X-Git-Url: https://git.donarmstrong.com/?a=blobdiff_plain;f=unweighted.cpp;h=a9515d160a93c50813f61f5a54c1da6423cd4da3;hb=a98eb683e17d8e49583bf2d215ab7562a4cdca75;hp=c39e6629df1c2749928833fe24af65d48e853d64;hpb=4e3fb2fef1f40eb8224abeaed0f42c33a89b6e5e;p=mothur.git diff --git a/unweighted.cpp b/unweighted.cpp index c39e662..a9515d1 100644 --- a/unweighted.cpp +++ b/unweighted.cpp @@ -11,103 +11,28 @@ /**************************************************************************************************/ -EstOutput Unweighted::getValues(Tree* t) { +EstOutput Unweighted::getValues(Tree* t, int p, string o) { try { globaldata = GlobalData::getInstance(); - - vector groups; - double UniqueBL; //a branch length is unique if it's chidren are from the same group - double totalBL; //all branch lengths - double UW; //Unweighted Value = UniqueBL / totalBL; - map::iterator it; //iterator to traverse pgroups - map copyIpcount; - - + processors = p; + outputDir = o; + //if the users enters no groups then give them the score of all groups int numGroups = globaldata->Groups.size(); //calculate number of comparsions int numComp = 0; + vector< vector > namesOfGroupCombos; for (int r=0; r groups; groups.push_back(globaldata->Groups[r]); groups.push_back(globaldata->Groups[l]); + namesOfGroupCombos.push_back(groups); } } - - //numComp+1 for AB, AC, BC, ABC - data.resize(numComp+1,0); - - int count = 0; - for (int a=0; aGroups[a]); groups.push_back(globaldata->Groups[l]); - - for(int i=t->getNumLeaves();igetNumNodes();i++){ - - int lc = t->tree[i].getLChild(); //lc = vector index of left child - int rc = t->tree[i].getRChild(); //rc = vector index of right child - - /**********************************************************************/ - //This section adds in all lengths that are non leaf - - copyIpcount = t->tree[i].pcount; - for (it = copyIpcount.begin(); it != copyIpcount.end();) { - if (inUsersGroups(it->first, groups) != true) { - copyIpcount.erase(it++); - }else { it++; } - } - - //if i's children are from the same group then i's pcount size will be 1 - //if copyIpcount.size() = 0 they are from a branch that is entirely from a group the user doesn't want - if (copyIpcount.size() == 0) { } - else if ((t->tree[i].getBranchLength() != -1) && (copyIpcount.size() == 1)) { UniqueBL += abs(t->tree[i].getBranchLength()); } - - //add i's BL to total if it is from the groups the user wants - if ((t->tree[i].getBranchLength() != -1) && (copyIpcount.size() != 0)) { - totalBL += abs(t->tree[i].getBranchLength()); - } - - /**********************************************************************/ - //This section adds in all lengths that are leaf - - //if i's chidren are leaves - if (t->tree[rc].getRChild() == -1) { - //if rc is a valid group and rc has a BL - if ((inUsersGroups(t->tree[rc].getGroup(), groups) == true) && (t->tree[rc].getBranchLength() != -1)) { - UniqueBL += abs(t->tree[rc].getBranchLength()); - totalBL += abs(t->tree[rc].getBranchLength()); - } - } - - if (t->tree[lc].getLChild() == -1) { - //if lc is a valid group and lc has a BL - if ((inUsersGroups(t->tree[lc].getGroup(), groups) == true) && (t->tree[lc].getBranchLength() != -1)) { - UniqueBL += abs(t->tree[lc].getBranchLength()); - totalBL += abs(t->tree[lc].getBranchLength()); - } - } - - /**********************************************************************/ - } - - UW = (UniqueBL / totalBL); - - if (isnan(UW) || isinf(UW)) { UW = 0; } - - data[count] = UW; - count++; - groups.clear(); - } - } - if (numComp != 1) { + vector groups; if (numGroups == 0) { //get score for all users groups for (int i = 0; i < tmap->namesOfGroups.size(); i++) { @@ -115,187 +40,216 @@ EstOutput Unweighted::getValues(Tree* t) { groups.push_back(tmap->namesOfGroups[i]); } } + namesOfGroupCombos.push_back(groups); }else { for (int i = 0; i < globaldata->Groups.size(); i++) { groups.push_back(globaldata->Groups[i]); } + namesOfGroupCombos.push_back(groups); } - - UniqueBL=0.0000; //a branch length is unique if it's chidren are from the same group - totalBL = 0.00; //all branch lengths - UW = 0.00; //Unweighted Value = UniqueBL / totalBL; - copyIpcount.clear(); + } + + #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) + if(processors == 1){ + data = driver(t, namesOfGroupCombos, 0, namesOfGroupCombos.size()); + }else{ + int numPairs = namesOfGroupCombos.size(); - for(int i=t->getNumLeaves();igetNumNodes();i++){ - - int lc = t->tree[i].getLChild(); //lc = vector index of left child - int rc = t->tree[i].getRChild(); //rc = vector index of right child - - /**********************************************************************/ - //This section adds in all lengths that are non leaf + int numPairsPerProcessor = numPairs / processors; + + for (int i = 0; i < processors; i++) { + int startPos = i * numPairsPerProcessor; - copyIpcount = t->tree[i].pcount; - for (it = copyIpcount.begin(); it != copyIpcount.end();) { - if (inUsersGroups(it->first, groups) != true) { - copyIpcount.erase(it++); - }else { it++; } + if(i == processors - 1){ + numPairsPerProcessor = numPairs - i * numPairsPerProcessor; + } + + lines.push_back(linePair(startPos, numPairsPerProcessor)); } + + data = createProcesses(t, namesOfGroupCombos); + lines.clear(); + } + #else + data = driver(t, namesOfGroupCombos, 0, namesOfGroupCombos.size()); + #endif + + return data; + } + catch(exception& e) { + m->errorOut(e, "Unweighted", "getValues"); + exit(1); + } +} +/**************************************************************************************************/ + +EstOutput Unweighted::createProcesses(Tree* t, vector< vector > namesOfGroupCombos) { + try { +#if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) + int process = 1; + int num = 0; + vector processIDS; + + EstOutput results; + + //loop through and create all the processes you want + while (process != processors) { + int pid = fork(); - //if i's children are from the same group then i's pcount size will be 1 - //if copyIpcount.size() = 0 they are from a branch that is entirely from a group the user doesn't want - if (copyIpcount.size() == 0) { } - else if ((t->tree[i].getBranchLength() != -1) && (copyIpcount.size() == 1)) { UniqueBL += abs(t->tree[i].getBranchLength()); } + 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){ + EstOutput myresults; + myresults = driver(t, namesOfGroupCombos, lines[process].start, lines[process].num); + + if (m->control_pressed) { exit(0); } + + m->mothurOut("Merging results."); m->mothurOutEndLine(); + + //pass numSeqs to parent + ofstream out; + string tempFile = outputDir + toString(getpid()) + ".unweighted.results.temp"; + m->openOutputFile(tempFile, out); + out << myresults.size() << endl; + for (int i = 0; i < myresults.size(); i++) { out << myresults[i] << '\t'; } out << endl; + out.close(); + + exit(0); + }else { m->mothurOut("unable to spawn the necessary processes."); m->mothurOutEndLine(); exit(0); } + } + + results = driver(t, namesOfGroupCombos, lines[0].start, lines[0].num); + + //force parent to wait until all the processes are done + for (int i=0;i<(processors-1);i++) { + int temp = processIDS[i]; + wait(&temp); + } + + if (m->control_pressed) { return results; } + + //get data created by processes + for (int i=0;i<(processors-1);i++) { + ifstream in; + string s = outputDir + toString(processIDS[i]) + ".unweighted.results.temp"; + m->openInputFile(s, in); - //add i's BL to total if it is from the groups the user wants - if ((t->tree[i].getBranchLength() != -1) && (copyIpcount.size() != 0)) { - totalBL += abs(t->tree[i].getBranchLength()); + //get quantiles + if (!in.eof()) { + int num; + in >> num; m->gobble(in); + + if (m->control_pressed) { break; } + + double w; + for (int j = 0; j < num; j++) { + in >> w; + results.push_back(w); } + m->gobble(in); + } + in.close(); + remove(s.c_str()); + } + + m->mothurOut("DONE."); m->mothurOutEndLine(); m->mothurOutEndLine(); + + return results; +#endif + } + catch(exception& e) { + m->errorOut(e, "Unweighted", "createProcesses"); + exit(1); + } +} +/**************************************************************************************************/ +EstOutput Unweighted::driver(Tree* t, vector< vector > namesOfGroupCombos, int start, int num) { + try { + + EstOutput results; results.resize(num); + + int count = 0; + int total = num; + int twentyPercent = (total * 0.20); + if (twentyPercent == 0) { twentyPercent = 1; } + + for (int h = start; h < (start+num); h++) { + + if (m->control_pressed) { return results; } + + double UniqueBL=0.0000; //a branch length is unique if it's chidren are from the same group + double totalBL = 0.00; //all branch lengths + double UW = 0.00; //Unweighted Value = UniqueBL / totalBL; + + for(int i=0;igetNumNodes();i++){ - /**********************************************************************/ - //This section adds in all lengths that are leaf - - //if i's chidren are leaves - if (t->tree[rc].getRChild() == -1) { - //if rc is a valid group and rc has a BL - if ((inUsersGroups(t->tree[rc].getGroup(), groups) == true) && (t->tree[rc].getBranchLength() != -1)) { - UniqueBL += abs(t->tree[rc].getBranchLength()); - totalBL += abs(t->tree[rc].getBranchLength()); - } + if (m->control_pressed) { return data; } + + //pcountSize = 0, they are from a branch that is entirely from a group the user doesn't want + //pcountSize = 2, not unique to one group + //pcountSize = 1, unique to one group + + int pcountSize = 0; + for (int j = 0; j < namesOfGroupCombos[h].size(); j++) { + map::iterator itGroup = t->tree[i].pcount.find(namesOfGroupCombos[h][j]); + if (itGroup != t->tree[i].pcount.end()) { pcountSize++; if (pcountSize > 1) { break; } } } - - if (t->tree[lc].getLChild() == -1) { - //if lc is a valid group and lc has a BL - if ((inUsersGroups(t->tree[lc].getGroup(), groups) == true) && (t->tree[lc].getBranchLength() != -1)) { - UniqueBL += abs(t->tree[lc].getBranchLength()); - totalBL += abs(t->tree[lc].getBranchLength()); - } + + if (pcountSize == 0) { } + else if ((t->tree[i].getBranchLength() != -1) && (pcountSize == 1)) { UniqueBL += abs(t->tree[i].getBranchLength()); } + + if ((t->tree[i].getBranchLength() != -1) && (pcountSize != 0)) { + totalBL += abs(t->tree[i].getBranchLength()); } - - /**********************************************************************/ } - + UW = (UniqueBL / totalBL); if (isnan(UW) || isinf(UW)) { UW = 0; } - data[count] = UW; - } + results[count] = UW; + count++; - return data; - + //report progress + if((count % twentyPercent) == 0) { float tempOut = (count / (float)total); if (isnan(tempOut) || isinf(tempOut)) { tempOut = 0.0; } m->mothurOut("Percentage complete: " + toString((int(tempOut) * 100.0))); m->mothurOutEndLine(); } + } + + //report progress + if((count % twentyPercent) != 0) { float tempOut = (count / (float)total); if (isnan(tempOut) || isinf(tempOut)) { tempOut = 0.0; } m->mothurOut("Percentage complete: " + toString((int(tempOut) * 100.0))); m->mothurOutEndLine(); } + + return results; } catch(exception& e) { - errorOut(e, "Unweighted", "getValues"); + m->errorOut(e, "Unweighted", "driver"); exit(1); } } - /**************************************************************************************************/ -EstOutput Unweighted::getValues(Tree* t, string groupA, string groupB) { +EstOutput Unweighted::getValues(Tree* t, string groupA, string groupB, int p, string o) { try { - globaldata = GlobalData::getInstance(); - - vector groups; - double UniqueBL; //a branch length is unique if it's chidren are from the same group - double totalBL; //all branch lengths - double UW; //Unweighted Value = UniqueBL / totalBL; - map::iterator it; //iterator to traverse pgroups - map copyIpcount; - copyTree = new Tree; - + globaldata = GlobalData::getInstance(); + processors = p; + outputDir = o; + + //if the users enters no groups then give them the score of all groups int numGroups = globaldata->Groups.size(); - + //calculate number of comparsions int numComp = 0; + vector< vector > namesOfGroupCombos; for (int r=0; r groups; groups.push_back(globaldata->Groups[r]); groups.push_back(globaldata->Groups[l]); + namesOfGroupCombos.push_back(groups); } } - - //numComp+1 for AB, AC, BC, ABC - data.resize(numComp+1,0); - - int count = 0; - for (int a=0; agetCopy(t); - - //groups in this combo - groups.push_back(globaldata->Groups[a]); groups.push_back(globaldata->Groups[l]); - - //swap labels in the groups you want to compare - copyTree->assembleRandomUnifracTree(groups[0], groups[1]); - - //copyTree->createNewickFile("random"+groupA+toString(count)); - - for(int i=copyTree->getNumLeaves();igetNumNodes();i++){ - - int lc = copyTree->tree[i].getLChild(); //lc = vector index of left child - int rc = copyTree->tree[i].getRChild(); //rc = vector index of right child - - /**********************************************************************/ - //This section adds in all lengths that are non leaf - copyIpcount = copyTree->tree[i].pcount; - for (it = copyIpcount.begin(); it != copyIpcount.end();) { - if (inUsersGroups(it->first, groups) != true) { - copyIpcount.erase(it++); - }else { it++; } - } - - //if i's children are from the same group then i's pcount size will be 1 - //if copyIpcount.size() = 0 they are from a branch that is entirely from a group the user doesn't want - if (copyIpcount.size() == 0) { } - else if ((copyTree->tree[i].getBranchLength() != -1) && (copyIpcount.size() == 1)) { UniqueBL += abs(copyTree->tree[i].getBranchLength()); } - - //add i's BL to total if it is from the groups the user wants - if ((copyTree->tree[i].getBranchLength() != -1) && (copyIpcount.size() != 0)) { - totalBL += abs(copyTree->tree[i].getBranchLength()); - } - - /**********************************************************************/ - //This section adds in all lengths that are leaf - - //if i's chidren are leaves - if (copyTree->tree[rc].getRChild() == -1) { - //if rc is a valid group and rc has a BL - if ((inUsersGroups(copyTree->tree[rc].getGroup(), groups) == true) && (copyTree->tree[rc].getBranchLength() != -1)) { - UniqueBL += abs(copyTree->tree[rc].getBranchLength()); - totalBL += abs(copyTree->tree[rc].getBranchLength()); - } - } - - if (copyTree->tree[lc].getLChild() == -1) { - //if lc is a valid group and lc has a BL - if ((inUsersGroups(copyTree->tree[lc].getGroup(), groups) == true) && (copyTree->tree[lc].getBranchLength() != -1)) { - UniqueBL += abs(copyTree->tree[lc].getBranchLength()); - totalBL += abs(copyTree->tree[lc].getBranchLength()); - } - } - - /**********************************************************************/ - } - - UW = (UniqueBL / totalBL); - - if (isnan(UW) || isinf(UW)) { UW = 0; } - - data[count] = UW; - count++; - groups.clear(); - } - } - if (numComp != 1) { + vector groups; if (numGroups == 0) { //get score for all users groups for (int i = 0; i < tmap->namesOfGroups.size(); i++) { @@ -303,89 +257,190 @@ EstOutput Unweighted::getValues(Tree* t, string groupA, string groupB) { groups.push_back(tmap->namesOfGroups[i]); } } + namesOfGroupCombos.push_back(groups); }else { for (int i = 0; i < globaldata->Groups.size(); i++) { groups.push_back(globaldata->Groups[i]); } + namesOfGroupCombos.push_back(groups); } - - UniqueBL=0.0000; //a branch length is unique if it's chidren are from the same group - totalBL = 0.00; //all branch lengths - UW = 0.00; //Unweighted Value = UniqueBL / totalBL; - copyIpcount.clear(); - - //copy random tree passed in - copyTree->getCopy(t); + } + + #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) + if(processors == 1){ + data = driver(t, namesOfGroupCombos, 0, namesOfGroupCombos.size(), true); + }else{ + int numPairs = namesOfGroupCombos.size(); - //swap labels in all the groups you want to compare - copyTree->assembleRandomUnifracTree(groups); + int numPairsPerProcessor = numPairs / processors; + + for (int i = 0; i < processors; i++) { + int startPos = i * numPairsPerProcessor; + if(i == processors - 1){ + numPairsPerProcessor = numPairs - i * numPairsPerProcessor; + } + lines.push_back(linePair(startPos, numPairsPerProcessor)); + } - for(int i=copyTree->getNumLeaves();igetNumNodes();i++){ + data = createProcesses(t, namesOfGroupCombos, true); + + lines.clear(); + } + #else + data = driver(t, namesOfGroupCombos, 0, namesOfGroupCombos.size(), true); + #endif + + return data; + } + catch(exception& e) { + m->errorOut(e, "Unweighted", "getValues"); + exit(1); + } +} +/**************************************************************************************************/ + +EstOutput Unweighted::createProcesses(Tree* t, vector< vector > namesOfGroupCombos, bool usingGroups) { + try { +#if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) + int process = 1; + int num = 0; + vector processIDS; - int lc = copyTree->tree[i].getLChild(); //lc = vector index of left child - int rc = copyTree->tree[i].getRChild(); //rc = vector index of right child - - /**********************************************************************/ - //This section adds in all lengths that are non leaf - - copyIpcount = copyTree->tree[i].pcount; - for (it = copyIpcount.begin(); it != copyIpcount.end();) { - if (inUsersGroups(it->first, groups) != true) { - copyIpcount.erase(it++); - }else { it++; } - } + EstOutput results; + + //loop through and create all the processes you want + while (process != processors) { + int pid = fork(); - //if i's children are from the same group then i's pcount size will be 1 - //if copyIpcount.size() = 0 they are from a branch that is entirely from a group the user doesn't want - if (copyIpcount.size() == 0) { } - else if ((copyTree->tree[i].getBranchLength() != -1) && (copyIpcount.size() == 1)) { abs(UniqueBL += copyTree->tree[i].getBranchLength()); } + 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){ + EstOutput myresults; + myresults = driver(t, namesOfGroupCombos, lines[process].start, lines[process].num, usingGroups); + + if (m->control_pressed) { exit(0); } + + //pass numSeqs to parent + ofstream out; + string tempFile = outputDir + toString(getpid()) + ".unweighted.results.temp"; + m->openOutputFile(tempFile, out); + out << myresults.size() << endl; + for (int i = 0; i < myresults.size(); i++) { out << myresults[i] << '\t'; } out << endl; + out.close(); + + exit(0); + }else { m->mothurOut("unable to spawn the necessary processes."); m->mothurOutEndLine(); exit(0); } + } + + results = driver(t, namesOfGroupCombos, lines[0].start, lines[0].num, usingGroups); + + //force parent to wait until all the processes are done + for (int i=0;i<(processors-1);i++) { + int temp = processIDS[i]; + wait(&temp); + } + + if (m->control_pressed) { return results; } + + //get data created by processes + for (int i=0;i<(processors-1);i++) { + ifstream in; + string s = outputDir + toString(processIDS[i]) + ".unweighted.results.temp"; + m->openInputFile(s, in); - //add i's BL to total if it is from the groups the user wants - if ((copyTree->tree[i].getBranchLength() != -1) && (copyIpcount.size() != 0)) { - totalBL += abs(copyTree->tree[i].getBranchLength()); + //get quantiles + if (!in.eof()) { + int num; + in >> num; m->gobble(in); + + if (m->control_pressed) { break; } + + double w; + for (int j = 0; j < num; j++) { + in >> w; + results.push_back(w); } + m->gobble(in); + } + in.close(); + remove(s.c_str()); + } + + return results; +#endif + } + catch(exception& e) { + m->errorOut(e, "Unweighted", "createProcesses"); + exit(1); + } +} +/**************************************************************************************************/ +EstOutput Unweighted::driver(Tree* t, vector< vector > namesOfGroupCombos, int start, int num, bool usingGroups) { + try { + + EstOutput results; results.resize(num); + + int count = 0; + int total = num; + int twentyPercent = (total * 0.20); + + Tree* copyTree = new Tree; + + for (int h = start; h < (start+num); h++) { + + if (m->control_pressed) { return results; } + + //copy random tree passed in + copyTree->getCopy(t); + + //swap labels in the groups you want to compare + copyTree->assembleRandomUnifracTree(namesOfGroupCombos[h]); - /**********************************************************************/ - //This section adds in all lengths that are leaf + double UniqueBL=0.0000; //a branch length is unique if it's chidren are from the same group + double totalBL = 0.00; //all branch lengths + double UW = 0.00; //Unweighted Value = UniqueBL / totalBL; + + for(int i=0;igetNumNodes();i++){ - //if i's chidren are leaves - if (copyTree->tree[rc].getRChild() == -1) { - //if rc is a valid group and rc has a BL - if ((inUsersGroups(copyTree->tree[rc].getGroup(), groups) == true) && (copyTree->tree[rc].getBranchLength() != -1)) { - UniqueBL += abs(copyTree->tree[rc].getBranchLength()); - totalBL += abs(copyTree->tree[rc].getBranchLength()); - } + if (m->control_pressed) { return data; } + + //pcountSize = 0, they are from a branch that is entirely from a group the user doesn't want + //pcountSize = 2, not unique to one group + //pcountSize = 1, unique to one group + + int pcountSize = 0; + for (int j = 0; j < namesOfGroupCombos[h].size(); j++) { + map::iterator itGroup = t->tree[i].pcount.find(namesOfGroupCombos[h][j]); + if (itGroup != t->tree[i].pcount.end()) { pcountSize++; if (pcountSize > 1) { break; } } } - - if (copyTree->tree[lc].getLChild() == -1) { - //if lc is a valid group and lc has a BL - if ((inUsersGroups(copyTree->tree[lc].getGroup(), groups) == true) && (copyTree->tree[lc].getBranchLength() != -1)) { - UniqueBL += abs(copyTree->tree[lc].getBranchLength()); - totalBL += abs(copyTree->tree[lc].getBranchLength()); - } + + if (pcountSize == 0) { } + else if ((t->tree[i].getBranchLength() != -1) && (pcountSize == 1)) { UniqueBL += abs(t->tree[i].getBranchLength()); } + + if ((t->tree[i].getBranchLength() != -1) && (pcountSize != 0)) { + totalBL += abs(t->tree[i].getBranchLength()); } - - /**********************************************************************/ } UW = (UniqueBL / totalBL); if (isnan(UW) || isinf(UW)) { UW = 0; } - data[count] = UW; + results[count] = UW; + count++; + } delete copyTree; - return data; - + return results; } catch(exception& e) { - errorOut(e, "Unweighted", "getValues"); + m->errorOut(e, "Unweighted", "driver"); exit(1); } } - - +/**************************************************************************************************/