I'm making a pong game with c++ that currently has two states, GameState and MenuState. The way it is now, both states are methods of the Game class. The Game class has all the instance data that both methods/states use (so paddles, ball, menu buttons, etc). Is it bad practice to make a gamestate a method rather than its own class?

This is how my main function is right now

#include "Game.h"

int main( int argc, char* args[] )
{
	Game game;

	game.Run();

    return 0;
}

And the Run function

int Game::Run()
{
	if (error)
	{
		return 1;
	}

    //The frame rate regulator, passed to GameState
    Basic_SDL::Timer fps;

	//Which player scored (left or right), passed to GameState
	int scoreKeeper = 0;

	rightPaddle.setAI(true);
	
	//While the suer hasn't quit
	while (quit == false)
	{
		if (!paused)
		{
			GameState(fps, scoreKeeper);
		}
		else
		{
			MenuState();
		}
	}
	
    //Clean up
    clean_up();
}

It works fine this way, I'm just not sure if making each state a method is looked down upon. Any other advice on how to implement game states would be appreciated also .

Is it bad practice to make a gamestate a method rather than its own class?

For games with only a handful of states and simple transitions between states, it isn't bad practice at all.

When you start developing more advanced games with a large number of states, state machines within states, and complex interactions between states, it becomes more important to develop a more advance state machine framework.

Have you looked at Lazy Foo's article on State Machines? It's uses enums and a case/switch function, and is relatively simple to understand.

Basically, you have a base GameState class that is abstract. It has three virtual functions - handleInput(), logic(), and draw().

Then, your other states - in this case, the Menu and the actual Pong state - will inherit from this base class. They'll inherit the three vital functions, and then you can put what you want in each respective state (e.g; a c_Player in the Pong state, or a c_Button in the Menu state).

The states are switched first by preparing the change with a set_next_state() function, then calling change_state() to actually change states.

I'd recommend going through the article, it really simplifies the whole state machine topic (at least it did for me )

http://lazyfoo.net/articles/article06/index.php

You can even apply the same guidelines for an AI state machine too, which is pretty neat.

Thanks, all the answers are great! I don't think I'll need the more complicated state machines for a while, but it's nice to at least have an idea of how people do it. I can see from the responses that there're a lot of ways to implement game states. Are there reasons to use one way over another in specific situations, or is it more just a matter of preference?

To me implementing a state-machine feels equivalent to making a Rube-Goldberg-Machine. There are variables already in a programming language, why implement a god-variable to combine various things from the whole program only to try to interpret its value in a huge switch case statement a moment later to separate the things you just combined to essentially just simulate how a variable works using a variable? Or even worse build dozens of vapor classes with virtual methods, one per case, to disguise that switch?

Why not just make decisions at those points where they naturally appear with a bit of local data?