X-Git-Url: https://git.donarmstrong.com/?a=blobdiff_plain;f=aligncommand.cpp;h=f03a9017d4867f1d8634a0ef72a25f1590bf13b6;hb=0a1f333a778804cb8d7e019eeb40192d96023502;hp=a32ec4df279b2fa1158ec5e818ce03545de4ed0c;hpb=19fcbbdba99658f5eca244803280f9ee7f9f6607;p=mothur.git diff --git a/aligncommand.cpp b/aligncommand.cpp index a32ec4d..f03a901 100644 --- a/aligncommand.cpp +++ b/aligncommand.cpp @@ -15,15 +15,6 @@ */ #include "aligncommand.h" -#include "sequence.hpp" - -#include "gotohoverlap.hpp" -#include "needlemanoverlap.hpp" -#include "blastalign.hpp" -#include "noalign.hpp" - -#include "nast.hpp" -#include "nastreport.hpp" #include "referencedb.h" //********************************************************************************************************************** @@ -233,23 +224,23 @@ AlignCommand::AlignCommand(string option) { // ...at some point should added some additional type checking... string temp; temp = validParameter.validFile(parameters, "ksize", false); if (temp == "not found"){ temp = "8"; } - convert(temp, kmerSize); + m->mothurConvert(temp, kmerSize); temp = validParameter.validFile(parameters, "match", false); if (temp == "not found"){ temp = "1.0"; } - convert(temp, match); + m->mothurConvert(temp, match); temp = validParameter.validFile(parameters, "mismatch", false); if (temp == "not found"){ temp = "-1.0"; } - convert(temp, misMatch); + m->mothurConvert(temp, misMatch); temp = validParameter.validFile(parameters, "gapopen", false); if (temp == "not found"){ temp = "-2.0"; } - convert(temp, gapOpen); + m->mothurConvert(temp, gapOpen); temp = validParameter.validFile(parameters, "gapextend", false); if (temp == "not found"){ temp = "-1.0"; } - convert(temp, gapExtend); + m->mothurConvert(temp, gapExtend); temp = validParameter.validFile(parameters, "processors", false); if (temp == "not found"){ temp = m->getProcessors(); } m->setProcessors(temp); - convert(temp, processors); + m->mothurConvert(temp, processors); temp = validParameter.validFile(parameters, "flip", false); if (temp == "not found"){ temp = "f"; } flip = m->isTrue(temp); @@ -276,7 +267,7 @@ AlignCommand::AlignCommand(string option) { else { if (save) { rdb->setSavedReference(templateFileName); } } temp = validParameter.validFile(parameters, "threshold", false); if (temp == "not found"){ temp = "0.50"; } - convert(temp, threshold); + m->mothurConvert(temp, threshold); search = validParameter.validFile(parameters, "search", false); if (search == "not found"){ search = "kmer"; } if ((search != "suffix") && (search != "kmer") && (search != "blast")) { m->mothurOut("invalid search option: choices are kmer, suffix or blast."); m->mothurOutEndLine(); abort=true; } @@ -298,7 +289,6 @@ AlignCommand::~AlignCommand(){ if (abort == false) { for (int i = 0; i < lines.size(); i++) { delete lines[i]; } lines.clear(); delete templateDB; - delete alignment; } } //********************************************************************************************************************** @@ -307,18 +297,7 @@ int AlignCommand::execute(){ try { if (abort == true) { if (calledHelp) { return 0; } return 2; } - templateDB = new AlignmentDB(templateFileName, search, kmerSize, gapOpen, gapExtend, match, misMatch); - int longestBase = templateDB->getLongestBase(); - - if(align == "gotoh") { alignment = new GotohOverlap(gapOpen, gapExtend, match, misMatch, longestBase); } - else if(align == "needleman") { alignment = new NeedlemanOverlap(gapOpen, match, misMatch, longestBase); } - else if(align == "blast") { alignment = new BlastAlignment(gapOpen, gapExtend, match, misMatch); } - else if(align == "noalign") { alignment = new NoAlign(); } - else { - m->mothurOut(align + " is not a valid alignment option. I will run the command using needleman."); - m->mothurOutEndLine(); - alignment = new NeedlemanOverlap(gapOpen, match, misMatch, longestBase); - } + templateDB = new AlignmentDB(templateFileName, search, kmerSize, gapOpen, gapExtend, match, misMatch, rand()); for (int s = 0; s < candidateFileNames.size(); s++) { if (m->control_pressed) { outputTypes.clear(); return 0; } @@ -338,7 +317,7 @@ int AlignCommand::execute(){ #ifdef USE_MPI int pid, numSeqsPerProcessor; int tag = 2001; - vector MPIPos; + vector MPIPos; MPIWroteAccnos = false; MPI_Status status; @@ -442,20 +421,32 @@ int AlignCommand::execute(){ #else - vector positions = m->divideFile(candidateFileNames[s], processors); - for (int i = 0; i < (positions.size()-1); i++) { - lines.push_back(new linePair(positions[i], positions[(i+1)])); - } - #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) + vector positions; + #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) + positions = m->divideFile(candidateFileNames[s], processors); + for (int i = 0; i < (positions.size()-1); i++) { lines.push_back(new linePair(positions[i], positions[(i+1)])); } + #else + if (processors == 1) { + lines.push_back(new linePair(0, 1000)); + }else { + positions = m->setFilePosFasta(candidateFileNames[s], numFastaSeqs); + + //figure out how many sequences you have to process + int numSeqsPerProcessor = numFastaSeqs / processors; + for (int i = 0; i < processors; i++) { + int startIndex = i * numSeqsPerProcessor; + if(i == (processors - 1)){ numSeqsPerProcessor = numFastaSeqs - i * numSeqsPerProcessor; } + lines.push_back(new linePair(positions[startIndex], numSeqsPerProcessor)); + } + } + #endif + if(processors == 1){ numFastaSeqs = driver(lines[0], alignFileName, reportFileName, accnosFileName, candidateFileNames[s]); }else{ numFastaSeqs = createProcesses(alignFileName, reportFileName, accnosFileName, candidateFileNames[s]); } - #else - numFastaSeqs = driver(lines[0], alignFileName, reportFileName, accnosFileName, candidateFileNames[s]); - #endif - + if (m->control_pressed) { m->mothurRemove(accnosFileName); m->mothurRemove(alignFileName); m->mothurRemove(reportFileName); outputTypes.clear(); return 0; } //delete accnos file if its blank else report to user @@ -511,7 +502,6 @@ int AlignCommand::execute(){ } //********************************************************************************************************************** - int AlignCommand::driver(linePair* filePos, string alignFName, string reportFName, string accnosFName, string filename){ try { ofstream alignmentFile; @@ -529,10 +519,23 @@ int AlignCommand::driver(linePair* filePos, string alignFName, string reportFNam bool done = false; int count = 0; + + //moved this into driver to avoid deep copies in windows paralellized version + Alignment* alignment; + int longestBase = templateDB->getLongestBase(); + if(align == "gotoh") { alignment = new GotohOverlap(gapOpen, gapExtend, match, misMatch, longestBase); } + else if(align == "needleman") { alignment = new NeedlemanOverlap(gapOpen, match, misMatch, longestBase); } + else if(align == "blast") { alignment = new BlastAlignment(gapOpen, gapExtend, match, misMatch); } + else if(align == "noalign") { alignment = new NoAlign(); } + else { + m->mothurOut(align + " is not a valid alignment option. I will run the command using needleman."); + m->mothurOutEndLine(); + alignment = new NeedlemanOverlap(gapOpen, match, misMatch, longestBase); + } while (!done) { - if (m->control_pressed) { return 0; } + if (m->control_pressed) { break; } Sequence* candidateSeq = new Sequence(inFASTA); m->gobble(inFASTA); report.setCandidate(candidateSeq); @@ -548,7 +551,7 @@ int AlignCommand::driver(linePair* filePos, string alignFName, string reportFNam Sequence temp = templateDB->findClosestSequence(candidateSeq); Sequence* templateSeq = &temp; - + float searchScore = templateDB->getSearchScore(); Nast* nast = new Nast(alignment, candidateSeq, templateSeq); @@ -615,7 +618,7 @@ int AlignCommand::driver(linePair* filePos, string alignFName, string reportFNam delete candidateSeq; #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) - unsigned long int pos = inFASTA.tellg(); + unsigned long long pos = inFASTA.tellg(); if ((pos == -1) || (pos >= filePos->end)) { break; } #else if (inFASTA.eof()) { break; } @@ -628,6 +631,7 @@ int AlignCommand::driver(linePair* filePos, string alignFName, string reportFNam //report progress if((count) % 100 != 0){ m->mothurOut(toString(count)); m->mothurOutEndLine(); } + delete alignment; alignmentFile.close(); inFASTA.close(); accnosFile.close(); @@ -641,7 +645,7 @@ int AlignCommand::driver(linePair* filePos, string alignFName, string reportFNam } //********************************************************************************************************************** #ifdef USE_MPI -int AlignCommand::driverMPI(int start, int num, MPI_File& inMPI, MPI_File& alignFile, MPI_File& reportFile, MPI_File& accnosFile, vector& MPIPos){ +int AlignCommand::driverMPI(int start, int num, MPI_File& inMPI, MPI_File& alignFile, MPI_File& reportFile, MPI_File& accnosFile, vector& MPIPos){ try { string outputString = ""; MPI_Status statusReport; @@ -665,9 +669,22 @@ int AlignCommand::driverMPI(int start, int num, MPI_File& inMPI, MPI_File& align delete buf; } + Alignment* alignment; + int longestBase = templateDB->getLongestBase(); + if(align == "gotoh") { alignment = new GotohOverlap(gapOpen, gapExtend, match, misMatch, longestBase); } + else if(align == "needleman") { alignment = new NeedlemanOverlap(gapOpen, match, misMatch, longestBase); } + else if(align == "blast") { alignment = new BlastAlignment(gapOpen, gapExtend, match, misMatch); } + else if(align == "noalign") { alignment = new NoAlign(); } + else { + m->mothurOut(align + " is not a valid alignment option. I will run the command using needleman."); + m->mothurOutEndLine(); + alignment = new NeedlemanOverlap(gapOpen, match, misMatch, longestBase); + } + + for(int i=0;icontrol_pressed) { return 0; } + if (m->control_pressed) { delete alignment; return 0; } //read next sequence int length = MPIPos[start+i+1] - MPIPos[start+i]; @@ -807,11 +824,10 @@ int AlignCommand::driverMPI(int start, int num, MPI_File& inMPI, MPI_File& align int AlignCommand::createProcesses(string alignFileName, string reportFileName, string accnosFName, string filename) { try { -#if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) + int num = 0; processIDS.resize(0); +#if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) int process = 1; - int num = 0; - // processIDS.resize(0); //loop through and create all the processes you want while (process != processors) { @@ -883,9 +899,90 @@ int AlignCommand::createProcesses(string alignFileName, string reportFileName, s m->openOutputFile(accnosFName, out); out.close(); } +#else + ////////////////////////////////////////////////////////////////////////////////////////////////////// + //Windows version shared memory, so be careful when passing variables through the alignData struct. + //Above fork() will clone, so memory is separate, but that's not the case with windows, + ////////////////////////////////////////////////////////////////////////////////////////////////////// + + vector pDataArray; + DWORD dwThreadIdArray[processors-1]; + HANDLE hThreadArray[processors-1]; + + //Create processor worker threads. + for( int i=0; istart, lines[i]->end, flip, match, misMatch, gapOpen, gapExtend, threshold, i); + pDataArray.push_back(tempalign); + processIDS.push_back(i); + + //MySeqSumThreadFunction is in header. It must be global or static to work with the threads. + //default security attributes, thread function name, argument to thread function, use default creation flags, returns the thread identifier + hThreadArray[i] = CreateThread(NULL, 0, MyAlignThreadFunction, pDataArray[i], 0, &dwThreadIdArray[i]); + } + + //need to check for line ending error + ifstream inFASTA; + m->openInputFile(filename, inFASTA); + inFASTA.seekg(lines[processors-1]->start-1); + char c = inFASTA.peek(); + + if (c != '>') { //we need to move back + lines[processors-1]->start--; + } + + //using the main process as a worker saves time and memory + //do my part - do last piece because windows is looking for eof + num = driver(lines[processors-1], (alignFileName + toString(processors-1) + ".temp"), (reportFileName + toString(processors-1) + ".temp"), (accnosFName + toString(processors-1) + ".temp"), filename); + + //Wait until all threads have terminated. + WaitForMultipleObjects(processors-1, hThreadArray, TRUE, INFINITE); + + //Close all thread handles and free memory allocations. + for(int i=0; i < pDataArray.size(); i++){ + num += pDataArray[i]->count; + CloseHandle(hThreadArray[i]); + delete pDataArray[i]; + } + + vector nonBlankAccnosFiles; + if (!(m->isBlank(accnosFName))) { nonBlankAccnosFiles.push_back(accnosFName); } + else { m->mothurRemove(accnosFName); } //remove so other files can be renamed to it + + for (int i = 1; i < processors; i++) { + appendAlignFiles((alignFileName + toString(i) + ".temp"), alignFileName); + m->mothurRemove((alignFileName + toString(i) + ".temp")); + + appendReportFiles((reportFileName + toString(i) + ".temp"), reportFileName); + m->mothurRemove((reportFileName + toString(i) + ".temp")); + + if (!(m->isBlank(accnosFName + toString(i) + ".temp"))) { + nonBlankAccnosFiles.push_back(accnosFName + toString(i) + ".temp"); + }else { m->mothurRemove((accnosFName + toString(i) + ".temp")); } + } + + //append accnos files + if (nonBlankAccnosFiles.size() != 0) { + rename(nonBlankAccnosFiles[0].c_str(), accnosFName.c_str()); + + for (int h=1; h < nonBlankAccnosFiles.size(); h++) { + appendAlignFiles(nonBlankAccnosFiles[h], accnosFName); + m->mothurRemove(nonBlankAccnosFiles[h]); + } + }else { //recreate the accnosfile if needed + ofstream out; + m->openOutputFile(accnosFName, out); + out.close(); + } +#endif return num; -#endif } catch(exception& e) { m->errorOut(e, "AlignCommand", "createProcesses");