X-Git-Url: https://git.donarmstrong.com/?a=blobdiff_plain;f=weighted.cpp;h=cf1291dfa91af84e4d821541b2a8f43a169d64f7;hb=4a760c2d164aa955dee7d3d38da323822763d906;hp=1ac274619bf2c8971ea7f17ddb03c8cf98dde0ed;hpb=cbb336d5650cfe1f83a398a837794dcfbbbd3756;p=mothur.git diff --git a/weighted.cpp b/weighted.cpp index 1ac2746..cf1291d 100644 --- a/weighted.cpp +++ b/weighted.cpp @@ -13,15 +13,15 @@ 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(); + numGroups = m->getNumGroups(); if (m->control_pressed) { return data; } @@ -31,14 +31,14 @@ EstOutput Weighted::getValues(Tree* t, int p, string o) { for (int l = 0; l < i; l++) { //initialize weighted scores //WScore[globaldata->Groups[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 defined (__APPLE__) || (__MACH__) || (linux) || (__linux) || (__linux__) || (__unix__) || (__unix) if(processors == 1){ - data = driver(t, namesOfGroupCombos, 0, namesOfGroupCombos.size()); + data = driver(t, namesOfGroupCombos, 0, namesOfGroupCombos.size(), ct); }else{ int numPairs = namesOfGroupCombos.size(); @@ -52,12 +52,12 @@ EstOutput Weighted::getValues(Tree* t, int p, string o) { lines.push_back(linePair(startPos, numPairsPerProcessor)); } - data = createProcesses(t, namesOfGroupCombos); + data = createProcesses(t, namesOfGroupCombos, ct); lines.clear(); } #else - data = driver(t, namesOfGroupCombos, 0, namesOfGroupCombos.size()); + data = driver(t, namesOfGroupCombos, 0, namesOfGroupCombos.size(), ct); #endif return data; @@ -69,11 +69,10 @@ EstOutput Weighted::getValues(Tree* t, int p, string o) { } /**************************************************************************************************/ -EstOutput Weighted::createProcesses(Tree* t, vector< vector > namesOfGroupCombos) { +EstOutput Weighted::createProcesses(Tree* t, vector< vector > namesOfGroupCombos, CountTable* ct) { try { -#if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) +#if defined (__APPLE__) || (__MACH__) || (linux) || (__linux) || (__linux__) || (__unix__) || (__unix) int process = 1; - int num = 0; vector processIDS; EstOutput results; @@ -88,9 +87,9 @@ EstOutput Weighted::createProcesses(Tree* t, vector< vector > namesOfGro }else if (pid == 0){ EstOutput Myresults; - Myresults = driver(t, namesOfGroupCombos, lines[process].start, lines[process].num); + Myresults = driver(t, namesOfGroupCombos, lines[process].start, lines[process].num, ct); - m->mothurOut("Merging results."); m->mothurOutEndLine(); + //m->mothurOut("Merging results."); m->mothurOutEndLine(); //pass numSeqs to parent ofstream out; @@ -111,7 +110,7 @@ EstOutput Weighted::createProcesses(Tree* t, vector< vector > namesOfGro } } - results = driver(t, namesOfGroupCombos, lines[0].start, lines[0].num); + 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++) { @@ -142,10 +141,10 @@ EstOutput Weighted::createProcesses(Tree* t, vector< vector > namesOfGro m->gobble(in); } in.close(); - remove(s.c_str()); + m->mothurRemove(s); } - m->mothurOut("DONE."); m->mothurOutEndLine(); m->mothurOutEndLine(); + //m->mothurOut("DONE."); m->mothurOutEndLine(); m->mothurOutEndLine(); return results; #endif @@ -156,7 +155,7 @@ EstOutput Weighted::createProcesses(Tree* t, vector< vector > namesOfGro } } /**************************************************************************************************/ -EstOutput Weighted::driver(Tree* t, vector< vector > namesOfGroupCombos, int start, int num) { +EstOutput Weighted::driver(Tree* t, vector< vector > namesOfGroupCombos, int start, int num, CountTable* ct) { try { EstOutput results; vector D; @@ -180,7 +179,7 @@ EstOutput Weighted::driver(Tree* t, vector< vector > namesOfGroupCombos, int numSeqsInGroupI = it->second; double sum = getLengthToRoot(t, t->groupNodeInfo[groupA][j], groupA, groupB); - double weightedSum = ((numSeqsInGroupI * sum) / (double)tmap->seqsPerGroup[groupA]); + double weightedSum = ((numSeqsInGroupI * sum) / (double)ct->getGroupCount(groupA)); D[count] += weightedSum; } @@ -191,23 +190,18 @@ EstOutput Weighted::driver(Tree* t, vector< vector > namesOfGroupCombos, int numSeqsInGroupL = it->second; double sum = getLengthToRoot(t, t->groupNodeInfo[groupB][j], groupA, groupB); - double weightedSum = ((numSeqsInGroupL * sum) / (double)tmap->seqsPerGroup[groupB]); + double weightedSum = ((numSeqsInGroupL * sum) / (double)ct->getGroupCount(groupB)); D[count] += weightedSum; } count++; } - + //calculate u for the group comb - for (int h = start; h < (start+num); h++) { //report progress - m->mothurOut("Processing combo: " + toString(h)); m->mothurOutEndLine(); - - int numLeaves = t->getNumLeaves(); - map tempTotals; //maps node to total Branch Length - map nodePcountSize; //maps node to pcountSize - + //m->mothurOut("Processing combo: " + toString(h)); m->mothurOutEndLine(); + string groupA = namesOfGroupCombos[h][0]; string groupB = namesOfGroupCombos[h][1]; @@ -217,56 +211,39 @@ EstOutput Weighted::driver(Tree* t, vector< vector > namesOfGroupCombos, if (m->control_pressed) { return data; } double u; - int pcountSize = 0; + //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]; - pcountSize++; + 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]; - pcountSize++; + u -= (double) t->tree[i].pcount[groupB] / (double) ct->getGroupCount(groupB); } - u = abs(u * t->tree[i].getBranchLength()); - - nodePcountSize[i] = pcountSize; - - //if you are a leaf from a users group add to total - if (i < numLeaves) { - if ((t->tree[i].getBranchLength() != -1) && pcountSize != 0) { - //cout << "added to total" << endl; + if (includeRoot) { + if (t->tree[i].getBranchLength() != -1) { + u = abs(u * t->tree[i].getBranchLength()); WScore[(groupA+groupB)] += u; } - tempTotals[i] = 0.0; //we don't care about you, or we have already added you - }else{ //if you are not a leaf - //do both your chidren have have descendants from the users groups? - int lc = t->tree[i].getLChild(); - int rc = t->tree[i].getRChild(); - - //if yes, add your childrens tempTotals - if ((nodePcountSize[lc] != 0) && (nodePcountSize[rc] != 0)) { - WScore[(groupA+groupB)] += tempTotals[lc] + tempTotals[rc]; - //cout << "added to total " << tempTotals[lc] << '\t' << tempTotals[rc] << endl; + }else { + //if this is not the root then add it + if (rootForGrouping[namesOfGroupCombos[h]].count(i) == 0) { if (t->tree[i].getBranchLength() != -1) { - tempTotals[i] = u; - }else { - tempTotals[i] = 0.0; + u = abs(u * t->tree[i].getBranchLength()); + WScore[(groupA+groupB)] += u; } - }else if ((nodePcountSize[lc] == 0) && (nodePcountSize[rc] == 0)) { tempTotals[i] = 0.0; //we don't care about you - }else { //if no, your tempTotal is your childrens temp totals + your branch length - tempTotals[i] = tempTotals[lc] + tempTotals[rc] + u; } - //cout << "temptotal = "<< tempTotals[i] << endl; } } + } /********************************************************/ @@ -275,7 +252,6 @@ EstOutput Weighted::driver(Tree* t, vector< vector > namesOfGroupCombos, 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++; @@ -291,9 +267,10 @@ EstOutput Weighted::driver(Tree* t, vector< vector > namesOfGroupCombos, /**************************************************************************************************/ 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; } @@ -310,7 +287,7 @@ EstOutput Weighted::getValues(Tree* t, string groupA, string groupB) { int numSeqsInGroupI = it->second; double sum = getLengthToRoot(t, t->groupNodeInfo[groups[0]][j], groups[0], groups[1]); - double weightedSum = ((numSeqsInGroupI * sum) / (double)tmap->seqsPerGroup[groups[0]]); + double weightedSum = ((numSeqsInGroupI * sum) / (double)ct->getGroupCount(groups[0])); D += weightedSum; } @@ -321,74 +298,50 @@ EstOutput Weighted::getValues(Tree* t, string groupA, string groupB) { int numSeqsInGroupL = it->second; double sum = getLengthToRoot(t, t->groupNodeInfo[groups[1]][j], groups[0], groups[1]); - double weightedSum = ((numSeqsInGroupL * sum) / (double)tmap->seqsPerGroup[groups[1]]); + double weightedSum = ((numSeqsInGroupL * sum) / (double)ct->getGroupCount(groups[1])); D += weightedSum; } - - int numLeaves = t->getNumLeaves(); - map tempTotals; //maps node to total Branch Length - map nodePcountSize; //maps node to pcountSize - + //calculate u for the group comb for(int i=0;igetNumNodes();i++){ - + if (m->control_pressed) { return data; } double u; - int pcountSize = 0; + //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]; - pcountSize++; + 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]; - pcountSize++; + u -= (double) t->tree[i].pcount[groupB] / (double) ct->getGroupCount(groupB); } - - u = abs(u * t->tree[i].getBranchLength()); - nodePcountSize[i] = pcountSize; - - //if you are a leaf from a users group add to total - if (i < numLeaves) { - if ((t->tree[i].getBranchLength() != -1) && pcountSize != 0) { - //cout << "added to total" << endl; - WScore[(groupA+groupB)] += u; + if (includeRoot) { + if (t->tree[i].getBranchLength() != -1) { + u = abs(u * t->tree[i].getBranchLength()); + WScore[(groupA+groupB)] += u; } - tempTotals[i] = 0.0; //we don't care about you, or we have already added you - }else{ //if you are not a leaf - //do both your chidren have have descendants from the users groups? - int lc = t->tree[i].getLChild(); - int rc = t->tree[i].getRChild(); - - //if yes, add your childrens tempTotals - if ((nodePcountSize[lc] != 0) && (nodePcountSize[rc] != 0)) { - WScore[(groupA+groupB)] += tempTotals[lc] + tempTotals[rc]; - //cout << "added to total " << tempTotals[lc] << '\t' << tempTotals[rc] << endl; + }else{ + //if this is not the root then add it + if (rootForGrouping[groups].count(i) == 0) { if (t->tree[i].getBranchLength() != -1) { - tempTotals[i] = u; - }else { - tempTotals[i] = 0.0; + u = abs(u * t->tree[i].getBranchLength()); + WScore[(groupA+groupB)] += u; } - }else if ((nodePcountSize[lc] == 0) && (nodePcountSize[rc] == 0)) { tempTotals[i] = 0.0; //we don't care about you - }else { //if no, your tempTotal is your childrens temp totals + your branch length - tempTotals[i] = tempTotals[lc] + tempTotals[rc] + u; } - //cout << "temptotal = "<< tempTotals[i] << endl; } - } - + } /********************************************************/ - + //calculate weighted score for the group combination double UN; UN = (WScore[(groupA+groupB)] / D); @@ -408,72 +361,72 @@ double Weighted::getLengthToRoot(Tree* t, int v, string groupA, string groupB) { try { double sum = 0.0; - map tempTotals; //maps node to total Branch Length - map nodePcountSize; //maps node to pcountSize - map::iterator itCount; - int index = v; //you are a leaf if(t->tree[index].getBranchLength() != -1){ sum += abs(t->tree[index].getBranchLength()); } - tempTotals[index] = 0.0; + double tempTotal = 0.0; index = t->tree[index].getParent(); + + 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; } - - int pcountSize = 0; - map::iterator itGroup = t->tree[index].pcount.find(groupA); - if (itGroup != t->tree[index].pcount.end()) { pcountSize++; } - itGroup = t->tree[index].pcount.find(groupB); - if (itGroup != t->tree[index].pcount.end()) { pcountSize++; } - - nodePcountSize[index] = pcountSize; - - //do both your chidren have have descendants from the users groups? - int lc = t->tree[index].getLChild(); - int rc = t->tree[index].getRChild(); - itCount = nodePcountSize.find(lc); - if (itCount == nodePcountSize.end()) { - int LpcountSize = 0; - itGroup = t->tree[lc].pcount.find(groupA); - if (itGroup != t->tree[lc].pcount.end()) { LpcountSize++; } - itGroup = t->tree[lc].pcount.find(groupB); - if (itGroup != t->tree[lc].pcount.end()) { LpcountSize++; } - nodePcountSize[lc] = LpcountSize; - } + int parent = t->tree[index].getParent(); - itCount = nodePcountSize.find(rc); - if (itCount == nodePcountSize.end()) { - int RpcountSize = 0; - itGroup = t->tree[rc].pcount.find(groupA); - if (itGroup != t->tree[rc].pcount.end()) { RpcountSize++; } - itGroup = t->tree[rc].pcount.find(groupB); - if (itGroup != t->tree[rc].pcount.end()) { RpcountSize++; } - nodePcountSize[rc] = RpcountSize; - } + if (includeRoot) { //add everyone + if(t->tree[index].getBranchLength() != -1){ sum += abs(t->tree[index].getBranchLength()); } + }else { - //if yes, add your childrens tempTotals - if ((nodePcountSize[lc] != 0) && (nodePcountSize[rc] != 0)) { - sum += tempTotals[lc] + tempTotals[rc]; - - //cout << "added to total " << tempTotals[lc] << '\t' << tempTotals[rc] << endl; - if (t->tree[index].getBranchLength() != -1) { - tempTotals[index] = abs(t->tree[index].getBranchLength()); - }else { - tempTotals[index] = 0.0; + //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()); + } } - }else { //if no, your tempTotal is your childrens temp totals + your branch length - tempTotals[index] = tempTotals[lc] + tempTotals[rc] + abs(t->tree[index].getBranchLength()); } - //cout << "temptotal = "<< tempTotals[i] << endl; - index = t->tree[index].getParent(); + index = parent; + } + + //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) { @@ -484,4 +437,3 @@ double Weighted::getLengthToRoot(Tree* t, int v, string groupA, string groupB) { /**************************************************************************************************/ -