Last night I had a moment of inspiration which will either sound like utter nonsense to experienced folks or be on a right path. So I'm putting it out for review. Note that this is a purely conceptual design so far, I'm making the topic before I begin working on proper design to see if its feasible, or common sense . . . or nonsense. For a summery skip the next block.
System is the root class, it has a structure composed of containers such as vectors called GameState.
System runs the game loop.
First 'windows messages' is checked and inputs are passed to the class InputClass
System runs frame() in which:
InputUpdate() is called passing GameState into InputClass
InputClass has the current frames input messages which it uses to update GameState
InputUpdate returns the updated GameState
GameLogicClass is called with GameLogicUpdate() which takes GameState
GameLogicClass does two things: it breaks down GameState to AiGameState and does none entity processes with GameState
GameLogicClass calls AiUpdate() passing AiGameState to AiClass
AiClass breaks down AiGameState to EntityGameState calling EntityUpdate() to pass EntityGameState where there are multiple entities and thus multiple EntityGameStates (JoesGameState, JanesGameState)
Each entity process its own AiGameState subset and returns it, the returned subsets then update the superset AiGameState
AiGameState then updates GameState inside GameLogicClass after AiUpdate returns it
GameLogicClass returns to System with an update representing the computers choices and other updates like adding to a construction timer
GraphicsLogicClass then does its thing with GameState, particle calculations ect.
FinalUpdateClass messes some more with GameState (placeholder class, UI ect would go here)
GraphicsClass finally gets GameState and renders it when it returns the game loop restarts with updated GameState
To save, System calls a function that writes GameState to a file and before game loop begins at runtime the user can load it or start new.
My thinking behind this is the GameState is like the consciousness of the game all of the classes are mental faculties. Entities get subsets of GameSate because they shouldn't know everything in the game.
Also as classes breakdown GameState the resulting structure is faster to update. When all of the subsets in a class are updated they can all be used to update the superset in one go. Each minor function does not have to deal with the whole GameState. By using a master GameState to pass to core functions I never have to worry about any class calling another except for system, everything is an independent module depending only on GameSate, and only affecting GameState. To add physics for example I just write the class and pass it GameState.
Comments, suggestions, criticisms, additions are greatly appreciated. A design philosophy is what I'm after and I'm hoping Ive hit it on the head with this. If I haven't Id really benefit from knowing now while the summer semester is young.
GameState as a game's conciousness
There is no need to provide additional encapsulation over a game state, and in fact doing so short-sightedly will likely cause need for changes down the road.
Beyond that, I agree with the ideas you have in a vague sense; but I think you're over-complicating your interface.
Here is what I usually do for my "GameMode" class (generally the same concept):
It's a method I learned a few years ago and it's so simple and flexible that I still use it today. You'll see it occasionally in open-source games in one form or another, too.
Beyond that, I agree with the ideas you have in a vague sense; but I think you're over-complicating your interface.
Here is what I usually do for my "GameMode" class (generally the same concept):
/**
@class GameMode
@brief
Encapsulates all game-implementation specific things, and allows for vast changes in
how the game might be played or how game logic might be performed between various
states, without requiring a bunch of conditionals etc.
**/
class GameMode
{
public:
/**
@fn Update
@brief
Handles player input, game logic and physics, and prepares for the next frame.
@param elapsedtime
Time in milliseconds between this frame and the last frame.
@param next
Next gamemode to be run. "this" to continue with this mode, "NULL" to quit.
@return true if the caller is responsible for freeing the gamemode.
**/
virtual bool Update(uint64_t elapsedtime, GameMode*& next) = 0;
/**
@fn Draw
@brief
Does nothing but draw the game.
@param elapsedtime
Time in milliseconds between this frame and the last frame.
@param context
The rendering context for this game -- Direct3DDevice9, GL context, SDL_Renderer, w.e you're using here.
**/
virtual void Draw(uint64_t elapsedtime, RenderContext& context) = 0;
};
class GM_TitleScreen: public GameMode
{
/* implement the gamemode here */
};
...
int main(int argc, char** argv)
{
// get a rendering context
RenderContext context = SetupRendering();
// get the time of starting the game
uint64_t lasttime = GetTimeMS();
// create the title/loading screen
GameMode* game = new GM_TitleScreen(...);
// loop as long as we have a game state
while(game)
{
// calculate elapsed time
uint64_t now = GetTimeMS();
uint64_t elapsedtime = now - lasttime;
GameMode* next = game;
// draw the game
game->Draw(elapsedtime, context);
// update the game
bool freegame = game->Update(elapsedtime, next);
// change gamemode as requested by Update
if(freegame)
{
delete game;
}
game = next;
}
Quit(context);
return 0;
}
It's a method I learned a few years ago and it's so simple and flexible that I still use it today. You'll see it occasionally in open-source games in one form or another, too.
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement