X-Git-Url: https://git.donarmstrong.com/?p=mothur.git;a=blobdiff_plain;f=weighted.cpp;h=b0d06fb0078e0201975e7738d1321e7932230253;hp=c9ad81295aaacb3bd4ae7eb8f7e8ca7c1cef9b80;hb=df7e3ff9f68ef157b0328a2d353c3258c5d45d89;hpb=bd93b1a6f9fe9a6a4a7ac2e9f106e5c83a438856 diff --git a/weighted.cpp b/weighted.cpp index c9ad812..b0d06fb 100644 --- a/weighted.cpp +++ b/weighted.cpp @@ -13,55 +13,42 @@ EstOutput Weighted::getValues(Tree* t, int p, string o) { try { - globaldata = GlobalData::getInstance(); - data.clear(); //clear out old values int numGroups; vector D; processors = p; outputDir = o; + + CountTable* ct = t->getCountTable(); - numGroups = globaldata->Groups.size(); - - vector sums = getBranchLengthSums(t); + numGroups = m->getNumGroups(); if (m->control_pressed) { return data; } //calculate number of comparisons i.e. with groups A,B,C = AB, AC, BC = 3; vector< vector > namesOfGroupCombos; for (int i=0; iGroups[i]+globaldata->Groups[l]] = 0.0; - vector groups; groups.push_back(globaldata->Groups[i]); groups.push_back(globaldata->Groups[l]); + vector groups; groups.push_back((m->getGroups())[i]); groups.push_back((m->getGroups())[l]); namesOfGroupCombos.push_back(groups); } } - #if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) - if(processors == 1){ - data = driver(t, namesOfGroupCombos, 0, namesOfGroupCombos.size(), sums); - }else{ - int numPairs = namesOfGroupCombos.size(); - - 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(new linePair(startPos, numPairsPerProcessor)); - } + int numPairs = namesOfGroupCombos.size(); + 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)); + } + + data = createProcesses(t, namesOfGroupCombos, ct); + + lines.clear(); - data = createProcesses(t, namesOfGroupCombos, sums); - - for (int i = 0; i < lines.size(); i++) { delete lines[i]; } lines.clear(); - } - #else - data = driver(t, namesOfGroupCombos, 0, namesOfGroupCombos.size(), sums); - #endif - return data; } catch(exception& e) { @@ -71,14 +58,12 @@ EstOutput Weighted::getValues(Tree* t, int p, string o) { } /**************************************************************************************************/ -EstOutput Weighted::createProcesses(Tree* t, vector< vector > namesOfGroupCombos, vector& sums) { +EstOutput Weighted::createProcesses(Tree* t, vector< vector > namesOfGroupCombos, CountTable* ct) { try { -#if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) - int process = 1; - int num = 0; - vector processIDS; - + vector processIDS; EstOutput results; +#if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) || (__linux__) || (__unix__) || (__unix) + int process = 1; //loop through and create all the processes you want while (process != processors) { @@ -89,32 +74,33 @@ EstOutput Weighted::createProcesses(Tree* t, vector< vector > namesOfGro process++; }else if (pid == 0){ EstOutput Myresults; - Myresults = driver(t, namesOfGroupCombos, lines[process]->start, lines[process]->num, sums); - - if (m->control_pressed) { exit(0); } - - m->mothurOut("Merging results."); m->mothurOutEndLine(); - + Myresults = driver(t, namesOfGroupCombos, lines[process].start, lines[process].num, ct); + //pass numSeqs to parent ofstream out; string tempFile = outputDir + toString(getpid()) + ".weighted.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); } + }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); + } } - results = driver(t, namesOfGroupCombos, lines[0]->start, lines[0]->num, sums); - + results = driver(t, namesOfGroupCombos, lines[0].start, lines[0].num, ct); + //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 @@ -138,13 +124,50 @@ EstOutput Weighted::createProcesses(Tree* t, vector< vector > namesOfGro m->gobble(in); } in.close(); - remove(s.c_str()); + m->mothurRemove(s); + } +#else + + //fill in functions + vector pDataArray; + DWORD dwThreadIdArray[processors-1]; + HANDLE hThreadArray[processors-1]; + vector cts; + vector trees; + + //Create processor worker threads. + for( int i=1; icopy(ct); + Tree* copyTree = new Tree(copyCount); + copyTree->getCopy(t); + + cts.push_back(copyCount); + trees.push_back(copyTree); + + weightedData* tempweighted = new weightedData(m, lines[i].start, lines[i].num, namesOfGroupCombos, copyTree, copyCount, includeRoot); + pDataArray.push_back(tempweighted); + processIDS.push_back(i); + + hThreadArray[i-1] = CreateThread(NULL, 0, MyWeightedThreadFunction, pDataArray[i-1], 0, &dwThreadIdArray[i-1]); } - m->mothurOut("DONE."); m->mothurOutEndLine(); m->mothurOutEndLine(); - - return results; -#endif + results = driver(t, namesOfGroupCombos, lines[0].start, lines[0].num, ct); + + //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++){ + for (int j = 0; j < pDataArray[i]->results.size(); j++) { results.push_back(pDataArray[i]->results[j]); } + delete cts[i]; + delete trees[i]; + CloseHandle(hThreadArray[i]); + delete pDataArray[i]; + } +#endif + + return results; } catch(exception& e) { m->errorOut(e, "Weighted", "createProcesses"); @@ -152,17 +175,12 @@ EstOutput Weighted::createProcesses(Tree* t, vector< vector > namesOfGro } } /**************************************************************************************************/ -EstOutput Weighted::driver(Tree* t, vector< vector > namesOfGroupCombos, int start, int num, vector& sums) { +EstOutput Weighted::driver(Tree* t, vector< vector > namesOfGroupCombos, int start, int num, CountTable* ct) { try { - globaldata = GlobalData::getInstance(); - EstOutput results; vector D; int count = 0; - int total = num; - int twentyPercent = (total * 0.20); - for (int h = start; h < (start+num); h++) { if (m->control_pressed) { return results; } @@ -171,6 +189,7 @@ EstOutput Weighted::driver(Tree* t, vector< vector > namesOfGroupCombos, string groupA = namesOfGroupCombos[h][0]; string groupB = namesOfGroupCombos[h][1]; + set validBranches; WScore[groupA+groupB] = 0.0; D.push_back(0.0000); //initialize a spot in D for each combination @@ -179,7 +198,8 @@ EstOutput Weighted::driver(Tree* t, vector< vector > namesOfGroupCombos, map::iterator it = t->tree[t->groupNodeInfo[groupA][j]].pcount.find(groupA); int numSeqsInGroupI = it->second; - double weightedSum = ((numSeqsInGroupI * sums[t->groupNodeInfo[groupA][j]]) / (double)tmap->seqsPerGroup[groupA]); + double sum = getLengthToRoot(t, t->groupNodeInfo[groupA][j], groupA, groupB); + double weightedSum = ((numSeqsInGroupI * sum) / (double)ct->getGroupCount(groupA)); D[count] += weightedSum; } @@ -189,58 +209,69 @@ EstOutput Weighted::driver(Tree* t, vector< vector > namesOfGroupCombos, map::iterator it = t->tree[t->groupNodeInfo[groupB][j]].pcount.find(groupB); int numSeqsInGroupL = it->second; - double weightedSum = ((numSeqsInGroupL * sums[t->groupNodeInfo[groupB][j]]) / (double)tmap->seqsPerGroup[groupB]); + double sum = getLengthToRoot(t, t->groupNodeInfo[groupB][j], groupA, groupB); + double weightedSum = ((numSeqsInGroupL * sum) / (double)ct->getGroupCount(groupB)); D[count] += weightedSum; } count++; - - //report progress - if((count) % twentyPercent == 0){ m->mothurOut("Percentage complete: " + toString(int((count / (float)total) * 100.0))); m->mothurOutEndLine(); } } - - m->mothurOut("Percentage complete: 100"); m->mothurOutEndLine(); - + //calculate u for the group comb - for(int i=0;igetNumNodes();i++){ - for (int h = start; h < (start+num); h++) { + for (int h = start; h < (start+num); h++) { + //report progress + //m->mothurOut("Processing combo: " + toString(h)); m->mothurOutEndLine(); + + string groupA = namesOfGroupCombos[h][0]; + string groupB = namesOfGroupCombos[h][1]; + + //calculate u for the group comb + for(int i=0;igetNumNodes();i++){ - string groupA = namesOfGroupCombos[h][0]; - string groupB = namesOfGroupCombos[h][1]; - - if (m->control_pressed) { return results; } + if (m->control_pressed) { return data; } double u; + //int pcountSize = 0; //does this node have descendants from groupA it = t->tree[i].pcount.find(groupA); //if it does u = # of its descendants with a certain group / total number in tree with a certain group if (it != t->tree[i].pcount.end()) { - u = (double) t->tree[i].pcount[groupA] / (double) tmap->seqsPerGroup[groupA]; + u = (double) t->tree[i].pcount[groupA] / (double) ct->getGroupCount(groupA); }else { u = 0.00; } - - + + //does this node have descendants from group l it = t->tree[i].pcount.find(groupB); + //if it does subtract their percentage from u if (it != t->tree[i].pcount.end()) { - u -= (double) t->tree[i].pcount[groupB] / (double) tmap->seqsPerGroup[groupB]; + u -= (double) t->tree[i].pcount[groupB] / (double) ct->getGroupCount(groupB); + } + + if (includeRoot) { + if (t->tree[i].getBranchLength() != -1) { + u = abs(u * t->tree[i].getBranchLength()); + WScore[(groupA+groupB)] += u; + } + }else { + //if this is not the root then add it + if (rootForGrouping[namesOfGroupCombos[h]].count(i) == 0) { + if (t->tree[i].getBranchLength() != -1) { + u = abs(u * t->tree[i].getBranchLength()); + WScore[(groupA+groupB)] += u; + } + } } - - u = abs(u * t->tree[i].getBranchLength()); - - //save groupcombs u value - WScore[(groupA+groupB)] += u; } + } /********************************************************/ - //calculate weighted score for the group combination double UN; count = 0; for (int h = start; h < (start+num); h++) { UN = (WScore[namesOfGroupCombos[h][0]+namesOfGroupCombos[h][1]] / D[count]); - if (isnan(UN) || isinf(UN)) { UN = 0; } results.push_back(UN); count++; @@ -254,17 +285,19 @@ EstOutput Weighted::driver(Tree* t, vector< vector > namesOfGroupCombos, } } /**************************************************************************************************/ -EstOutput Weighted::getValues(Tree* t, string groupA, string groupB, vector& sums) { +EstOutput Weighted::getValues(Tree* t, string groupA, string groupB) { try { - globaldata = GlobalData::getInstance(); data.clear(); //clear out old values + + CountTable* ct = t->getCountTable(); if (m->control_pressed) { return data; } //initialize weighted score WScore[(groupA+groupB)] = 0.0; double D = 0.0; + set validBranches; vector groups; groups.push_back(groupA); groups.push_back(groupB); @@ -273,7 +306,8 @@ EstOutput Weighted::getValues(Tree* t, string groupA, string groupB, vector::iterator it = t->tree[t->groupNodeInfo[groups[0]][j]].pcount.find(groups[0]); int numSeqsInGroupI = it->second; - double weightedSum = ((numSeqsInGroupI * sums[t->groupNodeInfo[groups[0]][j]]) / (double)tmap->seqsPerGroup[groups[0]]); + double sum = getLengthToRoot(t, t->groupNodeInfo[groups[0]][j], groups[0], groups[1]); + double weightedSum = ((numSeqsInGroupI * sum) / (double)ct->getGroupCount(groups[0])); D += weightedSum; } @@ -283,40 +317,51 @@ EstOutput Weighted::getValues(Tree* t, string groupA, string groupB, vector::iterator it = t->tree[t->groupNodeInfo[groups[1]][j]].pcount.find(groups[1]); int numSeqsInGroupL = it->second; - double weightedSum = ((numSeqsInGroupL * sums[t->groupNodeInfo[groups[1]][j]]) / (double)tmap->seqsPerGroup[groups[1]]); + double sum = getLengthToRoot(t, t->groupNodeInfo[groups[1]][j], groups[0], groups[1]); + double weightedSum = ((numSeqsInGroupL * sum) / (double)ct->getGroupCount(groups[1])); D += weightedSum; } - + //calculate u for the group comb for(int i=0;igetNumNodes();i++){ - + if (m->control_pressed) { return data; } double u; + //int pcountSize = 0; //does this node have descendants from groupA it = t->tree[i].pcount.find(groupA); //if it does u = # of its descendants with a certain group / total number in tree with a certain group if (it != t->tree[i].pcount.end()) { - u = (double) t->tree[i].pcount[groupA] / (double) tmap->seqsPerGroup[groupA]; + u = (double) t->tree[i].pcount[groupA] / (double) ct->getGroupCount(groupA); }else { u = 0.00; } - - + + //does this node have descendants from group l it = t->tree[i].pcount.find(groupB); //if it does subtract their percentage from u if (it != t->tree[i].pcount.end()) { - u -= (double) t->tree[i].pcount[groupB] / (double) tmap->seqsPerGroup[groupB]; + u -= (double) t->tree[i].pcount[groupB] / (double) ct->getGroupCount(groupB); } - - u = abs(u * t->tree[i].getBranchLength()); - - //save groupcombs u value - WScore[(groupA+groupB)] += u; - } - + + if (includeRoot) { + if (t->tree[i].getBranchLength() != -1) { + u = abs(u * t->tree[i].getBranchLength()); + WScore[(groupA+groupB)] += u; + } + }else{ + //if this is not the root then add it + if (rootForGrouping[groups].count(i) == 0) { + if (t->tree[i].getBranchLength() != -1) { + u = abs(u * t->tree[i].getBranchLength()); + WScore[(groupA+groupB)] += u; + } + } + } + } /********************************************************/ - + //calculate weighted score for the group combination double UN; UN = (WScore[(groupA+groupB)] / D); @@ -332,37 +377,77 @@ EstOutput Weighted::getValues(Tree* t, string groupA, string groupB, vector Weighted::getBranchLengthSums(Tree* t) { +double Weighted::getLengthToRoot(Tree* t, int v, string groupA, string groupB) { try { - vector sums; + double sum = 0.0; + int index = v; + + //you are a leaf + if(t->tree[index].getBranchLength() != -1){ sum += abs(t->tree[index].getBranchLength()); } + double tempTotal = 0.0; + index = t->tree[index].getParent(); - for(int v=0;vgetNumLeaves();v++){ + vector grouping; grouping.push_back(groupA); grouping.push_back(groupB); + + rootForGrouping[grouping].insert(index); + + //while you aren't at root + while(t->tree[index].getParent() != -1){ + + if (m->control_pressed) { return sum; } - if (m->control_pressed) { return sums; } + int parent = t->tree[index].getParent(); + + if (includeRoot) { //add everyone + if(t->tree[index].getBranchLength() != -1){ sum += abs(t->tree[index].getBranchLength()); } + }else { - int index = v; - double sum = 0.0000; - - //while you aren't at root - while(t->tree[index].getParent() != -1){ - - //if you have a BL - if(t->tree[index].getBranchLength() != -1){ - sum += abs(t->tree[index].getBranchLength()); + //am I the root for this grouping? if so I want to stop "early" + //does my sibling have descendants from the users groups? + int lc = t->tree[parent].getLChild(); + int rc = t->tree[parent].getRChild(); + + int sib = lc; + if (lc == index) { sib = rc; } + + map::iterator itGroup; + int pcountSize = 0; + itGroup = t->tree[sib].pcount.find(groupA); + if (itGroup != t->tree[sib].pcount.end()) { pcountSize++; } + itGroup = t->tree[sib].pcount.find(groupB); + if (itGroup != t->tree[sib].pcount.end()) { pcountSize++; } + + //if yes, I am not the root so add me + if (pcountSize != 0) { + if (t->tree[index].getBranchLength() != -1) { + sum += abs(t->tree[index].getBranchLength()) + tempTotal; + tempTotal = 0.0; + }else { + sum += tempTotal; + tempTotal = 0.0; + } + rootForGrouping[grouping].clear(); + rootForGrouping[grouping].insert(parent); + }else { //if no, I may be the root so add my br to tempTotal until I am proven innocent + if (t->tree[index].getBranchLength() != -1) { + tempTotal += abs(t->tree[index].getBranchLength()); + } } - index = t->tree[index].getParent(); - } - - //get last breanch length added - if(t->tree[index].getBranchLength() != -1){ - sum += abs(t->tree[index].getBranchLength()); } - sums.push_back(sum); + index = parent; } - return sums; + //get all nodes above the root to add so we don't add their u values above + index = *(rootForGrouping[grouping].begin()); + + while(t->tree[index].getParent() != -1){ + int parent = t->tree[index].getParent(); + rootForGrouping[grouping].insert(parent); + index = parent; + } + return sum; } catch(exception& e) { m->errorOut(e, "Weighted", "getBranchLengthSums"); @@ -372,4 +457,3 @@ vector Weighted::getBranchLengthSums(Tree* t) { /**************************************************************************************************/ -