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 = 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 = 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 = 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                                 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 = 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                                         #ifdef USE_MPI
 102                                                 int pid;
 103                                                 MPI_Comm_size(MPI_COMM_WORLD, &processors); //set processors to the number of mpi processes running
 104                                                 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 105
 106                                                 if (pid == 0) {
 107                                         #endif
 108
 109                                         ifstream in;
 110                                         ableToOpen = openInputFile(fastaFileNames[i], in, "noerror");
 111
 112                                         //if you can't open it, try default location
 113                                         if (ableToOpen == 1) {
 114                                                 if (m->getDefaultPath() != "") { //default path is set
 115                                                         string tryPath = m->getDefaultPath() + getSimpleName(fastaFileNames[i]);
 116                                                         m->mothurOut("Unable to open " + fastaFileNames[i] + ". Trying default " + tryPath); m->mothurOutEndLine();
 117                                                         ableToOpen = openInputFile(tryPath, in, "noerror");
 118                                                         fastaFileNames[i] = tryPath;
 119                                                 }
 120                                         }
 121                                         in.close();
 122
 123                                         #ifdef USE_MPI
 124                                                         for (int j = 1; j < processors; j++) {
 125                                                                 MPI_Send(&ableToOpen, 1, MPI_INT, j, 2001, MPI_COMM_WORLD);
 126                                                         }
 127                                                 }else{
 128                                                         MPI_Status status;
 129                                                         MPI_Recv(&ableToOpen, 1, MPI_INT, 0, 2001, MPI_COMM_WORLD, &status);
 130                                                 }
 131
 132                                         #endif
 133
 134                                         if (ableToOpen == 1) {
 135                                                 m->mothurOut("Unable to open " + fastaFileNames[i] + ". It will be disregarded."); m->mothurOutEndLine();
 136                                                 //erase from file list
 137                                                 fastaFileNames.erase(fastaFileNames.begin()+i);
 138                                                 i--;
 139                                         }
 140
 141                                 }
 142
 143                                 //make sure there is at least one valid file left
 144                                 if (fastaFileNames.size() == 0) { m->mothurOut("no valid files."); m->mothurOutEndLine(); abort = true; }
 145                         }
 146
 147
 148                         taxonomyFileName = validParameter.validFile(parameters, "taxonomy", true);
 149                         if (taxonomyFileName == "not found") {
 150                                 m->mothurOut("taxonomy is a required parameter for the classify.seqs command.");
 151                                 m->mothurOutEndLine();
 152                                 abort = true;
 153                         }
 154                         else if (taxonomyFileName == "not open") { abort = true; }
 155
 156
 157                         namefile = validParameter.validFile(parameters, "name", false);
 158                         if (namefile == "not found") { namefile = "";  }
 159
 160                         else {
 161                                 splitAtDash(namefile, namefileNames);
 162
 163                                 //go through files and make sure they are good, if not, then disregard them
 164                                 for (int i = 0; i < namefileNames.size(); i++) {
 165                                         if (inputDir != "") {
 166                                                 string path = hasPath(namefileNames[i]);
 167                                                 //if the user has not given a path then, add inputdir. else leave path alone.
 168                                                 if (path == "") {       namefileNames[i] = inputDir + namefileNames[i];         }
 169                                         }
 170                                         int ableToOpen;
 171
 172                                         #ifdef USE_MPI
 173                                                 int pid;
 174                                                 MPI_Comm_size(MPI_COMM_WORLD, &processors); //set processors to the number of mpi processes running
 175                                                 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 176
 177                                                 if (pid == 0) {
 178                                         #endif
 179
 180                                         ifstream in;
 181                                         ableToOpen = openInputFile(namefileNames[i], in, "noerror");
 182
 183                                         //if you can't open it, try default location
 184                                         if (ableToOpen == 1) {
 185                                                 if (m->getDefaultPath() != "") { //default path is set
 186                                                         string tryPath = m->getDefaultPath() + getSimpleName(namefileNames[i]);
 187                                                         m->mothurOut("Unable to open " + namefileNames[i] + ". Trying default " + tryPath); m->mothurOutEndLine();
 188                                                         ableToOpen = openInputFile(tryPath, in, "noerror");
 189                                                         namefileNames[i] = tryPath;
 190                                                 }
 191                                         }
 192                                         in.close();
 193
 194                                         #ifdef USE_MPI
 195                                                         for (int j = 1; j < processors; j++) {
 196                                                                 MPI_Send(&ableToOpen, 1, MPI_INT, j, 2001, MPI_COMM_WORLD);
 197                                                         }
 198                                                 }else{
 199                                                         MPI_Status status;
 200                                                         MPI_Recv(&ableToOpen, 1, MPI_INT, 0, 2001, MPI_COMM_WORLD, &status);
 201                                                 }
 202
 203                                         #endif
 204
 205                                         if (ableToOpen == 1) {
 206                                                 m->mothurOut("Unable to open " + namefileNames[i] + ". It will be disregarded."); m->mothurOutEndLine();  abort = true;
 207                                                 //erase from file list
 208                                                 namefileNames.erase(namefileNames.begin()+i);
 209                                                 i--;
 210                                         }
 211
 212                                 }
 213                         }
 214
 215                         if (namefile != "") {
 216                                 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(); }
 217                         }
 218
 219                         groupfile = validParameter.validFile(parameters, "group", false);
 220                         if (groupfile == "not found") { groupfile = "";  }
 221                         else {
 222                                 splitAtDash(groupfile, groupfileNames);
 223
 224                                 //go through files and make sure they are good, if not, then disregard them
 225                                 for (int i = 0; i < groupfileNames.size(); i++) {
 226                                         if (inputDir != "") {
 227                                                 string path = hasPath(groupfileNames[i]);
 228                                                 //if the user has not given a path then, add inputdir. else leave path alone.
 229                                                 if (path == "") {       groupfileNames[i] = inputDir + groupfileNames[i];               }
 230                                         }
 231                                         int ableToOpen;
 232
 233                                         #ifdef USE_MPI
 234                                                 int pid;
 235                                                 MPI_Comm_size(MPI_COMM_WORLD, &processors); //set processors to the number of mpi processes running
 236                                                 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 237
 238                                                 if (pid == 0) {
 239                                         #endif
 240
 241                                         ifstream in;
 242                                         ableToOpen = openInputFile(groupfileNames[i], in, "noerror");
 243
 244                                         //if you can't open it, try default location
 245                                         if (ableToOpen == 1) {
 246                                                 if (m->getDefaultPath() != "") { //default path is set
 247                                                         string tryPath = m->getDefaultPath() + getSimpleName(groupfileNames[i]);
 248                                                         m->mothurOut("Unable to open " + groupfileNames[i] + ". Trying default " + tryPath); m->mothurOutEndLine();
 249                                                         ableToOpen = openInputFile(tryPath, in, "noerror");
 250                                                         groupfileNames[i] = tryPath;
 251                                                 }
 252                                         }
 253                                         in.close();
 254
 255                                         #ifdef USE_MPI
 256                                                         for (int j = 1; j < processors; j++) {
 257                                                                 MPI_Send(&ableToOpen, 1, MPI_INT, j, 2001, MPI_COMM_WORLD);
 258                                                         }
 259                                                 }else{
 260                                                         MPI_Status status;
 261                                                         MPI_Recv(&ableToOpen, 1, MPI_INT, 0, 2001, MPI_COMM_WORLD, &status);
 262                                                 }
 263
 264                                         #endif
 265
 266                                         if (ableToOpen == 1) {
 267                                                 m->mothurOut("Unable to open " + groupfileNames[i] + ". It will be disregarded."); m->mothurOutEndLine(); groupfileNames[i] = "";
 268                                                 //erase from file list
 269                                                 groupfileNames.erase(groupfileNames.begin()+i);
 270                                                 i--;
 271                                         }
 272                                 }
 273                         }
 274
 275                         if (groupfile != "") {
 276                                 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(); }
 277                         }else {
 278                                 for (int i = 0; i < fastaFileNames.size(); i++) {  groupfileNames.push_back("");  }
 279                         }
 280
 281                         //check for optional parameter and set defaults
 282                         // ...at some point should added some additional type checking...
 283                         string temp;
 284                         temp = validParameter.validFile(parameters, "ksize", false);            if (temp == "not found"){       temp = "8";                             }
 285                         convert(temp, kmerSize);
 286
 287                         temp = validParameter.validFile(parameters, "processors", false);       if (temp == "not found"){       temp = "1";                             }
 288                         convert(temp, processors);
 289
 290                         search = validParameter.validFile(parameters, "search", false);         if (search == "not found"){     search = "kmer";                }
 291
 292                         method = validParameter.validFile(parameters, "method", false);         if (method == "not found"){     method = "bayesian";    }
 293
 294                         temp = validParameter.validFile(parameters, "match", false);            if (temp == "not found"){       temp = "1.0";                   }
 295                         convert(temp, match);
 296
 297                         temp = validParameter.validFile(parameters, "mismatch", false);         if (temp == "not found"){       temp = "-1.0";                  }
 298                         convert(temp, misMatch);
 299
 300                         temp = validParameter.validFile(parameters, "gapopen", false);          if (temp == "not found"){       temp = "-2.0";                  }
 301                         convert(temp, gapOpen);
 302
 303                         temp = validParameter.validFile(parameters, "gapextend", false);        if (temp == "not found"){       temp = "-1.0";                  }
 304                         convert(temp, gapExtend);
 305
 306                         temp = validParameter.validFile(parameters, "numwanted", false);        if (temp == "not found"){       temp = "10";                    }
 307                         convert(temp, numWanted);
 308
 309                         temp = validParameter.validFile(parameters, "cutoff", false);           if (temp == "not found"){       temp = "0";                             }
 310                         convert(temp, cutoff);
 311
 312                         temp = validParameter.validFile(parameters, "probs", false);            if (temp == "not found"){       temp = "true";                  }
 313                         probs = isTrue(temp);
 314
 315                         temp = validParameter.validFile(parameters, "iters", false);            if (temp == "not found") { temp = "100";                        }
 316                         convert(temp, iters);
 317
 318
 319
 320                         if ((method == "bayesian") && (search != "kmer"))  {
 321                                 m->mothurOut("The bayesian method requires the kmer search." + search + "will be disregarded." ); m->mothurOutEndLine();
 322                                 search = "kmer";
 323                         }
 324                 }
 325
 326         }
 327         catch(exception& e) {
 328                 m->errorOut(e, "ClassifySeqsCommand", "ClassifySeqsCommand");
 329                 exit(1);
 330         }
 331 }
 332
 333 //**********************************************************************************************************************
 334
 335 ClassifySeqsCommand::~ClassifySeqsCommand(){
 336
 337         if (abort == false) {
 338                 for (int i = 0; i < lines.size(); i++) {  delete lines[i];  }  lines.clear();
 339         }
 340 }
 341
 342 //**********************************************************************************************************************
 343
 344 void ClassifySeqsCommand::help(){
 345         try {
 346                 m->mothurOut("The classify.seqs command reads a fasta file containing sequences and creates a .taxonomy file and a .tax.summary file.\n");
 347                 m->mothurOut("The classify.seqs command parameters are template, fasta, name, search, ksize, method, taxonomy, processors, match, mismatch, gapopen, gapextend, numwanted and probs.\n");
 348                 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");
 349                 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");
 350                 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");
 351                 m->mothurOut("The group parameter allows you add a group file so you can have the summary totals broken up by group.\n");
 352                 m->mothurOut("The method parameter allows you to specify classification method to use.  Your options are: bayesian and knn. The default is bayesian.\n");
 353                 m->mothurOut("The ksize parameter allows you to specify the kmer size for finding most similar template to candidate.  The default is 8.\n");
 354                 m->mothurOut("The processors parameter allows you to specify the number of processors to use. The default is 1.\n");
 355                 #ifdef USE_MPI
 356                 m->mothurOut("When using MPI, the processors parameter is set to the number of MPI processes running. \n");
 357                 #endif
 358                 m->mothurOut("The match parameter allows you to specify the bonus for having the same base. The default is 1.0.\n");
 359                 m->mothurOut("The mistmatch parameter allows you to specify the penalty for having different bases.  The default is -1.0.\n");
 360                 m->mothurOut("The gapopen parameter allows you to specify the penalty for opening a gap in an alignment. The default is -2.0.\n");
 361                 m->mothurOut("The gapextend parameter allows you to specify the penalty for extending a gap in an alignment.  The default is -1.0.\n");
 362                 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");
 363                 m->mothurOut("The cutoff parameter allows you to specify a bootstrap confidence threshold for your taxonomy.  The default is 0.\n");
 364                 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");
 365                 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");
 366                 m->mothurOut("The classify.seqs command should be in the following format: \n");
 367                 m->mothurOut("classify.seqs(template=yourTemplateFile, fasta=yourFastaFile, method=yourClassificationMethod, search=yourSearchmethod, ksize=yourKmerSize, taxonomy=yourTaxonomyFile, processors=yourProcessors) \n");
 368                 m->mothurOut("Example classify.seqs(fasta=amazon.fasta, template=core.filtered, method=knn, search=gotoh, ksize=8, processors=2)\n");
 369                 m->mothurOut("The .taxonomy file consists of 2 columns: 1 = your sequence name, 2 = the taxonomy for your sequence. \n");
 370                 m->mothurOut("The .tax.summary is a summary of the different taxonomies represented in your fasta file. \n");
 371                 m->mothurOut("Note: No spaces between parameter labels (i.e. fasta), '=' and parameters (i.e.yourFastaFile).\n\n");
 372         }
 373         catch(exception& e) {
 374                 m->errorOut(e, "ClassifySeqsCommand", "help");
 375                 exit(1);
 376         }
 377 }
 378
 379
 380 //**********************************************************************************************************************
 381
 382 int ClassifySeqsCommand::execute(){
 383         try {
 384                 if (abort == true) {    return 0;       }
 385
 386                 if(method == "bayesian"){       classify = new Bayesian(taxonomyFileName, templateFileName, search, kmerSize, cutoff, iters);           }
 387                 else if(method == "knn"){       classify = new Knn(taxonomyFileName, templateFileName, search, kmerSize, gapOpen, gapExtend, match, misMatch, numWanted);                               }
 388                 else {
 389                         m->mothurOut(search + " is not a valid method option. I will run the command using bayesian.");
 390                         m->mothurOutEndLine();
 391                         classify = new Bayesian(taxonomyFileName, templateFileName, search, kmerSize, cutoff, iters);
 392                 }
 393
 394                 if (m->control_pressed) { delete classify; return 0; }
 395
 396                 vector<string> outputNames;
 397
 398                 for (int s = 0; s < fastaFileNames.size(); s++) {
 399
 400                         m->mothurOut("Classifying sequences from " + fastaFileNames[s] + " ..." ); m->mothurOutEndLine();
 401
 402                         if (outputDir == "") { outputDir += hasPath(fastaFileNames[s]); }
 403                         string newTaxonomyFile = outputDir + getRootName(getSimpleName(fastaFileNames[s])) + getRootName(getSimpleName(taxonomyFileName)) + "taxonomy";
 404                         string tempTaxonomyFile = outputDir + getRootName(getSimpleName(fastaFileNames[s])) + "taxonomy.temp";
 405                         string taxSummary = outputDir + getRootName(getSimpleName(fastaFileNames[s])) + getRootName(getSimpleName(taxonomyFileName)) + "tax.summary";
 406
 407                         outputNames.push_back(newTaxonomyFile);
 408                         outputNames.push_back(taxSummary);
 409
 410                         int start = time(NULL);
 411                         int numFastaSeqs = 0;
 412                         for (int i = 0; i < lines.size(); i++) {  delete lines[i];  }  lines.clear();
 413
 414 #ifdef USE_MPI
 415                                 int pid, end, numSeqsPerProcessor;
 416                                 int tag = 2001;
 417                                 vector<unsigned long int> MPIPos;
 418
 419                                 MPI_Status status;
 420                                 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 421                                 MPI_Comm_size(MPI_COMM_WORLD, &processors);
 422
 423                                 MPI_File inMPI;
 424                                 MPI_File outMPINewTax;
 425                                 MPI_File outMPITempTax;
 426
 427                                 int outMode=MPI_MODE_CREATE|MPI_MODE_WRONLY;
 428                                 int inMode=MPI_MODE_RDONLY;
 429
 430                                 //char* outNewTax = new char[newTaxonomyFile.length()];
 431                                 //memcpy(outNewTax, newTaxonomyFile.c_str(), newTaxonomyFile.length());
 432
 433                                 char outNewTax[1024];
 434                                 strcpy(outNewTax, newTaxonomyFile.c_str());
 435
 436                                 //char* outTempTax = new char[tempTaxonomyFile.length()];
 437                                 //memcpy(outTempTax, tempTaxonomyFile.c_str(), tempTaxonomyFile.length());
 438
 439                                 char outTempTax[1024];
 440                                 strcpy(outTempTax, tempTaxonomyFile.c_str());
 441
 442                                 //char* inFileName = new char[fastaFileNames[s].length()];
 443                                 //memcpy(inFileName, fastaFileNames[s].c_str(), fastaFileNames[s].length());
 444
 445                                 char inFileName[1024];
 446                                 strcpy(inFileName, fastaFileNames[s].c_str());
 447
 448                                 MPI_File_open(MPI_COMM_WORLD, inFileName, inMode, MPI_INFO_NULL, &inMPI);  //comm, filename, mode, info, filepointer
 449                                 MPI_File_open(MPI_COMM_WORLD, outNewTax, outMode, MPI_INFO_NULL, &outMPINewTax);
 450                                 MPI_File_open(MPI_COMM_WORLD, outTempTax, outMode, MPI_INFO_NULL, &outMPITempTax);
 451
 452                                 //delete outNewTax;
 453                                 //delete outTempTax;
 454                                 //delete inFileName;
 455
 456                                 if (m->control_pressed) {  MPI_File_close(&inMPI);  MPI_File_close(&outMPINewTax);   MPI_File_close(&outMPITempTax);  delete classify; return 0;  }
 457
 458                                 if (pid == 0) { //you are the root process
 459
 460                                         MPIPos = setFilePosFasta(fastaFileNames[s], numFastaSeqs); //fills MPIPos, returns numSeqs
 461
 462                                         //send file positions to all processes
 463                                         for(int i = 1; i < processors; i++) {
 464                                                 MPI_Send(&numFastaSeqs, 1, MPI_INT, i, tag, MPI_COMM_WORLD);
 465                                                 MPI_Send(&MPIPos[0], (numFastaSeqs+1), MPI_LONG, i, tag, MPI_COMM_WORLD);
 466                                         }
 467
 468                                         //figure out how many sequences you have to align
 469                                         numSeqsPerProcessor = numFastaSeqs / processors;
 470                                         int startIndex =  pid * numSeqsPerProcessor;
 471                                         if(pid == (processors - 1)){    numSeqsPerProcessor = numFastaSeqs - pid * numSeqsPerProcessor;         }
 472
 473
 474                                         //align your part
 475                                         driverMPI(startIndex, numSeqsPerProcessor, inMPI, outMPINewTax, outMPITempTax, MPIPos);
 476
 477                                         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;  }
 478
 479                                         for (int i = 1; i < processors; i++) {
 480                                                 int done;
 481                                                 MPI_Recv(&done, 1, MPI_INT, i, tag, MPI_COMM_WORLD, &status);
 482                                         }
 483                                 }else{ //you are a child process
 484                                         MPI_Recv(&numFastaSeqs, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
 485                                         MPIPos.resize(numFastaSeqs+1);
 486                                         MPI_Recv(&MPIPos[0], (numFastaSeqs+1), MPI_LONG, 0, tag, MPI_COMM_WORLD, &status);
 487
 488                                         //figure out how many sequences you have to align
 489                                         numSeqsPerProcessor = numFastaSeqs / processors;
 490                                         int startIndex =  pid * numSeqsPerProcessor;
 491                                         if(pid == (processors - 1)){    numSeqsPerProcessor = numFastaSeqs - pid * numSeqsPerProcessor;         }
 492
 493
 494                                         //align your part
 495                                         driverMPI(startIndex, numSeqsPerProcessor, inMPI, outMPINewTax, outMPITempTax, MPIPos);
 496
 497                                         if (m->control_pressed) {  MPI_File_close(&inMPI);  MPI_File_close(&outMPINewTax);   MPI_File_close(&outMPITempTax);  delete classify; return 0;  }
 498
 499                                         int done = 0;
 500                                         MPI_Send(&done, 1, MPI_INT, 0, tag, MPI_COMM_WORLD);
 501                                 }
 502
 503                                 //close files
 504                                 MPI_File_close(&inMPI);
 505                                 MPI_File_close(&outMPINewTax);
 506                                 MPI_File_close(&outMPITempTax);
 507                                 MPI_Barrier(MPI_COMM_WORLD); //make everyone wait - just in case
 508
 509 #else
 510                 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
 511                         if(processors == 1){
 512                                 ifstream inFASTA;
 513                                 openInputFile(fastaFileNames[s], inFASTA);
 514                                 getNumSeqs(inFASTA, numFastaSeqs);
 515                                 inFASTA.close();
 516
 517                                 lines.push_back(new linePair(0, numFastaSeqs));
 518
 519                                 driver(lines[0], newTaxonomyFile, tempTaxonomyFile, fastaFileNames[s]);
 520                         }
 521                         else{
 522                                 vector<unsigned long int> positions;
 523                                 processIDS.resize(0);
 524
 525                                 ifstream inFASTA;
 526                                 openInputFile(fastaFileNames[s], inFASTA);
 527
 528                                 string input;
 529                                 while(!inFASTA.eof()){
 530                                         input = getline(inFASTA);
 531                                         if (input.length() != 0) {
 532                                                 if(input[0] == '>'){    unsigned long int pos = inFASTA.tellg(); positions.push_back(pos - input.length() - 1); }
 533                                         }
 534                                 }
 535                                 inFASTA.close();
 536
 537                                 numFastaSeqs = positions.size();
 538
 539                                 int numSeqsPerProcessor = numFastaSeqs / processors;
 540
 541                                 for (int i = 0; i < processors; i++) {
 542                                         unsigned long int startPos = positions[ i * numSeqsPerProcessor ];
 543                                         if(i == processors - 1){
 544                                                 numSeqsPerProcessor = numFastaSeqs - i * numSeqsPerProcessor;
 545                                         }
 546                                         lines.push_back(new linePair(startPos, numSeqsPerProcessor));
 547                                 }
 548                                 createProcesses(newTaxonomyFile, tempTaxonomyFile, fastaFileNames[s]);
 549
 550                                 rename((newTaxonomyFile + toString(processIDS[0]) + ".temp").c_str(), newTaxonomyFile.c_str());
 551                                 rename((tempTaxonomyFile + toString(processIDS[0]) + ".temp").c_str(), tempTaxonomyFile.c_str());
 552
 553                                 for(int i=1;i<processors;i++){
 554                                         appendTaxFiles((newTaxonomyFile + toString(processIDS[i]) + ".temp"), newTaxonomyFile);
 555                                         appendTaxFiles((tempTaxonomyFile + toString(processIDS[i]) + ".temp"), tempTaxonomyFile);
 556                                         remove((newTaxonomyFile + toString(processIDS[i]) + ".temp").c_str());
 557                                         remove((tempTaxonomyFile + toString(processIDS[i]) + ".temp").c_str());
 558                                 }
 559
 560                         }
 561         #else
 562                         ifstream inFASTA;
 563                         openInputFile(fastaFileNames[s], inFASTA);
 564                         getNumSeqs(inFASTA, numFastaSeqs);
 565                         inFASTA.close();
 566
 567                         lines.push_back(new linePair(0, numFastaSeqs));
 568
 569                         driver(lines[0], newTaxonomyFile, tempTaxonomyFile, fastaFileNames[s]);
 570         #endif
 571 #endif
 572
 573                 m->mothurOutEndLine();
 574                 m->mothurOut("It took " + toString(time(NULL) - start) + " secs to classify " + toString(numFastaSeqs) + " sequences."); m->mothurOutEndLine(); m->mothurOutEndLine();
 575                 start = time(NULL);
 576
 577
 578                 #ifdef USE_MPI
 579                         if (pid == 0) {  //this part does not need to be paralellized
 580
 581                                 if(namefile != "") { m->mothurOut("Reading " + namefileNames[s] + "..."); cout.flush();  MPIReadNamesFile(namefileNames[s]);  m->mothurOut("  Done."); m->mothurOutEndLine(); }
 582                 #else
 583                         //read namefile
 584                         if(namefile != "") {
 585
 586                             m->mothurOut("Reading " + namefileNames[s] + "..."); cout.flush();
 587
 588                                 nameMap.clear(); //remove old names
 589
 590                                 ifstream inNames;
 591                                 openInputFile(namefileNames[s], inNames);
 592
 593                                 string firstCol, secondCol;
 594                                 while(!inNames.eof()) {
 595                                         inNames >> firstCol >> secondCol; gobble(inNames);
 596
 597                                         vector<string> temp;
 598                                         splitAtComma(secondCol, temp);
 599
 600                                         nameMap[firstCol] = temp;
 601                                 }
 602                                 inNames.close();
 603
 604                                 m->mothurOut("  Done."); m->mothurOutEndLine();
 605                         }
 606                 #endif
 607
 608                         string group = "";
 609                         if (groupfile != "") {  group = groupfileNames[s]; }
 610
 611                         PhyloSummary taxaSum(taxonomyFileName, group);
 612
 613                         if (m->control_pressed) {  for (int i = 0; i < outputNames.size(); i++) {       remove(outputNames[i].c_str()); } delete classify; return 0; }
 614
 615                         if (namefile == "") {  taxaSum.summarize(tempTaxonomyFile);  }
 616                         else {
 617                                 ifstream in;
 618                                 openInputFile(tempTaxonomyFile, in);
 619
 620                                 //read in users taxonomy file and add sequences to tree
 621                                 string name, taxon;
 622
 623                                 while(!in.eof()){
 624                                         in >> name >> taxon; gobble(in);
 625
 626                                         itNames = nameMap.find(name);
 627
 628                                         if (itNames == nameMap.end()) {
 629                                                 m->mothurOut(name + " is not in your name file please correct."); m->mothurOutEndLine(); exit(1);
 630                                         }else{
 631                                                 for (int i = 0; i < itNames->second.size(); i++) {
 632                                                         taxaSum.addSeqToTree(itNames->second[i], taxon);  //add it as many times as there are identical seqs
 633                                                 }
 634                                                 itNames->second.clear();
 635                                                 nameMap.erase(itNames->first);
 636                                         }
 637                                 }
 638                                 in.close();
 639                         }
 640                         remove(tempTaxonomyFile.c_str());
 641
 642                         if (m->control_pressed) {  for (int i = 0; i < outputNames.size(); i++) {       remove(outputNames[i].c_str()); } delete classify; return 0; }
 643
 644                         //print summary file
 645                         ofstream outTaxTree;
 646                         openOutputFile(taxSummary, outTaxTree);
 647                         taxaSum.print(outTaxTree);
 648                         outTaxTree.close();
 649
 650                         //output taxonomy with the unclassified bins added
 651                         ifstream inTax;
 652                         openInputFile(newTaxonomyFile, inTax);
 653
 654                         ofstream outTax;
 655                         string unclass = newTaxonomyFile + ".unclass.temp";
 656                         openOutputFile(unclass, outTax);
 657
 658                         //get maxLevel from phylotree so you know how many 'unclassified's to add
 659                         int maxLevel = taxaSum.getMaxLevel();
 660
 661                         //read taxfile - this reading and rewriting is done to preserve the confidence scores.
 662                         string name, taxon;
 663                         while (!inTax.eof()) {
 664                                 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; }
 665
 666                                 inTax >> name >> taxon; gobble(inTax);
 667
 668                                 string newTax = addUnclassifieds(taxon, maxLevel);
 669
 670                                 outTax << name << '\t' << newTax << endl;
 671                         }
 672                         inTax.close();
 673                         outTax.close();
 674
 675                         remove(newTaxonomyFile.c_str());
 676                         rename(unclass.c_str(), newTaxonomyFile.c_str());
 677
 678                         m->mothurOutEndLine();
 679                         m->mothurOut("It took " + toString(time(NULL) - start) + " secs to create the summary file for " + toString(numFastaSeqs) + " sequences."); m->mothurOutEndLine(); m->mothurOutEndLine();
 680
 681                         #ifdef USE_MPI
 682                                 }
 683                         #endif
 684
 685                         m->mothurOutEndLine();
 686                         m->mothurOut("Output File Names: "); m->mothurOutEndLine();
 687                         for (int i = 0; i < outputNames.size(); i++) {  m->mothurOut(outputNames[i]); m->mothurOutEndLine();    }
 688                         m->mothurOutEndLine();
 689                 }
 690
 691                 delete classify;
 692                 return 0;
 693         }
 694         catch(exception& e) {
 695                 m->errorOut(e, "ClassifySeqsCommand", "execute");
 696                 exit(1);
 697         }
 698 }
 699
 700 /**************************************************************************************************/
 701 string ClassifySeqsCommand::addUnclassifieds(string tax, int maxlevel) {
 702         try{
 703                 string newTax, taxon;
 704                 int level = 0;
 705
 706                 //keep what you have counting the levels
 707                 while (tax.find_first_of(';') != -1) {
 708                         //get taxon
 709                         taxon = tax.substr(0,tax.find_first_of(';'))+';';
 710                         tax = tax.substr(tax.find_first_of(';')+1, tax.length());
 711                         newTax += taxon;
 712                         level++;
 713                 }
 714
 715                 //add "unclassified" until you reach maxLevel
 716                 while (level < maxlevel) {
 717                         newTax += "unclassified;";
 718                         level++;
 719                 }
 720
 721                 return newTax;
 722         }
 723         catch(exception& e) {
 724                 m->errorOut(e, "ClassifySeqsCommand", "addUnclassifieds");
 725                 exit(1);
 726         }
 727 }
 728
 729 /**************************************************************************************************/
 730
 731 void ClassifySeqsCommand::createProcesses(string taxFileName, string tempTaxFile, string filename) {
 732         try {
 733 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
 734                 int process = 0;
 735                 //              processIDS.resize(0);
 736
 737                 //loop through and create all the processes you want
 738                 while (process != processors) {
 739                         int pid = fork();
 740
 741                         if (pid > 0) {
 742                                 processIDS.push_back(pid);  //create map from line number to pid so you can append files in correct order later
 743                                 process++;
 744                         }else if (pid == 0){
 745                                 driver(lines[process], taxFileName + toString(getpid()) + ".temp", tempTaxFile + toString(getpid()) + ".temp", filename);
 746                                 exit(0);
 747                         }else { m->mothurOut("unable to spawn the necessary processes."); m->mothurOutEndLine(); exit(0); }
 748                 }
 749
 750                 //force parent to wait until all the processes are done
 751                 for (int i=0;i<processors;i++) {
 752                         int temp = processIDS[i];
 753                         wait(&temp);
 754                 }
 755 #endif
 756         }
 757         catch(exception& e) {
 758                 m->errorOut(e, "ClassifySeqsCommand", "createProcesses");
 759                 exit(1);
 760         }
 761 }
 762 /**************************************************************************************************/
 763
 764 void ClassifySeqsCommand::appendTaxFiles(string temp, string filename) {
 765         try{
 766
 767                 ofstream output;
 768                 ifstream input;
 769                 openOutputFileAppend(filename, output);
 770                 openInputFile(temp, input);
 771
 772                 while(char c = input.get()){
 773                         if(input.eof())         {       break;                  }
 774                         else                            {       output << c;    }
 775                 }
 776
 777                 input.close();
 778                 output.close();
 779         }
 780         catch(exception& e) {
 781                 m->errorOut(e, "ClassifySeqsCommand", "appendTaxFiles");
 782                 exit(1);
 783         }
 784 }
 785
 786 //**********************************************************************************************************************
 787
 788 int ClassifySeqsCommand::driver(linePair* line, string taxFName, string tempTFName, string filename){
 789         try {
 790                 ofstream outTax;
 791                 openOutputFile(taxFName, outTax);
 792
 793                 ofstream outTaxSimple;
 794                 openOutputFile(tempTFName, outTaxSimple);
 795
 796                 ifstream inFASTA;
 797                 openInputFile(filename, inFASTA);
 798
 799                 inFASTA.seekg(line->start);
 800
 801                 string taxonomy;
 802
 803                 for(int i=0;i<line->numSeqs;i++){
 804                         if (m->control_pressed) { return 0; }
 805
 806                         Sequence* candidateSeq = new Sequence(inFASTA); gobble(inFASTA);
 807
 808                         if (candidateSeq->getName() != "") {
 809                                 taxonomy = classify->getTaxonomy(candidateSeq);
 810
 811                                 if (m->control_pressed) { delete candidateSeq; return 0; }
 812
 813                                 if (taxonomy != "bad seq") {
 814                                         //output confidence scores or not
 815                                         if (probs) {
 816                                                 outTax << candidateSeq->getName() << '\t' << taxonomy << endl;
 817                                         }else{
 818                                                 outTax << candidateSeq->getName() << '\t' << classify->getSimpleTax() << endl;
 819                                         }
 820
 821                                         outTaxSimple << candidateSeq->getName() << '\t' << classify->getSimpleTax() << endl;
 822                                 }
 823                         }
 824                         delete candidateSeq;
 825
 826                         if((i+1) % 100 == 0){
 827                                 m->mothurOut("Classifying sequence " + toString(i+1)); m->mothurOutEndLine();
 828                         }
 829                 }
 830
 831                 inFASTA.close();
 832                 outTax.close();
 833                 outTaxSimple.close();
 834
 835                 return 1;
 836         }
 837         catch(exception& e) {
 838                 m->errorOut(e, "ClassifySeqsCommand", "driver");
 839                 exit(1);
 840         }
 841 }
 842 //**********************************************************************************************************************
 843 #ifdef USE_MPI
 844 int ClassifySeqsCommand::driverMPI(int start, int num, MPI_File& inMPI, MPI_File& newFile, MPI_File& tempFile, vector<unsigned long int>& MPIPos){
 845         try {
 846                 MPI_Status statusNew;
 847                 MPI_Status statusTemp;
 848                 MPI_Status status;
 849
 850                 int pid;
 851                 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 852
 853                 string taxonomy;
 854                 string outputString;
 855
 856                 for(int i=0;i<num;i++){
 857
 858                         if (m->control_pressed) { return 0; }
 859
 860                         //read next sequence
 861                         int length = MPIPos[start+i+1] - MPIPos[start+i];
 862                         char* buf4 = new char[length];
 863                         MPI_File_read_at(inMPI, MPIPos[start+i], buf4, length, MPI_CHAR, &status);
 864
 865                         string tempBuf = buf4;
 866                         if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length);  }
 867                         istringstream iss (tempBuf,istringstream::in);
 868                         delete buf4;
 869
 870                         Sequence* candidateSeq = new Sequence(iss);
 871
 872                         if (candidateSeq->getName() != "") {
 873                                 taxonomy = classify->getTaxonomy(candidateSeq);
 874
 875                                 if (taxonomy != "bad seq") {
 876                                         //output confidence scores or not
 877                                         if (probs) {
 878                                                 outputString =  candidateSeq->getName() + "\t" + taxonomy + "\n";
 879                                         }else{
 880                                                 outputString =  candidateSeq->getName() + "\t" + classify->getSimpleTax() + "\n";
 881                                         }
 882
 883                                         int length = outputString.length();
 884                                         char* buf2 = new char[length];
 885                                         memcpy(buf2, outputString.c_str(), length);
 886
 887                                         MPI_File_write_shared(newFile, buf2, length, MPI_CHAR, &statusNew);
 888                                         delete buf2;
 889
 890                                         outputString =  candidateSeq->getName() + "\t" + classify->getSimpleTax() + "\n";
 891                                         length = outputString.length();
 892                                         char* buf = new char[length];
 893                                         memcpy(buf, outputString.c_str(), length);
 894
 895                                         MPI_File_write_shared(tempFile, buf, length, MPI_CHAR, &statusTemp);
 896                                         delete buf;
 897                                 }
 898                         }
 899                         delete candidateSeq;
 900
 901                         if((i+1) % 100 == 0){   cout << "Classifying sequence " << (i+1) << endl;       }
 902                 }
 903
 904                 if(num % 100 != 0){     cout << "Classifying sequence " << (num) << endl;       }
 905
 906
 907                 return 1;
 908         }
 909         catch(exception& e) {
 910                 m->errorOut(e, "ClassifySeqsCommand", "driverMPI");
 911                 exit(1);
 912         }
 913 }
 914
 915 //**********************************************************************************************************************
 916 int ClassifySeqsCommand::MPIReadNamesFile(string nameFilename){
 917         try {
 918
 919                 nameMap.clear(); //remove old names
 920
 921                 MPI_File inMPI;
 922                 MPI_Offset size;
 923                 MPI_Status status;
 924
 925                 //char* inFileName = new char[nameFilename.length()];
 926                 //memcpy(inFileName, nameFilename.c_str(), nameFilename.length());
 927
 928                 char inFileName[1024];
 929                 strcpy(inFileName, nameFilename.c_str());
 930
 931                 MPI_File_open(MPI_COMM_WORLD, inFileName, MPI_MODE_RDONLY, MPI_INFO_NULL, &inMPI);
 932                 MPI_File_get_size(inMPI, &size);
 933                 //delete inFileName;
 934
 935                 char* buffer = new char[size];
 936                 MPI_File_read(inMPI, buffer, size, MPI_CHAR, &status);
 937
 938                 string tempBuf = buffer;
 939                 if (tempBuf.length() > size) { tempBuf = tempBuf.substr(0, size);  }
 940                 istringstream iss (tempBuf,istringstream::in);
 941                 delete buffer;
 942
 943                 string firstCol, secondCol;
 944                 while(!iss.eof()) {
 945                         iss >> firstCol >> secondCol; gobble(iss);
 946
 947                         vector<string> temp;
 948                         splitAtComma(secondCol, temp);
 949
 950                         nameMap[firstCol] = temp;
 951                 }
 952
 953                 MPI_File_close(&inMPI);
 954
 955                 return 1;
 956         }
 957         catch(exception& e) {
 958                 m->errorOut(e, "ClassifySeqsCommand", "MPIReadNamesFile");
 959                 exit(1);
 960         }
 961 }
 962 #endif
 963 /**************************************************************************************************/