// converts a space separated quality string into a compressed quality string
// NOTE: this function has horrible amounts of overhead, but lean and mean code that I had before
// failed some of the unit tests.
void CRegexUtilities::ConvertQualities(string& qualities, CMosaikString& compQualities) {
string::iterator strIte = qualities.end() - 1;
while ( *strIte == ' ' ) {
qualities.erase( strIte );
strIte--;
}
vector<string> columns;
vector<string>::const_iterator sIter;
char* pQualities = (char*)qualities.c_str();
Chomp(pQualities);
back_insert_iterator<vector<string> > backiter(columns);
SplitString(backiter, " ", pQualities);
const unsigned int numQualities = (unsigned int)columns.size();
compQualities.Reserve(numQualities);
compQualities.SetLength(numQualities);
unsigned char* pCompQualities = (unsigned char*)compQualities.Data();
for(sIter = columns.begin(); sIter != columns.end(); ++sIter, ++pCompQualities) {
if(sIter->empty()) continue;
*pCompQualities = GetUnsignedChar((char*)sIter->c_str());
}
}
// extracts the genome assembly ID from a FASTA/FASTQ header
void CRegexUtilities::ExtractGenomeAssemblyID(const string& line, CMosaikString& genomeAssemblyID) {
#ifdef WIN32
cmatch results;
if(!regex_search(line.c_str(), results, mGenomeAssemblyIDRegex)) {
genomeAssemblyID.SetLength(0);
return;
}
genomeAssemblyID = results[1].str().c_str();
#else
// TODO: replace this with the TR1 regex above when it finally works in gcc. It doesn't work in gcc 4.3.3
// find the GA tag
const string gaTag = "GA(";
string::size_type gaPos = line.find(gaTag.c_str());
if(gaPos == string::npos) {
genomeAssemblyID.SetLength(0);
return;
}
// find the matching end parenthesis
const unsigned int start = gaPos + gaTag.size();
unsigned int stop = start;
const char* pBuffer = line.data();
unsigned int lineLen = line.size();
if(stop < lineLen) {
while(pBuffer[stop] != ')') {
stop++;
if(stop == lineLen) break;
}
}
if(start == stop) {
cout << "ERROR: could not parse genome assembly ID from FASTA header." << endl;
cout << " " << line << endl;
exit(1);
}
genomeAssemblyID = line.substr(start, stop - start).c_str();
#endif
}