Question

I have a binary search on a file. The file is filled with log messages where each line begins with a date (dates or sorted based on event occurs)

example:

• 2011-09-18 09.38.20.123
• 2011-09-18 09.38.20.245
• 2011-09-18 09.38.20.393
• 2011-09-18 09.38.20.400
• 2011-09-18 09.38.20.785

If i need to find this date for example: 2011-09-18 09.38.20.390 my binary search will not find an exact match – but i don’t need exact match, i need to find the closest date to it (there is my position).

The current code will jump between 2011-09-18 09.38.20.245 and 2011-09-18 09.38.20.393.

I need some help how to modify the below code so that i get the closest number. In the above situation i would like to have: 2011-09-18 09.38.20.245 (better more than less)

BinarySearch::BinarySearch(QString strFileName,QDateTime dtFrom_target,QDateTime dtTo_target)
{

    QFile file(strFileName);
    qint64 nFileSize = file.size();

    int nNewFromPos;
    int nNewToPos;

    nNewFromPos = Search(file, dtFrom_target, nFileSize);
    nNewToPos  = Search(file, dtFrom_target, nFileSize);

    if(nNewFromPos!=-1 && nNewToPos!=-1){
        // now parse the new range

    }
    else{
        // dates out of bound

    }   
}

int BinarySearch::Search(QFile &file, QDateTime dtKey, int nMax) 
{  
    file.open(QIODevice::ReadOnly);
    char lineBuffer[1024];  
    qint64 lineLength;
    QDateTime dtMid;        
    int mid;
    int min; 
    if(!min) min = 0;


   while (min <= nMax) 
   {
       mid=(min+nMax)/2;    // compute mid point.                               
       file.seek(mid);  // seek to middle of file (position based on bytes)
       qint64 lineLength=file.readLine(lineBuffer,sizeof(lineBuffer)); // read until \n or error

       if (lineLength != -1) //something is read
       {
           // validate string begin (pos = 0) starts with date

           lineLength = file.readLine(lineBuffer, 24); //read exactly enough chars for the date from the beginning of the log file


           if(lineLength == 23)
           {
            dtMid = QDateTime::fromString(QString(lineBuffer),"yyyy-MM-dd HH.mm.ss.zzz"); //2011-09-15 09.38.20.192

                if(dtMid.isValid())
                {
                    if(dtKey > dtMid){
                        min = mid + 1; 
                    }
                    else if(dtKey < dtMid){
                        max = mid - 1; // repeat search in bottom half.
                    }
                    else{
                        return mid;     // found it. return position
                    }
                }
            }
       }
   }
   return -1;    // failed to find key
}

Answer 1

Try to implement an algorithm equivalent to std::equal_range which returns a pair of result of std::lower_bound and std::upper_bound

1. Find the position of the first element in the sorted range which does not compare less than the value (lower bound)
2. Find the position of the first element in the sorted range which compares greater than the value (upper bound)

—

template<typename OutIter>
void ReadLogsInRange(
    std::istream& logStream, Log::Date from, Log::Date to, OutIter out)
{
    Log l = LowerBound(logStream, from);
    *out++ = l;
    while(logStream >> l && l.date < to)
        *out++ = l;
}

Full example: http://ideone.com/CvIYm