enable simplified mode if processing big FASTQ for many mutations

Author: sfchen

src/mutation.cpp

@@ -110,6 +110,7 @@ Match* Mutation::searchInRead(Read* r, char* simplifiedBuf, int distanceReq, int
 }
 
 vector<Mutation> Mutation::parseCsv(string filename) {
+    int num = 0;
     if(GlobalSettings::verbose) {
         cerr << "Parsing target mutations from CSV file: " << filename << endl;
     }
@@ -160,7 +161,8 @@ vector<Mutation> Mutation::parseCsv(string filename) {
         else {
             mutations.push_back(mut);
             if(GlobalSettings::verbose) {
-                cerr <<name<<" "<<left<<" "<<center<<" "<<right<< " "<<chr <<endl;
+                num++;
+                cerr <<num<<", "<<name<<" "<<left<<" "<<center<<" "<<right<< " "<<chr <<endl;
             }
         }
     }
@@ -172,6 +174,7 @@ vector<Mutation> Mutation::parseCsv(string filename) {
 }
 
 vector<Mutation> Mutation::parseBuiltIn() {
+    int num = 0;
     if(GlobalSettings::verbose) {
         cerr << "Using built-in target mutations" << endl;
     }
@@ -198,7 +201,8 @@ vector<Mutation> Mutation::parseBuiltIn() {
         Mutation mut(name, left, center, right, chr);
         mutations.push_back(mut);
         if(GlobalSettings::verbose) {
-            cerr <<name<<" "<<left<<" "<<center<<" "<<right<< " "<<chr <<endl;
+            num++;
+            cerr <<num<<", "<<name<<" "<<left<<" "<<center<<" "<<right<< " "<<chr <<endl;
         }
     }
     if(mutations.size() <= 0){
@@ -208,6 +212,7 @@ vector<Mutation> Mutation::parseBuiltIn() {
 }
 
 vector<Mutation> Mutation::parseVcf(string vcfFile, string refFile) {
+    int num = 0;
     if(GlobalSettings::verbose) {
         cerr << "Parsing target mutations from VCF file: " << vcfFile << endl;
         cerr << "With reference genome: " << refFile << endl; 
@@ -272,7 +277,8 @@ vector<Mutation> Mutation::parseVcf(string vcfFile, string refFile) {
             mut.setSmallIndel(true);
         mutations.push_back(mut);
         if(GlobalSettings::verbose) {
-            cerr <<name<<" "<<left<<" "<<center<<" "<<right<< " "<<chrom <<endl;
+            num++;
+            cerr <<num<<", "<<name<<" "<<left<<" "<<center<<" "<<right<< " "<<chrom <<endl;
         }
     }
     if(mutations.size() <= 0){

src/mutscan.cpp

@@ -5,6 +5,7 @@
 #include "sescanner.h"
 #include "pescanner.h"
 #include "util.h"
+#include "globalsettings.h"
 
 MutScan::MutScan(string mutationFile, string refFile, string read1File, string read2File, string html, int threadNum){
     mRead1File = read1File;
@@ -28,6 +29,35 @@ bool MutScan::scan(){
     }
 }
 
+void MutScan::evaluateSimplifiedMode(string r1file, string r2file, int mutationNum) {
+    if(mutationNum < 10000) {
+        GlobalSettings::setSimplifiedMode(false);
+        return ;
+    }
+    // use another ifstream to not affect current reader
+    ifstream is(r1file);
+    is.seekg (0, is.end);
+    long bytes = is.tellg();
+
+    if(r2file != "")
+        bytes *= 2;
+
+    // here we consider gz file for FASTQ has a compression rate of 3
+    if(ends_with(r1file, ".gz"))
+        bytes *= 3;
+
+    // enable simplified mode for over 50G FASTQ + 10,000 mutations
+    if(bytes > 50L * 1024L * 1024L * 1024L) {
+        if(GlobalSettings::verbose)
+            cerr << "Simplified mode is enabled automatically..."<<endl;
+        GlobalSettings::setSimplifiedMode(true);
+    }
+    else {
+        GlobalSettings::setSimplifiedMode(false);
+    }
+
+}
+
 bool MutScan::scanPairEnd(){
     FastqReader reader1(mRead1File);
     FastqReader reader2(mRead2File);

src/mutscan.h

@@ -15,6 +15,7 @@ class MutScan{
     bool scan();
     void textReport(vector<Mutation>& mutationList, vector<Match*> *mutationMatches);
     void htmlReport(vector<Mutation>& mutationList, vector<Match*> *mutationMatches);
+    static void evaluateSimplifiedMode(string r1file, string r2file, int mutationNum);
 
 private:
     bool scanPairEnd();

src/pescanner.cpp

@@ -9,6 +9,7 @@
 #include <memory.h>
 #include "util.h"
 #include "globalsettings.h"
+#include "mutscan.h"
 
 PairEndScanner::PairEndScanner(string mutationFile, string refFile, string read1File, string read2File, string html, int threadNum){
     mRead1File = read1File;
@@ -48,6 +49,12 @@ bool PairEndScanner::scan(){
     else
         mutationList = Mutation::parseBuiltIn();
 
+    if(GlobalSettings::verbose)
+        cerr << "Scanning "<< mutationList.size() << " mutations..."<<endl;
+
+    if(GlobalSettings::simplifiedModeToEvaluate)
+        MutScan::evaluateSimplifiedMode(mRead1File, mRead2File, mutationList.size());
+
     if(!GlobalSettings::legacyMode && mRollingHash==NULL){
         mRollingHash = new RollingHash();
         mRollingHash->initMutations(mutationList);

src/sescanner.cpp

@@ -9,6 +9,7 @@
 #include <memory.h>
 #include "util.h"
 #include "globalsettings.h"
+#include "mutscan.h"
 
 SingleEndScanner::SingleEndScanner(string mutationFile, string refFile, string read1File, string html, int threadNum){
     mRead1File = read1File;
@@ -46,6 +47,12 @@ bool SingleEndScanner::scan(){
     else
         mutationList = Mutation::parseBuiltIn();
 
+    if(GlobalSettings::verbose)
+        cerr << "Scanning "<< mutationList.size() << " mutations..."<<endl;
+
+    if(GlobalSettings::simplifiedModeToEvaluate)
+        MutScan::evaluateSimplifiedMode(mRead1File, mRead2File, mutationList.size());
+
     if(!GlobalSettings::legacyMode && mRollingHash==NULL){
         mRollingHash = new RollingHash();
         mRollingHash->initMutations(mutationList);