Improve reading data from a file

[vckngs7]

I am working with comma separated value (.csv) files which contain information that is used in later calculations. I have been storing the data in two dimensional vectors with success but I feel that my code is somewhat clunky. I am looking for improvements to my approach, either changes to my current approach or suggestions for a better approach all together.

void FileData::setDataVector (string file){
   	//Opens .csv file and checks that it can be read.
	std::ifstream infile;
    std::string inf=file+".csv";
    infile.open(inf.c_str(),ios::in);
    if (!infile.is_open()){
      cout << "Unable to open "+inf << endl;
      exit(1);
    }

    //Sets the size of the vector
    int i=0;
    while (infile.good()){
      getline(infile,inf);
      if (inf.size()!=0){
		i++; 
      }
    }
    dataVector.resize(i);
    infile.clear();
    infile.seekg(0, ios::beg);

    //Writes each line to the two dimensional vector
    std::string line;
    int start,stop;
    for (int n=0; infile.good(); n++){
      getline(infile,line);
      if ((line.size())==0) {
		return;
      }
      start=line.find_first_not_of(", ");
      stop=line.find_first_of(", ",start);
      //Fills out the columns of the current row of dataVector
      while(start!=-1) {
		dataVector[n].push_back(line.substr(start,stop-start));
		start=line.find_first_not_of(", ",stop);
		stop=line.find_first_of(", ",start);
      }
    }
    infile.close(); 
}

[Texan40]

I'm assuming dataVector is a vector<vector<string> > member in your class. I'm also assuming that the csv data in the file will look something like:

ABC,123,DEF,456,GHI,789
987,IHG,654,FED,321,CBA

I think that getting the number of lines in the file just to resize the vector in dataVector is potentially wasteful. Why not just read them directly from the file into dataVector as you encounter them? vectors are dynamic and thus perfect for this. Perhaps something like this?

#include <fstream>
#include <string>
#include <vector>

using namespace std;

vector<vector<string> > dataVector;

int main()
{
  ifstream in("test.csv"); // open file
  string buf; // buffer for getline
  while(getline(in,buf)) // read lines in file till error or eof
  {
    vector<string> tmp; // temp vector to tokenize csv lines
    size_t pos = buf.find(","); // find first delimiter
    while(!buf.empty()) // we will cannibalize the string as we tokenize it
    {
      tmp.push_back(buf.substr(0,pos)); // get string before delimiter
      buf.erase(0,pos+1); // delete to delimiter
      pos = buf.find(","); // find next delimiter
      if(pos == string::npos) // no more delimiters
      {
        tmp.push_back(buf);
        buf.clear(); // empty out string
      }
    }
    dataVector.push_back(tmp);
  }
  return 0;
}

[vckngs7]

I like that a lot. Thank you so much.

If anyone else has other ideas or different approaches that might be worth looking into I'd love to hear them.

[vckngs7]

A few changes had to be made in order to use the above code. They changes I used are as follows.

void setVector (){
   //Opens .csv file and checks that it can be read.
    std::ifstream infile(("file.csv").c_str()); //the file name must be a c-str
    std::string buf; 
    if (!infile.is_open()){
      cout << "Unable to open +"file.csv" << endl;
      exit(1);
    }
    std::vector<std::string> tmp; //Declaring temp vector and pos as an int 
    size_t pos;
    getline(infile,buf); //Skips the label line
	while(getline(infile,buf)){
		pos=buf.find(',');
		if (pos==string::npos) continue; //Skips blank lines
		while(!buf.empty()){
			tmp.push_back(buf.substr(0,pos));
			buf.erase(0,pos+1);
			pos = buf.find(',');
			if (pos==string::npos){
				tmp.push_back(buf);
				buf.clear();
			}
		}
		dataVector.push_back(tmp);
		tmp.clear(); //Resets tmp vector to be used again
	}
    infile.close(); 

} 

*edited to add in tmp.clear()