distancecommand.cpp

   1 /*
   2  *  distancecommand.cpp
   3  *  Mothur
   4  *
   5  *  Created by Sarah Westcott on 5/7/09.
   6  *  Copyright 2009 Schloss Lab UMASS Amherst. All rights reserved.
   7  *
   8  */
   9
  10 #include "distancecommand.h"
  11 #include "ignoregaps.h"
  12 #include "eachgapdist.h"
  13 #include "eachgapignore.h"
  14 #include "onegapdist.h"
  15 #include "onegapignore.h"
  16
  17 //**********************************************************************************************************************
  18
  19 DistanceCommand::DistanceCommand(string option) {
  20         try {
  21                 abort = false;
  22                 Estimators.clear();
  23
  24                 //allow user to run help
  25                 if(option == "help") { help(); abort = true; }
  26
  27                 else {
  28                         //valid paramters for this command
  29                         string Array[] =  {"fasta", "output", "calc", "countends", "cutoff", "processors", "outputdir","inputdir"};
  30                         vector<string> myArray (Array, Array+(sizeof(Array)/sizeof(string)));
  31
  32                         OptionParser parser(option);
  33                         map<string, string> parameters = parser.getParameters();
  34
  35                         ValidParameters validParameter;
  36                         map<string, string>::iterator it2;
  37
  38                         //check to make sure all parameters are valid for command
  39                         for (it2 = parameters.begin(); it2 != parameters.end(); it2++) {
  40                                 if (validParameter.isValidParameter(it2->first, myArray, it2->second) != true) {  abort = true;  }
  41                         }
  42
  43                         //if the user changes the input directory command factory will send this info to us in the output parameter
  44                         string inputDir = validParameter.validFile(parameters, "inputdir", false);
  45                         if (inputDir == "not found"){   inputDir = "";          }
  46                         else {
  47                                 string path;
  48                                 it2 = parameters.find("fasta");
  49                                 //user has given a template file
  50                                 if(it2 != parameters.end()){
  51                                         path = hasPath(it2->second);
  52                                         //if the user has not given a path then, add inputdir. else leave path alone.
  53                                         if (path == "") {       parameters["fasta"] = inputDir + it2->second;           }
  54                                 }
  55                         }
  56
  57                         //check for required parameters
  58                         fastafile = validParameter.validFile(parameters, "fasta", true);
  59                         if (fastafile == "not found") { m->mothurOut("fasta is a required parameter for the dist.seqs command."); m->mothurOutEndLine(); abort = true; }
  60                         else if (fastafile == "not open") { abort = true; }
  61                         else{
  62                                 ifstream inFASTA;
  63                                 openInputFile(fastafile, inFASTA);
  64                                 alignDB = SequenceDB(inFASTA);
  65                                 inFASTA.close();
  66                         }
  67
  68                         //if the user changes the output directory command factory will send this info to us in the output parameter
  69                         outputDir = validParameter.validFile(parameters, "outputdir", false);           if (outputDir == "not found"){
  70                                 outputDir = "";
  71                                 outputDir += hasPath(fastafile); //if user entered a file with a path then preserve it
  72                         }
  73
  74                         //check for optional parameter and set defaults
  75                         // ...at some point should added some additional type checking...
  76                         calc = validParameter.validFile(parameters, "calc", false);
  77                         if (calc == "not found") { calc = "onegap";  }
  78                         else {
  79                                  if (calc == "default")  {  calc = "onegap";  }
  80                         }
  81                         splitAtDash(calc, Estimators);
  82
  83                         string temp;
  84                         temp = validParameter.validFile(parameters, "countends", false);        if(temp == "not found"){        temp = "T";     }
  85                         convert(temp, countends);
  86
  87                         temp = validParameter.validFile(parameters, "cutoff", false);           if(temp == "not found"){        temp = "1.0"; }
  88                         convert(temp, cutoff);
  89
  90                         temp = validParameter.validFile(parameters, "processors", false);       if(temp == "not found"){        temp = "1"; }
  91                         convert(temp, processors);
  92
  93                         output = validParameter.validFile(parameters, "output", false);         if(output == "not found"){      output = "column"; }
  94
  95                         if ((output != "column") && (output != "lt") && (output != "square")) { m->mothurOut(output + " is not a valid output form. Options are column, lt and square. I will use column."); m->mothurOutEndLine(); output = "column"; }
  96
  97                         ValidCalculators validCalculator;
  98
  99                         if (isTrue(countends) == true) {
 100                                 for (int i=0; i<Estimators.size(); i++) {
 101                                         if (validCalculator.isValidCalculator("distance", Estimators[i]) == true) {
 102                                                 if (Estimators[i] == "nogaps")                  {       distCalculator = new ignoreGaps();      }
 103                                                 else if (Estimators[i] == "eachgap")    {       distCalculator = new eachGapDist();     }
 104                                                 else if (Estimators[i] == "onegap")             {       distCalculator = new oneGapDist();      }
 105                                         }
 106                                 }
 107                         }else {
 108                                 for (int i=0; i<Estimators.size(); i++) {
 109                                         if (validCalculator.isValidCalculator("distance", Estimators[i]) == true) {
 110                                                 if (Estimators[i] == "nogaps")          {       distCalculator = new ignoreGaps();                                      }
 111                                                 else if (Estimators[i] == "eachgap"){   distCalculator = new eachGapIgnoreTermGapDist();        }
 112                                                 else if (Estimators[i] == "onegap")     {       distCalculator = new oneGapIgnoreTermGapDist();         }
 113                                         }
 114                                 }
 115                         }
 116
 117                 }
 118
 119         }
 120         catch(exception& e) {
 121                 m->errorOut(e, "DistanceCommand", "DistanceCommand");
 122                 exit(1);
 123         }
 124 }
 125
 126 //**********************************************************************************************************************
 127
 128 DistanceCommand::~DistanceCommand(){
 129
 130         for(int i=0;i<lines.size();i++){
 131                 delete lines[i];
 132         }
 133
 134 }
 135
 136 //**********************************************************************************************************************
 137
 138 void DistanceCommand::help(){
 139         try {
 140                 m->mothurOut("The dist.seqs command reads a file containing sequences and creates a distance file.\n");
 141                 m->mothurOut("The dist.seqs command parameters are fasta, calc, countends, output, cutoff and processors.  \n");
 142                 m->mothurOut("The fasta parameter is required.\n");
 143                 m->mothurOut("The calc parameter allows you to specify the method of calculating the distances.  Your options are: nogaps, onegap or eachgap. The default is onegap.\n");
 144                 m->mothurOut("The countends parameter allows you to specify whether to include terminal gaps in distance.  Your options are: T or F. The default is T.\n");
 145                 m->mothurOut("The cutoff parameter allows you to specify maximum distance to keep. The default is 1.0.\n");
 146                 m->mothurOut("The output parameter allows you to specify format of your distance matrix. Options are column, lt, and square. The default is column.\n");
 147                 m->mothurOut("The processors parameter allows you to specify number of processors to use.  The default is 1.\n");
 148                 m->mothurOut("The dist.seqs command should be in the following format: \n");
 149                 m->mothurOut("dist.seqs(fasta=yourFastaFile, calc=yourCalc, countends=yourEnds, cutoff= yourCutOff, processors=yourProcessors) \n");
 150                 m->mothurOut("Example dist.seqs(fasta=amazon.fasta, calc=eachgap, countends=F, cutoff= 2.0, processors=3).\n");
 151                 m->mothurOut("Note: No spaces between parameter labels (i.e. calc), '=' and parameters (i.e.yourCalc).\n\n");
 152         }
 153         catch(exception& e) {
 154                 m->errorOut(e, "DistanceCommand", "help");
 155                 exit(1);
 156         }
 157 }
 158 //**********************************************************************************************************************
 159
 160 int DistanceCommand::execute(){
 161         try {
 162
 163                 if (abort == true) { return 0; }
 164
 165                 int numSeqs = alignDB.getNumSeqs();
 166                 cutoff += 0.005;
 167
 168                 string outputFile;
 169
 170                 if (output == "lt") { //does the user want lower triangle phylip formatted file
 171                         outputFile = outputDir + getRootName(getSimpleName(fastafile)) + "phylip.dist";
 172                         remove(outputFile.c_str());
 173
 174                         //output numSeqs to phylip formatted dist file
 175                 }else if (output == "column") { //user wants column format
 176                         outputFile = outputDir + getRootName(getSimpleName(fastafile)) + "dist";
 177                         remove(outputFile.c_str());
 178                 }else { //assume square
 179                         outputFile = outputDir + getRootName(getSimpleName(fastafile)) + "square.dist";
 180                         remove(outputFile.c_str());
 181                 }
 182
 183
 184 #ifdef USE_MPI
 185
 186                 int pid, start, end;
 187                 int tag = 2001;
 188
 189                 MPI_Status status;
 190                 MPI_Comm_size(MPI_COMM_WORLD, &processors); //set processors to the number of mpi processes running
 191                 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
 192
 193                 //each process gets where it should start and stop in the file
 194                 start = int (sqrt(float(pid)/float(processors)) * numSeqs);
 195                 end = int (sqrt(float(pid+1)/float(processors)) * numSeqs);
 196
 197                 if (pid == 0) { //you are the root process
 198                         //do your part
 199                         string outputMyPart;
 200                         driverMPI(start, end, outputMyPart, cutoff);
 201
 202                         ofstream out;
 203                         openOutputFile(outputFile, out);
 204
 205                         out << outputMyPart;
 206
 207                         //get the childrens parts
 208                         for(int i = 1; i < processors; i++) {
 209                                 int length;
 210                                 MPI_Recv(&length, 1, MPI_INT, i, tag, MPI_COMM_WORLD, &status);
 211
 212                                 char buf[length];
 213
 214                                 MPI_Recv(buf, length, MPI_CHAR, i, tag, MPI_COMM_WORLD, &status);
 215
 216                                 outputMyPart = buf;
 217                                 out << outputMyPart;
 218                         }
 219
 220                         out.close();
 221
 222                 }else { //you are a child process
 223                         //do your part
 224                         string outputMyPart;
 225                         driverMPI(start, end, outputMyPart, cutoff);
 226
 227                         //send results to parent
 228                         int length = outputMyPart.length();
 229                         char buf[length];
 230                         strcpy(buf, outputMyPart.c_str());
 231
 232                         MPI_Send( &length, 1, MPI_INT, 0, tag, MPI_COMM_WORLD);
 233                         MPI_Send(buf, length, MPI_CHAR, 0, tag, MPI_COMM_WORLD);
 234                 }
 235
 236
 237 #else
 238
 239         #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
 240                 //if you don't need to fork anything
 241                 if(processors == 1){
 242                         driver(0, numSeqs, outputFile, cutoff);
 243                 }else{ //you have multiple processors
 244
 245                         for (int i = 0; i < processors; i++) {
 246                                 lines.push_back(new linePair());
 247                                 lines[i]->start = int (sqrt(float(i)/float(processors)) * numSeqs);
 248                                 lines[i]->end = int (sqrt(float(i+1)/float(processors)) * numSeqs);
 249                         }
 250
 251                         createProcesses(outputFile);
 252
 253                         map<int, int>::iterator it = processIDS.begin();
 254                         rename((outputFile + toString(it->second) + ".temp").c_str(), outputFile.c_str());
 255                         it++;
 256
 257                         //append and remove temp files
 258                         for (; it != processIDS.end(); it++) {
 259                                 appendFiles((outputFile + toString(it->second) + ".temp"), outputFile);
 260                                 remove((outputFile + toString(it->second) + ".temp").c_str());
 261                         }
 262                 }
 263         #else
 264                 ifstream inFASTA;
 265                 driver(0, numSeqs, outputFile, cutoff);
 266         #endif
 267
 268 #endif
 269                 if (m->control_pressed) { delete distCalculator; remove(outputFile.c_str()); return 0; }
 270
 271                 #ifdef USE_MPI
 272                         MPI_Comm_rank(MPI_COMM_WORLD, &pid);
 273
 274                         if (pid == 0) { //only one process should output to screen
 275                 #endif
 276
 277                 if (output == "square") {  convertMatrix(outputFile); }
 278
 279                 #ifdef USE_MPI
 280                         }
 281                 #endif
 282
 283                 if (m->control_pressed) { delete distCalculator; remove(outputFile.c_str()); return 0; }
 284
 285                 delete distCalculator;
 286
 287                 m->mothurOutEndLine();
 288                 m->mothurOut("Output File Name: "); m->mothurOutEndLine();
 289                 m->mothurOut(outputFile); m->mothurOutEndLine();
 290                 m->mothurOutEndLine();
 291
 292                 return 0;
 293
 294         }
 295         catch(exception& e) {
 296                 m->errorOut(e, "DistanceCommand", "execute");
 297                 exit(1);
 298         }
 299 }
 300 /**************************************************************************************************/
 301 void DistanceCommand::createProcesses(string filename) {
 302         try {
 303 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
 304                 int process = 0;
 305                 processIDS.clear();
 306
 307                 //loop through and create all the processes you want
 308                 while (process != processors) {
 309                         int pid = fork();
 310
 311                         if (pid > 0) {
 312                                 processIDS[lines[process]->end] = pid;  //create map from line number to pid so you can append files in correct order later
 313                                 process++;
 314                         }else if (pid == 0){
 315                                 driver(lines[process]->start, lines[process]->end, filename + toString(getpid()) + ".temp", cutoff);
 316                                 exit(0);
 317                         }else { m->mothurOut("unable to spawn the necessary processes."); m->mothurOutEndLine(); exit(0); }
 318                 }
 319
 320                 //force parent to wait until all the processes are done
 321                 for (map<int, int>::iterator it = processIDS.begin(); it != processIDS.end(); it++) {
 322                         int temp = it->second;
 323                         wait(&temp);
 324                 }
 325 #endif
 326         }
 327         catch(exception& e) {
 328                 m->errorOut(e, "DistanceCommand", "createProcesses");
 329                 exit(1);
 330         }
 331 }
 332
 333 /**************************************************************************************************/
 334 /////// need to fix to work with calcs and sequencedb
 335 int DistanceCommand::driver(int startLine, int endLine, string dFileName, float cutoff){
 336         try {
 337
 338                 int startTime = time(NULL);
 339
 340                 //column file
 341                 ofstream outFile(dFileName.c_str(), ios::trunc);
 342                 outFile.setf(ios::fixed, ios::showpoint);
 343                 outFile << setprecision(4);
 344
 345                 if((output == "lt") && startLine == 0){ outFile << alignDB.getNumSeqs() << endl;        }
 346
 347                 for(int i=startLine;i<endLine;i++){
 348                         if(output == "lt")      {
 349                                 string name = alignDB.get(i).getName();
 350                                 if (name.length() < 10) { //pad with spaces to make compatible
 351                                         while (name.length() < 10) {  name += " ";  }
 352                                 }
 353                                 outFile << name << '\t';
 354                         }
 355                         for(int j=0;j<i;j++){
 356
 357                                 if (m->control_pressed) { outFile.close(); return 0;  }
 358
 359                                 distCalculator->calcDist(alignDB.get(i), alignDB.get(j));
 360                                 double dist = distCalculator->getDist();
 361
 362                                 if(dist <= cutoff){
 363                                         if (output == "column") { outFile << alignDB.get(i).getName() << ' ' << alignDB.get(j).getName() << ' ' << dist << endl; }
 364                                 }
 365                                 if (output == "lt") {  outFile << dist << '\t'; }
 366
 367                                 if (output == "square") { //make a square column you can convert to square phylip
 368                                         outFile << alignDB.get(i).getName() << '\t' << alignDB.get(j).getName() << '\t' << dist << endl;
 369                                         outFile << alignDB.get(j).getName() << '\t' << alignDB.get(i).getName() << '\t' << dist << endl;
 370                                 }
 371
 372                         }
 373
 374                         if (output == "lt") { outFile << endl; }
 375
 376                         if(i % 100 == 0){
 377                                 m->mothurOut(toString(i) + "\t" + toString(time(NULL) - startTime)); m->mothurOutEndLine();
 378                         }
 379
 380                 }
 381                 m->mothurOut(toString(endLine-1) + "\t" + toString(time(NULL) - startTime)); m->mothurOutEndLine();
 382
 383                 outFile.close();
 384
 385                 return 1;
 386         }
 387         catch(exception& e) {
 388                 m->errorOut(e, "DistanceCommand", "driver");
 389                 exit(1);
 390         }
 391 }
 392 /**************************************************************************************************/
 393 /////// need to fix to work with calcs and sequencedb
 394 int DistanceCommand::driverMPI(int startLine, int endLine, string& outputString, float cutoff){
 395         try {
 396
 397                 int startTime = time(NULL);
 398
 399                 outputString = "";
 400
 401                 if((output == "lt") && startLine == 0){ outputString += (toString(alignDB.getNumSeqs()) + '\n');        }
 402
 403                 for(int i=startLine;i<endLine;i++){
 404
 405                         if(output == "lt")      {
 406                                 string name = alignDB.get(i).getName();
 407                                 if (name.length() < 10) { //pad with spaces to make compatible
 408                                         while (name.length() < 10) {  name += " ";  }
 409                                 }
 410                                 outputString += (name + '\t');
 411                         }
 412                         for(int j=0;j<i;j++){
 413
 414                                 if (m->control_pressed) {  return 0;  }
 415
 416                                 distCalculator->calcDist(alignDB.get(i), alignDB.get(j));
 417                                 double dist = distCalculator->getDist();
 418
 419                                 if(dist <= cutoff){
 420                                         if (output == "column") { outputString += (alignDB.get(i).getName() + ' ' + alignDB.get(j).getName() + ' ' + toString(dist) + '\n'); }
 421                                 }
 422                                 if (output == "lt") {   outputString += (toString(dist) + '\t'); }
 423
 424                                 if (output == "square") { //make a square column you can convert to square phylip
 425                                         outputString += (alignDB.get(i).getName() + ' ' + alignDB.get(j).getName() + ' ' + toString(dist) + '\n');
 426                                         outputString += (alignDB.get(j).getName() + ' ' + alignDB.get(i).getName() + ' ' + toString(dist) + '\n');
 427                                 }
 428
 429                         }
 430
 431                         if (output == "lt") { outputString += '\n'; }
 432
 433                         if(i % 100 == 0){
 434                                 m->mothurOut(toString(i) + "\t" + toString(time(NULL) - startTime)); m->mothurOutEndLine();
 435                         }
 436
 437                 }
 438                 m->mothurOut(toString(endLine-1) + "\t" + toString(time(NULL) - startTime)); m->mothurOutEndLine();
 439
 440                 return 1;
 441         }
 442         catch(exception& e) {
 443                 m->errorOut(e, "DistanceCommand", "driver");
 444                 exit(1);
 445         }
 446 }
 447
 448 /**************************************************************************************************/
 449 int DistanceCommand::convertMatrix(string outputFile) {
 450         try{
 451
 452                 //sort file by first column so the distances for each row are together
 453                 string outfile = getRootName(outputFile) + "sorted.dist.temp";
 454
 455                 //use the unix sort
 456                 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
 457                         string command = "sort -n " + outputFile + " -o " + outfile;
 458                         system(command.c_str());
 459                 #else //sort using windows sort
 460                         string command = "sort " + outputFile + " /O " + outfile;
 461                         system(command.c_str());
 462                 #endif
 463
 464
 465                 //output to new file distance for each row and save positions in file where new row begins
 466                 ifstream in;
 467                 openInputFile(outfile, in);
 468
 469                 ofstream out;
 470                 openOutputFile(outputFile, out);
 471
 472                 out.setf(ios::fixed, ios::floatfield); out.setf(ios::showpoint);
 473
 474                 out << alignDB.getNumSeqs() << endl;
 475
 476                 //get first currentRow
 477                 string first, currentRow, second;
 478                 float dist;
 479                 map<string, float> rowDists; //take advantage of the fact that maps are already sorted by key
 480                 map<string, float>::iterator it;
 481
 482                 in >> first;
 483                 currentRow = first;
 484
 485                 rowDists[first] = 0.00; //distance to yourself is 0.0
 486
 487                 in.seekg(0);
 488                 //openInputFile(outfile, in);
 489
 490                 while(!in.eof()) {
 491                         if (m->control_pressed) { in.close(); remove(outfile.c_str()); out.close(); return 0; }
 492
 493                         in >> first >> second >> dist; gobble(in);
 494
 495                         if (first != currentRow) {
 496                                 //print out last row
 497                                 out << currentRow << '\t'; //print name
 498
 499                                 //print dists
 500                                 for (it = rowDists.begin(); it != rowDists.end(); it++) {
 501                                         out << it->second << '\t';
 502                                 }
 503                                 out << endl;
 504
 505                                 //start new row
 506                                 currentRow = first;
 507                                 rowDists.clear();
 508                                 rowDists[first] = 0.00;
 509                                 rowDists[second] = dist;
 510                         }else{
 511                                 rowDists[second] = dist;
 512                         }
 513                 }
 514                 //print out last row
 515                 out << currentRow << '\t'; //print name
 516
 517                 //print dists
 518                 for (it = rowDists.begin(); it != rowDists.end(); it++) {
 519                         out << it->second << '\t';
 520                 }
 521                 out << endl;
 522
 523                 in.close();
 524                 out.close();
 525
 526                 remove(outfile.c_str());
 527
 528                 return 1;
 529
 530         }
 531         catch(exception& e) {
 532                 m->errorOut(e, "DistanceCommand", "convertMatrix");
 533                 exit(1);
 534         }
 535 }
 536 /**************************************************************************************************
 537 void DistanceCommand::appendFiles(string temp, string filename) {
 538         try{
 539                 ofstream output;
 540                 ifstream input;
 541
 542                 //open output file in append mode
 543                 openOutputFileAppend(filename, output);
 544                 openInputFile(temp, input);
 545
 546                 while(char c = input.get()){
 547                         if(input.eof())         {       break;                  }
 548                         else                            {       output << c;    }
 549                 }
 550
 551                 input.close();
 552                 output.close();
 553         }
 554         catch(exception& e) {
 555                 m->errorOut(e, "DistanceCommand", "appendFiles");
 556                 exit(1);
 557         }
 558 }
 559 /**************************************************************************************************/
 560
 561