modified filter.seqs to not store all seqs in memory but to read off disc

diff --git a/Mothur.xcodeproj/project.pbxproj b/Mothur.xcodeproj/project.pbxproj
index ec6d06cfceaaa938c834bf17be5a77bca9d14aea..2bdb5b97e8c572bdf7f23440dd7e80c9fcb8c10a 100644 (file)
--- a/Mothur.xcodeproj/project.pbxproj
+++ b/Mothur.xcodeproj/project.pbxproj
@@ -189,8 +189,8 @@
                21E859D70FC4632E005E1A48 /* matrixoutputcommand.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = matrixoutputcommand.cpp; sourceTree = SOURCE_ROOT; };
                370B88050F8A4EE4005AB382 /* getoturepcommand.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = getoturepcommand.h; sourceTree = SOURCE_ROOT; };
                370B88060F8A4EE4005AB382 /* getoturepcommand.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = getoturepcommand.cpp; sourceTree = SOURCE_ROOT; };
-               371B30B20FD7EE67000414CA /* screenseqscommand.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = screenseqscommand.cpp; sourceTree = "<group>"; };
-               371B30B30FD7EE67000414CA /* screenseqscommand.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = screenseqscommand.h; sourceTree = "<group>"; };
+               371B30B20FD7EE67000414CA /* screenseqscommand.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = screenseqscommand.cpp; sourceTree = SOURCE_ROOT; };
+               371B30B30FD7EE67000414CA /* screenseqscommand.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = screenseqscommand.h; sourceTree = SOURCE_ROOT; };
                372E126E0F26365B0095CF7E /* readotucommand.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = readotucommand.h; sourceTree = SOURCE_ROOT; };
                372E126F0F26365B0095CF7E /* readotucommand.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = readotucommand.cpp; sourceTree = SOURCE_ROOT; };
                372E12940F263D5A0095CF7E /* readdistcommand.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = readdistcommand.h; sourceTree = SOURCE_ROOT; };
@@ -236,10 +236,10 @@
                374610820F40652400460C57 /* unweighted.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = unweighted.cpp; sourceTree = SOURCE_ROOT; };
                3746109B0F40657600460C57 /* unifracunweightedcommand.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = unifracunweightedcommand.h; sourceTree = SOURCE_ROOT; };
                3746109C0F40657600460C57 /* unifracunweightedcommand.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = unifracunweightedcommand.cpp; sourceTree = SOURCE_ROOT; };
-               3749271B0FD58C840031C06B /* getsabundcommand.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = getsabundcommand.h; sourceTree = "<group>"; };
-               3749271C0FD58C840031C06B /* getsabundcommand.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = getsabundcommand.cpp; sourceTree = "<group>"; };
-               3749273D0FD5956B0031C06B /* getrabundcommand.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = getrabundcommand.h; sourceTree = "<group>"; };
-               3749273E0FD5956B0031C06B /* getrabundcommand.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = getrabundcommand.cpp; sourceTree = "<group>"; };
+               3749271B0FD58C840031C06B /* getsabundcommand.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = getsabundcommand.h; sourceTree = SOURCE_ROOT; };
+               3749271C0FD58C840031C06B /* getsabundcommand.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = getsabundcommand.cpp; sourceTree = SOURCE_ROOT; };
+               3749273D0FD5956B0031C06B /* getrabundcommand.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = getrabundcommand.h; sourceTree = SOURCE_ROOT; };
+               3749273E0FD5956B0031C06B /* getrabundcommand.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = getrabundcommand.cpp; sourceTree = SOURCE_ROOT; };
                37519A690F80E6EB00FED5E8 /* sharedanderbergs.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = sharedanderbergs.h; sourceTree = SOURCE_ROOT; };
                37519A6A0F80E6EB00FED5E8 /* sharedanderbergs.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = sharedanderbergs.cpp; sourceTree = SOURCE_ROOT; };
                37519A9F0F810D0200FED5E8 /* venncommand.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = venncommand.h; sourceTree = SOURCE_ROOT; };
@@ -308,10 +308,10 @@
                3792948D0F2E271100B9034A /* treecalculator.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = treecalculator.h; sourceTree = SOURCE_ROOT; };
                379643EB0FB9B5A80081FDB6 /* readseqs.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = readseqs.cpp; sourceTree = SOURCE_ROOT; };
                379643EC0FB9B5A80081FDB6 /* readseqs.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = readseqs.h; sourceTree = SOURCE_ROOT; };
-               3799A94C0FD6A58C00E33EDE /* distancedb.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = distancedb.cpp; sourceTree = "<group>"; };
-               3799A94D0FD6A58C00E33EDE /* distancedb.hpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.h; path = distancedb.hpp; sourceTree = "<group>"; };
-               3799A94E0FD6A58C00E33EDE /* seqsummarycommand.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = seqsummarycommand.cpp; sourceTree = "<group>"; };
-               3799A94F0FD6A58C00E33EDE /* seqsummarycommand.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = seqsummarycommand.h; sourceTree = "<group>"; };
+               3799A94C0FD6A58C00E33EDE /* distancedb.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = distancedb.cpp; sourceTree = SOURCE_ROOT; };
+               3799A94D0FD6A58C00E33EDE /* distancedb.hpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.h; path = distancedb.hpp; sourceTree = SOURCE_ROOT; };
+               3799A94E0FD6A58C00E33EDE /* seqsummarycommand.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = seqsummarycommand.cpp; sourceTree = SOURCE_ROOT; };
+               3799A94F0FD6A58C00E33EDE /* seqsummarycommand.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = seqsummarycommand.h; sourceTree = SOURCE_ROOT; };
                37AD4CE20F28AEA300AA2D49 /* sharedlistvector.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = sharedlistvector.h; sourceTree = SOURCE_ROOT; };
                37AD4CE30F28AEA300AA2D49 /* sharedlistvector.cpp */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.cpp.cpp; path = sharedlistvector.cpp; sourceTree = SOURCE_ROOT; };
                37AD4DB90F28E2FE00AA2D49 /* tree.h */ = {isa = PBXFileReference; fileEncoding = 4; lastKnownFileType = sourcecode.c.h; path = tree.h; sourceTree = SOURCE_ROOT; };

diff --git a/filterseqscommand.cpp b/filterseqscommand.cpp
index e8c96b3a42b88b6dbd4a003fe3a2454d2bbb4cb3..e493cbfbf84f90fb0b2c30c58dc7619ca21c437b 100644 (file)
--- a/filterseqscommand.cpp
+++ b/filterseqscommand.cpp
@@ -12,18 +12,13 @@
 /**************************************************************************************/
 
 FilterSeqsCommand::FilterSeqsCommand(){
+
        globaldata = GlobalData::getInstance();
        
        if(globaldata->getFastaFile() == "")            {       cout << "You must enter a fasta formatted file" << endl;        }
        trump = globaldata->getTrump()[0];
-//     vertical = 
-//     readSeqs->read();
-//     db = readSeqs->getDB();
-//     numSeqs = db->size();
-//     
-//     alignmentLength = db->get(0).getAlignLength();
-//
-//     filter = string(alignmentLength, '1');
+       numSeqs = 0;
+
 }
 
 /**************************************************************************************/
@@ -36,82 +31,66 @@ void FilterSeqsCommand::doHard() {
        ifstream fileHandle;
        openInputFile(hardName, fileHandle);
        
-       fileHandle >> hardFilter;
+       fileHandle >> filter;
+
+}
+
+/**************************************************************************************/
+
+void FilterSeqsCommand::doTrump(Sequence seq) {
        
-       if(hardFilter.length() != filter.length()){
-               cout << "The hard filter is not the same length as the alignment: Hard filter will not be applied." << endl;
-       }
-       else{
-               filter = hardFilter;
+       string curAligned = seq.getAligned();
+       
+       for(int j = 0; j < alignmentLength; j++) {
+               if(curAligned[j] == trump){
+                       filter[j] = '0';
+               }
        }
 
 }
 
 /**************************************************************************************/
 
-void FilterSeqsCommand::doTrump(Sequence seq) {
+void FilterSeqsCommand::doVertical() {
 
+       for(int i=0;i<alignmentLength;i++){
+               if(gap[i] == numSeqs)   {       filter[i] = '0';        }
+       }
        
-//     for(int i = 0; i < numSeqs; i++) {
-//             string curAligned = db->get(i).getAligned();;
-//
-//             for(int j = 0; j < alignmentLength; j++) {
-//                     if(curAligned[j] == trump){
-//                             filter[j] = '0';
-//                     }
-//             }
-//     }
-
 }
 
 /**************************************************************************************/
 
-void FilterSeqsCommand::doVertical(Sequence seq) {
-
-//     vector<int> counts(alignmentLength, 0);
-//     
-//     for(int i = 0; i < numSeqs; i++) {
-//             string curAligned = db->get(i).getAligned();;
-//             
-//             for(int j = 0; j < alignmentLength; j++) {
-//                     if(curAligned[j] == '-' || curAligned[j] == '.'){
-//                             counts[j]++;
-//                     }
-//             }
-//     }
-//     for(int i=0;i<alignmentLength;i++){
-//             if(counts[i] == numSeqs)        {       filter[i] = '0';                }
-//     }
+void FilterSeqsCommand::doSoft() {
+       
+       int threshold = int (soft * numSeqs);
+       bool keep = 0;
+       
+       for(int i=0;i<alignmentLength;i++){
+               if(a[i] >= threshold)           {       keep = 1;       }
+               else if(t[i] >= threshold)      {       keep = 1;       }
+               else if(g[i] >= threshold)      {       keep = 1;       }
+               else if(c[i] >= threshold)      {       keep = 1;       }
+               
+               if(keep == 0)   {       filter[i] = 0;          }
+       }
+       
 }
 
 /**************************************************************************************/
 
-void FilterSeqsCommand::doSoft(Sequence seq) {
-
-//     int softThreshold = numSeqs * (float)atoi(globaldata->getSoft().c_str()) / 100.0;
-//
-//     vector<int> a(alignmentLength, 0);
-//     vector<int> t(alignmentLength, 0);
-//     vector<int> g(alignmentLength, 0);
-//     vector<int> c(alignmentLength, 0);
-//     vector<int> x(alignmentLength, 0);
-//     
-//     for(int i=0;i<numSeqs;i++){
-//             string curAligned = db->get(i).getAligned();;
-//
-//             for(int j=0;j<alignmentLength;j++){
-//                     if(toupper(curAligned[j]) == 'A')                                                                               {       a[j]++; }
-//                     else if(toupper(curAligned[j]) == 'T' || toupper(curAligned[i]) == 'U') {       t[j]++; }
-//                     else if(toupper(curAligned[j]) == 'G')                                                                  {       g[j]++; }
-//                     else if(toupper(curAligned[j]) == 'C')                                                                  {       c[j]++; }
-//             }
-//     }
-//
-//     for(int i=0;i<alignmentLength;i++){
-//             if(a[i] < softThreshold && t[i] < softThreshold && g[i] < softThreshold && c[i] < softThreshold){
-//                     filter[i] = '0';                        
-//             }
-//     }
+void FilterSeqsCommand::getFreqs(Sequence seq) {
+
+       string curAligned = seq.getAligned();;
+       
+       for(int j=0;j<alignmentLength;j++){
+               if(toupper(curAligned[j]) == 'A')                                                                               {       a[j]++;         }
+               else if(toupper(curAligned[j]) == 'T' || toupper(curAligned[j]) == 'U') {       t[j]++;         }
+               else if(toupper(curAligned[j]) == 'G')                                                                  {       g[j]++;         }
+               else if(toupper(curAligned[j]) == 'C')                                                                  {       c[j]++;         }
+               else if(curAligned[j] == '-' || curAligned[j] == '.')                                   {       gap[j]++;       }
+       }
+       
 }
 
 /**************************************************************************************/
@@ -124,39 +103,68 @@ int FilterSeqsCommand::execute() {
                Sequence testSeq(inFASTA);
                alignmentLength = testSeq.getAlignLength();
                inFASTA.seekg(0);
-               int numSeqs = 0;
+               
+               if(globaldata->getSoft() != "" || isTrue(globaldata->getVertical())){
+                       a.assign(alignmentLength, 0);
+                       t.assign(alignmentLength, 0);
+                       g.assign(alignmentLength, 0);
+                       c.assign(alignmentLength, 0);
+                       gap.assign(alignmentLength, 0);
+               }
+               if(globaldata->getSoft() != ""){
+                       soft = (float)atoi(globaldata->getSoft().c_str()) / 100.0;
+               }
                
                if(globaldata->getHard().compare("") != 0)      {       doHard();                                                               }
                else                                                                            {       filter = string(alignmentLength, '1');  }
-               while(!inFASTA.eof()){
-                       Sequence seq(inFASTA);
-                       if(globaldata->getTrump().compare("") != 0)     {       doTrump(seq);           }
-                       if(isTrue(globaldata->getVertical()) == 1)      {       doVertical(seq);        }
-//                     if(globaldata->getSoft().compare("") != 0)      {       doSoft(seq);            }                       
-                       numSeqs++;
+
+               if(globaldata->getTrump().compare("") != 0 || isTrue(globaldata->getVertical()) || globaldata->getSoft().compare("") != 0){
+               
+                       while(!inFASTA.eof()){
+                               Sequence seq(inFASTA);
+                               if(globaldata->getTrump().compare("") != 0)     {       doTrump(seq);           }
+                               if(isTrue(globaldata->getVertical()) || globaldata->getSoft().compare("") != 0){        getFreqs(seq);  }
+                               numSeqs++;
+                               cout.flush();
+                       }
+               
                }
+               inFASTA.close();
                
-               ofstream outfile;
-               string filterFile = getRootName(globaldata->inputFileName) + "filter";
-               openOutputFile(filterFile, outfile);
+               if(isTrue(globaldata->getVertical()) == 1)      {       doVertical();   }
+               if(globaldata->getSoft().compare("") != 0)      {       doSoft();               }                       
 
-               outfile << filter << endl;
-               outfile.close();
+               ofstream outFilter;
+               string filterFile = getRootName(globaldata->inputFileName) + "filter";
+               openOutputFile(filterFile, outFilter);
+               outFilter << filter << endl;
+               outFilter.close();
                
+
+               openInputFile(globaldata->getFastaFile(), inFASTA);
                string filteredFasta = getRootName(globaldata->inputFileName) + "filter.fasta";
-               openOutputFile(filteredFasta, outfile);
-
-//             for(int i=0;i<numSeqs;i++){
-//                     string curAligned = db->get(i).getAligned();
-//                     outfile << '>' << db->get(i).getName() << endl;
-//                     for(int j=0;j<alignmentLength;j++){
-//                             if(filter[j] == '1'){
-//                                     outfile << curAligned[j];
-//                             }
-//                     }
-//                     outfile << endl;
-//             }
-//             outfile.close();
+               ofstream outFASTA;
+               openOutputFile(filteredFasta, outFASTA);
+
+               numSeqs = 0;
+               while(!inFASTA.eof()){
+                       Sequence seq(inFASTA);
+                       string align = seq.getAligned();
+                       string filterSeq = "";
+       
+                       for(int j=0;j<alignmentLength;j++){
+                               if(filter[j] == '1'){
+                                       filterSeq += align[j];
+                               }
+                       }
+
+                       outFASTA << '>' << seq.getName() << endl << filterSeq << endl;
+                       numSeqs++;
+                       gobble(inFASTA);
+               }
+               outFASTA.close();
+               inFASTA.close();
+               
                
                int filteredLength = 0;
                for(int i=0;i<alignmentLength;i++){

diff --git a/filterseqscommand.h b/filterseqscommand.h
index 924c36627fec65abd777c5ae530bf154583e954f..4b7d0111f11f5712dc9075bf313ec695222c7d7c 100644 (file)
--- a/filterseqscommand.h
+++ b/filterseqscommand.h
@@ -28,19 +28,19 @@ public:
 private:
        void doHard();
        void doTrump(Sequence);
-       void doSoft(Sequence);
-       void doVertical(Sequence);
+       void doVertical();
+       void doSoft();
+       void getFreqs(Sequence);
        string filter;  
        int alignmentLength;
 
        char trump;
        bool vertical;
+       float soft;
+       int numSeqs;
        
        GlobalData* globaldata; 
-//     ReadSeqs* readSeqs;
-//     SequenceDB* db;
-       
-
+       vector<int> a, t, g, c, gap;
 };
 
 #endif