C++ fstream輸出錯誤的數據

Question

上下文首先：

我的程序做一些並行計算，它們記錄在一個文件中。線程按塊分組（我正在使用CUDA）。日誌文件格式化是這樣的：

#begin run 
({blockIdx,threadIdx}) {thread_info} 
({blockIdx,threadIdx}) {thread_info} 
... 
#end run 

我寫的應該讀取日誌文件，並通過線程運行的每個郵件進行排序的功能。

//------------------------------------------------------------------------------ 
// 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(); 
} 

問題是生成的文件有很多垃圾數據。例如，一個帶有6KB的輸入日誌文件會生成一個192KB的輸出日誌！看起來輸出文件有很多重複的輸入文件。雖然，在調試代碼時，deque在排序前後顯示正確的值。我認爲流媒體寫作本身有問題。

編輯：該功能沒有並行運行。

Answer 1

只是爲了顯示最終的代碼。注意substr上的變化，現在而不是它接收長度的索引。

//------------------------------------------------------------------------------------ 
// 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(); 
}