*/
#include "preclustercommand.h"
-#include "sequenceparser.h"
#include "deconvolutecommand.h"
-//**********************************************************************************************************************
-inline bool comparePriority(seqPNode first, seqPNode second) { return (first.numIdentical > second.numIdentical); }
-
//**********************************************************************************************************************
vector<string> PreClusterCommand::setParameters(){
try {
CommandParameter pname("name", "InputTypes", "", "", "none", "none", "none",false,false); parameters.push_back(pname);
CommandParameter pgroup("group", "InputTypes", "", "", "none", "none", "none",false,false); parameters.push_back(pgroup);
CommandParameter pdiffs("diffs", "Number", "", "0", "", "", "",false,false); parameters.push_back(pdiffs);
- CommandParameter pbygroup("bygroup", "Boolean", "", "T", "", "", "",false,false); parameters.push_back(pbygroup);
+ CommandParameter pprocessors("processors", "Number", "", "1", "", "", "",false,false); parameters.push_back(pprocessors);
CommandParameter pinputdir("inputdir", "String", "", "", "", "", "",false,false); parameters.push_back(pinputdir);
CommandParameter poutputdir("outputdir", "String", "", "", "", "", "",false,false); parameters.push_back(poutputdir);
helpString += "The pre.cluster command outputs a new fasta and name file.\n";
helpString += "The pre.cluster command parameters are fasta, names and diffs. The fasta parameter is required. \n";
helpString += "The names parameter allows you to give a list of seqs that are identical. This file is 2 columns, first column is name or representative sequence, second column is a list of its identical sequences separated by commas.\n";
- helpString += "The group parameter allows you to provide a group file. \n";
+ helpString += "The group parameter allows you to provide a group file so you can cluster by group. \n";
helpString += "The diffs parameter allows you to specify maximum number of mismatched bases allowed between sequences in a grouping. The default is 1.\n";
- helpString += "The bygroup parameter allows you to indicate you whether you want to cluster sequences only within each group, default=T, when a groupfile is given. \n";
helpString += "The pre.cluster command should be in the following format: \n";
helpString += "pre.cluster(fasta=yourFastaFile, names=yourNamesFile, diffs=yourMaxDiffs) \n";
helpString += "Example pre.cluster(fasta=amazon.fasta, diffs=2).\n";
else { m->setNameFile(namefile); }
groupfile = validParameter.validFile(parameters, "group", true);
- if (groupfile == "not found") { groupfile = ""; }
+ if (groupfile == "not found") { groupfile = ""; bygroup = false; }
else if (groupfile == "not open") { abort = true; groupfile = ""; }
- else { m->setGroupFile(groupfile); }
+ else { m->setGroupFile(groupfile); bygroup = true; }
string temp = validParameter.validFile(parameters, "diffs", false); if(temp == "not found"){ temp = "1"; }
convert(temp, diffs);
- temp = validParameter.validFile(parameters, "bygroup", false);
- if (temp == "not found") {
- if (groupfile != "") { temp = "T"; }
- else { temp = "F"; }
- }
- bygroup = m->isTrue(temp);
+ temp = validParameter.validFile(parameters, "processors", false); if (temp == "not found"){ temp = m->getProcessors(); }
+ m->setProcessors(temp);
+ convert(temp, processors);
+
- if (bygroup && (groupfile == "")) { m->mothurOut("You cannot set bygroup=T, unless you provide a groupfile."); m->mothurOutEndLine(); abort=true; }
-
}
}
vector<string> groups = parser->getNamesOfGroups();
- //precluster each group
- for (int i = 0; i < groups.size(); i++) {
-
- start = time(NULL);
-
- if (m->control_pressed) { delete parser; m->mothurRemove(newFastaFile); m->mothurRemove(newNamesFile); return 0; }
-
- m->mothurOutEndLine(); m->mothurOut("Processing group " + groups[i] + ":"); m->mothurOutEndLine();
-
- map<string, string> thisNameMap;
- if (namefile != "") { thisNameMap = parser->getNameMap(groups[i]); }
- vector<Sequence> thisSeqs = parser->getSeqs(groups[i]);
-
- //fill alignSeqs with this groups info.
- int numSeqs = loadSeqs(thisNameMap, thisSeqs);
-
- if (m->control_pressed) { delete parser; m->mothurRemove(newFastaFile); m->mothurRemove(newNamesFile); return 0; }
-
- if (diffs > length) { m->mothurOut("Error: diffs is greater than your sequence length."); m->mothurOutEndLine(); return 0; }
-
- int count = process();
-
- if (m->control_pressed) { delete parser; m->mothurRemove(newFastaFile); m->mothurRemove(newNamesFile); return 0; }
-
- m->mothurOut("Total number of sequences before pre.cluster was " + toString(alignSeqs.size()) + "."); m->mothurOutEndLine();
- m->mothurOut("pre.cluster removed " + toString(count) + " sequences."); m->mothurOutEndLine(); m->mothurOutEndLine();
- printData(newFastaFile, newNamesFile);
-
- m->mothurOut("It took " + toString(time(NULL) - start) + " secs to cluster " + toString(numSeqs) + " sequences."); m->mothurOutEndLine();
-
- }
+//#if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
+ if(processors == 1) { driverGroups(parser, newFastaFile, newNamesFile, 0, groups.size(), groups); }
+ else { createProcessesGroups(parser, newFastaFile, newNamesFile, groups); }
+//#else
+// driverGroups(parser, newFastaFile, newNamesFile, 0, groups.size(), groups);
+//#endif
delete parser;
+ if (m->control_pressed) { m->mothurRemove(newFastaFile); m->mothurRemove(newNamesFile); return 0; }
+
//run unique.seqs for deconvolute results
string inputString = "fasta=" + newFastaFile;
if (namefile != "") { inputString += ", name=" + newNamesFile; }
m->mothurOut("/******************************************/"); m->mothurOutEndLine();
m->renameFile(filenames["fasta"][0], newFastaFile);
+
+ m->mothurOut("It took " + toString(time(NULL) - start) + " secs to run pre.cluster."); m->mothurOutEndLine();
}else {
if (namefile != "") { readNameFile(); }
}
}
/**************************************************************************************************/
+int PreClusterCommand::createProcessesGroups(SequenceParser* parser, string newFName, string newNName, vector<string> groups) {
+ try {
+
+ vector<int> processIDS;
+ int process = 1;
+ int num = 0;
+
+ //sanity check
+ if (groups.size() < processors) { processors = groups.size(); }
+
+ //divide the groups between the processors
+ vector<linePair> lines;
+ int numGroupsPerProcessor = groups.size() / processors;
+ for (int i = 0; i < processors; i++) {
+ int startIndex = i * numGroupsPerProcessor;
+ int endIndex = (i+1) * numGroupsPerProcessor;
+ if(i == (processors - 1)){ endIndex = groups.size(); }
+ lines.push_back(linePair(startIndex, endIndex));
+ }
+
+#if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
+
+ //loop through and create all the processes you want
+ while (process != processors) {
+ int pid = fork();
+
+ 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){
+ num = driverGroups(parser, newFName + toString(getpid()) + ".temp", newNName + toString(getpid()) + ".temp", lines[process].start, lines[process].end, groups);
+ 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);
+ }
+ }
+
+ //do my part
+ num = driverGroups(parser, newFName, newNName, lines[0].start, lines[0].end, groups);
+
+ //force parent to wait until all the processes are done
+ for (int i=0;i<processIDS.size();i++) {
+ int temp = processIDS[i];
+ wait(&temp);
+ }
+
+#else
+
+ //////////////////////////////////////////////////////////////////////////////////////////////////////
+ //Windows version shared memory, so be careful when passing variables through the preClusterData struct.
+ //Above fork() will clone, so memory is separate, but that's not the case with windows,
+ //////////////////////////////////////////////////////////////////////////////////////////////////////
+
+ vector<preClusterData*> pDataArray;
+ DWORD dwThreadIdArray[processors-1];
+ HANDLE hThreadArray[processors-1];
+
+ //Create processor worker threads.
+ for( int i=1; i<processors; i++ ){
+ // Allocate memory for thread data.
+ string extension = toString(i) + ".temp";
+
+ preClusterData* tempPreCluster = new preClusterData(fastafile, namefile, groupfile, (newFName+extension), (newNName+extension), groups, m, lines[i].start, lines[i].end, diffs, i);
+ pDataArray.push_back(tempPreCluster);
+ processIDS.push_back(i);
+
+ //MySeqSumThreadFunction is in header. It must be global or static to work with the threads.
+ //default security attributes, thread function name, argument to thread function, use default creation flags, returns the thread identifier
+ hThreadArray[i-1] = CreateThread(NULL, 0, MyPreclusterThreadFunction, pDataArray[i-1], 0, &dwThreadIdArray[i-1]);
+ }
+
+
+ //using the main process as a worker saves time and memory
+ num = driverGroups(parser, newFName, newNName, lines[0].start, lines[0].end, groups);
+
+ //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++){
+ CloseHandle(hThreadArray[i]);
+ delete pDataArray[i];
+ }
+
+#endif
+
+ //append output files
+ for(int i=0;i<processIDS.size();i++){
+ m->appendFiles((newFName + toString(processIDS[i]) + ".temp"), newFName);
+ m->mothurRemove((newFName + toString(processIDS[i]) + ".temp"));
+
+ m->appendFiles((newNName + toString(processIDS[i]) + ".temp"), newNName);
+ m->mothurRemove((newNName + toString(processIDS[i]) + ".temp"));
+ }
+
+ return num;
+
+ }
+ catch(exception& e) {
+ m->errorOut(e, "PreClusterCommand", "createProcessesGroups");
+ exit(1);
+ }
+}
+/**************************************************************************************************/
+int PreClusterCommand::driverGroups(SequenceParser* parser, string newFFile, string newNFile, int start, int end, vector<string> groups){
+ try {
+
+ int numSeqs = 0;
+
+ //precluster each group
+ for (int i = start; i < end; i++) {
+
+ start = time(NULL);
+
+ if (m->control_pressed) { return 0; }
+
+ m->mothurOutEndLine(); m->mothurOut("Processing group " + groups[i] + ":"); m->mothurOutEndLine();
+
+ map<string, string> thisNameMap;
+ if (namefile != "") { thisNameMap = parser->getNameMap(groups[i]); }
+ vector<Sequence> thisSeqs = parser->getSeqs(groups[i]);
+
+ //fill alignSeqs with this groups info.
+ numSeqs = loadSeqs(thisNameMap, thisSeqs);
+
+ if (m->control_pressed) { return 0; }
+
+ if (diffs > length) { m->mothurOut("Error: diffs is greater than your sequence length."); m->mothurOutEndLine(); m->control_pressed = true; return 0; }
+
+ int count = process();
+
+ if (m->control_pressed) { return 0; }
+
+ m->mothurOut("Total number of sequences before pre.cluster was " + toString(alignSeqs.size()) + "."); m->mothurOutEndLine();
+ m->mothurOut("pre.cluster removed " + toString(count) + " sequences."); m->mothurOutEndLine(); m->mothurOutEndLine();
+ printData(newFFile, newNFile);
+
+ m->mothurOut("It took " + toString(time(NULL) - start) + " secs to cluster " + toString(numSeqs) + " sequences."); m->mothurOutEndLine();
+
+ }
+
+ return numSeqs;
+ }
+ catch(exception& e) {
+ m->errorOut(e, "PreClusterCommand", "driverGroups");
+ exit(1);
+ }
+}
+/**************************************************************************************************/
int PreClusterCommand::process(){
try {
#include "command.hpp"
#include "sequence.hpp"
+#include "sequenceparser.h"
/************************************************************/
struct seqPNode {
~seqPNode() {}
};
/************************************************************/
+inline bool comparePriority(seqPNode first, seqPNode second) { return (first.numIdentical > second.numIdentical); }
+//************************************************************/
class PreClusterCommand : public Command {
void help() { m->mothurOut(getHelpString()); }
private:
- int diffs, length;
+
+ struct linePair {
+ int start;
+ int end;
+ linePair(int i, int j) : start(i), end(j) {}
+ };
+
+ int diffs, length, processors;
bool abort, bygroup;
string fastafile, namefile, outputDir, groupfile;
vector<seqPNode> alignSeqs; //maps the number of identical seqs to a sequence
void printData(string, string); //fasta filename, names file name
int process();
int loadSeqs(map<string, string>&, vector<Sequence>&);
+ int driverGroups(SequenceParser*, string, string, int, int, vector<string> groups);
+ int createProcessesGroups(SequenceParser*, string, string, vector<string>);
};
-/************************************************************/
+/**************************************************************************************************/
+//custom data structure for threads to use.
+// This is passed by void pointer so it can be any data type
+// that can be passed using a single void pointer (LPVOID).
+typedef struct preClusterData {
+ string fastafile;
+ string namefile;
+ string groupfile;
+ string newFName, newNName;
+ MothurOut* m;
+ int start;
+ int end;
+ int diffs, threadID;
+ vector<string> groups;
+
+ preClusterData(){}
+ preClusterData(string f, string n, string g, string nff, string nnf, vector<string> gr, MothurOut* mout, int st, int en, int d, int tid) {
+ fastafile = f;
+ namefile = n;
+ groupfile = g;
+ newFName = nff;
+ newNName = nnf;
+ m = mout;
+ start = st;
+ end = en;
+ diffs = d;
+ threadID = tid;
+ groups = gr;
+ }
+};
+/**************************************************************************************************/
+#if defined (__APPLE__) || (__MACH__) || (linux) || (__linux)
+#else
+static DWORD WINAPI MyPreclusterThreadFunction(LPVOID lpParam){
+ preClusterData* pDataArray;
+ pDataArray = (preClusterData*)lpParam;
+
+ try {
+
+ //parse fasta and name file by group
+ SequenceParser* parser;
+ if (pDataArray->namefile != "") { parser = new SequenceParser(pDataArray->groupfile, pDataArray->fastafile, pDataArray->namefile); }
+ else { parser = new SequenceParser(pDataArray->groupfile, pDataArray->fastafile); }
+
+ int numSeqs = 0;
+ vector<seqPNode> alignSeqs;
+ //clear out old files
+ ofstream outF; pDataArray->m->openOutputFile(pDataArray->newFName, outF); outF.close();
+ ofstream outN; pDataArray->m->openOutputFile(pDataArray->newNName, outN); outN.close();
+
+ //precluster each group
+ for (int k = pDataArray->start; k < pDataArray->end; k++) {
+
+ int start = time(NULL);
+
+ if (pDataArray->m->control_pressed) { delete parser; return 0; }
+
+ pDataArray->m->mothurOutEndLine(); pDataArray->m->mothurOut("Processing group " + pDataArray->groups[k] + ":"); pDataArray->m->mothurOutEndLine();
+
+ map<string, string> thisNameMap;
+ if (pDataArray->namefile != "") { thisNameMap = parser->getNameMap(pDataArray->groups[k]); }
+ vector<Sequence> thisSeqs = parser->getSeqs(pDataArray->groups[k]);
+
+ //fill alignSeqs with this groups info.
+ ////////////////////////////////////////////////////
+ //numSeqs = loadSeqs(thisNameMap, thisSeqs); same function below
+
+ int length = 0;
+ alignSeqs.clear();
+ map<string, string>::iterator it;
+ bool error = false;
+
+ for (int i = 0; i < thisSeqs.size(); i++) {
+
+ if (pDataArray->m->control_pressed) { delete parser; return 0; }
+
+ if (pDataArray->namefile != "") {
+ it = thisNameMap.find(thisSeqs[i].getName());
+
+ //should never be true since parser checks for this
+ if (it == thisNameMap.end()) { pDataArray->m->mothurOut(thisSeqs[i].getName() + " is not in your names file, please correct."); pDataArray->m->mothurOutEndLine(); error = true; }
+ else{
+ //get number of reps
+ int numReps = 1;
+ for(int j=0;j<(it->second).length();j++){
+ if((it->second)[j] == ','){ numReps++; }
+ }
+
+ seqPNode tempNode(numReps, thisSeqs[i], it->second);
+ alignSeqs.push_back(tempNode);
+ if (thisSeqs[i].getAligned().length() > length) { length = thisSeqs[i].getAligned().length(); }
+ }
+ }else { //no names file, you are identical to yourself
+ seqPNode tempNode(1, thisSeqs[i], thisSeqs[i].getName());
+ alignSeqs.push_back(tempNode);
+ if (thisSeqs[i].getAligned().length() > length) { length = thisSeqs[i].getAligned().length(); }
+ }
+ }
+
+ //sanity check
+ if (error) { pDataArray->m->control_pressed = true; }
+
+ thisSeqs.clear();
+ numSeqs = alignSeqs.size();
+
+ ////////////////////////////////////////////////////
+
+ if (pDataArray->m->control_pressed) { delete parser; return 0; }
+
+ if (pDataArray->diffs > length) { pDataArray->m->mothurOut("Error: diffs is greater than your sequence length."); pDataArray->m->mothurOutEndLine(); pDataArray->m->control_pressed = true; return 0; }
+
+ ////////////////////////////////////////////////////
+ //int count = process(); - same function below
+
+ //sort seqs by number of identical seqs
+ sort(alignSeqs.begin(), alignSeqs.end(), comparePriority);
+
+ int count = 0;
+
+ //think about running through twice...
+ for (int i = 0; i < numSeqs; i++) {
+
+ //are you active
+ // itActive = active.find(alignSeqs[i].seq.getName());
+
+ if (alignSeqs[i].active) { //this sequence has not been merged yet
+
+ //try to merge it with all smaller seqs
+ for (int j = i+1; j < numSeqs; j++) {
+
+ if (pDataArray->m->control_pressed) { delete parser; return 0; }
+
+ if (alignSeqs[j].active) { //this sequence has not been merged yet
+ //are you within "diff" bases
+ //int mismatch = calcMisMatches(alignSeqs[i].seq.getAligned(), alignSeqs[j].seq.getAligned());
+ int mismatch = 0;
+
+ for (int k = 0; k < alignSeqs[i].seq.getAligned().length(); k++) {
+ //do they match
+ if (alignSeqs[i].seq.getAligned()[k] != alignSeqs[j].seq.getAligned()[k]) { mismatch++; }
+ if (mismatch > pDataArray->diffs) { mismatch = length; break; } //to far to cluster
+ }
+
+ if (mismatch <= pDataArray->diffs) {
+ //merge
+ alignSeqs[i].names += ',' + alignSeqs[j].names;
+ alignSeqs[i].numIdentical += alignSeqs[j].numIdentical;
+
+ alignSeqs[j].active = 0;
+ alignSeqs[j].numIdentical = 0;
+ count++;
+ }
+ }//end if j active
+ }//end if i != j
+
+ //remove from active list
+ alignSeqs[i].active = 0;
+
+ }//end if active i
+ if(i % 100 == 0) { pDataArray->m->mothurOut(toString(i) + "\t" + toString(numSeqs - count) + "\t" + toString(count)); pDataArray->m->mothurOutEndLine(); }
+ }
+
+ if(numSeqs % 100 != 0) { pDataArray->m->mothurOut(toString(numSeqs) + "\t" + toString(numSeqs - count) + "\t" + toString(count)); pDataArray->m->mothurOutEndLine(); }
+ ////////////////////////////////////////////////////
+
+ if (pDataArray->m->control_pressed) { delete parser; return 0; }
+
+ pDataArray->m->mothurOut("Total number of sequences before pre.cluster was " + toString(alignSeqs.size()) + ".");pDataArray-> m->mothurOutEndLine();
+ pDataArray->m->mothurOut("pre.cluster removed " + toString(count) + " sequences."); pDataArray->m->mothurOutEndLine(); pDataArray->m->mothurOutEndLine();
+
+ ////////////////////////////////////////////////////
+ //printData(pDataArray->newFFile, pDataArray->newNFile); - same as below
+ ofstream outFasta;
+ ofstream outNames;
+
+ pDataArray->m->openOutputFileAppend(pDataArray->newFName, outFasta);
+ pDataArray->m->openOutputFileAppend(pDataArray->newNName, outNames);
+
+ for (int i = 0; i < alignSeqs.size(); i++) {
+ if (alignSeqs[i].numIdentical != 0) {
+ alignSeqs[i].seq.printSequence(outFasta);
+ outNames << alignSeqs[i].seq.getName() << '\t' << alignSeqs[i].names << endl;
+ }
+ }
+
+ outFasta.close();
+ outNames.close();
+ ////////////////////////////////////////////////////
+
+ pDataArray->m->mothurOut("It took " + toString(time(NULL) - start) + " secs to cluster " + toString(numSeqs) + " sequences."); pDataArray->m->mothurOutEndLine();
+
+ }
+
+ return numSeqs;
+
+ }
+ catch(exception& e) {
+ pDataArray->m->errorOut(e, "AlignCommand", "MyPreclusterThreadFunction");
+ exit(1);
+ }
+}
+#endif
+/**************************************************************************************************/
#endif