EstOutput Weighted::getValues(Tree* t, int p, string o) {
try {
- globaldata = GlobalData::getInstance();
-
data.clear(); //clear out old values
int numGroups;
vector<double> D;
processors = p;
outputDir = o;
+
+ CountTable* ct = t->getCountTable();
- numGroups = globaldata->Groups.size();
-
- vector<double> 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<string> > namesOfGroupCombos;
for (int i=0; i<numGroups; i++) {
- for (int l = i+1; l < numGroups; l++) {
+ for (int l = 0; l < i; l++) {
//initialize weighted scores
//WScore[globaldata->Groups[i]+globaldata->Groups[l]] = 0.0;
- vector<string> groups; groups.push_back(globaldata->Groups[i]); groups.push_back(globaldata->Groups[l]);
+ vector<string> 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(), sums);
+ data = driver(t, namesOfGroupCombos, 0, namesOfGroupCombos.size(), ct);
}else{
int numPairs = namesOfGroupCombos.size();
if(i == processors - 1){
numPairsPerProcessor = numPairs - i * numPairsPerProcessor;
}
- lines.push_back(new linePair(startPos, numPairsPerProcessor));
+ lines.push_back(linePair(startPos, numPairsPerProcessor));
}
- data = createProcesses(t, namesOfGroupCombos, sums);
+ data = createProcesses(t, namesOfGroupCombos, ct);
- for (int i = 0; i < lines.size(); i++) { delete lines[i]; } lines.clear();
+ lines.clear();
}
#else
- data = driver(t, namesOfGroupCombos, 0, namesOfGroupCombos.size(), sums);
+ data = driver(t, namesOfGroupCombos, 0, namesOfGroupCombos.size(), ct);
#endif
return data;
}
/**************************************************************************************************/
-EstOutput Weighted::createProcesses(Tree* t, vector< vector<string> > namesOfGroupCombos, vector<double>& sums) {
+EstOutput Weighted::createProcesses(Tree* t, vector< vector<string> > 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<int> processIDS;
EstOutput results;
}else if (pid == 0){
EstOutput Myresults;
- Myresults = driver(t, namesOfGroupCombos, lines[process]->start, lines[process]->num, sums);
+ 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;
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++) {
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
}
}
/**************************************************************************************************/
-EstOutput Weighted::driver(Tree* t, vector< vector<string> > namesOfGroupCombos, int start, int num, vector<double>& sums) {
+EstOutput Weighted::driver(Tree* t, vector< vector<string> > namesOfGroupCombos, int start, int num, CountTable* ct) {
try {
EstOutput results;
vector<double> D;
string groupA = namesOfGroupCombos[h][0];
string groupB = namesOfGroupCombos[h][1];
+ set<int> validBranches;
WScore[groupA+groupB] = 0.0;
D.push_back(0.0000); //initialize a spot in D for each combination
map<string, int>::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;
}
map<string, int>::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++;
}
-
+
//calculate u for the group comb
- int total = t->getNumNodes();
- int twentyPercent = (total * 0.20);
- for(int i=0;i<t->getNumNodes();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;i<t->getNumNodes();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;
}
- //report progress
- if((i) % twentyPercent == 0){ m->mothurOut("Percentage complete: " + toString(int((i / (float)total) * 100.0))); m->mothurOutEndLine(); }
+
}
- m->mothurOut("Percentage complete: 100"); m->mothurOutEndLine();
-
/********************************************************/
//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++;
}
}
/**************************************************************************************************/
-EstOutput Weighted::getValues(Tree* t, string groupA, string groupB, vector<double>& 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<int> validBranches;
vector<string> groups; groups.push_back(groupA); groups.push_back(groupB);
map<string, int>::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;
}
map<string, int>::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;i<t->getNumNodes();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);
}
}
/**************************************************************************************************/
-vector<double> Weighted::getBranchLengthSums(Tree* t) {
+double Weighted::getLengthToRoot(Tree* t, int v, string groupA, string groupB) {
try {
- vector<double> 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();
+
+ vector<string> grouping; grouping.push_back(groupA); grouping.push_back(groupB);
- for(int v=0;v<t->getNumLeaves();v++){
+ rootForGrouping[grouping].insert(index);
- if (m->control_pressed) { return sums; }
+ //while you aren't at root
+ while(t->tree[index].getParent() != -1){
+
+ if (m->control_pressed) { return sum; }
+
+ 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<string, int>::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");
/**************************************************************************************************/
-
projects / mothur.git / blobdiff