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 vector<string> ClassifySeqsCommand::getValidParameters(){
  20         try {
  21                 string AlignArray[] =  {"template","fasta","name","group","search","ksize","method","processors","taxonomy","match","mismatch","gapopen","gapextend","numwanted","cutoff","probs","iters", "outputdir","inputdir"};
  22                 vector<string> myArray (AlignArray, AlignArray+(sizeof(AlignArray)/sizeof(string)));
  23                 return myArray;
  24         }
  25         catch(exception& e) {
  26                 m->errorOut(e, "ClassifySeqsCommand", "getValidParameters");
  27                 exit(1);
  28         }
  29 }
  30 //**********************************************************************************************************************
  31 ClassifySeqsCommand::ClassifySeqsCommand(){
  32         try {
  33                 abort = true;
  34                 //initialize outputTypes
  35                 vector<string> tempOutNames;
  36                 outputTypes["taxonomy"] = tempOutNames;
  37                 outputTypes["taxsummary"] = tempOutNames;
  38                 outputTypes["matchdist"] = tempOutNames;
  39         }
  40         catch(exception& e) {
  41                 m->errorOut(e, "ClassifySeqsCommand", "ClassifySeqsCommand");
  42                 exit(1);
  43         }
  44 }
  45 //**********************************************************************************************************************
  46 vector<string> ClassifySeqsCommand::getRequiredParameters(){
  47         try {
  48                 string Array[] =  {"fasta","template","taxonomy"};
  49                 vector<string> myArray (Array, Array+(sizeof(Array)/sizeof(string)));
  50                 return myArray;
  51         }
  52         catch(exception& e) {
  53                 m->errorOut(e, "ClassifySeqsCommand", "getRequiredParameters");
  54                 exit(1);
  55         }
  56 }
  57 //**********************************************************************************************************************
  58 vector<string> ClassifySeqsCommand::getRequiredFiles(){
  59         try {
  60                 vector<string> myArray;
  61                 return myArray;
  62         }
  63         catch(exception& e) {
  64                 m->errorOut(e, "ClassifySeqsCommand", "getRequiredFiles");
  65                 exit(1);
  66         }
  67 }
  68 //**********************************************************************************************************************
  69 ClassifySeqsCommand::ClassifySeqsCommand(string option)  {
  70         try {
  71                 abort = false;
  72
  73                 //allow user to run help
  74                 if(option == "help") { help(); abort = true; }
  75
  76                 else {
  77
  78                         //valid paramters for this command
  79                         string AlignArray[] =  {"template","fasta","name","group","search","ksize","method","processors","taxonomy","match","mismatch","gapopen","gapextend","numwanted","cutoff","probs","iters", "outputdir","inputdir"};
  80                         vector<string> myArray (AlignArray, AlignArray+(sizeof(AlignArray)/sizeof(string)));
  81
  82                         OptionParser parser(option);
  83                         map<string, string> parameters = parser.getParameters();
  84
  85                         ValidParameters validParameter("classify.seqs");
  86                         map<string, string>::iterator it;
  87
  88                         //check to make sure all parameters are valid for command
  89                         for (it = parameters.begin(); it != parameters.end(); it++) {
  90                                 if (validParameter.isValidParameter(it->first, myArray, it->second) != true) {  abort = true;  }
  91                         }
  92
  93                         //initialize outputTypes
  94                         vector<string> tempOutNames;
  95                         outputTypes["taxonomy"] = tempOutNames;
  96                         outputTypes["taxsummary"] = tempOutNames;
  97                         outputTypes["matchdist"] = tempOutNames;
  98
  99                         //if the user changes the output directory command factory will send this info to us in the output parameter
 100                         outputDir = validParameter.validFile(parameters, "outputdir", false);           if (outputDir == "not found"){  outputDir = "";         }
 101
 102                         //if the user changes the input directory command factory will send this info to us in the output parameter
 103                         string inputDir = validParameter.validFile(parameters, "inputdir", false);
 104                         if (inputDir == "not found"){   inputDir = "";          }
 105                         else {
 106                                 string path;
 107                                 it = parameters.find("template");
 108                                 //user has given a template file
 109                                 if(it != parameters.end()){
 110                                         path = m->hasPath(it->second);
 111                                         //if the user has not given a path then, add inputdir. else leave path alone.
 112                                         if (path == "") {       parameters["template"] = inputDir + it->second;         }
 113                                 }
 114
 115                                 it = parameters.find("taxonomy");
 116                                 //user has given a template file
 117                                 if(it != parameters.end()){
 118                                         path = m->hasPath(it->second);
 119                                         //if the user has not given a path then, add inputdir. else leave path alone.
 120                                         if (path == "") {       parameters["taxonomy"] = inputDir + it->second;         }
 121                                 }
 122
 123                                 it = parameters.find("group");
 124                                 //user has given a template file
 125                                 if(it != parameters.end()){
 126                                         path = m->hasPath(it->second);
 127                                         //if the user has not given a path then, add inputdir. else leave path alone.
 128                                         if (path == "") {       parameters["group"] = inputDir + it->second;            }
 129                                 }
 130                         }
 131
 132                         //check for required parameters
 133                         templateFileName = validParameter.validFile(parameters, "template", true);
 134                         if (templateFileName == "not found") {
 135                                 m->mothurOut("template is a required parameter for the classify.seqs command.");
 136                                 m->mothurOutEndLine();
 137                                 abort = true;
 138                         }
 139                         else if (templateFileName == "not open") { abort = true; }
 140
 141
 142                         fastaFileName = validParameter.validFile(parameters, "fasta", false);
 143                         if (fastaFileName == "not found") { m->mothurOut("fasta is a required parameter for the classify.seqs command."); m->mothurOutEndLine(); abort = true;  }
 144                         else {
 145                                 m->splitAtDash(fastaFileName, fastaFileNames);
 146
 147                                 //go through files and make sure they are good, if not, then disregard them
 148                                 for (int i = 0; i < fastaFileNames.size(); i++) {
 149                                         if (inputDir != "") {
 150                                                 string path = m->hasPath(fastaFileNames[i]);
 151                                                 //if the user has not given a path then, add inputdir. else leave path alone.
 152                                                 if (path == "") {       fastaFileNames[i] = inputDir + fastaFileNames[i];               }
 153                                         }
 154
 155                                         int ableToOpen;
 156
 157                                         ifstream in;
 158                                         ableToOpen = m->openInputFile(fastaFileNames[i], in, "noerror");
 159
 160                                         //if you can't open it, try default location
 161                                         if (ableToOpen == 1) {
 162                                                 if (m->getDefaultPath() != "") { //default path is set
 163                                                         string tryPath = m->getDefaultPath() + m->getSimpleName(fastaFileNames[i]);
 164                                                         m->mothurOut("Unable to open " + fastaFileNames[i] + ". Trying default " + tryPath); m->mothurOutEndLine();
 165                                                         ifstream in2;
 166                                                         ableToOpen = m->openInputFile(tryPath, in2, "noerror");
 167                                                         in2.close();
 168                                                         fastaFileNames[i] = tryPath;
 169                                                 }
 170                                         }
 171
 172                                         if (ableToOpen == 1) {
 173                                                 if (m->getOutputDir() != "") { //default path is set
 174                                                         string tryPath = m->getOutputDir() + m->getSimpleName(fastaFileNames[i]);
 175                                                         m->mothurOut("Unable to open " + fastaFileNames[i] + ". Trying output directory " + tryPath); m->mothurOutEndLine();
 176                                                         ifstream in2;
 177                                                         ableToOpen = m->openInputFile(tryPath, in2, "noerror");
 178                                                         in2.close();
 179                                                         fastaFileNames[i] = tryPath;
 180                                                 }
 181                                         }
 182
 183                                         in.close();
 184
 185                                         if (ableToOpen == 1) {
 186                                                 m->mothurOut("Unable to open " + fastaFileNames[i] + ". It will be disregarded."); m->mothurOutEndLine();
 187                                                 //erase from file list
 188                                                 fastaFileNames.erase(fastaFileNames.begin()+i);
 189                                                 i--;
 190                                         }
 191
 192                                 }
 193
 194                                 //make sure there is at least one valid file left
 195                                 if (fastaFileNames.size() == 0) { m->mothurOut("no valid files."); m->mothurOutEndLine(); abort = true; }
 196                         }
 197
 198
 199                         taxonomyFileName = validParameter.validFile(parameters, "taxonomy", true);
 200                         if (taxonomyFileName == "not found") {
 201                                 m->mothurOut("taxonomy is a required parameter for the classify.seqs command.");
 202                                 m->mothurOutEndLine();
 203                                 abort = true;
 204                         }
 205                         else if (taxonomyFileName == "not open") { abort = true; }
 206
 207
 208                         namefile = validParameter.validFile(parameters, "name", false);
 209                         if (namefile == "not found") { namefile = "";  }
 210
 211                         else {
 212                                 m->splitAtDash(namefile, namefileNames);
 213
 214                                 //go through files and make sure they are good, if not, then disregard them
 215                                 for (int i = 0; i < namefileNames.size(); i++) {
 216                                         if (inputDir != "") {
 217                                                 string path = m->hasPath(namefileNames[i]);
 218                                                 //if the user has not given a path then, add inputdir. else leave path alone.
 219                                                 if (path == "") {       namefileNames[i] = inputDir + namefileNames[i];         }
 220                                         }
 221                                         int ableToOpen;
 222
 223                                         ifstream in;
 224                                         ableToOpen = m->openInputFile(namefileNames[i], in, "noerror");
 225
 226                                         //if you can't open it, try default location
 227                                         if (ableToOpen == 1) {
 228                                                 if (m->getDefaultPath() != "") { //default path is set
 229                                                         string tryPath = m->getDefaultPath() + m->getSimpleName(namefileNames[i]);
 230                                                         m->mothurOut("Unable to open " + namefileNames[i] + ". Trying default " + tryPath); m->mothurOutEndLine();
 231                                                         ifstream in2;
 232                                                         ableToOpen = m->openInputFile(tryPath, in2, "noerror");
 233                                                         in2.close();
 234                                                         namefileNames[i] = tryPath;
 235                                                 }
 236                                         }
 237
 238                                         if (ableToOpen == 1) {
 239                                                 if (m->getOutputDir() != "") { //default path is set
 240                                                         string tryPath = m->getOutputDir() + m->getSimpleName(namefileNames[i]);
 241                                                         m->mothurOut("Unable to open " + namefileNames[i] + ". Trying output directory " + tryPath); m->mothurOutEndLine();
 242                                                         ifstream in2;
 243                                                         ableToOpen = m->openInputFile(tryPath, in2, "noerror");
 244                                                         in2.close();
 245                                                         namefileNames[i] = tryPath;
 246                                                 }
 247                                         }
 248                                         in.close();
 249
 250                                         if (ableToOpen == 1) {
 251                                                 m->mothurOut("Unable to open " + namefileNames[i] + ". It will be disregarded."); m->mothurOutEndLine();  abort = true;
 252                                                 //erase from file list
 253                                                 namefileNames.erase(namefileNames.begin()+i);
 254                                                 i--;
 255                                         }
 256
 257                                 }
 258                         }
 259
 260                         if (namefile != "") {
 261                                 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(); }
 262                         }
 263
 264                         groupfile = validParameter.validFile(parameters, "group", false);
 265                         if (groupfile == "not found") { groupfile = "";  }
 266                         else {
 267                                 m->splitAtDash(groupfile, groupfileNames);
 268
 269                                 //go through files and make sure they are good, if not, then disregard them
 270                                 for (int i = 0; i < groupfileNames.size(); i++) {
 271                                         if (inputDir != "") {
 272                                                 string path = m->hasPath(groupfileNames[i]);
 273                                                 //if the user has not given a path then, add inputdir. else leave path alone.
 274                                                 if (path == "") {       groupfileNames[i] = inputDir + groupfileNames[i];               }
 275                                         }
 276                                         int ableToOpen;
 277
 278                                         ifstream in;
 279                                         ableToOpen = m->openInputFile(groupfileNames[i], in, "noerror");
 280
 281                                         //if you can't open it, try default location
 282                                         if (ableToOpen == 1) {
 283                                                 if (m->getDefaultPath() != "") { //default path is set
 284                                                         string tryPath = m->getDefaultPath() + m->getSimpleName(groupfileNames[i]);
 285                                                         m->mothurOut("Unable to open " + groupfileNames[i] + ". Trying default " + tryPath); m->mothurOutEndLine();
 286                                                         ifstream in2;
 287                                                         ableToOpen = m->openInputFile(tryPath, in2, "noerror");
 288                                                         in2.close();
 289                                                         groupfileNames[i] = tryPath;
 290                                                 }
 291                                         }
 292
 293                                         if (ableToOpen == 1) {
 294                                                 if (m->getOutputDir() != "") { //default path is set
 295                                                         string tryPath = m->getOutputDir() + m->getSimpleName(groupfileNames[i]);
 296                                                         m->mothurOut("Unable to open " + groupfileNames[i] + ". Trying output directory " + tryPath); m->mothurOutEndLine();
 297                                                         ifstream in2;
 298                                                         ableToOpen = m->openInputFile(tryPath, in2, "noerror");
 299                                                         in2.close();
 300                                                         groupfileNames[i] = tryPath;
 301                                                 }
 302                                         }
 303
 304                                         in.close();
 305
 306                                         if (ableToOpen == 1) {
 307                                                 m->mothurOut("Unable to open " + groupfileNames[i] + ". It will be disregarded."); m->mothurOutEndLine(); groupfileNames[i] = "";
 308                                                 //erase from file list
 309                                                 groupfileNames.erase(groupfileNames.begin()+i);
 310                                                 i--;
 311                                         }
 312                                 }
 313                         }
 314
 315                         if (groupfile != "") {
 316                                 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(); }
 317                         }else {
 318                                 for (int i = 0; i < fastaFileNames.size(); i++) {  groupfileNames.push_back("");  }
 319                         }
 320
 321                         //check for optional parameter and set defaults
 322                         // ...at some point should added some additional type checking...
 323                         string temp;
 324                         temp = validParameter.validFile(parameters, "ksize", false);            if (temp == "not found"){       temp = "8";                             }
 325                         convert(temp, kmerSize);
 326
 327                         temp = validParameter.validFile(parameters, "processors", false);       if (temp == "not found"){       temp = "1";                             }
 328                         convert(temp, processors);
 329
 330                         search = validParameter.validFile(parameters, "search", false);         if (search == "not found"){     search = "kmer";                }
 331
 332                         method = validParameter.validFile(parameters, "method", false);         if (method == "not found"){     method = "bayesian";    }
 333
 334                         temp = validParameter.validFile(parameters, "match", false);            if (temp == "not found"){       temp = "1.0";                   }
 335                         convert(temp, match);
 336
 337                         temp = validParameter.validFile(parameters, "mismatch", false);         if (temp == "not found"){       temp = "-1.0";                  }
 338                         convert(temp, misMatch);
 339
 340                         temp = validParameter.validFile(parameters, "gapopen", false);          if (temp == "not found"){       temp = "-2.0";                  }
 341                         convert(temp, gapOpen);
 342
 343                         temp = validParameter.validFile(parameters, "gapextend", false);        if (temp == "not found"){       temp = "-1.0";                  }
 344                         convert(temp, gapExtend);
 345
 346                         temp = validParameter.validFile(parameters, "numwanted", false);        if (temp == "not found"){       temp = "10";                    }
 347                         convert(temp, numWanted);
 348
 349                         temp = validParameter.validFile(parameters, "cutoff", false);           if (temp == "not found"){       temp = "0";                             }
 350                         convert(temp, cutoff);
 351
 352                         temp = validParameter.validFile(parameters, "probs", false);            if (temp == "not found"){       temp = "true";                  }
 353                         probs = m->isTrue(temp);
 354
 355                         temp = validParameter.validFile(parameters, "iters", false);            if (temp == "not found") { temp = "100";                        }
 356                         convert(temp, iters);
 357
 358
 359
 360                         if ((method == "bayesian") && (search != "kmer"))  {
 361                                 m->mothurOut("The bayesian method requires the kmer search." + search + "will be disregarded." ); m->mothurOutEndLine();
 362                                 search = "kmer";
 363                         }
 364                 }
 365
 366         }
 367         catch(exception& e) {
 368                 m->errorOut(e, "ClassifySeqsCommand", "ClassifySeqsCommand");
 369                 exit(1);
 370         }
 371 }
 372
 373 //**********************************************************************************************************************
 374
 375 ClassifySeqsCommand::~ClassifySeqsCommand(){
 376
 377         if (abort == false) {
 378                 for (int i = 0; i < lines.size(); i++) {  delete lines[i];  }  lines.clear();
 379         }
 380 }
 381
 382 //**********************************************************************************************************************
 383
 384 void ClassifySeqsCommand::help(){
 385         try {
 386                 m->mothurOut("The classify.seqs command reads a fasta file containing sequences and creates a .taxonomy file and a .tax.summary file.\n");
 387                 m->mothurOut("The classify.seqs command parameters are template, fasta, name, search, ksize, method, taxonomy, processors, match, mismatch, gapopen, gapextend, numwanted and probs.\n");
 388                 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");
 389                 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");
 390                 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");
 391                 m->mothurOut("The group parameter allows you add a group file so you can have the summary totals broken up by group.\n");
 392                 m->mothurOut("The method parameter allows you to specify classification method to use.  Your options are: bayesian and knn. The default is bayesian.\n");
 393                 m->mothurOut("The ksize parameter allows you to specify the kmer size for finding most similar template to candidate.  The default is 8.\n");
 394                 m->mothurOut("The processors parameter allows you to specify the number of processors to use. The default is 1.\n");
 395                 #ifdef USE_MPI
 396                 m->mothurOut("When using MPI, the processors parameter is set to the number of MPI processes running. \n");
 397                 #endif
 398                 m->mothurOut("The match parameter allows you to specify the bonus for having the same base. The default is 1.0.\n");
 399                 m->mothurOut("The mistmatch parameter allows you to specify the penalty for having different bases.  The default is -1.0.\n");
 400                 m->mothurOut("The gapopen parameter allows you to specify the penalty for opening a gap in an alignment. The default is -2.0.\n");
 401                 m->mothurOut("The gapextend parameter allows you to specify the penalty for extending a gap in an alignment.  The default is -1.0.\n");
 402                 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");
 403                 m->mothurOut("The cutoff parameter allows you to specify a bootstrap confidence threshold for your taxonomy.  The default is 0.\n");
 404                 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");
 405                 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");
 406                 m->mothurOut("The classify.seqs command should be in the following format: \n");
 407                 m->mothurOut("classify.seqs(template=yourTemplateFile, fasta=yourFastaFile, method=yourClassificationMethod, search=yourSearchmethod, ksize=yourKmerSize, taxonomy=yourTaxonomyFile, processors=yourProcessors) \n");
 408                 m->mothurOut("Example classify.seqs(fasta=amazon.fasta, template=core.filtered, method=knn, search=gotoh, ksize=8, processors=2)\n");
 409                 m->mothurOut("The .taxonomy file consists of 2 columns: 1 = your sequence name, 2 = the taxonomy for your sequence. \n");
 410                 m->mothurOut("The .tax.summary is a summary of the different taxonomies represented in your fasta file. \n");
 411                 m->mothurOut("Note: No spaces between parameter labels (i.e. fasta), '=' and parameters (i.e.yourFastaFile).\n\n");
 412         }
 413         catch(exception& e) {
 414                 m->errorOut(e, "ClassifySeqsCommand", "help");
 415                 exit(1);
 416         }
 417 }
 418
 419
 420 //**********************************************************************************************************************
 421
 422 int ClassifySeqsCommand::execute(){
 423         try {
 424                 if (abort == true) {    return 0;       }
 425
 426                 if(method == "bayesian"){       classify = new Bayesian(taxonomyFileName, templateFileName, search, kmerSize, cutoff, iters);           }
 427                 else if(method == "knn"){       classify = new Knn(taxonomyFileName, templateFileName, search, kmerSize, gapOpen, gapExtend, match, misMatch, numWanted);                               }
 428                 else {
 429                         m->mothurOut(search + " is not a valid method option. I will run the command using bayesian.");
 430                         m->mothurOutEndLine();
 431                         classify = new Bayesian(taxonomyFileName, templateFileName, search, kmerSize, cutoff, iters);
 432                 }
 433
 434                 if (m->control_pressed) { delete classify; return 0; }
 435
 436
 437                 for (int s = 0; s < fastaFileNames.size(); s++) {
 438
 439                         m->mothurOut("Classifying sequences from " + fastaFileNames[s] + " ..." ); m->mothurOutEndLine();
 440
 441                         string RippedTaxName = m->getRootName(m->getSimpleName(taxonomyFileName));
 442                         RippedTaxName = m->getExtension(RippedTaxName.substr(0, RippedTaxName.length()-1));
 443                         if (RippedTaxName[0] == '.') { RippedTaxName = RippedTaxName.substr(1, RippedTaxName.length()); }
 444                         RippedTaxName +=  ".";
 445
 446                         if (outputDir == "") { outputDir += m->hasPath(fastaFileNames[s]); }
 447                         string newTaxonomyFile = outputDir + m->getRootName(m->getSimpleName(fastaFileNames[s])) + RippedTaxName + "taxonomy";
 448                         string tempTaxonomyFile = outputDir + m->getRootName(m->getSimpleName(fastaFileNames[s])) + "taxonomy.temp";
 449                         string taxSummary = outputDir + m->getRootName(m->getSimpleName(fastaFileNames[s])) + RippedTaxName + "tax.summary";
 450
 451                         if ((method == "knn") && (search == "distance")) {
 452                                 string DistName = outputDir + m->getRootName(m->getSimpleName(fastaFileNames[s])) + "match.dist";
 453                                 classify->setDistName(DistName);  outputNames.push_back(DistName); outputTypes["matchdist"].push_back(DistName);
 454                         }
 455
 456                         outputNames.push_back(newTaxonomyFile); outputTypes["taxonomy"].push_back(newTaxonomyFile);
 457                         outputNames.push_back(taxSummary);      outputTypes["taxsummary"].push_back(taxSummary);
 458
 459                         int start = time(NULL);
 460                         int numFastaSeqs = 0;
 461                         for (int i = 0; i < lines.size(); i++) {  delete lines[i];  }  lines.clear();
 462
 463 #ifdef USE_MPI
 464                                 int pid, end, numSeqsPerProcessor;
 465                                 int tag = 2001;
 466                                 vector<unsigned long int> MPIPos;
 467
 468                                 MPI_Status status;
 469                                 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 470                                 MPI_Comm_size(MPI_COMM_WORLD, &processors);
 471
 472                                 MPI_File inMPI;
 473                                 MPI_File outMPINewTax;
 474                                 MPI_File outMPITempTax;
 475
 476                                 int outMode=MPI_MODE_CREATE|MPI_MODE_WRONLY;
 477                                 int inMode=MPI_MODE_RDONLY;
 478
 479                                 char outNewTax[1024];
 480                                 strcpy(outNewTax, newTaxonomyFile.c_str());
 481
 482                                 char outTempTax[1024];
 483                                 strcpy(outTempTax, tempTaxonomyFile.c_str());
 484
 485                                 char inFileName[1024];
 486                                 strcpy(inFileName, fastaFileNames[s].c_str());
 487
 488                                 MPI_File_open(MPI_COMM_WORLD, inFileName, inMode, MPI_INFO_NULL, &inMPI);  //comm, filename, mode, info, filepointer
 489                                 MPI_File_open(MPI_COMM_WORLD, outNewTax, outMode, MPI_INFO_NULL, &outMPINewTax);
 490                                 MPI_File_open(MPI_COMM_WORLD, outTempTax, outMode, MPI_INFO_NULL, &outMPITempTax);
 491
 492                                 if (m->control_pressed) { outputTypes.clear(); MPI_File_close(&inMPI);  MPI_File_close(&outMPINewTax);   MPI_File_close(&outMPITempTax);  delete classify; return 0;  }
 493
 494                                 if (pid == 0) { //you are the root process
 495
 496                                         MPIPos = m->setFilePosFasta(fastaFileNames[s], numFastaSeqs); //fills MPIPos, returns numSeqs
 497
 498                                         //send file positions to all processes
 499                                         for(int i = 1; i < processors; i++) {
 500                                                 MPI_Send(&numFastaSeqs, 1, MPI_INT, i, tag, MPI_COMM_WORLD);
 501                                                 MPI_Send(&MPIPos[0], (numFastaSeqs+1), MPI_LONG, i, tag, MPI_COMM_WORLD);
 502                                         }
 503
 504                                         //figure out how many sequences you have to align
 505                                         numSeqsPerProcessor = numFastaSeqs / processors;
 506                                         int startIndex =  pid * numSeqsPerProcessor;
 507                                         if(pid == (processors - 1)){    numSeqsPerProcessor = numFastaSeqs - pid * numSeqsPerProcessor;         }
 508
 509
 510                                         //align your part
 511                                         driverMPI(startIndex, numSeqsPerProcessor, inMPI, outMPINewTax, outMPITempTax, MPIPos);
 512
 513                                         if (m->control_pressed) {  outputTypes.clear(); 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;  }
 514
 515                                         for (int i = 1; i < processors; i++) {
 516                                                 int done;
 517                                                 MPI_Recv(&done, 1, MPI_INT, i, tag, MPI_COMM_WORLD, &status);
 518                                         }
 519                                 }else{ //you are a child process
 520                                         MPI_Recv(&numFastaSeqs, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
 521                                         MPIPos.resize(numFastaSeqs+1);
 522                                         MPI_Recv(&MPIPos[0], (numFastaSeqs+1), MPI_LONG, 0, tag, MPI_COMM_WORLD, &status);
 523
 524                                         //figure out how many sequences you have to align
 525                                         numSeqsPerProcessor = numFastaSeqs / processors;
 526                                         int startIndex =  pid * numSeqsPerProcessor;
 527                                         if(pid == (processors - 1)){    numSeqsPerProcessor = numFastaSeqs - pid * numSeqsPerProcessor;         }
 528
 529
 530                                         //align your part
 531                                         driverMPI(startIndex, numSeqsPerProcessor, inMPI, outMPINewTax, outMPITempTax, MPIPos);
 532
 533                                         if (m->control_pressed) {  outputTypes.clear(); MPI_File_close(&inMPI);  MPI_File_close(&outMPINewTax);   MPI_File_close(&outMPITempTax);  delete classify; return 0;  }
 534
 535                                         int done = 0;
 536                                         MPI_Send(&done, 1, MPI_INT, 0, tag, MPI_COMM_WORLD);
 537                                 }
 538
 539                                 //close files
 540                                 MPI_File_close(&inMPI);
 541                                 MPI_File_close(&outMPINewTax);
 542                                 MPI_File_close(&outMPITempTax);
 543                                 MPI_Barrier(MPI_COMM_WORLD); //make everyone wait - just in case
 544
 545 #else
 546
 547                         vector<unsigned long int> positions = m->divideFile(fastaFileNames[s], processors);
 548
 549                         for (int i = 0; i < (positions.size()-1); i++) {
 550                                 lines.push_back(new linePair(positions[i], positions[(i+1)]));
 551                         }
 552
 553                 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
 554                         if(processors == 1){
 555                                 numFastaSeqs = driver(lines[0], newTaxonomyFile, tempTaxonomyFile, fastaFileNames[s]);
 556                         }
 557                         else{
 558                                 processIDS.resize(0);
 559
 560                                 numFastaSeqs = createProcesses(newTaxonomyFile, tempTaxonomyFile, fastaFileNames[s]);
 561
 562                                 rename((newTaxonomyFile + toString(processIDS[0]) + ".temp").c_str(), newTaxonomyFile.c_str());
 563                                 rename((tempTaxonomyFile + toString(processIDS[0]) + ".temp").c_str(), tempTaxonomyFile.c_str());
 564
 565                                 for(int i=1;i<processors;i++){
 566                                         appendTaxFiles((newTaxonomyFile + toString(processIDS[i]) + ".temp"), newTaxonomyFile);
 567                                         appendTaxFiles((tempTaxonomyFile + toString(processIDS[i]) + ".temp"), tempTaxonomyFile);
 568                                         remove((newTaxonomyFile + toString(processIDS[i]) + ".temp").c_str());
 569                                         remove((tempTaxonomyFile + toString(processIDS[i]) + ".temp").c_str());
 570                                 }
 571
 572                         }
 573         #else
 574                         numFastaSeqs = driver(lines[0], newTaxonomyFile, tempTaxonomyFile, fastaFileNames[s]);
 575         #endif
 576 #endif
 577
 578                 m->mothurOutEndLine();
 579                 m->mothurOut("It took " + toString(time(NULL) - start) + " secs to classify " + toString(numFastaSeqs) + " sequences."); m->mothurOutEndLine(); m->mothurOutEndLine();
 580                 start = time(NULL);
 581
 582
 583                 #ifdef USE_MPI
 584                         if (pid == 0) {  //this part does not need to be paralellized
 585
 586                                 if(namefile != "") { m->mothurOut("Reading " + namefileNames[s] + "..."); cout.flush();  MPIReadNamesFile(namefileNames[s]);  m->mothurOut("  Done."); m->mothurOutEndLine(); }
 587                 #else
 588                         //read namefile
 589                         if(namefile != "") {
 590
 591                             m->mothurOut("Reading " + namefileNames[s] + "..."); cout.flush();
 592
 593                                 nameMap.clear(); //remove old names
 594
 595                                 ifstream inNames;
 596                                 m->openInputFile(namefileNames[s], inNames);
 597
 598                                 string firstCol, secondCol;
 599                                 while(!inNames.eof()) {
 600                                         inNames >> firstCol >> secondCol; m->gobble(inNames);
 601
 602                                         vector<string> temp;
 603                                         m->splitAtComma(secondCol, temp);
 604
 605                                         nameMap[firstCol] = temp;
 606                                 }
 607                                 inNames.close();
 608
 609                                 m->mothurOut("  Done."); m->mothurOutEndLine();
 610                         }
 611                 #endif
 612
 613                         string group = "";
 614                         if (groupfile != "") {  group = groupfileNames[s]; }
 615
 616                         PhyloSummary taxaSum(taxonomyFileName, group);
 617
 618                         if (m->control_pressed) { outputTypes.clear();  for (int i = 0; i < outputNames.size(); i++) {  remove(outputNames[i].c_str()); } delete classify; return 0; }
 619
 620                         if (namefile == "") {  taxaSum.summarize(tempTaxonomyFile);  }
 621                         else {
 622                                 ifstream in;
 623                                 m->openInputFile(tempTaxonomyFile, in);
 624
 625                                 //read in users taxonomy file and add sequences to tree
 626                                 string name, taxon;
 627
 628                                 while(!in.eof()){
 629                                         in >> name >> taxon; m->gobble(in);
 630
 631                                         itNames = nameMap.find(name);
 632
 633                                         if (itNames == nameMap.end()) {
 634                                                 m->mothurOut(name + " is not in your name file please correct."); m->mothurOutEndLine(); exit(1);
 635                                         }else{
 636                                                 for (int i = 0; i < itNames->second.size(); i++) {
 637                                                         taxaSum.addSeqToTree(itNames->second[i], taxon);  //add it as many times as there are identical seqs
 638                                                 }
 639                                                 itNames->second.clear();
 640                                                 nameMap.erase(itNames->first);
 641                                         }
 642                                 }
 643                                 in.close();
 644                         }
 645                         remove(tempTaxonomyFile.c_str());
 646
 647                         if (m->control_pressed) {  outputTypes.clear(); for (int i = 0; i < outputNames.size(); i++) {  remove(outputNames[i].c_str()); } delete classify; return 0; }
 648
 649                         //print summary file
 650                         ofstream outTaxTree;
 651                         m->openOutputFile(taxSummary, outTaxTree);
 652                         taxaSum.print(outTaxTree);
 653                         outTaxTree.close();
 654
 655                         //output taxonomy with the unclassified bins added
 656                         ifstream inTax;
 657                         m->openInputFile(newTaxonomyFile, inTax);
 658
 659                         ofstream outTax;
 660                         string unclass = newTaxonomyFile + ".unclass.temp";
 661                         m->openOutputFile(unclass, outTax);
 662
 663                         //get maxLevel from phylotree so you know how many 'unclassified's to add
 664                         int maxLevel = taxaSum.getMaxLevel();
 665
 666                         //read taxfile - this reading and rewriting is done to preserve the confidence scores.
 667                         string name, taxon;
 668                         while (!inTax.eof()) {
 669                                 if (m->control_pressed) { outputTypes.clear();  for (int i = 0; i < outputNames.size(); i++) {  remove(outputNames[i].c_str()); } remove(unclass.c_str()); delete classify; return 0; }
 670
 671                                 inTax >> name >> taxon; m->gobble(inTax);
 672
 673                                 string newTax = addUnclassifieds(taxon, maxLevel);
 674
 675                                 outTax << name << '\t' << newTax << endl;
 676                         }
 677                         inTax.close();
 678                         outTax.close();
 679
 680                         remove(newTaxonomyFile.c_str());
 681                         rename(unclass.c_str(), newTaxonomyFile.c_str());
 682
 683                         m->mothurOutEndLine();
 684                         m->mothurOut("It took " + toString(time(NULL) - start) + " secs to create the summary file for " + toString(numFastaSeqs) + " sequences."); m->mothurOutEndLine(); m->mothurOutEndLine();
 685
 686                         #ifdef USE_MPI
 687                                 }
 688                         #endif
 689
 690                         m->mothurOutEndLine();
 691                         m->mothurOut("Output File Names: "); m->mothurOutEndLine();
 692                         for (int i = 0; i < outputNames.size(); i++) {  m->mothurOut(outputNames[i]); m->mothurOutEndLine();    }
 693                         m->mothurOutEndLine();
 694                 }
 695
 696                 delete classify;
 697                 return 0;
 698         }
 699         catch(exception& e) {
 700                 m->errorOut(e, "ClassifySeqsCommand", "execute");
 701                 exit(1);
 702         }
 703 }
 704
 705 /**************************************************************************************************/
 706 string ClassifySeqsCommand::addUnclassifieds(string tax, int maxlevel) {
 707         try{
 708                 string newTax, taxon;
 709                 int level = 0;
 710
 711                 //keep what you have counting the levels
 712                 while (tax.find_first_of(';') != -1) {
 713                         //get taxon
 714                         taxon = tax.substr(0,tax.find_first_of(';'))+';';
 715                         tax = tax.substr(tax.find_first_of(';')+1, tax.length());
 716                         newTax += taxon;
 717                         level++;
 718                 }
 719
 720                 //add "unclassified" until you reach maxLevel
 721                 while (level < maxlevel) {
 722                         newTax += "unclassified;";
 723                         level++;
 724                 }
 725
 726                 return newTax;
 727         }
 728         catch(exception& e) {
 729                 m->errorOut(e, "ClassifySeqsCommand", "addUnclassifieds");
 730                 exit(1);
 731         }
 732 }
 733
 734 /**************************************************************************************************/
 735
 736 int ClassifySeqsCommand::createProcesses(string taxFileName, string tempTaxFile, string filename) {
 737         try {
 738 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
 739                 int process = 0;
 740                 int num = 0;
 741
 742                 //loop through and create all the processes you want
 743                 while (process != processors) {
 744                         int pid = fork();
 745
 746                         if (pid > 0) {
 747                                 processIDS.push_back(pid);  //create map from line number to pid so you can append files in correct order later
 748                                 process++;
 749                         }else if (pid == 0){
 750                                 num = driver(lines[process], taxFileName + toString(getpid()) + ".temp", tempTaxFile + toString(getpid()) + ".temp", filename);
 751
 752                                 //pass numSeqs to parent
 753                                 ofstream out;
 754                                 string tempFile = filename + toString(getpid()) + ".num.temp";
 755                                 m->openOutputFile(tempFile, out);
 756                                 out << num << endl;
 757                                 out.close();
 758
 759                                 exit(0);
 760                         }else {
 761                                 m->mothurOut("[ERROR]: unable to spawn the necessary processes."); m->mothurOutEndLine();
 762                                 for (int i = 0; i < processIDS.size(); i++) { kill (processIDS[i], SIGINT); }
 763                                 exit(0);
 764                         }
 765                 }
 766
 767                 //force parent to wait until all the processes are done
 768                 for (int i=0;i<processors;i++) {
 769                         int temp = processIDS[i];
 770                         wait(&temp);
 771                 }
 772
 773                 for (int i = 0; i < processIDS.size(); i++) {
 774                         ifstream in;
 775                         string tempFile =  filename + toString(processIDS[i]) + ".num.temp";
 776                         m->openInputFile(tempFile, in);
 777                         if (!in.eof()) { int tempNum = 0; in >> tempNum; num += tempNum; }
 778                         in.close(); remove(tempFile.c_str());
 779                 }
 780
 781                 return num;
 782 #endif
 783         }
 784         catch(exception& e) {
 785                 m->errorOut(e, "ClassifySeqsCommand", "createProcesses");
 786                 exit(1);
 787         }
 788 }
 789 /**************************************************************************************************/
 790
 791 void ClassifySeqsCommand::appendTaxFiles(string temp, string filename) {
 792         try{
 793
 794                 ofstream output;
 795                 ifstream input;
 796                 m->openOutputFileAppend(filename, output);
 797                 m->openInputFile(temp, input);
 798
 799                 while(char c = input.get()){
 800                         if(input.eof())         {       break;                  }
 801                         else                            {       output << c;    }
 802                 }
 803
 804                 input.close();
 805                 output.close();
 806         }
 807         catch(exception& e) {
 808                 m->errorOut(e, "ClassifySeqsCommand", "appendTaxFiles");
 809                 exit(1);
 810         }
 811 }
 812
 813 //**********************************************************************************************************************
 814
 815 int ClassifySeqsCommand::driver(linePair* filePos, string taxFName, string tempTFName, string filename){
 816         try {
 817                 ofstream outTax;
 818                 m->openOutputFile(taxFName, outTax);
 819
 820                 ofstream outTaxSimple;
 821                 m->openOutputFile(tempTFName, outTaxSimple);
 822
 823                 ifstream inFASTA;
 824                 m->openInputFile(filename, inFASTA);
 825
 826                 string taxonomy;
 827
 828                 inFASTA.seekg(filePos->start);
 829
 830                 bool done = false;
 831                 int count = 0;
 832
 833                 while (!done) {
 834                         if (m->control_pressed) { return 0; }
 835
 836                         Sequence* candidateSeq = new Sequence(inFASTA); m->gobble(inFASTA);
 837
 838                         if (candidateSeq->getName() != "") {
 839                                 taxonomy = classify->getTaxonomy(candidateSeq);
 840
 841                                 if (m->control_pressed) { delete candidateSeq; return 0; }
 842
 843                                 if (taxonomy != "bad seq") {
 844                                         //output confidence scores or not
 845                                         if (probs) {
 846                                                 outTax << candidateSeq->getName() << '\t' << taxonomy << endl;
 847                                         }else{
 848                                                 outTax << candidateSeq->getName() << '\t' << classify->getSimpleTax() << endl;
 849                                         }
 850
 851                                         outTaxSimple << candidateSeq->getName() << '\t' << classify->getSimpleTax() << endl;
 852                                 }
 853                                 count++;
 854                         }
 855                         delete candidateSeq;
 856
 857                         #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
 858                                 unsigned long int pos = inFASTA.tellg();
 859                                 if ((pos == -1) || (pos >= filePos->end)) { break; }
 860                         #else
 861                                 if (inFASTA.eof()) { break; }
 862                         #endif
 863
 864                         //report progress
 865                         if((count) % 100 == 0){ m->mothurOut("Processing sequence: " + toString(count)); m->mothurOutEndLine();         }
 866                 }
 867                 //report progress
 868                 if((count) % 100 != 0){ m->mothurOut("Processing sequence: " + toString(count)); m->mothurOutEndLine();         }
 869
 870                 inFASTA.close();
 871                 outTax.close();
 872                 outTaxSimple.close();
 873
 874                 return count;
 875         }
 876         catch(exception& e) {
 877                 m->errorOut(e, "ClassifySeqsCommand", "driver");
 878                 exit(1);
 879         }
 880 }
 881 //**********************************************************************************************************************
 882 #ifdef USE_MPI
 883 int ClassifySeqsCommand::driverMPI(int start, int num, MPI_File& inMPI, MPI_File& newFile, MPI_File& tempFile, vector<unsigned long int>& MPIPos){
 884         try {
 885                 MPI_Status statusNew;
 886                 MPI_Status statusTemp;
 887                 MPI_Status status;
 888
 889                 int pid;
 890                 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 891
 892                 string taxonomy;
 893                 string outputString;
 894
 895                 for(int i=0;i<num;i++){
 896
 897                         if (m->control_pressed) { return 0; }
 898
 899                         //read next sequence
 900                         int length = MPIPos[start+i+1] - MPIPos[start+i];
 901                         char* buf4 = new char[length];
 902                         MPI_File_read_at(inMPI, MPIPos[start+i], buf4, length, MPI_CHAR, &status);
 903
 904                         string tempBuf = buf4;
 905                         if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length);  }
 906                         istringstream iss (tempBuf,istringstream::in);
 907                         delete buf4;
 908
 909                         Sequence* candidateSeq = new Sequence(iss);
 910
 911                         if (candidateSeq->getName() != "") {
 912                                 taxonomy = classify->getTaxonomy(candidateSeq);
 913
 914                                 if (taxonomy != "bad seq") {
 915                                         //output confidence scores or not
 916                                         if (probs) {
 917                                                 outputString =  candidateSeq->getName() + "\t" + taxonomy + "\n";
 918                                         }else{
 919                                                 outputString =  candidateSeq->getName() + "\t" + classify->getSimpleTax() + "\n";
 920                                         }
 921
 922                                         int length = outputString.length();
 923                                         char* buf2 = new char[length];
 924                                         memcpy(buf2, outputString.c_str(), length);
 925
 926                                         MPI_File_write_shared(newFile, buf2, length, MPI_CHAR, &statusNew);
 927                                         delete buf2;
 928
 929                                         outputString =  candidateSeq->getName() + "\t" + classify->getSimpleTax() + "\n";
 930                                         length = outputString.length();
 931                                         char* buf = new char[length];
 932                                         memcpy(buf, outputString.c_str(), length);
 933
 934                                         MPI_File_write_shared(tempFile, buf, length, MPI_CHAR, &statusTemp);
 935                                         delete buf;
 936                                 }
 937                         }
 938                         delete candidateSeq;
 939
 940                         if((i+1) % 100 == 0){   cout << "Classifying sequence " << (i+1) << endl;       }
 941                 }
 942
 943                 if(num % 100 != 0){     cout << "Classifying sequence " << (num) << endl;       }
 944
 945
 946                 return 1;
 947         }
 948         catch(exception& e) {
 949                 m->errorOut(e, "ClassifySeqsCommand", "driverMPI");
 950                 exit(1);
 951         }
 952 }
 953
 954 //**********************************************************************************************************************
 955 int ClassifySeqsCommand::MPIReadNamesFile(string nameFilename){
 956         try {
 957
 958                 nameMap.clear(); //remove old names
 959
 960                 MPI_File inMPI;
 961                 MPI_Offset size;
 962                 MPI_Status status;
 963
 964                 //char* inFileName = new char[nameFilename.length()];
 965                 //memcpy(inFileName, nameFilename.c_str(), nameFilename.length());
 966
 967                 char inFileName[1024];
 968                 strcpy(inFileName, nameFilename.c_str());
 969
 970                 MPI_File_open(MPI_COMM_WORLD, inFileName, MPI_MODE_RDONLY, MPI_INFO_NULL, &inMPI);
 971                 MPI_File_get_size(inMPI, &size);
 972                 //delete inFileName;
 973
 974                 char* buffer = new char[size];
 975                 MPI_File_read(inMPI, buffer, size, MPI_CHAR, &status);
 976
 977                 string tempBuf = buffer;
 978                 if (tempBuf.length() > size) { tempBuf = tempBuf.substr(0, size);  }
 979                 istringstream iss (tempBuf,istringstream::in);
 980                 delete buffer;
 981
 982                 string firstCol, secondCol;
 983                 while(!iss.eof()) {
 984                         iss >> firstCol >> secondCol; m->gobble(iss);
 985
 986                         vector<string> temp;
 987                         m->splitAtComma(secondCol, temp);
 988
 989                         nameMap[firstCol] = temp;
 990                 }
 991
 992                 MPI_File_close(&inMPI);
 993
 994                 return 1;
 995         }
 996         catch(exception& e) {
 997                 m->errorOut(e, "ClassifySeqsCommand", "MPIReadNamesFile");
 998                 exit(1);
 999         }
1000 }
1001 #endif
1002 /**************************************************************************************************/