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