Parsing text file into list gives segmentation fault - c++

Question

I'm getting a segmentation fault while trying to parse a big text file. The file contains 91 529 mRNA transcripts and details about these transcripts. I've created a RefSeqTranscript object that will take these details. When I parse the file, I create a list of these objects and start putting the details into these lists. It works fine for the first 1829 transcripts and then crashes with a segmentation fault. The method I'm running is:
void TranscriptGBFFParser::ParseFile(list<RefSeqTranscript> &transcripts, const char* filepath)
{
cout << "Parsing " << filepath << "..." << endl;
ifstream infile;
infile.open(filepath);
int num = 0;
RefSeqTranscript *transcript = new RefSeqTranscript();
for(string line; getline(infile, line); )
{
in.clear();
in.str(line);
if (boost::starts_with(line, "LOCUS"))
{
if((*transcript).transcriptRefSeqAcc.size() > 0)
{
cout << (*transcript).transcriptRefSeqAcc << ":" << (*transcript).gi << ":" << (*transcript).gene.geneName << ":" << ++num << endl;
transcripts.push_back(*transcript);
delete transcript;
RefSeqTranscript *transcript = new RefSeqTranscript();
}
}
else if (boost::starts_with(line, " var"))
{
TranscriptVariation variant;
(*transcript).variations.push_back(variant);
}
//Store the definition of the transcript in the description attribute
else if (boost::starts_with(line, "DEFINITION"))
{
(*transcript).description = line.substr(12);
for(line; getline(infile, line); )
{
if(boost::starts_with(line, "ACCESSION "))
break;
(*transcript).description += line.substr(12);
}
}
//The accession number and GI number are obtained from the VERSION line
else if (boost::starts_with(line, "VERSION"))
{
string versions = line.substr(12);
vector<string> strs;
boost::split(strs, versions, boost::is_any_of( " GI:" ), boost::token_compress_on);
boost::trim_left(strs[0]);
(*transcript).transcriptRefSeqAcc = strs[0];
(*transcript).gi = atoi(strs[1].c_str());
}
//Gene information is obtained from the "gene" sections of each transcript
else if (boost::starts_with(line, " gene"))
{
for(line; getline(infile, line); )
{
if(boost::starts_with(line.substr(21), "/gene="))
{
Gene *gene = new Gene();
string name = line.substr(27);
Utilities::trim(name, '\"');
(*gene).geneName = name;
(*transcript).gene = *gene;
delete gene;
break;
}
}
(*transcript).gene.geneID = 0;
}
else if (boost::starts_with(line, " CDS"))
{
(*transcript).proteinRefSeqAcc = "";
}
else if (boost::starts_with(line, "ORIGIN"))
{
(*transcript).sequence = "";
}
}
cout << (*transcript).transcriptRefSeqAcc << ":" << (*transcript).gi << ":" << (*transcript).gene.geneName << endl;
transcripts.push_back(*transcript);
delete transcript;
cout << "No. transcripts: " << transcripts.size() << endl;
cout << flush;
infile.close();
cout << "Finished parsing " << filepath << "." << endl;
}
I'm new to C++ and don't have a great understanding of how to work with pointers etc so I'm guessing I might have done something wrong there. I don't understand why it would work for almost 2000 objects before cutting out though.
The file I'm parsing is 2.1 GB and consists of about 44 000 000 lines so any tips on how to improve the efficiency would also be much appreciated.

Answer 1

This is probably not the only answer, but you have a leak...
if (boost::starts_with(line, "LOCUS"))
{
if((*transcript).transcriptRefSeqAcc.size() > 0)
{
cout << (*transcript).transcriptRefSeqAcc << ":" << (*transcript).gi << ":" << (*transcript).gene.geneName << ":" << ++num << endl;
transcripts.push_back(*transcript);
delete transcript;
// LEAK!
RefSeqTranscript *transcript = new RefSeqTranscript();
}
}
You probably mean:
transcript = new RefSeqTranscript();

Answer 2

It's hard to say anything specific unless you provide some more details:
What line does it crashed in?
Do you really need all of those transcripts at the same time?
But I would suggest you a couple improvements:
Don't use pointer (or at least use smart pointer) for the RefSeqTranscript *transcript;
Don't use pointer for the Gene *gene;
Generally, don't use pointers unless you realy need them;
And you have a bug here:
delete transcript;
RefSeqTranscript *transcript = new RefSeqTranscript();
Since you've laready declared transcript outside the loop's body, here you hide it with new variable with the same name. This causes memory leak, and moreover, you delete an outer transcript and do not replace it with anything. So, you probably get a crash on the next iteration.