++) {       remove(outputNames[i].c_str()); } delete classify; return 0; }
 481
 482                         ifstream in;
 483                         openInputFile(tempTaxonomyFile, in);
 484
 485                         //read in users taxonomy file and add sequences to tree
 486                         string name, taxon;
 487                         while(!in.eof()){
 488                                 in >> name >> taxon; gobble(in);
 489
 490                                 if (m->control_pressed) {  for (int i = 0; i < outputNames.size(); i++) {       remove(outputNames[i].c_str()); } remove(tempTaxonomyFile.c_str()); delete classify; return 0; }
 491
 492                                 if (namefile != "") {
 493                                         itNames = nameMap.find(name);
 494
 495                                         if (itNames == nameMap.end()) {
 496                                                 m->mothurOut(name + " is not in your name file please correct."); m->mothurOutEndLine(); exit(1);
 497                                         }else{
 498                                                 for (int i = 0; i < itNames->second; i++) {
 499                                                         taxaBrowser.addSeqToTree(name+toString(i), taxon);  //add it as many times as there are identical seqs
 500                                                 }
 501                                         }
 502                                 }else {  taxaBrowser.addSeqToTree(name, taxon);  } //add it once
 503                         }
 504                         in.close();
 505
 506                         taxaBrowser.assignHeirarchyIDs(0);
 507
 508                         if (m->control_pressed) {  for (int i = 0; i < outputNames.size(); i++) {       remove(outputNames[i].c_str()); } remove(tempTaxonomyFile.c_str()); delete classify; return 0; }
 509
 510                         taxaBrowser.binUnclassified();
 511
 512                         remove(tempTaxonomyFile.c_str());
 513
 514                         if (m->control_pressed) {  for (int i = 0; i < outputNames.size(); i++) {       remove(outputNames[i].c_str()); } delete classify; return 0; }
 515
 516
 517                         //print summary file
 518                         ofstream outTaxTree;
 519                         openOutputFile(taxSummary, outTaxTree);
 520                         taxaBrowser.print(outTaxTree);
 521                         outTaxTree.close();
 522
 523                         //output taxonomy with the unclassified bins added
 524                         ifstream inTax;
 525                         openInputFile(newTaxonomyFile, inTax);
 526
 527                         ofstream outTax;
 528                         string unclass = newTaxonomyFile + ".unclass.temp";
 529                         openOutputFile(unclass, outTax);
 530
 531                         //get maxLevel from phylotree so you know how many 'unclassified's to add
 532                         int maxLevel = taxaBrowser.getMaxLevel();
 533
 534                         //read taxfile - this reading and rewriting is done to preserve the confidence scores.
 535                         while (!inTax.eof()) {
 536                                 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; }
 537
 538                                 inTax >> name >> taxon; gobble(inTax);
 539
 540                                 string newTax = addUnclassifieds(taxon, maxLevel);
 541
 542                                 outTax << name << '\t' << newTax << endl;
 543                         }
 544                         inTax.close();
 545                         outTax.close();
 546
 547                         remove(newTaxonomyFile.c_str());
 548                         rename(unclass.c_str(), newTaxonomyFile.c_str());
 549
 550                         #ifdef USE_MPI
 551                                 }
 552                         #endif
 553
 554                         m->mothurOutEndLine();
 555                         m->mothurOut("Output File Names: "); m->mothurOutEndLine();
 556                         for (int i = 0; i < outputNames.size(); i++) {  m->mothurOut(outputNames[i]); m->mothurOutEndLine();    }
 557                         m->mothurOutEndLine();
 558
 559
 560                         m->mothurOutEndLine();
 561                         m->mothurOut("It took " + toString(time(NULL) - start) + " secs to classify " + toString(numFastaSeqs) + " sequences."); m->mothurOutEndLine(); m->mothurOutEndLine();
 562                 }
 563
 564                 delete classify;
 565                 return 0;
 566         }
 567         catch(exception& e) {
 568                 m->errorOut(e, "ClassifySeqsCommand", "execute");
 569                 exit(1);
 570         }
 571 }
 572
 573 /**************************************************************************************************/
 574 string ClassifySeqsCommand::addUnclassifieds(string tax, int maxlevel) {
 575         try{
 576                 string newTax, taxon;
 577                 int level = 0;
 578
 579                 //keep what you have counting the levels
 580                 while (tax.find_first_of(';') != -1) {
 581                         //get taxon
 582                         taxon = tax.substr(0,tax.find_first_of(';'))+';';
 583                         tax = tax.substr(tax.find_first_of(';')+1, tax.length());
 584                         newTax += taxon;
 585                         level++;
 586                 }
 587
 588                 //add "unclassified" until you reach maxLevel
 589                 while (level < maxlevel) {
 590                         newTax += "unclassified;";
 591                         level++;
 592                 }
 593
 594                 return newTax;
 595         }
 596         catch(exception& e) {
 597                 m->errorOut(e, "ClassifySeqsCommand", "addUnclassifieds");
 598                 exit(1);
 599         }
 600 }
 601
 602 /**************************************************************************************************/
 603
 604 void ClassifySeqsCommand::createProcesses(string taxFileName, string tempTaxFile, string filename) {
 605         try {
 606 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
 607                 int process = 0;
 608                 //              processIDS.resize(0);
 609
 610                 //loop through and create all the processes you want
 611                 while (process != processors) {
 612                         int pid = fork();
 613
 614                         if (pid > 0) {
 615                                 processIDS.push_back(pid);  //create map from line number to pid so you can append files in correct order later
 616                                 process++;
 617                         }else if (pid == 0){
 618                                 driver(lines[process], taxFileName + toString(getpid()) + ".temp", tempTaxFile + toString(getpid()) + ".temp", filename);
 619                                 exit(0);
 620                         }else { m->mothurOut("unable to spawn the necessary processes."); m->mothurOutEndLine(); exit(0); }
 621                 }
 622
 623                 //force parent to wait until all the processes are done
 624                 for (int i=0;i<processors;i++) {
 625                         int temp = processIDS[i];
 626                         wait(&temp);
 627                 }
 628 #endif
 629         }
 630         catch(exception& e) {
 631                 m->errorOut(e, "ClassifySeqsCommand", "createProcesses");
 632                 exit(1);
 633         }
 634 }
 635 /**************************************************************************************************/
 636
 637 void ClassifySeqsCommand::appendTaxFiles(string temp, string filename) {
 638         try{
 639
 640                 ofstream output;
 641                 ifstream input;
 642                 openOutputFileAppend(filename, output);
 643                 openInputFile(temp, input);
 644
 645                 while(char c = input.get()){
 646                         if(input.eof())         {       break;                  }
 647                         else                            {       output << c;    }
 648                 }
 649
 650                 input.close();
 651                 output.close();
 652         }
 653         catch(exception& e) {
 654                 m->errorOut(e, "ClassifySeqsCommand", "appendTaxFiles");
 655                 exit(1);
 656         }
 657 }
 658
 659 //**********************************************************************************************************************
 660
 661 int ClassifySeqsCommand::driver(linePair* line, string taxFName, string tempTFName, string filename){
 662         try {
 663                 ofstream outTax;
 664                 openOutputFile(taxFName, outTax);
 665
 666                 ofstream outTaxSimple;
 667                 openOutputFile(tempTFName, outTaxSimple);
 668
 669                 ifstream inFASTA;
 670                 openInputFile(filename, inFASTA);
 671
 672                 inFASTA.seekg(line->start);
 673
 674                 string taxonomy;
 675
 676                 for(int i=0;i<line->numSeqs;i++){
 677                         if (m->control_pressed) { return 0; }
 678
 679                         Sequence* candidateSeq = new Sequence(inFASTA);
 680
 681                         if (candidateSeq->getName() != "") {
 682                                 taxonomy = classify->getTaxonomy(candidateSeq);
 683
 684                                 if (m->control_pressed) { delete candidateSeq; return 0; }
 685
 686                                 if (taxonomy != "bad seq") {
 687                                         //output confidence scores or not
 688                                         if (probs) {
 689                                                 outTax << candidateSeq->getName() << '\t' << taxonomy << endl;
 690                                         }else{
 691                                                 outTax << candidateSeq->getName() << '\t' << classify->getSimpleTax() << endl;
 692                                         }
 693
 694                                         outTaxSimple << candidateSeq->getName() << '\t' << classify->getSimpleTax() << endl;
 695                                 }
 696                         }
 697                         delete candidateSeq;
 698
 699                         if((i+1) % 100 == 0){
 700                                 m->mothurOut("Classifying sequence " + toString(i+1)); m->mothurOutEndLine();
 701                         }
 702                 }
 703
 704                 inFASTA.close();
 705                 outTax.close();
 706                 outTaxSimple.close();
 707
 708                 return 1;
 709         }
 710         catch(exception& e) {
 711                 m->errorOut(e, "ClassifySeqsCommand", "driver");
 712                 exit(1);
 713         }
 714 }
 715 //**********************************************************************************************************************
 716 #ifdef USE_MPI
 717 int ClassifySeqsCommand::driverMPI(int start, int num, MPI_File& inMPI, MPI_File& newFile, MPI_File& tempFile, vector<long>& MPIPos){
 718         try {
 719                 MPI_Status statusNew;
 720                 MPI_Status statusTemp;
 721                 MPI_Status status;
 722
 723                 int pid;
 724                 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 725
 726                 string taxonomy;
 727                 string outputString;
 728
 729                 for(int i=0;i<num;i++){
 730
 731                         if (m->control_pressed) { return 0; }
 732
 733                         //read next sequence
 734                         int length = MPIPos[start+i+1] - MPIPos[start+i];
 735                         char buf4[length];
 736                         MPI_File_read_at(inMPI, MPIPos[start+i], buf4, length, MPI_CHAR, &status);
 737
 738                         string tempBuf = buf4;
 739                         if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length);  }
 740                         istringstream iss (tempBuf,istringstream::in);
 741
 742                         Sequence* candidateSeq = new Sequence(iss);
 743
 744                         if (candidateSeq->getName() != "") {
 745                                 taxonomy = classify->getTaxonomy(candidateSeq);
 746
 747                                 if (taxonomy != "bad seq") {
 748                                         //output confidence scores or not
 749                                         if (probs) {
 750                                                 outputString =  candidateSeq->getName() + "\t" + taxonomy + "\n";
 751                                         }else{
 752                                                 outputString =  candidateSeq->getName() + "\t" + classify->getSimpleTax() + "\n";
 753                                         }
 754
 755                                         int length = outputString.length();
 756                                         char buf2[length];
 757                                         strcpy(buf2, outputString.c_str());
 758
 759                                         MPI_File_write_shared(newFile, buf2, length, MPI_CHAR, &statusNew);
 760
 761                                         outputString =  candidateSeq->getName() + "\t" + classify->getSimpleTax() + "\n";
 762                                         length = outputString.length();
 763                                         char buf[length];
 764                                         strcpy(buf, outputString.c_str());
 765
 766                                         MPI_File_write_shared(tempFile, buf, length, MPI_CHAR, &statusTemp);
 767                                 }
 768                         }
 769                         delete candidateSeq;
 770
 771                         if((i+1) % 100 == 0){   cout << "Classifying sequence " << (i+1) << endl;       }
 772                 }
 773
 774                 if(num % 100 != 0){     cout << "Classifying sequence " << (num) << endl;       }
 775
 776
 777                 return 1;
 778         }
 779         catch(exception& e) {
 780                 m->errorOut(e, "ClassifySeqsCommand", "driverMPI");
 781                 exit(1);
 782         }
 783 }
 784
 785 //**********************************************************************************************************************
 786 int ClassifySeqsCommand::MPIReadNamesFile(string nameFilename){
 787         try {
 788
 789                 nameMap.clear(); //remove old names
 790
 791                 MPI_File inMPI;
 792                 MPI_Offset size;
 793                 MPI_Status status;
 794
 795                 char inFileName[nameFilename.length()];
 796                 strcpy(inFileName, nameFilename.c_str());
 797
 798                 MPI_File_open(MPI_COMM_WORLD, inFileName, MPI_MODE_RDONLY, MPI_INFO_NULL, &inMPI);
 799                 MPI_File_get_size(inMPI, &size);
 800
 801                 char buffer[size];
 802                 MPI_File_read(inMPI, buffer, size, MPI_CHAR, &status);
 803
 804                 string tempBuf = buffer;
 805                 if (tempBuf.length() > size) { tempBuf = tempBuf.substr(0, size);  }
 806                 istringstream iss (tempBuf,istringstream::in);
 807
 808                 string firstCol, secondCol;
 809                 while(!iss.eof()) {
 810                         iss >> firstCol >> secondCol; gobble(iss);
 811                         nameMap[firstCol] = getNumNames(secondCol);  //ex. seq1 seq1,seq3,seq5 -> seq1 = 3.
 812                 }
 813
 814                 MPI_File_close(&inMPI);
 815
 816                 return 1;
 817         }
 818         catch(exception& e) {
 819                 m->errorOut(e, "ClassifySeqsCommand", "MPIReadNamesFile");
 820                 exit(1);
 821         }
 822 }
 823 #endif
 824 /**************************************************************************************************/