2 * filterseqscommand.cpp
5 * Created by Thomas Ryabin on 5/4/09.
6 * Copyright 2009 Schloss Lab UMASS Amherst. All rights reserved.
10 #include "filterseqscommand.h"
11 #include "sequence.hpp"
14 //**********************************************************************************************************************
15 vector<string> FilterSeqsCommand::setParameters(){
17 CommandParameter pfasta("fasta", "InputTypes", "", "", "none", "none", "none",false,true); parameters.push_back(pfasta);
18 CommandParameter phard("hard", "InputTypes", "", "", "none", "none", "none",false,false); parameters.push_back(phard);
19 CommandParameter ptrump("trump", "String", "", "*", "", "", "",false,false); parameters.push_back(ptrump);
20 CommandParameter psoft("soft", "Number", "", "0", "", "", "",false,false); parameters.push_back(psoft);
21 CommandParameter pvertical("vertical", "Boolean", "", "T", "", "", "",false,false); parameters.push_back(pvertical);
22 CommandParameter pprocessors("processors", "Number", "", "1", "", "", "",false,false); parameters.push_back(pprocessors);
23 CommandParameter pinputdir("inputdir", "String", "", "", "", "", "",false,false); parameters.push_back(pinputdir);
24 CommandParameter poutputdir("outputdir", "String", "", "", "", "", "",false,false); parameters.push_back(poutputdir);
26 vector<string> myArray;
27 for (int i = 0; i < parameters.size(); i++) { myArray.push_back(parameters[i].name); }
31 m->errorOut(e, "FilterSeqsCommand", "setParameters");
35 //**********************************************************************************************************************
36 string FilterSeqsCommand::getHelpString(){
38 string helpString = "";
39 helpString += "The filter.seqs command reads a file containing sequences and creates a .filter and .filter.fasta file.\n";
40 helpString += "The filter.seqs command parameters are fasta, trump, soft, hard, processors and vertical. \n";
41 helpString += "The fasta parameter is required, unless you have a valid current fasta file. You may enter several fasta files to build the filter from and filter, by separating their names with -'s.\n";
42 helpString += "For example: fasta=abrecovery.fasta-amazon.fasta \n";
43 helpString += "The trump option will remove a column if the trump character is found at that position in any sequence of the alignment. Default=*, meaning no trump. \n";
44 helpString += "A soft mask removes any column where the dominant base (i.e. A, T, G, C, or U) does not occur in at least a designated percentage of sequences. Default=0.\n";
45 helpString += "The hard parameter allows you to enter a file containing the filter you want to use.\n";
46 helpString += "The vertical parameter removes columns where all sequences contain a gap character. The default is T.\n";
47 helpString += "The processors parameter allows you to specify the number of processors to use. The default is 1.\n";
48 helpString += "The filter.seqs command should be in the following format: \n";
49 helpString += "filter.seqs(fasta=yourFastaFile, trump=yourTrump) \n";
50 helpString += "Example filter.seqs(fasta=abrecovery.fasta, trump=.).\n";
51 helpString += "Note: No spaces between parameter labels (i.e. fasta), '=' and parameters (i.e.yourFasta).\n";
55 m->errorOut(e, "FilterSeqsCommand", "getHelpString");
59 //**********************************************************************************************************************
60 FilterSeqsCommand::FilterSeqsCommand(){
62 abort = true; calledHelp = true;
64 vector<string> tempOutNames;
65 outputTypes["fasta"] = tempOutNames;
66 outputTypes["filter"] = tempOutNames;
69 m->errorOut(e, "FilterSeqsCommand", "FilterSeqsCommand");
73 /**************************************************************************************/
74 FilterSeqsCommand::FilterSeqsCommand(string option) {
76 abort = false; calledHelp = false;
79 //allow user to run help
80 if(option == "help") { help(); abort = true; calledHelp = true; }
81 else if(option == "citation") { citation(); abort = true; calledHelp = true;}
84 vector<string> myArray = setParameters();
86 OptionParser parser(option);
87 map<string,string> parameters = parser.getParameters();
89 ValidParameters validParameter("filter.seqs");
90 map<string,string>::iterator it;
92 //check to make sure all parameters are valid for command
93 for (it = parameters.begin(); it != parameters.end(); it++) {
94 if (validParameter.isValidParameter(it->first, myArray, it->second) != true) { abort = true; }
97 //initialize outputTypes
98 vector<string> tempOutNames;
99 outputTypes["fasta"] = tempOutNames;
100 outputTypes["filter"] = tempOutNames;
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 = ""; }
107 it = parameters.find("fasta");
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["fasta"] = inputDir + it->second; }
115 it = parameters.find("hard");
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["hard"] = inputDir + it->second; }
124 //check for required parameters
125 fasta = validParameter.validFile(parameters, "fasta", false);
126 if (fasta == "not found") {
127 fasta = m->getFastaFile();
128 if (fasta != "") { fastafileNames.push_back(fasta); m->mothurOut("Using " + fasta + " as input file for the fasta parameter."); m->mothurOutEndLine(); }
129 else { m->mothurOut("You have no current fastafile and the fasta parameter is required."); m->mothurOutEndLine(); abort = true; }
132 m->splitAtDash(fasta, fastafileNames);
134 //go through files and make sure they are good, if not, then disregard them
135 for (int i = 0; i < fastafileNames.size(); i++) {
138 if (fastafileNames[i] == "current") {
139 fastafileNames[i] = m->getFastaFile();
140 if (fastafileNames[i] != "") { m->mothurOut("Using " + fastafileNames[i] + " as input file for the fasta parameter where you had given current."); m->mothurOutEndLine(); }
142 m->mothurOut("You have no current fastafile, ignoring current."); m->mothurOutEndLine(); ignore=true;
143 //erase from file list
144 fastafileNames.erase(fastafileNames.begin()+i);
150 if (inputDir != "") {
151 string path = m->hasPath(fastafileNames[i]);
152 //if the user has not given a path then, add inputdir. else leave path alone.
153 if (path == "") { fastafileNames[i] = inputDir + fastafileNames[i]; }
157 int ableToOpen = m->openInputFile(fastafileNames[i], in, "noerror");
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();
165 ableToOpen = m->openInputFile(tryPath, in2, "noerror");
167 fastafileNames[i] = tryPath;
171 //if you can't open it, try default location
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();
177 ableToOpen = m->openInputFile(tryPath, in2, "noerror");
179 fastafileNames[i] = tryPath;
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);
191 string simpleName = m->getSimpleName(fastafileNames[i]);
192 filterFileName += simpleName.substr(0, simpleName.find_first_of('.'));
193 m->setFastaFile(fastafileNames[i]);
199 //make sure there is at least one valid file left
200 if (fastafileNames.size() == 0) { m->mothurOut("no valid files."); m->mothurOutEndLine(); abort = true; }
204 //if the user changes the output directory command factory will send this info to us in the output parameter
205 outputDir = validParameter.validFile(parameters, "outputdir", false); if (outputDir == "not found"){
207 outputDir += m->hasPath(fastafileNames[0]); //if user entered a file with a path then preserve it
210 //check for optional parameter and set defaults
211 // ...at some point should added some additional type checking...
214 hard = validParameter.validFile(parameters, "hard", true); if (hard == "not found") { hard = ""; }
215 else if (hard == "not open") { hard = ""; abort = true; }
217 temp = validParameter.validFile(parameters, "trump", false); if (temp == "not found") { temp = "*"; }
220 temp = validParameter.validFile(parameters, "soft", false); if (temp == "not found") { soft = 0; }
221 else { soft = (float)atoi(temp.c_str()) / 100.0; }
223 temp = validParameter.validFile(parameters, "processors", false); if (temp == "not found"){ temp = m->getProcessors(); }
224 m->setProcessors(temp);
225 m->mothurConvert(temp, processors);
227 vertical = validParameter.validFile(parameters, "vertical", false);
228 if (vertical == "not found") {
229 if ((hard == "") && (trump == '*') && (soft == 0)) { vertical = "T"; } //you have not given a hard file or set the trump char.
230 else { vertical = "F"; }
237 catch(exception& e) {
238 m->errorOut(e, "FilterSeqsCommand", "FilterSeqsCommand");
242 /**************************************************************************************/
244 int FilterSeqsCommand::execute() {
247 if (abort == true) { if (calledHelp) { return 0; } return 2; }
250 m->openInputFile(fastafileNames[0], inFASTA);
252 Sequence testSeq(inFASTA);
253 alignmentLength = testSeq.getAlignLength();
256 ////////////create filter/////////////////
257 m->mothurOut("Creating Filter... "); m->mothurOutEndLine();
259 filter = createFilter();
261 m->mothurOutEndLine(); m->mothurOutEndLine();
263 if (m->control_pressed) { outputTypes.clear(); return 0; }
267 MPI_Comm_rank(MPI_COMM_WORLD, &pid);
269 if (pid == 0) { //only one process should output the filter
274 //prevent giantic file name
276 if (fastafileNames.size() > 3) { filterFile = outputDir + "merge.filter"; }
277 else { filterFile = outputDir + filterFileName + ".filter"; }
279 m->openOutputFile(filterFile, outFilter);
280 outFilter << filter << endl;
282 outputNames.push_back(filterFile); outputTypes["filter"].push_back(filterFile);
288 ////////////run filter/////////////////
290 m->mothurOut("Running Filter... "); m->mothurOutEndLine();
294 m->mothurOutEndLine(); m->mothurOutEndLine();
296 int filteredLength = 0;
297 for(int i=0;i<alignmentLength;i++){
298 if(filter[i] == '1'){ filteredLength++; }
301 if (m->control_pressed) { outputTypes.clear(); for(int i = 0; i < outputNames.size(); i++) { m->mothurRemove(outputNames[i]); } return 0; }
304 m->mothurOutEndLine();
305 m->mothurOut("Length of filtered alignment: " + toString(filteredLength)); m->mothurOutEndLine();
306 m->mothurOut("Number of columns removed: " + toString((alignmentLength-filteredLength))); m->mothurOutEndLine();
307 m->mothurOut("Length of the original alignment: " + toString(alignmentLength)); m->mothurOutEndLine();
308 m->mothurOut("Number of sequences used to construct filter: " + toString(numSeqs)); m->mothurOutEndLine();
310 //set fasta file as new current fastafile
312 itTypes = outputTypes.find("fasta");
313 if (itTypes != outputTypes.end()) {
314 if ((itTypes->second).size() != 0) { current = (itTypes->second)[0]; m->setFastaFile(current); }
317 m->mothurOutEndLine();
318 m->mothurOut("Output File Names: "); m->mothurOutEndLine();
319 for(int i = 0; i < outputNames.size(); i++) { m->mothurOut(outputNames[i]); m->mothurOutEndLine(); }
320 m->mothurOutEndLine();
325 catch(exception& e) {
326 m->errorOut(e, "FilterSeqsCommand", "execute");
330 /**************************************************************************************/
331 int FilterSeqsCommand::filterSequences() {
336 for (int s = 0; s < fastafileNames.size(); s++) {
338 for (int i = 0; i < lines.size(); i++) { delete lines[i]; } lines.clear();
340 string filteredFasta = outputDir + m->getRootName(m->getSimpleName(fastafileNames[s])) + "filter.fasta";
342 int pid, numSeqsPerProcessor, num;
344 vector<unsigned long long>MPIPos;
347 MPI_Comm_size(MPI_COMM_WORLD, &processors); //set processors to the number of mpi processes running
348 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
352 int outMode=MPI_MODE_CREATE|MPI_MODE_WRONLY;
353 int inMode=MPI_MODE_RDONLY;
355 char outFilename[1024];
356 strcpy(outFilename, filteredFasta.c_str());
358 char inFileName[1024];
359 strcpy(inFileName, fastafileNames[s].c_str());
361 MPI_File_open(MPI_COMM_WORLD, inFileName, inMode, MPI_INFO_NULL, &inMPI); //comm, filename, mode, info, filepointer
362 MPI_File_open(MPI_COMM_WORLD, outFilename, outMode, MPI_INFO_NULL, &outMPI);
364 if (m->control_pressed) { MPI_File_close(&inMPI); MPI_File_close(&outMPI); return 0; }
366 if (pid == 0) { //you are the root process
368 MPIPos = m->setFilePosFasta(fastafileNames[s], num); //fills MPIPos, returns numSeqs
371 //send file positions to all processes
372 for(int i = 1; i < processors; i++) {
373 MPI_Send(&num, 1, MPI_INT, i, tag, MPI_COMM_WORLD);
374 MPI_Send(&MPIPos[0], (num+1), MPI_LONG, i, tag, MPI_COMM_WORLD);
377 //figure out how many sequences you have to do
378 numSeqsPerProcessor = num / processors;
379 int startIndex = pid * numSeqsPerProcessor;
380 if(pid == (processors - 1)){ numSeqsPerProcessor = num - pid * numSeqsPerProcessor; }
384 driverMPIRun(startIndex, numSeqsPerProcessor, inMPI, outMPI, MPIPos);
386 if (m->control_pressed) { MPI_File_close(&inMPI); MPI_File_close(&outMPI); return 0; }
389 for(int i = 1; i < processors; i++) {
391 MPI_Recv(buf, 5, MPI_CHAR, i, tag, MPI_COMM_WORLD, &status);
394 }else { //you are a child process
395 MPI_Recv(&num, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
396 MPIPos.resize(num+1);
398 MPI_Recv(&MPIPos[0], (num+1), MPI_LONG, 0, tag, MPI_COMM_WORLD, &status);
400 //figure out how many sequences you have to align
401 numSeqsPerProcessor = num / processors;
402 int startIndex = pid * numSeqsPerProcessor;
403 if(pid == (processors - 1)){ numSeqsPerProcessor = num - pid * numSeqsPerProcessor; }
407 driverMPIRun(startIndex, numSeqsPerProcessor, inMPI, outMPI, MPIPos);
409 if (m->control_pressed) { MPI_File_close(&inMPI); MPI_File_close(&outMPI); return 0; }
414 //tell parent you are done.
415 MPI_Send(buf, 5, MPI_CHAR, 0, tag, MPI_COMM_WORLD);
418 MPI_File_close(&outMPI);
419 MPI_File_close(&inMPI);
420 MPI_Barrier(MPI_COMM_WORLD); //make everyone wait - just in case
424 vector<unsigned long long> positions;
425 if (savedPositions.size() != 0) { positions = savedPositions[s]; }
427 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
428 positions = m->divideFile(fastafileNames[s], processors);
431 int numFastaSeqs = 0;
432 positions = m->setFilePosFasta(fastafileNames[s], numFastaSeqs);
436 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
437 //vector<unsigned long long> positions = m->divideFile(fastafileNames[s], processors);
439 for (int i = 0; i < (positions.size()-1); i++) {
440 lines.push_back(new linePair(positions[i], positions[(i+1)]));
444 int numFastaSeqs = driverRunFilter(filter, filteredFasta, fastafileNames[s], lines[0]);
445 numSeqs += numFastaSeqs;
447 int numFastaSeqs = createProcessesRunFilter(filter, fastafileNames[s], filteredFasta);
448 numSeqs += numFastaSeqs;
451 if (m->control_pressed) { return 1; }
454 lines.push_back(new linePair(0, 1000));
455 int numFastaSeqs = driverRunFilter(filter, filteredFasta, fastafileNames[s], lines[0]);
456 numSeqs += numFastaSeqs;
458 int numFastaSeqs = positions.size()-1;
459 //positions = m->setFilePosFasta(fastafileNames[s], numFastaSeqs);
461 //figure out how many sequences you have to process
462 int numSeqsPerProcessor = numFastaSeqs / processors;
463 for (int i = 0; i < processors; i++) {
464 int startIndex = i * numSeqsPerProcessor;
465 if(i == (processors - 1)){ numSeqsPerProcessor = numFastaSeqs - i * numSeqsPerProcessor; }
466 lines.push_back(new linePair(positions[startIndex], numSeqsPerProcessor));
469 numFastaSeqs = createProcessesRunFilter(filter, fastafileNames[s], filteredFasta);
470 numSeqs += numFastaSeqs;
473 if (m->control_pressed) { return 1; }
476 outputNames.push_back(filteredFasta); outputTypes["fasta"].push_back(filteredFasta);
481 catch(exception& e) {
482 m->errorOut(e, "FilterSeqsCommand", "filterSequences");
487 /**************************************************************************************/
488 int FilterSeqsCommand::driverMPIRun(int start, int num, MPI_File& inMPI, MPI_File& outMPI, vector<unsigned long long>& MPIPos) {
490 string outputString = "";
494 for(int i=0;i<num;i++){
496 if (m->control_pressed) { return 0; }
499 int length = MPIPos[start+i+1] - MPIPos[start+i];
500 char* buf4 = new char[length];
501 MPI_File_read_at(inMPI, MPIPos[start+i], buf4, length, MPI_CHAR, &status);
503 string tempBuf = buf4;
504 if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); }
505 istringstream iss (tempBuf,istringstream::in);
508 Sequence seq(iss); m->gobble(iss);
510 if (seq.getName() != "") {
511 string align = seq.getAligned();
512 string filterSeq = "";
514 for(int j=0;j<alignmentLength;j++){
515 if(filter[j] == '1'){
516 filterSeq += align[j];
521 outputString += ">" + seq.getName() + "\n" + filterSeq + "\n";
523 if(count % 10 == 0){ //output to file
524 //send results to parent
525 int length = outputString.length();
526 char* buf = new char[length];
527 memcpy(buf, outputString.c_str(), length);
529 MPI_File_write_shared(outMPI, buf, length, MPI_CHAR, &status);
536 if((i+1) % 100 == 0){ cout << (i+1) << endl; m->mothurOutJustToLog(toString(i+1) + "\n"); }
539 if(outputString != ""){ //output to file
540 //send results to parent
541 int length = outputString.length();
542 char* buf = new char[length];
543 memcpy(buf, outputString.c_str(), length);
545 MPI_File_write_shared(outMPI, buf, length, MPI_CHAR, &status);
550 if((num) % 100 != 0){ cout << (num) << endl; m->mothurOutJustToLog(toString(num) + "\n"); }
554 catch(exception& e) {
555 m->errorOut(e, "FilterSeqsCommand", "driverRunFilter");
560 /**************************************************************************************/
561 int FilterSeqsCommand::driverRunFilter(string F, string outputFilename, string inputFilename, linePair* filePos) {
564 m->openOutputFile(outputFilename, out);
567 m->openInputFile(inputFilename, in);
569 in.seekg(filePos->start);
576 if (m->control_pressed) { in.close(); out.close(); return 0; }
578 Sequence seq(in); m->gobble(in);
579 if (seq.getName() != "") {
580 string align = seq.getAligned();
581 string filterSeq = "";
583 for(int j=0;j<alignmentLength;j++){
584 if(filter[j] == '1'){
585 filterSeq += align[j];
589 out << '>' << seq.getName() << endl << filterSeq << endl;
593 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
594 unsigned long long pos = in.tellg();
595 if ((pos == -1) || (pos >= filePos->end)) { break; }
597 if (in.eof()) { break; }
601 if((count) % 100 == 0){ m->mothurOut(toString(count)); m->mothurOutEndLine(); }
604 if((count) % 100 != 0){ m->mothurOut(toString(count)); m->mothurOutEndLine(); }
612 catch(exception& e) {
613 m->errorOut(e, "FilterSeqsCommand", "driverRunFilter");
617 /**************************************************************************************************/
619 int FilterSeqsCommand::createProcessesRunFilter(string F, string filename, string filteredFastaName) {
626 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
629 //loop through and create all the processes you want
630 while (process != processors) {
634 processIDS.push_back(pid); //create map from line number to pid so you can append files in correct order later
637 string filteredFasta = filename + toString(getpid()) + ".temp";
638 num = driverRunFilter(F, filteredFasta, filename, lines[process]);
640 //pass numSeqs to parent
642 string tempFile = filename + toString(getpid()) + ".num.temp";
643 m->openOutputFile(tempFile, out);
649 m->mothurOut("[ERROR]: unable to spawn the necessary processes."); m->mothurOutEndLine();
650 for (int i = 0; i < processIDS.size(); i++) { kill (processIDS[i], SIGINT); }
655 num = driverRunFilter(F, filteredFastaName, filename, lines[0]);
657 //force parent to wait until all the processes are done
658 for (int i=0;i<processIDS.size();i++) {
659 int temp = processIDS[i];
663 for (int i = 0; i < processIDS.size(); i++) {
665 string tempFile = filename + toString(processIDS[i]) + ".num.temp";
666 m->openInputFile(tempFile, in);
667 if (!in.eof()) { int tempNum = 0; in >> tempNum; num += tempNum; }
668 in.close(); m->mothurRemove(tempFile);
670 m->appendFiles((filename + toString(processIDS[i]) + ".temp"), filteredFastaName);
671 m->mothurRemove((filename + toString(processIDS[i]) + ".temp"));
676 //////////////////////////////////////////////////////////////////////////////////////////////////////
677 //Windows version shared memory, so be careful when passing variables through the filterData struct.
678 //Above fork() will clone, so memory is separate, but that's not the case with windows,
679 //Taking advantage of shared memory to allow both threads to add info to F.
680 //////////////////////////////////////////////////////////////////////////////////////////////////////
682 vector<filterRunData*> pDataArray;
683 DWORD dwThreadIdArray[processors-1];
684 HANDLE hThreadArray[processors-1];
686 //Create processor worker threads.
687 for( int i=0; i<processors-1; i++){
689 string extension = "";
690 if (i != 0) { extension = toString(i) + ".temp"; }
692 filterRunData* tempFilter = new filterRunData(filter, filename, (filteredFastaName + extension), m, lines[i]->start, lines[i]->end, alignmentLength, i);
693 pDataArray.push_back(tempFilter);
694 processIDS.push_back(i);
696 hThreadArray[i] = CreateThread(NULL, 0, MyRunFilterThreadFunction, pDataArray[i], 0, &dwThreadIdArray[i]);
699 num = driverRunFilter(F, (filteredFastaName + toString(processors-1) + ".temp"), filename, lines[processors-1]);
701 //Wait until all threads have terminated.
702 WaitForMultipleObjects(processors-1, hThreadArray, TRUE, INFINITE);
704 //Close all thread handles and free memory allocations.
705 for(int i=0; i < pDataArray.size(); i++){
706 num += pDataArray[i]->count;
707 CloseHandle(hThreadArray[i]);
708 delete pDataArray[i];
711 for (int i = 1; i < processors; i++) {
712 m->appendFiles((filteredFastaName + toString(i) + ".temp"), filteredFastaName);
713 m->mothurRemove((filteredFastaName + toString(i) + ".temp"));
720 catch(exception& e) {
721 m->errorOut(e, "FilterSeqsCommand", "createProcessesRunFilter");
725 /**************************************************************************************/
726 string FilterSeqsCommand::createFilter() {
728 string filterString = "";
731 if (soft != 0) { F.setSoft(soft); }
732 if (trump != '*') { F.setTrump(trump); }
734 F.setLength(alignmentLength);
736 if(trump != '*' || m->isTrue(vertical) || soft != 0){
740 if(hard.compare("") != 0) { F.doHard(hard); }
741 else { F.setFilter(string(alignmentLength, '1')); }
744 if(trump != '*' || m->isTrue(vertical) || soft != 0){
745 for (int s = 0; s < fastafileNames.size(); s++) {
747 for (int i = 0; i < lines.size(); i++) { delete lines[i]; } lines.clear();
750 int pid, numSeqsPerProcessor, num;
752 vector<unsigned long long> MPIPos;
756 MPI_Comm_size(MPI_COMM_WORLD, &processors);
757 MPI_Comm_rank(MPI_COMM_WORLD, &pid);
759 //char* tempFileName = new char(fastafileNames[s].length());
760 //tempFileName = &(fastafileNames[s][0]);
762 char tempFileName[1024];
763 strcpy(tempFileName, fastafileNames[s].c_str());
765 MPI_File_open(MPI_COMM_WORLD, tempFileName, MPI_MODE_RDONLY, MPI_INFO_NULL, &inMPI); //comm, filename, mode, info, filepointer
767 if (m->control_pressed) { MPI_File_close(&inMPI); return 0; }
769 if (pid == 0) { //you are the root process
770 MPIPos = m->setFilePosFasta(fastafileNames[s], num); //fills MPIPos, returns numSeqs
773 //send file positions to all processes
774 for(int i = 1; i < processors; i++) {
775 MPI_Send(&num, 1, MPI_INT, i, tag, MPI_COMM_WORLD);
776 MPI_Send(&MPIPos[0], (num+1), MPI_LONG, i, tag, MPI_COMM_WORLD);
779 //figure out how many sequences you have to do
780 numSeqsPerProcessor = num / processors;
781 int startIndex = pid * numSeqsPerProcessor;
782 if(pid == (processors - 1)){ numSeqsPerProcessor = num - pid * numSeqsPerProcessor; }
786 MPICreateFilter(startIndex, numSeqsPerProcessor, F, inMPI, MPIPos);
788 if (m->control_pressed) { MPI_File_close(&inMPI); return 0; }
790 }else { //i am the child process
791 MPI_Recv(&num, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
792 MPIPos.resize(num+1);
794 MPI_Recv(&MPIPos[0], (num+1), MPI_LONG, 0, tag, MPI_COMM_WORLD, &status);
796 //figure out how many sequences you have to align
797 numSeqsPerProcessor = num / processors;
798 int startIndex = pid * numSeqsPerProcessor;
799 if(pid == (processors - 1)){ numSeqsPerProcessor = num - pid * numSeqsPerProcessor; }
803 MPICreateFilter(startIndex, numSeqsPerProcessor, F, inMPI, MPIPos);
805 if (m->control_pressed) { MPI_File_close(&inMPI); return 0; }
808 MPI_File_close(&inMPI);
809 MPI_Barrier(MPI_COMM_WORLD); //make everyone wait - just in case
813 vector<unsigned long long> positions;
814 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
815 positions = m->divideFile(fastafileNames[s], processors);
816 for (int i = 0; i < (positions.size()-1); i++) {
817 lines.push_back(new linePair(positions[i], positions[(i+1)]));
821 int numFastaSeqs = driverCreateFilter(F, fastafileNames[s], lines[0]);
822 numSeqs += numFastaSeqs;
824 int numFastaSeqs = createProcessesCreateFilter(F, fastafileNames[s]);
825 numSeqs += numFastaSeqs;
829 lines.push_back(new linePair(0, 1000));
830 int numFastaSeqs = driverCreateFilter(F, fastafileNames[s], lines[0]);
831 numSeqs += numFastaSeqs;
833 int numFastaSeqs = 0;
834 positions = m->setFilePosFasta(fastafileNames[s], numFastaSeqs);
836 //figure out how many sequences you have to process
837 int numSeqsPerProcessor = numFastaSeqs / processors;
838 for (int i = 0; i < processors; i++) {
839 int startIndex = i * numSeqsPerProcessor;
840 if(i == (processors - 1)){ numSeqsPerProcessor = numFastaSeqs - i * numSeqsPerProcessor; }
841 lines.push_back(new linePair(positions[startIndex], numSeqsPerProcessor));
844 numFastaSeqs = createProcessesCreateFilter(F, fastafileNames[s]);
845 numSeqs += numFastaSeqs;
848 //save the file positions so we can reuse them in the runFilter function
849 savedPositions[s] = positions;
851 if (m->control_pressed) { return filterString; }
860 int Atag = 1; int Ttag = 2; int Ctag = 3; int Gtag = 4; int Gaptag = 5;
863 MPI_Comm_rank(MPI_COMM_WORLD, &pid);
865 if(trump != '*' || m->isTrue(vertical) || soft != 0){
867 if (pid == 0) { //only one process should output the filter
869 vector<int> temp; temp.resize(alignmentLength+1);
871 //get the frequencies from the child processes
872 for(int i = 1; i < processors; i++) {
874 for (int j = 0; j < 5; j++) {
876 MPI_Recv(&temp[0], (alignmentLength+1), MPI_INT, i, 2001, MPI_COMM_WORLD, &status);
877 int receiveTag = temp[temp.size()-1]; //child process added a int to the end to indicate what letter count this is for
879 if (receiveTag == Atag) { //you are recieveing the A frequencies
880 for (int k = 0; k < alignmentLength; k++) { F.a[k] += temp[k]; }
881 }else if (receiveTag == Ttag) { //you are recieveing the T frequencies
882 for (int k = 0; k < alignmentLength; k++) { F.t[k] += temp[k]; }
883 }else if (receiveTag == Ctag) { //you are recieveing the C frequencies
884 for (int k = 0; k < alignmentLength; k++) { F.c[k] += temp[k]; }
885 }else if (receiveTag == Gtag) { //you are recieveing the G frequencies
886 for (int k = 0; k < alignmentLength; k++) { F.g[k] += temp[k]; }
887 }else if (receiveTag == Gaptag) { //you are recieveing the gap frequencies
888 for (int k = 0; k < alignmentLength; k++) { F.gap[k] += temp[k]; }
894 //send my fequency counts
896 int ierr = MPI_Send(&(F.a[0]), (alignmentLength+1), MPI_INT, 0, 2001, MPI_COMM_WORLD);
898 ierr = MPI_Send (&(F.t[0]), (alignmentLength+1), MPI_INT, 0, 2001, MPI_COMM_WORLD);
900 ierr = MPI_Send(&(F.c[0]), (alignmentLength+1), MPI_INT, 0, 2001, MPI_COMM_WORLD);
902 ierr = MPI_Send(&(F.g[0]), (alignmentLength+1), MPI_INT, 0, 2001, MPI_COMM_WORLD);
903 F.gap.push_back(Gaptag);
904 ierr = MPI_Send(&(F.gap[0]), (alignmentLength+1), MPI_INT, 0, 2001, MPI_COMM_WORLD);
909 MPI_Barrier(MPI_COMM_WORLD); //make everyone wait - just in case
911 if (pid == 0) { //only one process should output the filter
914 F.setNumSeqs(numSeqs);
915 if(m->isTrue(vertical) == 1) { F.doVertical(); }
916 if(soft != 0) { F.doSoft(); }
917 filterString = F.getFilter();
920 //send filter string to kids
921 //for(int i = 1; i < processors; i++) {
922 // MPI_Send(&filterString[0], alignmentLength, MPI_CHAR, i, 2001, MPI_COMM_WORLD);
924 MPI_Bcast(&filterString[0], alignmentLength, MPI_CHAR, 0, MPI_COMM_WORLD);
926 //recieve filterString
927 char* tempBuf = new char[alignmentLength];
928 //MPI_Recv(&tempBuf[0], alignmentLength, MPI_CHAR, 0, 2001, MPI_COMM_WORLD, &status);
929 MPI_Bcast(tempBuf, alignmentLength, MPI_CHAR, 0, MPI_COMM_WORLD);
931 filterString = tempBuf;
932 if (filterString.length() > alignmentLength) { filterString = filterString.substr(0, alignmentLength); }
936 MPI_Barrier(MPI_COMM_WORLD);
941 catch(exception& e) {
942 m->errorOut(e, "FilterSeqsCommand", "createFilter");
946 /**************************************************************************************/
947 int FilterSeqsCommand::driverCreateFilter(Filters& F, string filename, linePair* filePos) {
951 m->openInputFile(filename, in);
953 in.seekg(filePos->start);
960 if (m->control_pressed) { in.close(); return 1; }
962 Sequence seq(in); m->gobble(in);
963 if (seq.getName() != "") {
964 if (seq.getAligned().length() != alignmentLength) { m->mothurOut("Sequences are not all the same length, please correct."); m->mothurOutEndLine(); m->control_pressed = true; }
966 if(trump != '*') { F.doTrump(seq); }
967 if(m->isTrue(vertical) || soft != 0) { F.getFreqs(seq); }
972 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
973 unsigned long long pos = in.tellg();
974 if ((pos == -1) || (pos >= filePos->end)) { break; }
976 if (in.eof()) { break; }
980 if((count) % 100 == 0){ m->mothurOut(toString(count)); m->mothurOutEndLine(); }
983 if((count) % 100 != 0){ m->mothurOut(toString(count)); m->mothurOutEndLine(); }
988 catch(exception& e) {
989 m->errorOut(e, "FilterSeqsCommand", "driverCreateFilter");
994 /**************************************************************************************/
995 int FilterSeqsCommand::MPICreateFilter(int start, int num, Filters& F, MPI_File& inMPI, vector<unsigned long long>& MPIPos) {
1000 MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
1002 for(int i=0;i<num;i++){
1004 if (m->control_pressed) { return 0; }
1006 //read next sequence
1007 int length = MPIPos[start+i+1] - MPIPos[start+i];
1009 char* buf4 = new char[length];
1010 MPI_File_read_at(inMPI, MPIPos[start+i], buf4, length, MPI_CHAR, &status);
1012 string tempBuf = buf4;
1013 if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); }
1014 istringstream iss (tempBuf,istringstream::in);
1019 if (seq.getAligned().length() != alignmentLength) { cout << "Alignment length is " << alignmentLength << " and sequence " << seq.getName() << " has length " << seq.getAligned().length() << ", please correct." << endl; exit(1); }
1021 if(trump != '*'){ F.doTrump(seq); }
1022 if(m->isTrue(vertical) || soft != 0){ F.getFreqs(seq); }
1026 if((i+1) % 100 == 0){ cout << (i+1) << endl; m->mothurOutJustToLog(toString(i+1) + "\n"); }
1030 if((num) % 100 != 0){ cout << num << endl; m->mothurOutJustToLog(toString(num) + "\n"); }
1034 catch(exception& e) {
1035 m->errorOut(e, "FilterSeqsCommand", "MPICreateFilter");
1040 /**************************************************************************************************/
1042 int FilterSeqsCommand::createProcessesCreateFilter(Filters& F, string filename) {
1048 #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
1050 //loop through and create all the processes you want
1051 while (process != processors) {
1055 processIDS.push_back(pid); //create map from line number to pid so you can append files in correct order later
1057 }else if (pid == 0){
1058 //reset child's filter counts to 0;
1059 F.a.clear(); F.a.resize(alignmentLength, 0);
1060 F.t.clear(); F.t.resize(alignmentLength, 0);
1061 F.g.clear(); F.g.resize(alignmentLength, 0);
1062 F.c.clear(); F.c.resize(alignmentLength, 0);
1063 F.gap.clear(); F.gap.resize(alignmentLength, 0);
1065 num = driverCreateFilter(F, filename, lines[process]);
1067 //write out filter counts to file
1068 filename += toString(getpid()) + "filterValues.temp";
1070 m->openOutputFile(filename, out);
1073 out << F.getFilter() << endl;
1074 for (int k = 0; k < alignmentLength; k++) { out << F.a[k] << '\t'; } out << endl;
1075 for (int k = 0; k < alignmentLength; k++) { out << F.t[k] << '\t'; } out << endl;
1076 for (int k = 0; k < alignmentLength; k++) { out << F.g[k] << '\t'; } out << endl;
1077 for (int k = 0; k < alignmentLength; k++) { out << F.c[k] << '\t'; } out << endl;
1078 for (int k = 0; k < alignmentLength; k++) { out << F.gap[k] << '\t'; } out << endl;
1080 //cout << F.getFilter() << endl;
1085 m->mothurOut("[ERROR]: unable to spawn the necessary processes."); m->mothurOutEndLine();
1086 for (int i = 0; i < processIDS.size(); i++) { kill (processIDS[i], SIGINT); }
1091 //parent do your part
1092 num = driverCreateFilter(F, filename, lines[0]);
1094 //force parent to wait until all the processes are done
1095 for (int i=0;i<(processors-1);i++) {
1096 int temp = processIDS[i];
1100 //parent reads in and combines Filter info
1101 for (int i = 0; i < processIDS.size(); i++) {
1102 string tempFilename = filename + toString(processIDS[i]) + "filterValues.temp";
1104 m->openInputFile(tempFilename, in);
1107 string tempFilterString;
1109 in >> tempNum; m->gobble(in); num += tempNum;
1111 in >> tempFilterString;
1112 F.mergeFilter(tempFilterString);
1114 for (int k = 0; k < alignmentLength; k++) { in >> temp; F.a[k] += temp; } m->gobble(in);
1115 for (int k = 0; k < alignmentLength; k++) { in >> temp; F.t[k] += temp; } m->gobble(in);
1116 for (int k = 0; k < alignmentLength; k++) { in >> temp; F.g[k] += temp; } m->gobble(in);
1117 for (int k = 0; k < alignmentLength; k++) { in >> temp; F.c[k] += temp; } m->gobble(in);
1118 for (int k = 0; k < alignmentLength; k++) { in >> temp; F.gap[k] += temp; } m->gobble(in);
1121 m->mothurRemove(tempFilename);
1127 //////////////////////////////////////////////////////////////////////////////////////////////////////
1128 //Windows version shared memory, so be careful when passing variables through the filterData struct.
1129 //Above fork() will clone, so memory is separate, but that's not the case with windows,
1130 //Taking advantage of shared memory to allow both threads to add info to F.
1131 //////////////////////////////////////////////////////////////////////////////////////////////////////
1133 vector<filterData*> pDataArray;
1134 DWORD dwThreadIdArray[processors];
1135 HANDLE hThreadArray[processors];
1137 //Create processor worker threads.
1138 for( int i=0; i<processors; i++ ){
1140 filterData* tempFilter = new filterData(filename, m, lines[i]->start, lines[i]->end, alignmentLength, trump, vertical, soft, hard, i);
1141 pDataArray.push_back(tempFilter);
1142 processIDS.push_back(i);
1144 hThreadArray[i] = CreateThread(NULL, 0, MyCreateFilterThreadFunction, pDataArray[i], 0, &dwThreadIdArray[i]);
1147 //Wait until all threads have terminated.
1148 WaitForMultipleObjects(processors, hThreadArray, TRUE, INFINITE);
1150 //Close all thread handles and free memory allocations.
1151 for(int i=0; i < pDataArray.size(); i++){
1152 num += pDataArray[i]->count;
1153 F.mergeFilter(pDataArray[i]->F.getFilter());
1155 for (int k = 0; k < alignmentLength; k++) { F.a[k] += pDataArray[i]->F.a[k]; }
1156 for (int k = 0; k < alignmentLength; k++) { F.t[k] += pDataArray[i]->F.t[k]; }
1157 for (int k = 0; k < alignmentLength; k++) { F.g[k] += pDataArray[i]->F.g[k]; }
1158 for (int k = 0; k < alignmentLength; k++) { F.c[k] += pDataArray[i]->F.c[k]; }
1159 for (int k = 0; k < alignmentLength; k++) { F.gap[k] += pDataArray[i]->F.gap[k]; }
1161 CloseHandle(hThreadArray[i]);
1162 delete pDataArray[i];
1169 catch(exception& e) {
1170 m->errorOut(e, "FilterSeqsCommand", "createProcessesCreateFilter");
1174 /**************************************************************************************/