X-Git-Url: https://git.donarmstrong.com/?a=blobdiff_plain;f=distancecommand.cpp;h=05ed87cb960c5c61fff158384723ff9d3b2a1e3a;hb=058715887611e00dc18324cb0bd4a4a8472530b3;hp=13f56eb90a422d43e821a9fa378f582691468cb8;hpb=74844a60d80c6dd06e3fb02ee9b928424f9019b0;p=mothur.git diff --git a/distancecommand.cpp b/distancecommand.cpp index 13f56eb..05ed87c 100644 --- a/distancecommand.cpp +++ b/distancecommand.cpp @@ -32,7 +32,7 @@ DistanceCommand::DistanceCommand(string option) { OptionParser parser(option); map parameters = parser.getParameters(); - ValidParameters validParameter; + ValidParameters validParameter("dist.seqs"); map::iterator it2; //check to make sure all parameters are valid for command @@ -162,11 +162,13 @@ int DistanceCommand::execute(){ if (abort == true) { return 0; } + int startTime = time(NULL); + int numSeqs = alignDB.getNumSeqs(); cutoff += 0.005; string outputFile; - + if (output == "lt") { //does the user want lower triangle phylip formatted file outputFile = outputDir + getRootName(getSimpleName(fastafile)) + "phylip.dist"; remove(outputFile.c_str()); @@ -179,8 +181,129 @@ int DistanceCommand::execute(){ outputFile = outputDir + getRootName(getSimpleName(fastafile)) + "square.dist"; remove(outputFile.c_str()); } + + +#ifdef USE_MPI + + int pid, start, end; + int tag = 2001; -#if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) + 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 + + //each process gets where it should start and stop in the file + start = int (sqrt(float(pid)/float(processors)) * numSeqs); + end = int (sqrt(float(pid+1)/float(processors)) * numSeqs); + + if (output != "lt") { + MPI_File outMPI; + int amode=MPI_MODE_CREATE|MPI_MODE_WRONLY; + + //char* filename = new char[outputFile.length()]; + //memcpy(filename, outputFile.c_str(), outputFile.length()); + + char filename[1024]; + strcpy(filename, outputFile.c_str()); + + MPI_File_open(MPI_COMM_WORLD, filename, amode, MPI_INFO_NULL, &outMPI); + //delete filename; + + if (pid == 0) { //you are the root process + + //do your part + string outputMyPart; + driverMPI(start, end, outMPI, cutoff); + + if (m->control_pressed) { MPI_File_close(&outMPI); delete distCalculator; return 0; } + + //wait on chidren + for(int i = 1; i < processors; i++) { + if (m->control_pressed) { MPI_File_close(&outMPI); delete distCalculator; return 0; } + + char buf[4]; + MPI_Recv(buf, 4, MPI_CHAR, i, tag, MPI_COMM_WORLD, &status); + } + }else { //you are a child process + //do your part + driverMPI(start, end, outMPI, cutoff); + + if (m->control_pressed) { MPI_File_close(&outMPI); delete distCalculator; return 0; } + + char buf[4]; + strcpy(buf, "done"); + //tell parent you are done. + MPI_Send(buf, 4, MPI_CHAR, 0, tag, MPI_COMM_WORLD); + } + + MPI_File_close(&outMPI); + + }else { //lower triangle format + if (pid == 0) { //you are the root process + + //do your part + string outputMyPart; + long mySize; + driverMPI(start, end, outputFile, mySize); + + if (m->control_pressed) { delete distCalculator; return 0; } + + int amode=MPI_MODE_APPEND|MPI_MODE_WRONLY|MPI_MODE_CREATE; // + MPI_File outMPI; + MPI_File inMPI; + + //char* filename = new char[outputFile.length()]; + //memcpy(filename, outputFile.c_str(), outputFile.length()); + + char filename[1024]; + strcpy(filename, outputFile.c_str()); + + MPI_File_open(MPI_COMM_SELF, filename, amode, MPI_INFO_NULL, &outMPI); + //delete filename; + + //wait on chidren + for(int b = 1; b < processors; b++) { + long fileSize; + + if (m->control_pressed) { MPI_File_close(&outMPI); delete distCalculator; return 0; } + + MPI_Recv(&fileSize, 1, MPI_LONG, b, tag, MPI_COMM_WORLD, &status); + + string outTemp = outputFile + toString(b) + ".temp"; + + char* buf = new char[outTemp.length()]; + memcpy(buf, outTemp.c_str(), outTemp.length()); + + MPI_File_open(MPI_COMM_SELF, buf, MPI_MODE_DELETE_ON_CLOSE|MPI_MODE_RDONLY, MPI_INFO_NULL, &inMPI); + delete buf; + + int count = 0; + while (count < fileSize) { + char buf2[1]; + MPI_File_read(inMPI, buf2, 1, MPI_CHAR, &status); + MPI_File_write(outMPI, buf2, 1, MPI_CHAR, &status); + count += 1; + } + + MPI_File_close(&inMPI); //deleted on close + } + + MPI_File_close(&outMPI); + }else { //you are a child process + //do your part + long size; + driverMPI(start, end, (outputFile + toString(pid) + ".temp"), size); + + if (m->control_pressed) { delete distCalculator; return 0; } + + //tell parent you are done. + MPI_Send(&size, 1, MPI_LONG, 0, tag, MPI_COMM_WORLD); + } + } + MPI_Barrier(MPI_COMM_WORLD); //make everyone wait - just in case +#else + + #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) //if you don't need to fork anything if(processors == 1){ driver(0, numSeqs, outputFile, cutoff); @@ -204,20 +327,43 @@ int DistanceCommand::execute(){ remove((outputFile + toString(it->second) + ".temp").c_str()); } } -#else + #else ifstream inFASTA; driver(0, numSeqs, outputFile, cutoff); + #endif + #endif + if (m->control_pressed) { delete distCalculator; remove(outputFile.c_str()); return 0; } + + #ifdef USE_MPI + MPI_Comm_rank(MPI_COMM_WORLD, &pid); + + if (pid == 0) { //only one process should output to screen + #endif if (output == "square") { convertMatrix(outputFile); } + ifstream fileHandle; + fileHandle.open(outputFile.c_str()); + if(fileHandle) { + gobble(fileHandle); + if (fileHandle.eof()) { m->mothurOut(outputFile + " is blank. This can result if there are no distances below your cutoff."); m->mothurOutEndLine(); } + } + + + #ifdef USE_MPI + } + #endif + + if (m->control_pressed) { delete distCalculator; remove(outputFile.c_str()); return 0; } + delete distCalculator; m->mothurOutEndLine(); m->mothurOut("Output File Name: "); m->mothurOutEndLine(); m->mothurOut(outputFile); m->mothurOutEndLine(); m->mothurOutEndLine(); - + m->mothurOut("It took " + toString(time(NULL) - startTime) + " to calculate the distances for " + toString(numSeqs) + " sequences."); m->mothurOutEndLine(); return 0; } @@ -282,6 +428,9 @@ int DistanceCommand::driver(int startLine, int endLine, string dFileName, float outFile << name << '\t'; } for(int j=0;jcontrol_pressed) { outFile.close(); return 0; } + distCalculator->calcDist(alignDB.get(i), alignDB.get(j)); double dist = distCalculator->getDist(); @@ -315,8 +464,139 @@ int DistanceCommand::driver(int startLine, int endLine, string dFileName, float exit(1); } } +#ifdef USE_MPI +/**************************************************************************************************/ +/////// need to fix to work with calcs and sequencedb +int DistanceCommand::driverMPI(int startLine, int endLine, MPI_File& outMPI, float cutoff){ + try { + MPI_Status status; + int startTime = time(NULL); + + string outputString = ""; + + for(int i=startLine;icontrol_pressed) { return 0; } + + distCalculator->calcDist(alignDB.get(i), alignDB.get(j)); + double dist = distCalculator->getDist(); + + if(dist <= cutoff){ + if (output == "column") { outputString += (alignDB.get(i).getName() + ' ' + alignDB.get(j).getName() + ' ' + toString(dist) + '\n'); } + } + + if (output == "square") { //make a square column you can convert to square phylip + outputString += (alignDB.get(i).getName() + ' ' + alignDB.get(j).getName() + ' ' + toString(dist) + '\n'); + outputString += (alignDB.get(j).getName() + ' ' + alignDB.get(i).getName() + ' ' + toString(dist) + '\n'); + } + } + + if(i % 100 == 0){ + //m->mothurOut(toString(i) + "\t" + toString(time(NULL) - startTime)); m->mothurOutEndLine(); + cout << i << '\t' << (time(NULL) - startTime) << endl; + } + + + //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; + + } + + //m->mothurOut(toString(endLine-1) + "\t" + toString(time(NULL) - startTime)); m->mothurOutEndLine(); + cout << (endLine-1) << '\t' << (time(NULL) - startTime) << endl; + return 1; + } + catch(exception& e) { + m->errorOut(e, "DistanceCommand", "driverMPI"); + exit(1); + } +} +/**************************************************************************************************/ +/////// need to fix to work with calcs and sequencedb +int DistanceCommand::driverMPI(int startLine, int endLine, string file, long& size){ + try { + MPI_Status status; + + MPI_File outMPI; + int amode=MPI_MODE_CREATE|MPI_MODE_WRONLY; + + //char* filename = new char[file.length()]; + //memcpy(filename, file.c_str(), file.length()); + + char filename[1024]; + strcpy(filename, file.c_str()); + + MPI_File_open(MPI_COMM_SELF, filename, amode, MPI_INFO_NULL, &outMPI); + //delete filename; + + int startTime = time(NULL); + + string outputString = ""; + size = 0; + + if((output == "lt") && startLine == 0){ outputString += toString(alignDB.getNumSeqs()) + "\n"; } + + for(int i=startLine;icontrol_pressed) { return 0; } + + distCalculator->calcDist(alignDB.get(i), alignDB.get(j)); + double dist = distCalculator->getDist(); + + if (output == "lt") { outputString += toString(dist) + "\t"; } + } + + if (output == "lt") { outputString += "\n"; } + + + if(i % 100 == 0){ + //m->mothurOut(toString(i) + "\t" + toString(time(NULL) - startTime)); m->mothurOutEndLine(); + cout << i << '\t' << (time(NULL) - startTime) << endl; + } + + + //send results to parent + int length = outputString.length(); + char* buf = new char[length]; + memcpy(buf, outputString.c_str(), length); + + MPI_File_write(outMPI, buf, length, MPI_CHAR, &status); + size += outputString.length(); + outputString = ""; + delete buf; + } + + //m->mothurOut(toString(endLine-1) + "\t" + toString(time(NULL) - startTime)); m->mothurOutEndLine(); + cout << (endLine-1) << '\t' << (time(NULL) - startTime) << endl; + MPI_File_close(&outMPI); + + return 1; + } + catch(exception& e) { + m->errorOut(e, "DistanceCommand", "driverMPI"); + exit(1); + } +} +#endif /**************************************************************************************************/ -void DistanceCommand::convertMatrix(string outputFile) { +int DistanceCommand::convertMatrix(string outputFile) { try{ //sort file by first column so the distances for each row are together @@ -358,6 +638,8 @@ void DistanceCommand::convertMatrix(string outputFile) { //openInputFile(outfile, in); while(!in.eof()) { + if (m->control_pressed) { in.close(); remove(outfile.c_str()); out.close(); return 0; } + in >> first >> second >> dist; gobble(in); if (first != currentRow) { @@ -393,12 +675,108 @@ void DistanceCommand::convertMatrix(string outputFile) { remove(outfile.c_str()); + return 1; + } catch(exception& e) { m->errorOut(e, "DistanceCommand", "convertMatrix"); exit(1); } } +/**************************************************************************************************/ +int DistanceCommand::convertToLowerTriangle(string outputFile) { + try{ + + //sort file by first column so the distances for each row are together + string outfile = getRootName(outputFile) + "sorted.dist.temp"; + + //use the unix sort + #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) + string command = "sort -n " + outputFile + " -o " + outfile; + system(command.c_str()); + #else //sort using windows sort + string command = "sort " + outputFile + " /O " + outfile; + system(command.c_str()); + #endif + + + //output to new file distance for each row and save positions in file where new row begins + ifstream in; + openInputFile(outfile, in); + + ofstream out; + openOutputFile(outputFile, out); + + out.setf(ios::fixed, ios::floatfield); out.setf(ios::showpoint); + + out << alignDB.getNumSeqs() << endl; + + //get first currentRow + string first, currentRow, second; + float dist; + int i, j; + i = 0; j = 0; + map rowDists; //take advantage of the fact that maps are already sorted by key + map::iterator it; + + in >> first; + currentRow = first; + + rowDists[first] = 0.00; //distance to yourself is 0.0 + + in.seekg(0); + //openInputFile(outfile, in); + + while(!in.eof()) { + if (m->control_pressed) { in.close(); remove(outfile.c_str()); out.close(); return 0; } + + in >> first >> second >> dist; gobble(in); + + if (first != currentRow) { + //print out last row + out << currentRow << '\t'; //print name + + //print dists + for (it = rowDists.begin(); it != rowDists.end(); it++) { + if (j >= i) { break; } + out << it->second << '\t'; + j++; + } + out << endl; + + //start new row + currentRow = first; + rowDists.clear(); + rowDists[first] = 0.00; + rowDists[second] = dist; + j = 0; + i++; + }else{ + rowDists[second] = dist; + } + } + //print out last row + out << currentRow << '\t'; //print name + + //print dists + for (it = rowDists.begin(); it != rowDists.end(); it++) { + out << it->second << '\t'; + } + out << endl; + + in.close(); + out.close(); + + remove(outfile.c_str()); + + return 1; + + } + catch(exception& e) { + m->errorOut(e, "DistanceCommand", "convertToLowerTriangle"); + exit(1); + } +} /************************************************************************************************** void DistanceCommand::appendFiles(string temp, string filename) { try{