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                         string RippedTaxName = getRootName(getSimpleName(taxonomyFileName));
 403                         RippedTaxName = getExtension(RippedTaxName.substr(0, RippedTaxName.length()-1));
 404                         if (RippedTaxName[0] == '.') { RippedTaxName = RippedTaxName.substr(1, RippedTaxName.length()); }
 405                         RippedTaxName +=  ".";
 406
 407                         if (outputDir == "") { outputDir += hasPath(fastaFileNames[s]); }
 408                         string newTaxonomyFile = outputDir + getRootName(getSimpleName(fastaFileNames[s])) + RippedTaxName + "taxonomy";
 409                         string tempTaxonomyFile = outputDir + getRootName(getSimpleName(fastaFileNames[s])) + "taxonomy.temp";
 410                         string taxSummary = outputDir + getRootName(getSimpleName(fastaFileNames[s])) + RippedTaxName + "tax.summary";
 411
 412                         if ((method == "knn") && (search == "distance")) {
 413                                 string DistName = outputDir + getRootName(getSimpleName(fastaFileNames[s])) + "match.dist";
 414                                 classify->setDistName(DistName);  outputNames.push_back(DistName);
 415                         }
 416
 417                         outputNames.push_back(newTaxonomyFile);
 418                         outputNames.push_back(taxSummary);
 419
 420                         int start = time(NULL);
 421                         int numFastaSeqs = 0;
 422                         for (int i = 0; i < lines.size(); i++) {  delete lines[i];  }  lines.clear();
 423
 424 #ifdef USE_MPI
 425                                 int pid, end, numSeqsPerProcessor;
 426                                 int tag = 2001;
 427                                 vector<unsigned long int> MPIPos;
 428
 429                                 MPI_Status status;
 430                                 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 431                                 MPI_Comm_size(MPI_COMM_WORLD, &processors);
 432
 433                                 MPI_File inMPI;
 434                                 MPI_File outMPINewTax;
 435                                 MPI_File outMPITempTax;
 436
 437                                 int outMode=MPI_MODE_CREATE|MPI_MODE_WRONLY;
 438                                 int inMode=MPI_MODE_RDONLY;
 439
 440                                 //char* outNewTax = new char[newTaxonomyFile.length()];
 441                                 //memcpy(outNewTax, newTaxonomyFile.c_str(), newTaxonomyFile.length());
 442
 443                                 char outNewTax[1024];
 444                                 strcpy(outNewTax, newTaxonomyFile.c_str());
 445
 446                                 //char* outTempTax = new char[tempTaxonomyFile.length()];
 447                                 //memcpy(outTempTax, tempTaxonomyFile.c_str(), tempTaxonomyFile.length());
 448
 449                                 char outTempTax[1024];
 450                                 strcpy(outTempTax, tempTaxonomyFile.c_str());
 451
 452                                 //char* inFileName = new char[fastaFileNames[s].length()];
 453                                 //memcpy(inFileName, fastaFileNames[s].c_str(), fastaFileNames[s].length());
 454
 455                                 char inFileName[1024];
 456                                 strcpy(inFileName, fastaFileNames[s].c_str());
 457
 458                                 MPI_File_open(MPI_COMM_WORLD, inFileName, inMode, MPI_INFO_NULL, &inMPI);  //comm, filename, mode, info, filepointer
 459                                 MPI_File_open(MPI_COMM_WORLD, outNewTax, outMode, MPI_INFO_NULL, &outMPINewTax);
 460                                 MPI_File_open(MPI_COMM_WORLD, outTempTax, outMode, MPI_INFO_NULL, &outMPITempTax);
 461
 462                                 //delete outNewTax;
 463                                 //delete outTempTax;
 464                                 //delete inFileName;
 465
 466                                 if (m->control_pressed) {  MPI_File_close(&inMPI);  MPI_File_close(&outMPINewTax);   MPI_File_close(&outMPITempTax);  delete classify; return 0;  }
 467
 468                                 if (pid == 0) { //you are the root process
 469
 470                                         MPIPos = setFilePosFasta(fastaFileNames[s], numFastaSeqs); //fills MPIPos, returns numSeqs
 471
 472                                         //send file positions to all processes
 473                                         for(int i = 1; i < processors; i++) {
 474                                                 MPI_Send(&numFastaSeqs, 1, MPI_INT, i, tag, MPI_COMM_WORLD);
 475                                                 MPI_Send(&MPIPos[0], (numFastaSeqs+1), MPI_LONG, i, tag, MPI_COMM_WORLD);
 476                                         }
 477
 478                                         //figure out how many sequences you have to align
 479                                         numSeqsPerProcessor = numFastaSeqs / processors;
 480                                         int startIndex =  pid * numSeqsPerProcessor;
 481                                         if(pid == (processors - 1)){    numSeqsPerProcessor = numFastaSeqs - pid * numSeqsPerProcessor;         }
 482
 483
 484                                         //align your part
 485                                         driverMPI(startIndex, numSeqsPerProcessor, inMPI, outMPINewTax, outMPITempTax, MPIPos);
 486
 487                                         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;  }
 488
 489                                         for (int i = 1; i < processors; i++) {
 490                                                 int done;
 491                                                 MPI_Recv(&done, 1, MPI_INT, i, tag, MPI_COMM_WORLD, &status);
 492                                         }
 493                                 }else{ //you are a child process
 494                                         MPI_Recv(&numFastaSeqs, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
 495                                         MPIPos.resize(numFastaSeqs+1);
 496                                         MPI_Recv(&MPIPos[0], (numFastaSeqs+1), MPI_LONG, 0, tag, MPI_COMM_WORLD, &status);
 497
 498                                         //figure out how many sequences you have to align
 499                                         numSeqsPerProcessor = numFastaSeqs / processors;
 500                                         int startIndex =  pid * numSeqsPerProcessor;
 501                                         if(pid == (processors - 1)){    numSeqsPerProcessor = numFastaSeqs - pid * numSeqsPerProcessor;         }
 502
 503
 504                                         //align your part
 505                                         driverMPI(startIndex, numSeqsPerProcessor, inMPI, outMPINewTax, outMPITempTax, MPIPos);
 506
 507                                         if (m->control_pressed) {  MPI_File_close(&inMPI);  MPI_File_close(&outMPINewTax);   MPI_File_close(&outMPITempTax);  delete classify; return 0;  }
 508
 509                                         int done = 0;
 510                                         MPI_Send(&done, 1, MPI_INT, 0, tag, MPI_COMM_WORLD);
 511                                 }
 512
 513                                 //close files
 514                                 MPI_File_close(&inMPI);
 515                                 MPI_File_close(&outMPINewTax);
 516                                 MPI_File_close(&outMPITempTax);
 517                                 MPI_Barrier(MPI_COMM_WORLD); //make everyone wait - just in case
 518
 519 #else
 520                 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
 521                         if(processors == 1){
 522                                 ifstream inFASTA;
 523                                 openInputFile(fastaFileNames[s], inFASTA);
 524                                 getNumSeqs(inFASTA, numFastaSeqs);
 525                                 inFASTA.close();
 526
 527                                 lines.push_back(new linePair(0, numFastaSeqs));
 528
 529                                 driver(lines[0], newTaxonomyFile, tempTaxonomyFile, fastaFileNames[s]);
 530                         }
 531                         else{
 532                                 vector<unsigned long int> positions;
 533                                 processIDS.resize(0);
 534
 535                                 ifstream inFASTA;
 536                                 openInputFile(fastaFileNames[s], inFASTA);
 537
 538                                 string input;
 539                                 while(!inFASTA.eof()){
 540                                         input = getline(inFASTA);
 541                                         if (input.length() != 0) {
 542                                                 if(input[0] == '>'){    unsigned long int pos = inFASTA.tellg(); positions.push_back(pos - input.length() - 1); }
 543                                         }
 544                                 }
 545                                 inFASTA.close();
 546
 547                                 numFastaSeqs = positions.size();
 548
 549                                 int numSeqsPerProcessor = numFastaSeqs / processors;
 550
 551                                 for (int i = 0; i < processors; i++) {
 552                                         unsigned long int startPos = positions[ i * numSeqsPerProcessor ];
 553                                         if(i == processors - 1){
 554                                                 numSeqsPerProcessor = numFastaSeqs - i * numSeqsPerProcessor;
 555                                         }
 556                                         lines.push_back(new linePair(startPos, numSeqsPerProcessor));
 557                                 }
 558                                 createProcesses(newTaxonomyFile, tempTaxonomyFile, fastaFileNames[s]);
 559
 560                                 rename((newTaxonomyFile + toString(processIDS[0]) + ".temp").c_str(), newTaxonomyFile.c_str());
 561                                 rename((tempTaxonomyFile + toString(processIDS[0]) + ".temp").c_str(), tempTaxonomyFile.c_str());
 562
 563                                 for(int i=1;i<processors;i++){
 564                                         appendTaxFiles((newTaxonomyFile + toString(processIDS[i]) + ".temp"), newTaxonomyFile);
 565                                         appendTaxFiles((tempTaxonomyFile + toString(processIDS[i]) + ".temp"), tempTaxonomyFile);
 566                                         remove((newTaxonomyFile + toString(processIDS[i]) + ".temp").c_str());
 567                                         remove((tempTaxonomyFile + toString(processIDS[i]) + ".temp").c_str());
 568                                 }
 569
 570                         }
 571         #else
 572                         ifstream inFASTA;
 573                         openInputFile(fastaFileNames[s], inFASTA);
 574                         getNumSeqs(inFASTA, numFastaSeqs);
 575                         inFASTA.close();
 576
 577                         lines.push_back(new linePair(0, numFastaSeqs));
 578
 579                         driver(lines[0], newTaxonomyFile, tempTaxonomyFile, fastaFileNames[s]);
 580         #endif
 581 #endif
 582
 583                 m->mothurOutEndLine();
 584                 m->mothurOut("It took " + toString(time(NULL) - start) + " secs to classify " + toString(numFastaSeqs) + " sequences."); m->mothurOutEndLine(); m->mothurOutEndLine();
 585                 start = time(NULL);
 586
 587
 588                 #ifdef USE_MPI
 589                         if (pid == 0) {  //this part does not need to be paralellized
 590
 591                                 if(namefile != "") { m->mothurOut("Reading " + namefileNames[s] + "..."); cout.flush();  MPIReadNamesFile(namefileNames[s]);  m->mothurOut("  Done."); m->mothurOutEndLine(); }
 592                 #else
 593                         //read namefile
 594                         if(namefile != "") {
 595
 596                             m->mothurOut("Reading " + namefileNames[s] + "..."); cout.flush();
 597
 598                                 nameMap.clear(); //remove old names
 599
 600                                 ifstream inNames;
 601                                 openInputFile(namefileNames[s], inNames);
 602
 603                                 string firstCol, secondCol;
 604                                 while(!inNames.eof()) {
 605                                         inNames >> firstCol >> secondCol; gobble(inNames);
 606
 607                                         vector<string> temp;
 608                                         splitAtComma(secondCol, temp);
 609
 610                                         nameMap[firstCol] = temp;
 611                                 }
 612                                 inNames.close();
 613
 614                                 m->mothurOut("  Done."); m->mothurOutEndLine();
 615                         }
 616                 #endif
 617
 618                         string group = "";
 619                         if (groupfile != "") {  group = groupfileNames[s]; }
 620
 621                         PhyloSummary taxaSum(taxonomyFileName, group);
 622
 623                         if (m->control_pressed) {  for (int i = 0; i < outputNames.size(); i++) {       remove(outputNames[i].c_str()); } delete classify; return 0; }
 624
 625                         if (namefile == "") {  taxaSum.summarize(tempTaxonomyFile);  }
 626                         else {
 627                                 ifstream in;
 628                                 openInputFile(tempTaxonomyFile, in);
 629
 630                                 //read in users taxonomy file and add sequences to tree
 631                                 string name, taxon;
 632
 633                                 while(!in.eof()){
 634                                         in >> name >> taxon; gobble(in);
 635
 636                                         itNames = nameMap.find(name);
 637
 638                                         if (itNames == nameMap.end()) {
 639                                                 m->mothurOut(name + " is not in your name file please correct."); m->mothurOutEndLine(); exit(1);
 640                                         }else{
 641                                                 for (int i = 0; i < itNames->second.size(); i++) {
 642                                                         taxaSum.addSeqToTree(itNames->second[i], taxon);  //add it as many times as there are identical seqs
 643                                                 }
 644                                                 itNames->second.clear();
 645                                                 nameMap.erase(itNames->first);
 646                                         }
 647                                 }
 648                                 in.close();
 649                         }
 650                         remove(tempTaxonomyFile.c_str());
 651
 652                         if (m->control_pressed) {  for (int i = 0; i < outputNames.size(); i++) {       remove(outputNames[i].c_str()); } delete classify; return 0; }
 653
 654                         //print summary file
 655                         ofstream outTaxTree;
 656                         openOutputFile(taxSummary, outTaxTree);
 657                         taxaSum.print(outTaxTree);
 658                         outTaxTree.close();
 659
 660                         //output taxonomy with the unclassified bins added
 661                         ifstream inTax;
 662                         openInputFile(newTaxonomyFile, inTax);
 663
 664                         ofstream outTax;
 665                         string unclass = newTaxonomyFile + ".unclass.temp";
 666                         openOutputFile(unclass, outTax);
 667
 668                         //get maxLevel from phylotree so you know how many 'unclassified's to add
 669                         int maxLevel = taxaSum.getMaxLevel();
 670
 671                         //read taxfile - this reading and rewriting is done to preserve the confidence scores.
 672                         string name, taxon;
 673                         while (!inTax.eof()) {
 674                                 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; }
 675
 676                                 inTax >> name >> taxon; gobble(inTax);
 677
 678                                 string newTax = addUnclassifieds(taxon, maxLevel);
 679
 680                                 outTax << name << '\t' << newTax << endl;
 681                         }
 682                         inTax.close();
 683                         outTax.close();
 684
 685                         remove(newTaxonomyFile.c_str());
 686                         rename(unclass.c_str(), newTaxonomyFile.c_str());
 687
 688                         m->mothurOutEndLine();
 689                         m->mothurOut("It took " + toString(time(NULL) - start) + " secs to create the summary file for " + toString(numFastaSeqs) + " sequences."); m->mothurOutEndLine(); m->mothurOutEndLine();
 690
 691                         #ifdef USE_MPI
 692                                 }
 693                         #endif
 694
 695                         m->mothurOutEndLine();
 696                         m->mothurOut("Output File Names: "); m->mothurOutEndLine();
 697                         for (int i = 0; i < outputNames.size(); i++) {  m->mothurOut(outputNames[i]); m->mothurOutEndLine();    }
 698                         m->mothurOutEndLine();
 699                 }
 700
 701                 delete classify;
 702                 return 0;
 703         }
 704         catch(exception& e) {
 705                 m->errorOut(e, "ClassifySeqsCommand", "execute");
 706                 exit(1);
 707         }
 708 }
 709
 710 /**************************************************************************************************/
 711 string ClassifySeqsCommand::addUnclassifieds(string tax, int maxlevel) {
 712         try{
 713                 string newTax, taxon;
 714                 int level = 0;
 715
 716                 //keep what you have counting the levels
 717                 while (tax.find_first_of(';') != -1) {
 718                         //get taxon
 719                         taxon = tax.substr(0,tax.find_first_of(';'))+';';
 720                         tax = tax.substr(tax.find_first_of(';')+1, tax.length());
 721                         newTax += taxon;
 722                         level++;
 723                 }
 724
 725                 //add "unclassified" until you reach maxLevel
 726                 while (level < maxlevel) {
 727                         newTax += "unclassified;";
 728                         level++;
 729                 }
 730
 731                 return newTax;
 732         }
 733         catch(exception& e) {
 734                 m->errorOut(e, "ClassifySeqsCommand", "addUnclassifieds");
 735                 exit(1);
 736         }
 737 }
 738
 739 /**************************************************************************************************/
 740
 741 void ClassifySeqsCommand::createProcesses(string taxFileName, string tempTaxFile, string filename) {
 742         try {
 743 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
 744                 int process = 0;
 745                 //              processIDS.resize(0);
 746
 747                 //loop through and create all the processes you want
 748                 while (process != processors) {
 749                         int pid = fork();
 750
 751                         if (pid > 0) {
 752                                 processIDS.push_back(pid);  //create map from line number to pid so you can append files in correct order later
 753                                 process++;
 754                         }else if (pid == 0){
 755                                 driver(lines[process], taxFileName + toString(getpid()) + ".temp", tempTaxFile + toString(getpid()) + ".temp", filename);
 756                                 exit(0);
 757                         }else { m->mothurOut("unable to spawn the necessary processes."); m->mothurOutEndLine(); exit(0); }
 758                 }
 759
 760                 //force parent to wait until all the processes are done
 761                 for (int i=0;i<processors;i++) {
 762                         int temp = processIDS[i];
 763                         wait(&temp);
 764                 }
 765 #endif
 766         }
 767         catch(exception& e) {
 768                 m->errorOut(e, "ClassifySeqsCommand", "createProcesses");
 769                 exit(1);
 770         }
 771 }
 772 /**************************************************************************************************/
 773
 774 void ClassifySeqsCommand::appendTaxFiles(string temp, string filename) {
 775         try{
 776
 777                 ofstream output;
 778                 ifstream input;
 779                 openOutputFileAppend(filename, output);
 780                 openInputFile(temp, input);
 781
 782                 while(char c = input.get()){
 783                         if(input.eof())         {       break;                  }
 784                         else                            {       output << c;    }
 785                 }
 786
 787                 input.close();
 788                 output.close();
 789         }
 790         catch(exception& e) {
 791                 m->errorOut(e, "ClassifySeqsCommand", "appendTaxFiles");
 792                 exit(1);
 793         }
 794 }
 795
 796 //**********************************************************************************************************************
 797
 798 int ClassifySeqsCommand::driver(linePair* line, string taxFName, string tempTFName, string filename){
 799         try {
 800                 ofstream outTax;
 801                 openOutputFile(taxFName, outTax);
 802
 803                 ofstream outTaxSimple;
 804                 openOutputFile(tempTFName, outTaxSimple);
 805
 806                 ifstream inFASTA;
 807                 openInputFile(filename, inFASTA);
 808
 809                 inFASTA.seekg(line->start);
 810
 811                 string taxonomy;
 812
 813                 for(int i=0;i<line->numSeqs;i++){
 814                         if (m->control_pressed) { return 0; }
 815
 816                         Sequence* candidateSeq = new Sequence(inFASTA); gobble(inFASTA);
 817
 818                         if (candidateSeq->getName() != "") {
 819                                 taxonomy = classify->getTaxonomy(candidateSeq);
 820
 821                                 if (m->control_pressed) { delete candidateSeq; return 0; }
 822
 823                                 if (taxonomy != "bad seq") {
 824                                         //output confidence scores or not
 825                                         if (probs) {
 826                                                 outTax << candidateSeq->getName() << '\t' << taxonomy << endl;
 827                                         }else{
 828                                                 outTax << candidateSeq->getName() << '\t' << classify->getSimpleTax() << endl;
 829                                         }
 830
 831                                         outTaxSimple << candidateSeq->getName() << '\t' << classify->getSimpleTax() << endl;
 832                                 }
 833                         }
 834                         delete candidateSeq;
 835
 836                         if((i+1) % 100 == 0){
 837                                 m->mothurOut("Classifying sequence " + toString(i+1)); m->mothurOutEndLine();
 838                         }
 839                 }
 840
 841                 inFASTA.close();
 842                 outTax.close();
 843                 outTaxSimple.close();
 844
 845                 return 1;
 846         }
 847         catch(exception& e) {
 848                 m->errorOut(e, "ClassifySeqsCommand", "driver");
 849                 exit(1);
 850         }
 851 }
 852 //**********************************************************************************************************************
 853 #ifdef USE_MPI
 854 int ClassifySeqsCommand::driverMPI(int start, int num, MPI_File& inMPI, MPI_File& newFile, MPI_File& tempFile, vector<unsigned long int>& MPIPos){
 855         try {
 856                 MPI_Status statusNew;
 857                 MPI_Status statusTemp;
 858                 MPI_Status status;
 859
 860                 int pid;
 861                 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 862
 863                 string taxonomy;
 864                 string outputString;
 865
 866                 for(int i=0;i<num;i++){
 867
 868                         if (m->control_pressed) { return 0; }
 869
 870                         //read next sequence
 871                         int length = MPIPos[start+i+1] - MPIPos[start+i];
 872                         char* buf4 = new char[length];
 873                         MPI_File_read_at(inMPI, MPIPos[start+i], buf4, length, MPI_CHAR, &status);
 874
 875                         string tempBuf = buf4;
 876                         if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length);  }
 877                         istringstream iss (tempBuf,istringstream::in);
 878                         delete buf4;
 879
 880                         Sequence* candidateSeq = new Sequence(iss);
 881
 882                         if (candidateSeq->getName() != "") {
 883                                 taxonomy = classify->getTaxonomy(candidateSeq);
 884
 885                                 if (taxonomy != "bad seq") {
 886                                         //output confidence scores or not
 887                                         if (probs) {
 888                                                 outputString =  candidateSeq->getName() + "\t" + taxonomy + "\n";
 889                                         }else{
 890                                                 outputString =  candidateSeq->getName() + "\t" + classify->getSimpleTax() + "\n";
 891                                         }
 892
 893                                         int length = outputString.length();
 894                                         char* buf2 = new char[length];
 895                                         memcpy(buf2, outputString.c_str(), length);
 896
 897                                         MPI_File_write_shared(newFile, buf2, length, MPI_CHAR, &statusNew);
 898                                         delete buf2;
 899
 900                                         outputString =  candidateSeq->getName() + "\t" + classify->getSimpleTax() + "\n";
 901                                         length = outputString.length();
 902                                         char* buf = new char[length];
 903                                         memcpy(buf, outputString.c_str(), length);
 904
 905                                         MPI_File_write_shared(tempFile, buf, length, MPI_CHAR, &statusTemp);
 906                                         delete buf;
 907                                 }
 908                         }
 909                         delete candidateSeq;
 910
 911                         if((i+1) % 100 == 0){   cout << "Classifying sequence " << (i+1) << endl;       }
 912                 }
 913
 914                 if(num % 100 != 0){     cout << "Classifying sequence " << (num) << endl;       }
 915
 916
 917                 return 1;
 918         }
 919         catch(exception& e) {
 920                 m->errorOut(e, "ClassifySeqsCommand", "driverMPI");
 921                 exit(1);
 922         }
 923 }
 924
 925 //**********************************************************************************************************************
 926 int ClassifySeqsCommand::MPIReadNamesFile(string nameFilename){
 927         try {
 928
 929                 nameMap.clear(); //remove old names
 930
 931                 MPI_File inMPI;
 932                 MPI_Offset size;
 933                 MPI_Status status;
 934
 935                 //char* inFileName = new char[nameFilename.length()];
 936                 //memcpy(inFileName, nameFilename.c_str(), nameFilename.length());
 937
 938                 char inFileName[1024];
 939                 strcpy(inFileName, nameFilename.c_str());
 940
 941                 MPI_File_open(MPI_COMM_WORLD, inFileName, MPI_MODE_RDONLY, MPI_INFO_NULL, &inMPI);
 942                 MPI_File_get_size(inMPI, &size);
 943                 //delete inFileName;
 944
 945                 char* buffer = new char[size];
 946                 MPI_File_read(inMPI, buffer, size, MPI_CHAR, &status);
 947
 948                 string tempBuf = buffer;
 949                 if (tempBuf.length() > size) { tempBuf = tempBuf.substr(0, size);  }
 950                 istringstream iss (tempBuf,istringstream::in);
 951                 delete buffer;
 952
 953                 string firstCol, secondCol;
 954                 while(!iss.eof()) {
 955                         iss >> firstCol >> secondCol; gobble(iss);
 956
 957                         vector<string> temp;
 958                         splitAtComma(secondCol, temp);
 959
 960                         nameMap[firstCol] = temp;
 961                 }
 962
 963                 MPI_File_close(&inMPI);
 964
 965                 return 1;
 966         }
 967         catch(exception& e) {
 968                 m->errorOut(e, "ClassifySeqsCommand", "MPIReadNamesFile");
 969                 exit(1);
 970         }
 971 }
 972 #endif
 973 /**************************************************************************************************/