how to build a vector of different objects after reading a file

Question

Hello, everyone! I am new for C++ and now work on a C++ project.The whole structure has been completed.But I have been wondering about

how to build the vector of different objects and how to read the file since the beginning.

In my problem, first I have to read a txt file with readObstacles(std::istream &fs) that has the following format:

RECTANGLE 1 2.5 2 2 0.2
CIRCLE 1 2.5 1.2
RECTANGLE 4 2 2 2 0.3

(each obstacle begins a new line)

I need to read the data information of different obstacles and store these obstacles in a vector.
Class Obstacle is the base class with two sub-classes CIRCLE and RECTANGLE.
I try to put the these different obstacles (that i think should have their data info) in a obstacle vector and then call the virtual function they both have.

Below is the code I try to do with:

vector<Obstacle> obsdata; 

Myworld::readObstacles(std::istream &fs)
{
    std::string shape;
    double num1,num2,num3,num4,num5;
    while(fs>>shape>>num1>>num2>>num3>>num4>>num5)
    {
        if(shape=="CIRCLE") {
            CIRCLE c;
            c.m_Xc=num1;
            c.m_Yc=num2;
            c.m_Radius=num3;       
            obsdata.push_back(c);
        }

        if(shape=="RECTANGLE") {
            RECTANGLE r;
            r.center_x=num1;
            r.center_y=num2;    
            r.width=num3;
            r.height=num4;
            r.angle=num5;
            obsdata.push_back(r);
        }
    }
}

MyWorld::writeMatlabDisplayCode(std::ostream &fs)
{
    for( i = 0; i < obsdata.size(); i++ )
        obsdata[i].writeMatlabDisplayCode(fs);
}

I know it doesn't work but i don't know what i should do.

Any words will help. Thanks !

Answer 1

You have a problem called slicing: you lose the data from the derived classes if you copy to an vector containing base class instances.

You can use a vector of pointers to the base though:

std::vector<Obstacle*> obsdata;
RECTANGLE r = new RECTANGLE;
// set properties
obsdata.push_back(r);
// use it later, e.g.:
obsdata[0]->writeMatlabDisplayCode(fs);
// clean up when you don't need obstacles anymore:
for(std::vector<Obstacle*>::iterator it = obsdata.begin(); it != obsdata.end(); ++it)
    delete *it;

You also should avoid trying to read in 5 numbers every time. Instead you could read in the obstacle description and then read as many numbers as the specific obstacle needs.

Answer 2

You can do it if you have a vector of pointers instead:

vector<Obstacle *> obsdata;

and then you "new" your subsequent CIRCLE and RECTANGLE:

    if(shape=="CIRCLE") {
        CIRCLE *c = new CIRCLE;
        c->m_Xc=num1;
        c->m_Yc=num2;
        c->m_Radius=num3;       
        obsdata.push_back(c);
    }

etc..

Answer 3

The condition of your while loop is a problem. Circle has 2 fewer entries than rectangle. I suggest you read the shape, then depending on type read the rest of the data. You know you can read 3 data points for circle and 5 for rectangle.

Some pseudo code

while( fs >> shape )
{
  if( shape == "CIRCLE" )
  { 
        CIRCLE c;
        fs >> c.m_Xc;
        fs >> c.m_Yc;
        fs >> c.m_Radius;       
        obsdata.push_back(c);
   }

  if( shape == "RECTANGLE" )
  {
      RECTANGLE r;
      fs >> r.center_x;
      fs >> r.center_y;    
      fs >> r.width;
      fs >> r.height;
      fs >> r.angle;
      obsdata.push_back(r);
  }
}

As gf suggested in the comment, a good C++ book or tutorial on the standard library would be helpful.

Answer 4

You could try just reading the whole line at a time then tokenize the values...

vector<Obstacle*> obsdata;

string line;
while(getline(fs, line)) {
 char *token = strtok(line.c_str(), " ");
 string shape(token);

 vector<double> numbers;
 stringstream ss;
 while(token = strtok(NULL, " ")) {
  double d;
  ss << token;
  ss >> d;

  numbers.push_back(d);
 }

 if(shape == "CIRCLE") {
  CIRCLE *c = new CIRCLE();
  c->m_Xc=numbers[0];
  c->m_Yc=numbers[1];
  c->m_Radius=numbers[2];       
  obsdata.push_back(c);
 } else if(shape == "RECTANGLE") {
  RECTANGLE *r = new RECTANGLE();
  r->center_x=numbers[0];
  r->center_y=numbers[1];    
  r->width=numbers[2];
  r->height=numbers[3];
  r->angle=numbers[4];
  obsdata.push_back(r);
 }
}

I would delete this vector of pointers in your MyWorld destructor:

~MyWorld() {
    // loop thanks to gf
    for(std::vector<Obstacle*>::iterator it = obsdata.begin(); it != obsdata.end(); ++it)
        delete *it;  
}

Answer 5

All of the advice here is good but I'd like to suggest two improvements:

• Managed your pointers (I'd pick shared_ptr but ptr_vector would also work)
• Use an object factory

This will make your solution more robust (and not leak memory) and more scalable (you remove the 'God' function that would become a compiler bottleneck as the number of Obstacles increases).

Your code may look something like:

vector<shared_ptr<Obstacle> > obsdata; 

Myworld::readObstacles(std::istream &fs)
{
    std::string shape;
    while(fs >> shape)
    {
        try
        {
            obsdata.push_back(shared_ptr<Obstacle>(m_ObstacleCreator.Create(shape, fs));
        }
        catch(ObstacleNotKnownException& e)
        {
            // Error handling here
        }
    }
}

The only thing left is to populate m_ObstacleCreator with the knowledge of how to create the various objects based on the shape name and the istream. An example:

Obstacle * CreateCircleObstacle(istream &fs)
{
    CIRCLE *c = new CIRCLE();
    fs >> c->m_Xc >> c->m_Yc >> c->m_Radius;
}

m_ObstacleCreator.Register("CIRCLE", &CreateCircleObstacle);

You may also want to make m_ObstacleCreator a singleton (Object Factories are one of the few genuine uses for singletons IMO).

For an excellent description on the details of Object Factory implementations pick up a copy of Alexandrescu's Modern C++ Design and check out chapter eight.