C++ fstream outputs wrong data - c++

Question

Context first:
My program do some parallel calculation which are logged in a file. Threads are grouped by blocks (I'm using CUDA). The log file is formated this way:
#begin run
({blockIdx,threadIdx}) {thread_info}
({blockIdx,threadIdx}) {thread_info}
...
#end run
I've wrote a function that should read the log file and sort each run messages by thread.
//------------------------------------------------------------------------------
// Comparison struct for log file sorting
//------------------------------------------------------------------------------
typedef struct
{
bool operator()(const string &rString1 , const string &rString2)
{
int closeParenthesisLocalition1 = rString1.find_first_of(')');
int closeParenthesisLocalition2 = rString2.find_first_of(')');
int compResult = rString1.compare(0 , closeParenthesisLocalition1 + 2 , rString2 , 0 , closeParenthesisLocalition2 + 2);
return (compResult < 0);
}
} comp;
//------------------------------------------------------------------------------------
// Sort the log file. Lines with same prefix (blockIdx,ThreadIdx) will be grouped in file per run.
//------------------------------------------------------------------------------------
void CudaUnitTest::sortFile()
{
comp comparison;
deque<string> threadsPrintfs;
ifstream inputFile(m_strInputFile);
assert(inputFile.is_open());
//Read whole input file and close it. Saves disk accesses.
string strContent((std::istreambuf_iterator<char>(inputFile)), std::istreambuf_iterator<char>());
inputFile.close();
ofstream outputFile(m_strOutputFile);
assert(outputFile.is_open());
string strLine;
int iBeginRunIdx = -10; //value just to addapt on while loop (to start on [0])
int iBeginRunNewLineOffset = 10; //"idx offset to a new line char in string. Starts with the offset of the string "#begin run\n".
int iEndRunIdx;
int iLastNewLineIdx;
int iNewLineIdx;
while((iBeginRunIdx = strContent.find("#begin run\n" , iBeginRunIdx + iBeginRunNewLineOffset)) != string::npos)
{
iEndRunIdx = strContent.find("#end run\n" , iBeginRunIdx + iBeginRunNewLineOffset);
assert(iEndRunIdx != string::npos);
iLastNewLineIdx = iBeginRunIdx + iBeginRunNewLineOffset;
while((iNewLineIdx = strContent.find("\n" , iLastNewLineIdx + 1)) < iEndRunIdx)
{
strLine = strContent.substr(iLastNewLineIdx + 1 , iNewLineIdx);
if(verifyPrefix(strLine))
threadsPrintfs.push_back(strLine);
iLastNewLineIdx = iNewLineIdx;
}
//sort last run info
sort(threadsPrintfs.begin() , threadsPrintfs.end() , comparison);
threadsPrintfs.push_front("#begin run\n");
threadsPrintfs.push_back("#end run\n");
//output it
for(deque<string>::iterator it = threadsPrintfs.begin() ; it != threadsPrintfs.end() ; ++it)
{
assert(outputFile.good());
outputFile.write(it->c_str() , it->size());
}
outputFile.flush();
threadsPrintfs.clear();
}
outputFile.close();
}
The problem is that the resulting file has a lot of trash data. For example an input log file with 6KB generated a output log of 192KB! It appears the output file has a lot of repetitions of the input file. When debugging code the deque showed the right values before and after sort, though. I think there is something wrong with the ofstream write itself.
Edit: The function isn't running in parallel.

Answer 1

Just to show the final code. Note the change on substr, now instead of an index it's receiving the lenght.
//------------------------------------------------------------------------------------
// Sort the log file. Lines with same prefix (blockIdx,ThreadIdx) will be grouped in file per run.
//------------------------------------------------------------------------------------
void CudaUnitTest::sortFile()
{
comp comparison;
deque<string> threadsPrintfs;
ifstream inputFile(m_strInputFile);
assert(inputFile.is_open());
//Read whole input file and close it. Saves disk accesses.
string strContent((std::istreambuf_iterator<char>(inputFile)), std::istreambuf_iterator<char>());
inputFile.close();
ofstream outputFile(m_strOutputFile);
assert(outputFile.is_open());
string strLine;
int iBeginRunIdx = -10; //value just to addapt on while loop (to start on [0])
int iBeginRunNewLineOffset = 10; //"idx offset to a new line char in string. Starts with the offset of the string "#begin run\n".
int iEndRunIdx;
int iLastNewLineIdx;
int iNewLineIdx;
while((iBeginRunIdx = strContent.find("#begin run\n" , iBeginRunIdx + iBeginRunNewLineOffset)) != string::npos)
{
iEndRunIdx = strContent.find("#end run\n" , iBeginRunIdx + iBeginRunNewLineOffset);
assert(iEndRunIdx != string::npos);
iLastNewLineIdx = iBeginRunIdx + iBeginRunNewLineOffset;
while((iNewLineIdx = strContent.find("\n" , iLastNewLineIdx + 1)) < iEndRunIdx)
{
strLine = strContent.substr(iLastNewLineIdx + 1 , iNewLineIdx - iLastNewLineIdx);
if(verifyPrefix(strLine))
threadsPrintfs.push_back(strLine);
iLastNewLineIdx = iNewLineIdx;
}
//sort last run info
sort(threadsPrintfs.begin() , threadsPrintfs.end() , comparison);
threadsPrintfs.push_front("#begin run\n");
threadsPrintfs.push_back("#end run\n");
//output it
for(deque<string>::iterator it = threadsPrintfs.begin() ; it != threadsPrintfs.end() ; ++it)
{
assert(outputFile.good());
outputFile.write(it->c_str() , it->size());
}
threadsPrintfs.clear();
}
outputFile.close();
}