OptionParser parser(option);
map<string,string> parameters = parser.getParameters();
- ValidParameters validParameter;
+ ValidParameters validParameter("filter.seqs");
map<string,string>::iterator it;
//check to make sure all parameters are valid for command
it = parameters.find("fasta");
//user has given a template file
if(it != parameters.end()){
- path = hasPath(it->second);
+ path = m->hasPath(it->second);
//if the user has not given a path then, add inputdir. else leave path alone.
if (path == "") { parameters["fasta"] = inputDir + it->second; }
}
it = parameters.find("hard");
//user has given a template file
if(it != parameters.end()){
- path = hasPath(it->second);
+ path = m->hasPath(it->second);
//if the user has not given a path then, add inputdir. else leave path alone.
if (path == "") { parameters["hard"] = inputDir + it->second; }
}
fasta = validParameter.validFile(parameters, "fasta", false);
if (fasta == "not found") { m->mothurOut("fasta is a required parameter for the filter.seqs command."); m->mothurOutEndLine(); abort = true; }
else {
- splitAtDash(fasta, fastafileNames);
+ m->splitAtDash(fasta, fastafileNames);
//go through files and make sure they are good, if not, then disregard them
for (int i = 0; i < fastafileNames.size(); i++) {
if (inputDir != "") {
- string path = hasPath(fastafileNames[i]);
+ string path = m->hasPath(fastafileNames[i]);
//if the user has not given a path then, add inputdir. else leave path alone.
if (path == "") { fastafileNames[i] = inputDir + fastafileNames[i]; }
}
- int ableToOpen;
ifstream in;
- ableToOpen = openInputFile(fastafileNames[i], in);
+ int ableToOpen = m->openInputFile(fastafileNames[i], in, "noerror");
+
+ //if you can't open it, try default location
+ if (ableToOpen == 1) {
+ if (m->getDefaultPath() != "") { //default path is set
+ string tryPath = m->getDefaultPath() + m->getSimpleName(fastafileNames[i]);
+ m->mothurOut("Unable to open " + fastafileNames[i] + ". Trying default " + tryPath); m->mothurOutEndLine();
+ ableToOpen = m->openInputFile(tryPath, in, "noerror");
+ fastafileNames[i] = tryPath;
+ }
+ }
+ in.close();
+
if (ableToOpen == 1) {
- m->mothurOut(fastafileNames[i] + " will be disregarded."); m->mothurOutEndLine();
+ m->mothurOut("Unable to open " + fastafileNames[i] + ". It will be disregarded."); m->mothurOutEndLine();
//erase from file list
fastafileNames.erase(fastafileNames.begin()+i);
i--;
}else{
- string simpleName = getSimpleName(fastafileNames[i]);
+ string simpleName = m->getSimpleName(fastafileNames[i]);
filterFileName += simpleName.substr(0, simpleName.find_first_of('.'));
}
in.close();
//if the user changes the output directory command factory will send this info to us in the output parameter
outputDir = validParameter.validFile(parameters, "outputdir", false); if (outputDir == "not found"){
outputDir = "";
- outputDir += hasPath(fastafileNames[0]); //if user entered a file with a path then preserve it
+ outputDir += m->hasPath(fastafileNames[0]); //if user entered a file with a path then preserve it
}
}
//check for optional parameter and set defaults
// ...at some point should added some additional type checking...
string temp;
+ hard = validParameter.validFile(parameters, "hard", true); if (hard == "not found") { hard = ""; }
+ else if (hard == "not open") { hard = ""; abort = true; }
+
temp = validParameter.validFile(parameters, "trump", false); if (temp == "not found") { temp = "*"; }
trump = temp[0];
temp = validParameter.validFile(parameters, "processors", false); if (temp == "not found"){ temp = "1"; }
convert(temp, processors);
- hard = validParameter.validFile(parameters, "hard", true); if (hard == "not found") { hard = ""; }
- else if (hard == "not open") { abort = true; }
-
- vertical = validParameter.validFile(parameters, "vertical", false); if (vertical == "not found") { vertical = "T"; }
+ vertical = validParameter.validFile(parameters, "vertical", false);
+ if (vertical == "not found") {
+ if ((hard == "") && (trump == '*') && (soft == 0)) { vertical = "T"; } //you have not given a hard file or set the trump char.
+ else { vertical = "F"; }
+ }
numSeqs = 0;
-
}
}
try {
m->mothurOut("The filter.seqs command reads a file containing sequences and creates a .filter and .filter.fasta file.\n");
- m->mothurOut("The filter.seqs command parameters are fasta, trump, soft, hard and vertical. \n");
+ m->mothurOut("The filter.seqs command parameters are fasta, trump, soft, hard, processors and vertical. \n");
m->mothurOut("The fasta parameter is required. You may enter several fasta files to build the filter from and filter, by separating their names with -'s.\n");
m->mothurOut("For example: fasta=abrecovery.fasta-amazon.fasta \n");
m->mothurOut("The trump parameter .... The default is ...\n");
m->mothurOut("The soft parameter .... The default is ....\n");
- m->mothurOut("The hard parameter .... The default is ....\n");
- m->mothurOut("The vertical parameter .... The default is T.\n");
+ m->mothurOut("The hard parameter allows you to enter a file containing the filter you want to use.\n");
+ m->mothurOut("The vertical parameter removes columns where all sequences contain a gap character. The default is T.\n");
+ m->mothurOut("The processors parameter allows you to specify the number of processors to use. The default is 1.\n");
m->mothurOut("The filter.seqs command should be in the following format: \n");
- m->mothurOut("filter.seqs(fasta=yourFastaFile, trump=yourTrump, soft=yourSoft, hard=yourHard, vertical=yourVertical) \n");
- m->mothurOut("Example filter.seqs(fasta=abrecovery.fasta, trump=..., soft=..., hard=..., vertical=T).\n");
+ m->mothurOut("filter.seqs(fasta=yourFastaFile, trump=yourTrump) \n");
+ m->mothurOut("Example filter.seqs(fasta=abrecovery.fasta, trump=.).\n");
m->mothurOut("Note: No spaces between parameter labels (i.e. fasta), '=' and parameters (i.e.yourFasta).\n\n");
}
try {
if (abort == true) { return 0; }
- vector<string> outputNames;
ifstream inFASTA;
- openInputFile(fastafileNames[0], inFASTA);
+ m->openInputFile(fastafileNames[0], inFASTA);
Sequence testSeq(inFASTA);
alignmentLength = testSeq.getAlignLength();
inFASTA.close();
////////////create filter/////////////////
+ m->mothurOut("Creating Filter... "); m->mothurOutEndLine();
filter = createFilter();
+ m->mothurOutEndLine(); m->mothurOutEndLine();
+
+ if (m->control_pressed) { return 0; }
+
+ #ifdef USE_MPI
+ int pid;
+ MPI_Comm_rank(MPI_COMM_WORLD, &pid);
+
+ if (pid == 0) { //only one process should output the filter
+ #endif
+
ofstream outFilter;
- string filterFile = outputDir + filterFileName + ".filter";
- openOutputFile(filterFile, outFilter);
+ //prevent giantic file name
+ string filterFile;
+ if (fastafileNames.size() > 3) { filterFile = outputDir + "merge.filter"; }
+ else { filterFile = outputDir + filterFileName + ".filter"; }
+
+ m->openOutputFile(filterFile, outFilter);
outFilter << filter << endl;
outFilter.close();
outputNames.push_back(filterFile);
+ #ifdef USE_MPI
+ }
+ #endif
////////////run filter/////////////////
- numSeqs = 0;
- for (int i = 0; i < fastafileNames.size(); i++) {
- ifstream in;
- openInputFile(fastafileNames[i], in);
- string filteredFasta = outputDir + getRootName(getSimpleName(fastafileNames[i])) + "filter.fasta";
- ofstream outFASTA;
- openOutputFile(filteredFasta, outFASTA);
- outputNames.push_back(filteredFasta);
-
-
- while(!in.eof()){
- if (m->control_pressed) { in.close(); outFASTA.close(); for(int i = 0; i < outputNames.size(); i++) { remove(outputNames[i].c_str()); } return 0; }
-
- Sequence seq(in);
- if (seq.getName() != "") {
- string align = seq.getAligned();
- string filterSeq = "";
-
- for(int j=0;j<alignmentLength;j++){
- if(filter[j] == '1'){
- filterSeq += align[j];
- }
- }
-
- outFASTA << '>' << seq.getName() << endl << filterSeq << endl;
- numSeqs++;
- }
- gobble(in);
- }
- outFASTA.close();
- in.close();
- }
+ m->mothurOut("Running Filter... "); m->mothurOutEndLine();
+ filterSequences();
+
+ m->mothurOutEndLine(); m->mothurOutEndLine();
+
int filteredLength = 0;
for(int i=0;i<alignmentLength;i++){
if(filter[i] == '1'){ filteredLength++; }
}
}
/**************************************************************************************/
-string FilterSeqsCommand::createFilter() {
+int FilterSeqsCommand::filterSequences() {
+ try {
+
+ 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, start, end, 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 tempMPI;
+ 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);
+ }
+
+ 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 {
- string filterString = "";
+#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("unable to spawn the necessary processes."); m->mothurOutEndLine(); 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 = "";
Filters F;
if (soft != 0) { F.setSoft(soft); }
F.setLength(alignmentLength);
- if(soft != 0 || isTrue(vertical)){
+ if(trump != '*' || m->isTrue(vertical) || soft != 0){
F.initialize();
}
else { F.setFilter(string(alignmentLength, '1')); }
numSeqs = 0;
-
- if(trump != '*' || isTrue(vertical) || soft != 0){
+ if(trump != '*' || m->isTrue(vertical) || soft != 0){
for (int s = 0; s < fastafileNames.size(); s++) {
for (int i = 0; i < lines.size(); i++) { delete lines[i]; } lines.clear();
#ifdef USE_MPI
- int pid, rc, ierr;
- int Atag = 1; int Ttag = 2; int Ctag = 3; int Gtag = 4; int Gaptag = 5;
+ int pid, numSeqsPerProcessor, num;
int tag = 2001;
+ vector<unsigned long int> MPIPos;
MPI_Status status;
- MPI_File in;
- rc = MPI_Comm_size(MPI_COMM_WORLD, &processors);
- rc = MPI_Comm_rank(MPI_COMM_WORLD, &pid);
-
+ MPI_File inMPI;
+ MPI_Comm_size(MPI_COMM_WORLD, &processors);
+ MPI_Comm_rank(MPI_COMM_WORLD, &pid);
- char* tempFileName = new char(fastafileNames[s].length());
- tempFileName = &(fastafileNames[s][0]);
+ //char* tempFileName = new char(fastafileNames[s].length());
+ //tempFileName = &(fastafileNames[s][0]);
+
+ char tempFileName[1024];
+ strcpy(tempFileName, fastafileNames[s].c_str());
+
+ MPI_File_open(MPI_COMM_WORLD, tempFileName, MPI_MODE_RDONLY, MPI_INFO_NULL, &inMPI); //comm, filename, mode, info, filepointer
- MPI_File_open(MPI_COMM_WORLD, tempFileName, MPI_MODE_RDONLY, MPI_INFO_NULL, &in); //comm, filename, mode, info, filepointer
+ if (m->control_pressed) { MPI_File_close(&inMPI); return 0; }
if (pid == 0) { //you are the root process
- setLines(fastafileNames[s]);
+ MPIPos = m->setFilePosFasta(fastafileNames[s], num); //fills MPIPos, returns numSeqs
+ numSeqs += num;
- for (int j = 0; j < lines.size(); j++) { //each process
- if (j != 0) { //don't send to yourself
- MPI_Send(&lines[j]->start, 1, MPI_INT, j, tag, MPI_COMM_WORLD); //start position in file
- MPI_Send(&lines[j]->numSeqs, 1, MPI_INT, j, tag, MPI_COMM_WORLD); //how many sequences we are sending
- MPI_Send(&bufferSizes[j], 1, MPI_INT, j, tag, MPI_COMM_WORLD); //how bytes for the read
- }
+ //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);
}
- //cout << "done sending" << endl;
- //cout << "parent = " << pid << " lines = " << lines[pid]->start << '\t' << lines[pid]->numSeqs << " size = " << lines.size() << endl;
-
- cout << "parent = " << pid << " address of Filter " << &F << " address of FilterString " << &filterString << " address of numSeqs = " << &numSeqs << " address of soft = " << &soft << endl;
-
- char* buf = new char(bufferSizes[0]);
- //cout << pid << '\t' << bufferSizes[0] << " line 1 start pos = " << lines[1]->start << " buffer size 0 " << bufferSizes[0] << " buffer size 1 " << bufferSizes[1] << endl;
- MPI_File_read_at(in, 0, buf, bufferSizes[0], MPI_CHAR, &status);
+
+ //figure out how many sequences you have to do
+ numSeqsPerProcessor = num / processors;
+ int startIndex = pid * numSeqsPerProcessor;
+ if(pid == (processors - 1)){ numSeqsPerProcessor = num - pid * numSeqsPerProcessor; }
- cout << pid << " done reading " << &buf << endl;
- string tempBuf = buf;
- delete buf;
- //cout << pid << '\t' << (tempBuf.substr(0, 10)) << endl;
-
- //parse buffer
- istringstream iss (tempBuf,istringstream::in);
- string name, seqstring;
- vector<string> seqs;
-
- while (iss) {
-
- if (m->control_pressed) { return filterString; }
- cout << "here" << endl;
- Sequence seq(iss);
- cout << "here1" << endl;
- gobble(iss);
- cout << seq.getName() << endl;
- if (seq.getName() != "") {
- seqs.push_back(seq.getAligned());
- }
-
- }
-
- for(int i=0;i<seqs.size();i++){
- if (m->control_pressed) { return filterString; }
-
- Sequence seq("", seqs[i]);
-
- if(trump != '*'){ F.doTrump(seq); }
- if(isTrue(vertical) || soft != 0){ F.getFreqs(seq); }
- cout.flush();
-
- //report progress
- if((i+1) % 100 == 0){ m->mothurOut(toString(i+1)); m->mothurOutEndLine(); }
- }
-
- //report progress
- if((seqs.size()) % 100 != 0){ m->mothurOut(toString(seqs.size())); m->mothurOutEndLine(); }
-
//do your part
- //MPICreateFilter(F, seqs);
-
- vector<int> temp; temp.resize(alignmentLength);
-
- //get the frequencies from the child processes
- for(int i = 0; i < ((processors-1)*5); i++) {
- cout << "i = " << i << endl;
- int ierr = MPI_Recv(&temp, alignmentLength, MPI_INT, MPI_ANY_SOURCE, tag, MPI_COMM_WORLD, &status);
-
- int receiveTag = temp[temp.size()-1]; //child process added a int to the end to indicate what letter count this is for
-
- int sender = status.MPI_SOURCE;
-
- if (receiveTag == Atag) { //you are recieveing the A frequencies
- for (int k = 0; k < alignmentLength; k++) { F.a[k] += temp[k]; }
- }else if (receiveTag == Ttag) { //you are recieveing the T frequencies
- for (int k = 0; k < alignmentLength; k++) { F.t[k] += temp[k]; }
- }else if (receiveTag == Ctag) { //you are recieveing the C frequencies
- for (int k = 0; k < alignmentLength; k++) { F.c[k] += temp[k]; }
- }else if (receiveTag == Gtag) { //you are recieveing the G frequencies
- for (int k = 0; k < alignmentLength; k++) { F.g[k] += temp[k]; }
- }else if (receiveTag == Gaptag) { //you are recieveing the gap frequencies
- for (int k = 0; k < alignmentLength; k++) { F.gap[k] += temp[k]; }
- }
-
- m->mothurOut("receive tag = " + toString(receiveTag) + " " + toString(sender) + " is complete."); m->mothurOutEndLine();
- }
-
+ MPICreateFilter(startIndex, numSeqsPerProcessor, F, inMPI, MPIPos);
+ if (m->control_pressed) { MPI_File_close(&inMPI); return 0; }
+
}else { //i am the child process
- int startPos, numLines, bufferSize;
- cout << "child = " << pid << " address of Filter " << &F << " address of FilterString " << &filterString << " address of numSeqs = " << &numSeqs << " address of soft = " << &soft<< endl;
- ierr = MPI_Recv(&startPos, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
- ierr = MPI_Recv(&numLines, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
- ierr = MPI_Recv(&bufferSize, 1, MPI_INT, 0, tag, MPI_COMM_WORLD, &status);
- //cout << "child = " << pid << " done recv messages startpos = " << startPos << " numLines = " << numLines << " buffersize = " << bufferSize << endl;
-
+ 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);
- //send freqs
- char* buf2 = new char(bufferSize);
- MPI_File_read_at( in, startPos, buf2, bufferSize, MPI_CHAR, &status);
- cout << pid << " done reading " << &buf2 << endl;
+ //figure out how many sequences you have to align
+ numSeqsPerProcessor = num / processors;
+ int startIndex = pid * numSeqsPerProcessor;
+ if(pid == (processors - 1)){ numSeqsPerProcessor = num - pid * numSeqsPerProcessor; }
- string tempBuf = buf2;
- delete buf2;
- // cout << pid << '\t' << (tempBuf.substr(0, 10)) << endl;
- istringstream iss (tempBuf,istringstream::in);
- string name, seqstring;
- vector<string> seqs;
-
- while (iss) {
-
- if (m->control_pressed) { return filterString; }
- cout << "here" << endl;
- Sequence seq(iss);
- cout << "here1" << endl;
- gobble(iss);
- cout << seq.getName() << endl;
-
- if (seq.getName() != "") {
- seqs.push_back(seq.getAligned());
- }
- }
-
- for(int i=0;i<seqs.size();i++){
-
- if (m->control_pressed) { return filterString; }
-
- Sequence seq("", seqs[i]);
-
- if(trump != '*'){ F.doTrump(seq); }
- if(isTrue(vertical) || soft != 0){ F.getFreqs(seq); }
- cout.flush();
-
- //report progress
- if((i+1) % 100 == 0){ m->mothurOut(toString(i+1)); m->mothurOutEndLine(); }
- }
-
- //report progress
- if((seqs.size()) % 100 != 0){ m->mothurOut(toString(seqs.size())); m->mothurOutEndLine(); }
-
- //MPICreateFilter(F, seqs);
-
- //send my fequency counts
- F.a.push_back(Atag);
- int ierr = MPI_Send( &F.a[0], alignmentLength, MPI_INT, 0, tag, MPI_COMM_WORLD);
- F.t.push_back(Ttag);
- ierr = MPI_Send( &F.t[0], alignmentLength, MPI_INT, 0, tag, MPI_COMM_WORLD);
- F.c.push_back(Ctag);
- ierr = MPI_Send( &F.c[0], alignmentLength, MPI_INT, 0, tag, MPI_COMM_WORLD);
- F.g.push_back(Gtag);
- ierr = MPI_Send( &F.g[0], alignmentLength, MPI_INT, 0, tag, MPI_COMM_WORLD);
- F.gap.push_back(Gaptag);
- ierr = MPI_Send( &F.gap[0], alignmentLength, MPI_INT, 0, tag, MPI_COMM_WORLD);
+ //do your part
+ MPICreateFilter(startIndex, numSeqsPerProcessor, F, inMPI, MPIPos);
- cout << "child " << pid << " done sending counts" << endl;
+ if (m->control_pressed) { MPI_File_close(&inMPI); return 0; }
}
- MPI_Barrier(MPI_COMM_WORLD);
+ 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){
- ifstream inFASTA;
- openInputFile(fastafileNames[s], inFASTA);
- int numFastaSeqs=count(istreambuf_iterator<char>(inFASTA),istreambuf_iterator<char>(), '>');
- inFASTA.close();
-
+ int numFastaSeqs = driverCreateFilter(F, fastafileNames[s], lines[0]);
numSeqs += numFastaSeqs;
-
- lines.push_back(new linePair(0, numFastaSeqs));
-
- driverCreateFilter(F, fastafileNames[s], lines[0]);
}else{
-
- setLines(fastafileNames[s]);
- createProcessesCreateFilter(F, fastafileNames[s]);
+ int numFastaSeqs = createProcessesCreateFilter(F, fastafileNames[s]);
+ numSeqs += numFastaSeqs;
}
- #else
- ifstream inFASTA;
- openInputFile(fastafileNames[s], inFASTA);
- int numFastaSeqs=count(istreambuf_iterator<char>(inFASTA),istreambuf_iterator<char>(), '>');
- inFASTA.close();
+ if (m->control_pressed) { return filterString; }
+ #else
+ int numFastaSeqs = driverCreateFilter(F, fastafileNames[s], lines[0]);
numSeqs += numFastaSeqs;
-
- lines.push_back(new linePair(0, numFastaSeqs));
-
- driverCreateFilter(F, lines[0], fastafileNames[s]);
+ if (m->control_pressed) { return filterString; }
#endif
#endif
}
}
-#ifdef USE_MPI
-//merge all frequency data and create filter string
- //int pid;
- //MPI_Comm_rank(MPI_COMM_WORLD, &pid);
+#ifdef USE_MPI
+ int pid;
+ int Atag = 1; int Ttag = 2; int Ctag = 3; int Gtag = 4; int Gaptag = 5;
+ MPI_Status status;
+
+ MPI_Comm_rank(MPI_COMM_WORLD, &pid);
+
+ if(trump != '*' || m->isTrue(vertical) || soft != 0){
+
+ if (pid == 0) { //only one process should output the filter
+
+ vector<int> temp; temp.resize(alignmentLength+1);
+
+ //get the frequencies from the child processes
+ for(int i = 1; i < processors; i++) {
+
+ for (int j = 0; j < 5; j++) {
- //if (pid == 0) { //only one process should output to screen
+ MPI_Recv(&temp[0], (alignmentLength+1), MPI_INT, i, 2001, MPI_COMM_WORLD, &status);
+ int receiveTag = temp[temp.size()-1]; //child process added a int to the end to indicate what letter count this is for
+
+ if (receiveTag == Atag) { //you are recieveing the A frequencies
+ for (int k = 0; k < alignmentLength; k++) { F.a[k] += temp[k]; }
+ }else if (receiveTag == Ttag) { //you are recieveing the T frequencies
+ for (int k = 0; k < alignmentLength; k++) { F.t[k] += temp[k]; }
+ }else if (receiveTag == Ctag) { //you are recieveing the C frequencies
+ for (int k = 0; k < alignmentLength; k++) { F.c[k] += temp[k]; }
+ }else if (receiveTag == Gtag) { //you are recieveing the G frequencies
+ for (int k = 0; k < alignmentLength; k++) { F.g[k] += temp[k]; }
+ }else if (receiveTag == Gaptag) { //you are recieveing the gap frequencies
+ for (int k = 0; k < alignmentLength; k++) { F.gap[k] += temp[k]; }
+ }
+ }
+ }
+ }else{
+
+ //send my fequency counts
+ F.a.push_back(Atag);
+ int ierr = MPI_Send(&(F.a[0]), (alignmentLength+1), MPI_INT, 0, 2001, MPI_COMM_WORLD);
+ F.t.push_back(Ttag);
+ ierr = MPI_Send (&(F.t[0]), (alignmentLength+1), MPI_INT, 0, 2001, MPI_COMM_WORLD);
+ F.c.push_back(Ctag);
+ ierr = MPI_Send(&(F.c[0]), (alignmentLength+1), MPI_INT, 0, 2001, MPI_COMM_WORLD);
+ F.g.push_back(Gtag);
+ ierr = MPI_Send(&(F.g[0]), (alignmentLength+1), MPI_INT, 0, 2001, MPI_COMM_WORLD);
+ F.gap.push_back(Gaptag);
+ ierr = MPI_Send(&(F.gap[0]), (alignmentLength+1), MPI_INT, 0, 2001, MPI_COMM_WORLD);
+ }
+
+ }
+
+ MPI_Barrier(MPI_COMM_WORLD); //make everyone wait - just in case
+
+ if (pid == 0) { //only one process should output the filter
#endif
- cout << "made it here" << endl;
F.setNumSeqs(numSeqs);
-
- if(isTrue(vertical) == 1) { F.doVertical(); }
+ if(m->isTrue(vertical) == 1) { F.doVertical(); }
if(soft != 0) { F.doSoft(); }
-//cout << "Filter String = " << F.getFilter() << endl;
filterString = F.getFilter();
-
+
+#ifdef USE_MPI
+ //send filter string to kids
+ //for(int i = 1; i < processors; i++) {
+ // MPI_Send(&filterString[0], alignmentLength, MPI_CHAR, i, 2001, MPI_COMM_WORLD);
+ //}
+ MPI_Bcast(&filterString[0], alignmentLength, MPI_CHAR, 0, MPI_COMM_WORLD);
+ }else{
+ //recieve filterString
+ char* tempBuf = new char[alignmentLength];
+ //MPI_Recv(&tempBuf[0], alignmentLength, MPI_CHAR, 0, 2001, MPI_COMM_WORLD, &status);
+ MPI_Bcast(tempBuf, alignmentLength, MPI_CHAR, 0, MPI_COMM_WORLD);
+
+ filterString = tempBuf;
+ if (filterString.length() > alignmentLength) { filterString = filterString.substr(0, alignmentLength); }
+ delete tempBuf;
+ }
+
+ MPI_Barrier(MPI_COMM_WORLD);
+#endif
+
return filterString;
}
catch(exception& e) {
}
}
/**************************************************************************************/
-int FilterSeqsCommand::driverCreateFilter(Filters& F, string filename, linePair* line) {
+int FilterSeqsCommand::driverCreateFilter(Filters& F, string filename, linePair* filePos) {
try {
ifstream in;
- openInputFile(filename, in);
+ m->openInputFile(filename, in);
- in.seekg(line->start);
-
- for(int i=0;i<line->numSeqs;i++){
+ in.seekg(filePos->start);
+
+ bool done = false;
+ int count = 0;
+
+ while (!done) {
if (m->control_pressed) { in.close(); return 1; }
- Sequence seq(in);
+ Sequence seq(in); m->gobble(in);
if (seq.getName() != "") {
- if(trump != '*'){ F.doTrump(seq); }
- if(isTrue(vertical) || soft != 0){ F.getFreqs(seq); }
+ if (seq.getAligned().length() != alignmentLength) { m->mothurOut("Sequences are not all the same length, please correct."); m->mothurOutEndLine(); m->control_pressed = true; }
+
+ if(trump != '*') { F.doTrump(seq); }
+ if(m->isTrue(vertical) || soft != 0) { F.getFreqs(seq); }
cout.flush();
+ 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((i+1) % 100 == 0){ m->mothurOut(toString(i+1)); m->mothurOutEndLine(); }
+ if((count) % 100 == 0){ m->mothurOut(toString(count)); m->mothurOutEndLine(); }
}
-
//report progress
- if((line->numSeqs) % 100 != 0){ m->mothurOut(toString(line->numSeqs)); m->mothurOutEndLine(); }
-
+ if((count) % 100 != 0){ m->mothurOut(toString(count)); m->mothurOutEndLine(); }
in.close();
- return 0;
+ return count;
}
catch(exception& e) {
m->errorOut(e, "FilterSeqsCommand", "driverCreateFilter");
exit(1);
}
}
+#ifdef USE_MPI
/**************************************************************************************/
-int FilterSeqsCommand::MPICreateFilter(Filters& F, vector<string>& seqStrings) {
+int FilterSeqsCommand::MPICreateFilter(int start, int num, Filters& F, MPI_File& inMPI, vector<unsigned long int>& MPIPos) {
try {
- for(int i=0;i<seqStrings.size();i++){
-
- if (m->control_pressed) { return 1; }
+ MPI_Status status;
+ int pid;
+ MPI_Comm_rank(MPI_COMM_WORLD, &pid); //find out who we are
+
+ for(int i=0;i<num;i++){
- Sequence seq("", seqStrings[0]);
+ 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);
+
+ if (seq.getAligned().length() != alignmentLength) { cout << "Alignment length is " << alignmentLength << " and sequence " << seq.getName() << " has length " << seq.getAligned().length() << ", please correct." << endl; exit(1); }
if(trump != '*'){ F.doTrump(seq); }
- if(isTrue(vertical) || soft != 0){ F.getFreqs(seq); }
+ if(m->isTrue(vertical) || soft != 0){ F.getFreqs(seq); }
cout.flush();
//report progress
- if((i+1) % 100 == 0){ m->mothurOut(toString(i+1)); m->mothurOutEndLine(); }
+ if((i+1) % 100 == 0){ cout << (i+1) << endl; m->mothurOutJustToLog(toString(i+1) + "\n"); }
}
//report progress
- if((seqStrings.size()) % 100 != 0){ m->mothurOut(toString(seqStrings.size())); m->mothurOutEndLine(); }
+ if((num) % 100 != 0){ cout << num << endl; m->mothurOutJustToLog(toString(num) + "\n"); }
return 0;
}
exit(1);
}
}
-
+#endif
/**************************************************************************************************/
int FilterSeqsCommand::createProcessesCreateFilter(Filters& F, string filename) {
try {
#if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
- int process = 0;
- int exitCommand = 1;
- vector<int> processIDS;
+ int process = 1;
+ int num = 0;
+ processIDS.clear();
//loop through and create all the processes you want
while (process != processors) {
- int pid = vfork();
+ 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){
- driverCreateFilter(F, filename, lines[process]);
+ //reset child's filter counts to 0;
+ F.a.clear(); F.a.resize(alignmentLength, 0);
+ F.t.clear(); F.t.resize(alignmentLength, 0);
+ F.g.clear(); F.g.resize(alignmentLength, 0);
+ F.c.clear(); F.c.resize(alignmentLength, 0);
+ F.gap.clear(); F.gap.resize(alignmentLength, 0);
+
+ num = driverCreateFilter(F, filename, lines[process]);
+
+ //write out filter counts to file
+ filename += toString(getpid()) + "filterValues.temp";
+ ofstream out;
+ m->openOutputFile(filename, out);
+
+ out << num << endl;
+ out << F.getFilter() << endl;
+ for (int k = 0; k < alignmentLength; k++) { out << F.a[k] << '\t'; } out << endl;
+ for (int k = 0; k < alignmentLength; k++) { out << F.t[k] << '\t'; } out << endl;
+ for (int k = 0; k < alignmentLength; k++) { out << F.g[k] << '\t'; } out << endl;
+ for (int k = 0; k < alignmentLength; k++) { out << F.c[k] << '\t'; } out << endl;
+ for (int k = 0; k < alignmentLength; k++) { out << F.gap[k] << '\t'; } out << endl;
+
+ //cout << F.getFilter() << endl;
+ out.close();
+
exit(0);
}else { m->mothurOut("unable to spawn the necessary processes."); m->mothurOutEndLine(); exit(0); }
}
+ //parent do your part
+ num = driverCreateFilter(F, filename, lines[0]);
+
//force parent to wait until all the processes are done
- for (int i=0;i<processors;i++) {
+ for (int i=0;i<(processors-1);i++) {
int temp = processIDS[i];
wait(&temp);
}
- return exitCommand;
-#endif
- }
- catch(exception& e) {
- m->errorOut(e, "FilterSeqsCommand", "createProcessesCreateFilter");
- exit(1);
- }
-}
-/**************************************************************************************************/
-
-int FilterSeqsCommand::setLines(string filename) {
- try {
- vector<int> positions;
- map<int, int> buf;
- bufferSizes.clear();
-
- int pid;
- MPI_Comm_rank(MPI_COMM_WORLD, &pid);
-
- ifstream inFASTA;
- openInputFile(filename, inFASTA);
+ //parent reads in and combines Filter info
+ for (int i = 0; i < processIDS.size(); i++) {
+ string tempFilename = filename + toString(processIDS[i]) + "filterValues.temp";
+ ifstream in;
+ m->openInputFile(tempFilename, in);
- string input;
- int numbuf = 0;
- while(!inFASTA.eof()){
- input = getline(inFASTA);
+ int temp, tempNum;
+ string tempFilterString;
- if (input.length() != 0) {
- numbuf += input.length();
- if(input[0] == '>'){ long int pos = inFASTA.tellg(); positions.push_back(pos - input.length() - 1); buf[(pos - input.length() - 1)] = numbuf; }
- }
- }
+ in >> tempNum; m->gobble(in); num += tempNum;
- inFASTA.close();
-
- int numFastaSeqs = positions.size();
-
- numSeqs += numFastaSeqs;
-
- int numSeqsPerProcessor = numFastaSeqs / processors;
-
- for (int i = 0; i < processors; i++) {
+ in >> tempFilterString;
+ F.mergeFilter(tempFilterString);
- long int startPos = positions[ i * numSeqsPerProcessor ];
- if(i == processors - 1){
- numSeqsPerProcessor = numFastaSeqs - i * numSeqsPerProcessor;
- bufferSizes.push_back(numbuf-buf[startPos]);
- }else{
- int myEnd = positions[ (i+1) * numSeqsPerProcessor ];
- bufferSizes.push_back(buf[myEnd]-buf[startPos]);
- }
- lines.push_back(new linePair(startPos, numSeqsPerProcessor));
- }
-
- return 0;
- }
- catch(exception& e) {
- m->errorOut(e, "FilterSeqsCommand", "setLines");
- exit(1);
- }
-}
-/**************************************************************************************************/
-int FilterSeqsCommand::parseBuffer(string file, vector<string>& seqs) {
- try {
-
- istringstream iss (file,istringstream::in);
- string name, seqstring;
-int pid;
-MPI_Comm_rank(MPI_COMM_WORLD, &pid);
- Sequence* seq34 = new Sequence();
-cout << "address of new sequence " << pid << '\t' << seq34 << endl;
-cout << "address of seqStrings " << pid << '\t' << &seqs << endl;
-
- while (iss) {
-
- if (m->control_pressed) { return 0; }
- cout << "here" << endl;
- Sequence* seq = new Sequence(iss);
- cout << "here1" << endl;
- gobble(iss);
- cout << seq->getName() << endl;
- if (seq->getName() != "") {
- seqs.push_back(seq->getAligned());
- }
- delete seq;
+ for (int k = 0; k < alignmentLength; k++) { in >> temp; F.a[k] += temp; } m->gobble(in);
+ for (int k = 0; k < alignmentLength; k++) { in >> temp; F.t[k] += temp; } m->gobble(in);
+ for (int k = 0; k < alignmentLength; k++) { in >> temp; F.g[k] += temp; } m->gobble(in);
+ for (int k = 0; k < alignmentLength; k++) { in >> temp; F.c[k] += temp; } m->gobble(in);
+ for (int k = 0; k < alignmentLength; k++) { in >> temp; F.gap[k] += temp; } m->gobble(in);
+
+ in.close();
+ remove(tempFilename.c_str());
}
- return 0;
+ return num;
+#endif
}
catch(exception& e) {
- m->errorOut(e, "FilterSeqsCommand", "parseBuffer");
+ m->errorOut(e, "FilterSeqsCommand", "createProcessesCreateFilter");
exit(1);
}
}
projects / mothur.git / blobdiff