made a small mod to the alignment parameters in trim.seqs

[mothur.git] / blastdb.cpp

diff --git a/blastdb.cpp b/blastdb.cpp
index 17db069cdedb8a019d980c32fc4b6e7767fab27b..760824dbb23f9b2a9d680ace769ccc1c6ffab96c 100644 (file)
--- a/blastdb.cpp
+++ b/blastdb.cpp
@@ -25,6 +25,19 @@ gapOpen(gO), gapExtend(gE), match(m), misMatch(mM) {
        queryFileName = toString(randNumber) + ".candidate.unaligned.fasta";
        blastFileName = toString(randNumber) + ".blast";
 
+}
+/**************************************************************************************************/
+
+BlastDB::BlastDB() : Database() {
+       
+       globaldata = GlobalData::getInstance();
+       count = 0;
+
+       int randNumber = rand();
+       dbFileName = toString(randNumber) + ".template.unaligned.fasta";
+       queryFileName = toString(randNumber) + ".candidate.unaligned.fasta";
+       blastFileName = toString(randNumber) + ".blast";
+
 }
 
 /**************************************************************************************************/
@@ -41,21 +54,22 @@ vector<int> BlastDB::findClosestSequences(Sequence* seq, int n) {
                vector<int> topMatches;
                
                ofstream queryFile;
-               openOutputFile(queryFileName, queryFile);
+               openOutputFile((queryFileName+seq->getName()), queryFile);
                queryFile << '>' << seq->getName() << endl;
                queryFile << seq->getUnaligned() << endl;
                queryFile.close();
+
                                
                //      the goal here is to quickly survey the database to find the closest match.  To do this we are using the default
                //      wordsize used in megablast.  I'm sure we're sacrificing accuracy for speed, but anyother way would take way too
                //      long.  With this setting, it seems comparable in speed to the suffix tree approach.
                
                string blastCommand = path + "blast/bin/blastall -p blastn -d " + dbFileName + " -m 8 -W 28 -v " + toString(n) + " -b " + toString(n);;
-               blastCommand += (" -i " + queryFileName + " -o " + blastFileName);
+               blastCommand += (" -i " + (queryFileName+seq->getName()) + " -o " + blastFileName+seq->getName());
                system(blastCommand.c_str());
                
                ifstream m8FileHandle;
-               openInputFile(blastFileName, m8FileHandle);
+               openInputFile(blastFileName+seq->getName(), m8FileHandle, "no error");
                
                string dummy;
                int templateAccession;
@@ -71,24 +85,9 @@ vector<int> BlastDB::findClosestSequences(Sequence* seq, int n) {
                        topMatches.push_back(templateAccession);
                }
                m8FileHandle.close();
-               
-               string root = dbFileName;
-               string temp = dbFileName + ".nsq";
-               remove(temp.c_str());   
-               temp = dbFileName + ".nsi";
-               remove(temp.c_str());
-               
-               temp = dbFileName + ".nsd";
-               remove(temp.c_str());   
-
-               temp = dbFileName + ".nin";
-               remove(temp.c_str());   
-
-               temp = dbFileName + ".nhr";
-               remove(temp.c_str());   
-       
+               remove((queryFileName+seq->getName()).c_str());
+               remove((blastFileName+seq->getName()).c_str());
 
-               
                return topMatches;
        }
        catch(exception& e) {
@@ -104,7 +103,7 @@ vector<int> BlastDB::findClosestMegaBlast(Sequence* seq, int n) {
                vector<int> topMatches;
                
                ofstream queryFile;
-               openOutputFile(queryFileName, queryFile);
+               openOutputFile((queryFileName+seq->getName()), queryFile);
                queryFile << '>' << seq->getName() << endl;
                queryFile << seq->getUnaligned() << endl;
                queryFile.close();
@@ -114,11 +113,11 @@ vector<int> BlastDB::findClosestMegaBlast(Sequence* seq, int n) {
                //      long.  With this setting, it seems comparable in speed to the suffix tree approach.
        
                string blastCommand = path + "blast/bin/megablast -e 1e-10 -d " + dbFileName + " -m 8 -b " + toString(n) + " -v " + toString(n); //-W 28 -p blastn
-               blastCommand += (" -i " + queryFileName + " -o " + blastFileName);
+               blastCommand += (" -i " + (queryFileName+seq->getName()) + " -o " + blastFileName+seq->getName());
                system(blastCommand.c_str());
-               
+
                ifstream m8FileHandle;
-               openInputFile(blastFileName, m8FileHandle, "no error");
+               openInputFile(blastFileName+seq->getName(), m8FileHandle, "no error");
        
                string dummy;
                int templateAccession;
@@ -135,6 +134,8 @@ vector<int> BlastDB::findClosestMegaBlast(Sequence* seq, int n) {
 //cout << templateAccession << endl;
                }
                m8FileHandle.close();
+               remove((queryFileName+seq->getName()).c_str());
+               remove((blastFileName+seq->getName()).c_str());
 //cout << "\n\n" ;             
                return topMatches;
        }
@@ -154,7 +155,7 @@ void BlastDB::addSequence(Sequence seq) {
                unalignedFastaFile << '>' << count << endl;                                     //      sequences, which will be input to formatdb
                unalignedFastaFile << seq.getUnaligned() << endl;
                unalignedFastaFile.close();
-               
+       
                count++;
        }
        catch(exception& e) {
@@ -181,6 +182,7 @@ void BlastDB::generateDB() {
                exit(1);
        }
 }
+/**************************************************************************************************/
 
 /**************************************************************************************************/