+ numSeqs = 0;
+
+ for (int s = 0; s < fastafileNames.size(); s++) {
+
+ for (int i = 0; i < lines.size(); i++) { delete lines[i]; } lines.clear();
+
+ string filteredFasta = outputDir + m->getRootName(m->getSimpleName(fastafileNames[s])) + "filter.fasta";
+#ifdef USE_MPI
+ int pid, numSeqsPerProcessor, num;
+ int tag = 2001;
+ vector<unsigned long int>MPIPos;
+
+ MPI_Status status;
+ MPI_Comm_size(MPI_COMM_WORLD, &processors); //set processors to the number of mpi processes running
+ MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
+
+ MPI_File outMPI;
+ MPI_File inMPI;
+ int outMode=MPI_MODE_CREATE|MPI_MODE_WRONLY;
+ int inMode=MPI_MODE_RDONLY;
+
+ char outFilename[1024];
+ strcpy(outFilename, filteredFasta.c_str());
+
+ char inFileName[1024];
+ strcpy(inFileName, fastafileNames[s].c_str());
+
+ MPI_File_open(MPI_COMM_WORLD, inFileName, inMode, MPI_INFO_NULL, &inMPI); //comm, filename, mode, info, filepointer
+ MPI_File_open(MPI_COMM_WORLD, outFilename, outMode, MPI_INFO_NULL, &outMPI);
+
+ if (m->control_pressed) { MPI_File_close(&inMPI); MPI_File_close(&outMPI); return 0; }
+
+ if (pid == 0) { //you are the root process
+
+ MPIPos = m->setFilePosFasta(fastafileNames[s], num); //fills MPIPos, returns numSeqs
+ numSeqs += num;
+
+ //send file positions to all processes
+ for(int i = 1; i < processors; i++) {
+ MPI_Send(&num, 1, MPI_INT, i, tag, MPI_COMM_WORLD);
+ MPI_Send(&MPIPos[0], (num+1), MPI_LONG, i, tag, MPI_COMM_WORLD);
+ }
+
+ //figure out how many sequences you have to do
+ numSeqsPerProcessor = num / processors;
+ int startIndex = pid * numSeqsPerProcessor;
+ if(pid == (processors - 1)){ numSeqsPerProcessor = num - pid * numSeqsPerProcessor; }
+
+
+ //do your part
+ driverMPIRun(startIndex, numSeqsPerProcessor, inMPI, outMPI, MPIPos);
+
+ if (m->control_pressed) { MPI_File_close(&inMPI); MPI_File_close(&outMPI); return 0; }
+
+ //wait on chidren
+ for(int i = 1; i < processors; i++) {
+ char buf[5];
+ MPI_Recv(buf, 5, MPI_CHAR, i, tag, MPI_COMM_WORLD, &status);
+ }
+
+ }else { //you are a child process
+ MPI_Recv(&num, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
+ MPIPos.resize(num+1);
+ numSeqs += num;
+ MPI_Recv(&MPIPos[0], (num+1), MPI_LONG, 0, tag, MPI_COMM_WORLD, &status);
+
+ //figure out how many sequences you have to align
+ numSeqsPerProcessor = num / processors;
+ int startIndex = pid * numSeqsPerProcessor;
+ if(pid == (processors - 1)){ numSeqsPerProcessor = num - pid * numSeqsPerProcessor; }
+
+
+ //align your part
+ driverMPIRun(startIndex, numSeqsPerProcessor, inMPI, outMPI, MPIPos);
+
+ if (m->control_pressed) { MPI_File_close(&inMPI); MPI_File_close(&outMPI); return 0; }
+
+ char buf[5];
+ strcpy(buf, "done");
+
+ //tell parent you are done.
+ MPI_Send(buf, 5, MPI_CHAR, 0, tag, MPI_COMM_WORLD);
+ }
+
+ MPI_File_close(&outMPI);
+ MPI_File_close(&inMPI);
+ MPI_Barrier(MPI_COMM_WORLD); //make everyone wait - just in case
+
+#else
+ vector<unsigned long int> positions = m->divideFile(fastafileNames[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)
+ if(processors == 1){
+ int numFastaSeqs = driverRunFilter(filter, filteredFasta, fastafileNames[s], lines[0]);
+ numSeqs += numFastaSeqs;
+ }else{
+ int numFastaSeqs = createProcessesRunFilter(filter, fastafileNames[s]);
+ numSeqs += numFastaSeqs;
+
+ rename((fastafileNames[s] + toString(processIDS[0]) + ".temp").c_str(), filteredFasta.c_str());
+
+ //append fasta files
+ for(int i=1;i<processors;i++){
+ m->appendFiles((fastafileNames[s] + toString(processIDS[i]) + ".temp"), filteredFasta);
+ remove((fastafileNames[s] + toString(processIDS[i]) + ".temp").c_str());
+ }
+ }
+
+ if (m->control_pressed) { return 1; }
+ #else
+ int numFastaSeqs = driverRunFilter(filter, filteredFasta, fastafileNames[s], lines[0]);
+ numSeqs += numFastaSeqs;
+
+ if (m->control_pressed) { return 1; }
+ #endif
+#endif
+ outputNames.push_back(filteredFasta); outputTypes["fasta"].push_back(filteredFasta);
+ }
+
+ return 0;
+ }
+ catch(exception& e) {
+ m->errorOut(e, "FilterSeqsCommand", "filterSequences");
+ exit(1);
+ }
+}
+#ifdef USE_MPI
+/**************************************************************************************/
+int FilterSeqsCommand::driverMPIRun(int start, int num, MPI_File& inMPI, MPI_File& outMPI, vector<unsigned long int>& MPIPos) {
+ try {
+ string outputString = "";
+ int count = 0;
+ MPI_Status status;
+
+ for(int i=0;i<num;i++){
+
+ if (m->control_pressed) { return 0; }
+
+ //read next sequence
+ int length = MPIPos[start+i+1] - MPIPos[start+i];
+ char* buf4 = new char[length];
+ MPI_File_read_at(inMPI, MPIPos[start+i], buf4, length, MPI_CHAR, &status);
+
+ string tempBuf = buf4;
+ if (tempBuf.length() > length) { tempBuf = tempBuf.substr(0, length); }
+ istringstream iss (tempBuf,istringstream::in);
+ delete buf4;
+
+ Sequence seq(iss); m->gobble(iss);
+
+ if (seq.getName() != "") {
+ string align = seq.getAligned();
+ string filterSeq = "";
+
+ for(int j=0;j<alignmentLength;j++){
+ if(filter[j] == '1'){
+ filterSeq += align[j];
+ }
+ }
+
+ count++;
+ outputString += ">" + seq.getName() + "\n" + filterSeq + "\n";
+
+ if(count % 10 == 0){ //output to file
+ //send results to parent
+ int length = outputString.length();
+ char* buf = new char[length];
+ memcpy(buf, outputString.c_str(), length);
+
+ MPI_File_write_shared(outMPI, buf, length, MPI_CHAR, &status);
+ outputString = "";
+ delete buf;
+ }
+
+ }
+
+ if((i+1) % 100 == 0){ cout << (i+1) << endl; m->mothurOutJustToLog(toString(i+1) + "\n"); }
+ }
+
+ if(outputString != ""){ //output to file
+ //send results to parent
+ int length = outputString.length();
+ char* buf = new char[length];
+ memcpy(buf, outputString.c_str(), length);
+
+ MPI_File_write_shared(outMPI, buf, length, MPI_CHAR, &status);
+ outputString = "";
+ delete buf;
+ }
+
+ if((num) % 100 != 0){ cout << (num) << endl; m->mothurOutJustToLog(toString(num) + "\n"); }
+
+ return 0;
+ }
+ catch(exception& e) {
+ m->errorOut(e, "FilterSeqsCommand", "driverRunFilter");
+ exit(1);
+ }
+}
+#endif
+/**************************************************************************************/
+int FilterSeqsCommand::driverRunFilter(string F, string outputFilename, string inputFilename, linePair* filePos) {
+ try {
+ ofstream out;
+ m->openOutputFile(outputFilename, out);
+
+ ifstream in;
+ m->openInputFile(inputFilename, in);
+
+ in.seekg(filePos->start);
+
+ bool done = false;
+ int count = 0;
+
+ while (!done) {
+
+ if (m->control_pressed) { in.close(); out.close(); return 0; }
+
+ Sequence seq(in); m->gobble(in);
+ if (seq.getName() != "") {
+ string align = seq.getAligned();
+ string filterSeq = "";
+
+ for(int j=0;j<alignmentLength;j++){
+ if(filter[j] == '1'){
+ filterSeq += align[j];
+ }
+ }
+
+ out << '>' << seq.getName() << endl << filterSeq << endl;
+ count++;
+ }
+
+ #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
+ unsigned long int pos = in.tellg();
+ if ((pos == -1) || (pos >= filePos->end)) { break; }
+ #else
+ if (in.eof()) { break; }
+ #endif
+
+ //report progress
+ if((count) % 100 == 0){ m->mothurOut(toString(count)); m->mothurOutEndLine(); }
+ }
+ //report progress
+ if((count) % 100 != 0){ m->mothurOut(toString(count)); m->mothurOutEndLine(); }
+
+
+ out.close();
+ in.close();
+
+ return count;
+ }
+ catch(exception& e) {
+ m->errorOut(e, "FilterSeqsCommand", "driverRunFilter");
+ exit(1);
+ }
+}
+/**************************************************************************************************/
+
+int FilterSeqsCommand::createProcessesRunFilter(string F, string filename) {
+ try {
+#if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
+ int process = 0;
+ int num = 0;
+ processIDS.clear();
+
+ //loop through and create all the processes you want
+ while (process != processors) {
+ int pid = fork();
+
+ if (pid > 0) {
+ processIDS.push_back(pid); //create map from line number to pid so you can append files in correct order later
+ process++;
+ }else if (pid == 0){
+ string filteredFasta = filename + toString(getpid()) + ".temp";
+ num = driverRunFilter(F, filteredFasta, filename, lines[process]);
+
+ //pass numSeqs to parent
+ ofstream out;
+ string tempFile = filename + toString(getpid()) + ".num.temp";
+ m->openOutputFile(tempFile, out);
+ out << num << endl;
+ out.close();
+
+ exit(0);
+ }else {
+ m->mothurOut("[ERROR]: unable to spawn the necessary processes."); m->mothurOutEndLine();
+ for (int i = 0; i < processIDS.size(); i++) { kill (processIDS[i], SIGINT); }
+ exit(0);
+ }
+ }
+
+ //force parent to wait until all the processes are done
+ for (int i=0;i<processors;i++) {
+ int temp = processIDS[i];
+ wait(&temp);
+ }
+
+ for (int i = 0; i < processIDS.size(); i++) {
+ ifstream in;
+ string tempFile = filename + toString(processIDS[i]) + ".num.temp";
+ m->openInputFile(tempFile, in);
+ if (!in.eof()) { int tempNum = 0; in >> tempNum; num += tempNum; }
+ in.close(); remove(tempFile.c_str());
+ }
+
+
+ return num;
+#endif
+ }
+ catch(exception& e) {
+ m->errorOut(e, "FilterSeqsCommand", "createProcessesRunFilter");
+ exit(1);
+ }
+}
+/**************************************************************************************/
+string FilterSeqsCommand::createFilter() {
+ try {
+ string filterString = "";