41 replies to this topic

[Icebone1000]

[source lang="cpp"]class Game{ list < GameObjects * > m_gos; void Update(float delta){ for(list::it goIt ..) (*goIt)->Update(delta); } ...};[/source]


Pseudo code..

Id like to make the Class Game the sole responsible for the creation and deletion of the gameobjects.
I always do this with a template function:
[source lang="cpp"]template< class derivedGO>derivedGO* Create(){ derivedGO *p = new derivedGO(); m_gos.add(p); return p;}[/source]
And this is the only way to give the Game class the objects (so if they got created externally, they will not be part of the game).
The only (really annoying imo) problem is that derivedGO must provide a compatible constructor, this sucks, cause I always, then, have to create a Init(params) function that is always called just after calling create..

Is this poor design?

Edited by Icebone1000, 22 October 2012 - 01:54 PM.

[Servant of the Lord]

I can't really comment on the design itself, but if you are using C++11, you can forward the arguments (without knowing how many arguments or what types) to the constructor like this:

template<typename derivedGO, typename Arg>
derivedGO* Create(Arg &&arg){

    derivedGO *p = new derivedGO(std::forward<Arg>(arg));
   m_gos.add(p);
   return p;
}

('Game' sounds like it's a monolith class, though)

[L. Spiro]

It is not particularly good design. You are trying to do too much for the users. There comes a point when you give users so little control that your engine/framework/whatever simply becomes useless.

Your game class is doing way too much. It is a mega-class.
Break it down into smaller classes each with a specific task.
For example, objects don’t exist inside the game, they exist inside scenes.
So for starters, get all the objects out of your Game class and put them into a Scene class.

On my first few engine attempts I also thought there should be at most only one scene active at a time.
This caused me to create hacky solutions when trying to render one set of objects with one camera and another set of objects with another camera.
The correct solution is not to limit how many scenes you can have.
So a scene can have any number of objects inside it. The Game class can have any number of scenes inside it.


Back to my first point, why do objects have to be created only through the Game (or Scene) class?
You probably think you are doing your users a favor and making their lives easier by doing some of the heavy lifting for them.
You’re not.
There is no logical reason why objects in the scene must strictly be created by the scene manager.
Of course, creating objects via the scene manager should be one option, but not the only one.

m_dmipInstance = m_sgpScene->CreateDrawableModelInstance( "Cayman.lsm" );
m_dmipInstance->Orientation().SetPos( CVector3( 0.0f, -m_dmipInstance->Aabb().m_vMin.y, 0.0f ) );
m_dmipInstance->SetCellShading( true );

CDrawableModelInstancePtr pdmipTemp = m_sgpScene->CreateDrawableModelInstance( "Ground.lsm" );
pdmipTemp->SetCastsShadow( false );

This is just one way to create models. You can also do this:

CDrawableModelInstancePtr dmipCustomLoad;
dmipCustomLoad.New();
dmipCustomLoad->LoadFile( "BMW X6" );
dmipCustomLoad->SetCookTorranceShading( true );
m_sgpScene->AddDrawableModelInstance( dmipCustomLoad );


dmipCustomLoad.New();  // Notice how this is a smart pointer.
// Custom setting of vertices, normals, etc. here.
m_sgpScene->AddDrawableModelInstance( dmipCustomLoad );

I am speaking from experience when I say that loading models strictly through the scene manager is a pain in the ass.
If you think you have to do it that way, you have a problem with your overall design/architecture that you need to fix.


If your goal is to make the engine easy to use, employ smart pointers and use them for the objects in your scene. And don’t be picky about how models end up in the scene.



L. Spiro

[uglybdavis]

All pseudo code here, but you should reconsider your approach to the general design on the game engine. I apologize in advance, i started typing all this out and got bored by the end, if i skipped something and you would like some explanation, just ask. But it should all be self-explanatory.

You have a game, which contains a stack of scenes. Only the top most scene is updated, but any scene can be rendered. This makes it so if you have an in game menu, you can pause the game (by no longer updating the game scene) and render it as well as the menu on top. Makes it pretty easy. The scene has game objects, which are basically just glue objects containing components (Transforms, character controllers, ui's etc). Each component does something specific, with variables you can set at any time. You can retrieve game objects from the scene, and components from game objects. All objects are created by Game, as it is the factory. When game starts, it creates the main menu scene, and you just take it from there. This may seem a bit fuzzy or abstract, but the implementation is simple, consider the following:

Some class definitions
[source lang="cpp"]class Component() { enum Type { RENDERER, ENEMY, PLAYER, UI, ANIMATION, SCRIPT } Type m_eComponentType; virtual void Awake() = 0; // Called when component is created virtual void Destroy() = 0; // Called when component is destroyed virtual void Start() = 0; // Called when component is activated virtual void Stop() = 0; // Called when component is deactivated virtual void Update(float dt) = 0; // I'm not going to explain}class Renderer : public Component { int m_nTextureHandle; // All sorts of component specific members int m_nBufferId; // Look, we even have a buffer! void Awake(); // Overload void Destroy(); // Overload void Start(); // Overload void Stop(); // Overload void Update(float dt); // Overload void Render(); // Again, component specific!}class Enemy : public Component { // Two components, we're on fire! int m_nHealth; // Enemy specific junk AStar* m_pPathFinder; // Specific junk void Awake(); // Override void Destroy(); // Override void Start(); // Override void Stop(); // Override void Update(float dt); // Override}class Player : public Component { // See above void Awake(); void Destroy(); void Start(); void Stop(); void Update(float dt);}class UI : public Component { // See above void Awake(); void Destroy(); void Start(); void Stop(); void Update(float dt);}class GameObject() { // Now for the fun bit std::vector<Component*> m_vComponents; // A game object has components std::string m_strGameObjectName; // So we can retrieve game objects later bool m_bIsActive; // Does not always need to be active void GameObjectWasCreated(); // Created void GameObjectDidBecomeActive(); // Became active void GameObjectDidBecomeInactive(); // Became inactive void GameObjectWasDestroyed(); // Destroyed void AddComponent(Component*); void RemoveComponent(Component*); Component* GetComponent(Component::Type); void Update(float dt); void Render();}class Scene() { std::vector<GameObject*> m_vGameObjectList; // You can be more creative bool m_bRenderWhenNotActive; void SceneWasCreated(); void SceneDidBecomeActive(); void SceneDidBecomeInactive(); void SceneWasDestroyed(); void AddGameObject(GameObject*); void RemoveGameObject(GameObject*); GameObject* GetGameObject(std::string); void Update(float dt); void Render();}class Game { std::vector<Scene*> m_vSceneStack; bool Initialize(); // Initializes the game (Keyboard, mouse, managers, etc) void Shutdown(); // Destroys the game (blows away scene stack, game objects etc, shuts down managers) Scene* CreateNewScene(); // Creates a scene object GameObject* CreateGameObject(); // Creates a game object Component* CreateComponent(Component::Type); // Creates a component void DestroyScene(Scene*); // Delete the scene void DestroyGameObject(GameObject*); // Delete the game object void DestroyComponent(Component*); // Delete the component void Update(float dt); // Updates the top most object void Render();}[/source]

A little bit of implementation
[source lang="cpp"]void Renderer::Awake() { m_eComponentType = RENDERER; m_bIsActive = true; } // Do other stuff to set up your buffers and texturesvoid Renderer::Destroy() { } // Release your buffers and textures herevoid Renderer::Start() { } // Don't carevoid Renderer::Stop() { } // Don't carevoid Renderer::Update(float dt) { } // Don't carevoid Renderer::Render() { BindBuffer(m_nBufferId); BindTexture(m_nTextureId); RenderBatch(); }void Enemy::Awake() { m_eComponentType = ENEMY; m_bIsActive = true; } // Do other enemy stuff, like ai initialization and whatnotvoid Enemy::Destroy() { } // Shutdownvoid Enemy::Start() { if (m_pAStar == 0) m_pAStar = AStar::Factory::NewPathFinder(); } // I don't know, just arbitrary shitvoid Enemy::Stop() { if (m_pAStart != 0) AStar::Factory::Destroy(m_pAStar); m_pAStar = 0; } // Make sure you shut shit downvoid Enemy::Update(float dt) { m_pAStar->Find("Player"); AttackPlayerWhenPossible(); }// Be creative, make your player and ui components and junk!// For extra credit, make a script component, being able to script your engine will make life easy.class GameObject() { std::vector<Component*> m_vComponents; std::string m_strGameObjectName; bool m_bIsActive; void GameObject::GameObjectWasCreated() { m_bIsActive = true; } // Depending on your game, go crazyvoid GameObject::GameObjectDidBecomeActive() { // Let all the active components know, they are active again! for (int i = 0, size = m_vComponents.size(); i < size; ++i ) { if (m_vComponents[i].m_bIsActive) m_vComponents[i].Start(); }} void GameObject::GameObjectDidBecomeInactive() { // Active components, go away! for (int i = 0, size = m_vComponents.size(); i < size; ++i ) { if (m_vComponents[i].m_bIsActive) m_vComponents[i].Stop(); }} void GameObject::GameObjectWasDestroyed() { DestroyAllComponents(); } // No more game object, no more componentsvoid GameObject::AddComponent(Component* p) { // Make sure no other component of this type exists (be creative) m_vComponents.push_back(p); p->Awake(); p->Start();}void GameObject::RemoveComponent(Component* p) { // Make sure p is in component list // Get the index of the component p->Stop(); p->Destroy(); m_vComponents.erase(m_vComponents.begin() + componentIndex);}Component* GameObject::GetComponent(Component::Type t) { for (int i = 0; i < m_vComponents.size(); i < size; ++i) if (m_vComponents[i].m_eComponentType == t) // Remember, add component is only allowing one of each type return m_vComponents[i]; return 0;} void GameObject::Update(float dt) { // Propogate update to all components for (int i = 0; i < m_vComponents.size(); i < size; ++i) if (m_vComponents[i].m_bIsActive) m_vComponents[i].Update(dt);}void GameObject::Render() { // Only render render components for (int i = 0; i < m_vComponents.size(); i < size; ++i) if (m_vComponents[i].m_eComponentType == RENDERER && m_vComponents[i].m_bIsActive) ((RendererComponent*)m_vComponents[i])->Render();}void Scene::SceneWasCreated() { } // Do junk void Scene::SceneDidBecomeActive() { for (int i = 0, size = m_vGameObjectList.size(); i < size; ++i) if (m_vGameObjectList[i]->m_bIsActive) // Only active objects need to wake up m_vGameObjectList[i]->GameObjectDidBecomeActive();} void Scene::SceneDidBecomeInactive() { for (int i = 0, size = m_vGameObjectList.size(); i < size; ++i) if (m_vGameObjectList[i]->m_bIsActive) // Only active objects need to go to sleep m_vGameObjectList[i]->GameObjectDidBecomeActive();}void Scene::SceneWasDestroyed() { } // Do the opposite of create void Scene::AddGameObject(GameObject* p) { // Should not have two addresses of the same object in the list m_vGameObjectList.push_back(p); p->GameObjectWasCreated(); // Game object created in scene p->GameObjectDidBecomeActive(); // Game object is now active} void Scene::RemoveGameObject(GameObject* p) { // Make sure that p is in game object list p->GameObjectDidBecomeInactive(); // Game object no longer active p->GameObjectWasDestroyed(); // Game object is no longer in scene}GameObject* Scene::GetGameObject(std::string name) { // Retrieve first instance of game object (or null) based on name string} void Scene::Update(float dt) { for (int i = 0, size = m_vGameObjectList.size(); i < size; ++i) if (m_vGameObjectList[i]->m_bIsActive) // Only propogate to active game objects m_vGameObjectList[i]->GameObjectDidBecomeActive();} void Scene::Render() { for (int i = 0, size = m_vGameObjectList.size(); i < size; ++i) if (m_vGameObjectList[i]->m_bIsActive && m_vGameObjectList[i]->GetComponent(RENDERER) != 0) m_vGameObjectList[i]->Render(); // Don't need the above check, but figured i'd demonstrate}bool Game::Initialize() { Scene* mainMenu = CreateNewScene(); mainMenu->m_bRenderWhenNotActive = true; GameObject* mainMenuUIObject = CreateGameObject(); mainMenuUIObject->m_strGameObjectName = "MainMenuUI"; Component* mainMenuUIComponent = CreateComponent(UI); // Set all sorts of textures and junk for the main menu ui mainMenu->AddGameObject(mainMenuUIObject); mainMenuUIObject->AddComponent(mainMenuUIComponent); // You can make other scenes here, or inside of main menu, the point is // each new scene will be pushed on top of the scene stack. // How the stack may be used: // Push main menu, push audio options, set volume, pop audio options // Press play, pushes the game, now your playing // Exit to menu pops the game, now your back to the main menu}void Game::Shutdown() { // Tedious cleanup} Scene* Game::CreateNewScene() { return new Scene(); // You have a factory, be more creative} GameObject* Game::CreateGameObject() { return new GameObject(); // Same as above} Component* Game::CreateComponent(Component::Type t) { switch (t) { // Return appropriate component subclass }} void Game::DestroyScene(Scene* s) { delete s; // I'm getting bored of typing all this} void Game::DestroyGameObject(GameObject* p) { delete p;}void Game::DestroyComponent(Component* p) { delete p;} void Game::Update(float dt) { // Only update the top most scene m_vSceneStack[m_vSceneStack.size() - 1]->Update(dt);}void Game::Render() { // Render all renderable scenes // So, if you have the following stack // main menu > game > options // and game and options are both rendering, and options are opaque // It just looks like game is paused under options. for (int i = 0; i < m_vSceneStack.size() - 1; ++i) if (m_vSceneStack[i]->m_bRenderWhenNotActive) m_vSceneStack[i]->Render(); m_vSceneStack[m_vSceneStack.size() - 1]->Render();}[/source]

The limitations of this system are obvious, but you could easily make script components, and add a lot of power to your engine.

Hope this helps!

Edit, didn't see L.Spyro's response when i wrote this. Consider this the long version.

Edited by uglybdavis, 22 October 2012 - 08:01 PM.

[rdragon1]

For example, objects don’t exist inside the game, they exist inside scenes.

What about game objects that need to persist across scenes? Like puzzle logic state that spans levels? Or the main player character + all of his stats/score/whatever?

[L. Spiro]

Data that needs to persist across scenes/states goes on the Game class. That is specifically its job if nothing else.
But that does not mean the 3D or 2D rendering data for your main character etc. That just means your current level, current HP, etc. The bare minimum that could be accessed by any part of the game at any time.

[EDIT]
Note that I failed to mention that all that data that belongs just to one game or another should be part of your “MyGame” class which inherits from “Game”.
Game itself is general across all games and should obviously not be the place for that kind of data.
[/EDIT]


L. Spiro

Edited by L. Spiro, 23 October 2012 - 01:53 PM.

[uart777]

To represent different scenes:

enum { SCENE_LOGO, SCENE_TITLE, SCENE_OPTIONS, SCENE_PLAY, SCENE_GAME_OVER };
int c_scene=SCENE_LOGO; // current scene

[nox_pp]

What about game objects that need to persist across scenes? Like puzzle logic state that spans levels? Or the main player character + all of his stats/score/whatever?

Segregate your shared data, so that you can better identify the character of it--what it means and where it comes from. Then, you don't need to expose it globally in your Game class, you can just pass it between scenes that need it (as parameters.) Yes, it's more work, but it's also more explicit, and ultimately much safer.

Is this poor design?

Yes, for a variety of reasons. What Servant of the Lord said is true, and will work well if you want to carry on with your design, but the design is flawed. Think to yourself, will a virtual Update method always be enough to fully achieve the purpose of a particular GameObject? The answer is no. Which means that you will unavoidably have to pluck GameObjects out of wherever they polymorphically reside, and cast them to their proper types. This is folly. How can you be sure of the types, except by experimentation with dynamic_cast or divine knowledge? Don't store heterogeneous types homogeneously. You might think that they're homogeneous types, because they derive from the same base class, but they're nearly useless if they're never casted.

It can seem messier to deal only with concrete types in this instance, but it's really only being more honest about what is actually happening. Furthermore, there should be a strict separation between GAME code and ENGINE code. This looks like Game code to me, so it doesn't really matter if it's directly oriented toward your game, and concretely references your derived GameObjects. If you've got too many derived GameObjects for this to be feasible, that's merely an indicator that something else is wrong.

[uart777]

nox_pp: I agree that OOP sucks C++ is for newbies who know nothing about real programming, but I must admit that uglybdavis presented some smart code. I admire that.

typedef struct {
void *p;
int x, y, w, h,
bpp, key, alpha;
} IMAGE;

int load_image(IMAGE *i, char *file);
void move_image(IMAGE *i, int x, int y);
void draw_image(IMAGE *i);

My ASM programming site: http://sungod777.zxq.net/

[L. Spiro]

To represent different scenes:

enum { SCENE_LOGO, SCENE_TITLE, SCENE_OPTIONS, SCENE_PLAY, SCENE_GAME_OVER };
int c_scene=SCENE_LOGO; // current scene

And then what? A switch case to handle the logic between different scenes based on just that integer?
General Game/Engine Structure

nox_pp: I agree that OOP sucks C++ is for newbies who know nothing about real programming, but I must admit that uglybdavis presented some smart code. I admire that.

typedef struct {
void *p;
int x, y, w, h,
bpp, key, alpha;
} IMAGE;

int load_image(IMAGE *i, char *file);
void move_image(IMAGE *i, int x, int y);
void draw_image(IMAGE *i);

I don’t get the point of this post.
No one asked for opinions on object-oriented programming nor do IMAGE structures (with C code trying to mimic objects with them) have anything to do with the topic.
If you are trying to imply that the original poster should abandon C++ and code in the style you proposed (because C++ is for newbies?), I would point out that your proposed style leaves much to be desired.

• const-correctness. Use “const char *”, not “char *”. Does draw_image() modify the image pointer? If so, why? If not, why is the pointer not const?
• Type-appropriateness. When will an image have a negative width or height? Don’t use signed types for things that are unsigned. You chose “int” out of laziness.
• Why do you have to move the image in a separate step from drawing it? This is superfluous. We aren’t working with lines where the previous position of the line might have a desirable side effect on the current draw operation.
• You basically just manually implemented objects in C. The class is “IMAGE” and it has 3 methods:
  
• bool load_image( const char * file )
• void move_image( int x, int y )
• void draw_image()
  Why would you not just use C++ instead? What exactly is the benefit from doing it the long “non-newbie” way? It is more verbose, less obvious that load_image() was actually supposed to return bool, and you don’t get the extra benefits of virtual functions, templates, etc.

I just don’t get the point.


L. Spiro

[uglybdavis]

What about game objects that need to persist across scenes? Like puzzle logic state that spans levels? Or the main player character + all of his stats/score/whatever?

Switching betwen scenes also presents a pretty good chance to save your data to disk, then re-read it when the next scene loads. Rather than using globals to track things you could just treat data as data. This might not be a good case for a lot of things, but it works out very well with iCloud.

My ASM programming site: http://sungod777.zxq.net/

Assemblers are dumb and they suck. ASM is is for newbies who know nothing about how to code. Real programmers use butterflies. They open their hands and let the butterflies delicate wing flap once. The disturbance ripples outwards, changing the flow of the eddy currents in the upper atmosphere; these cause momentary pockets of high pressure air to form. These pockets act as lenses that deflect incoming cosmic rays focusing them to strike the drive platter and flip the desired bits. It's all explained right here: http://xkcd.com/378/

[uart777]

"I don’t get the point of this post" - I merely demonstrated that OOP is not neccessary.

"When will an image have a negative width or height?" - When it's invalid/inverted (-1/0xFFFFFFFF).

"Why do you have to move the image in a separate step from drawing it?" - Because it's faster to send less parameters, but you don't know anything about push-call sequences.

L. Spiro: Let me show how to draw/paint/airbrush/sculpt anything: http://www.facebook....7/photos_stream You think you're so right, but you don't even know what your code converts to. How can you expect anyone to use your library?

uglybdavis: Forgive me for complimenting you "Assemblers are dumb and they suck. ASM is is for newbies who know nothing about how to code" - This statement shows how little you know about programming. You defined a PIXEL wrong. Alpha should be in leftmost byte (0x*AA*BBCCDD). Stop changing byte orders You disrespect ASM because you don't know anything about the processor's language.

"It's all explained right here" - Dummy

Edited by uart777, 25 October 2012 - 02:07 AM.

[jbadams]

uglybdavis: Forgive me for complimenting you [...] This statement shows how little you know about programming.

He was joking... very obviously joking -- he even linked the comic he was referencing. Did you really think someone was seriously recommending that you should use the disturbances caused by the flapping of a butterfly's wings to deflect cosmic rays at a drive platter rather than programming?

It is however a tongue-in-cheek observation of the fact that assembly simply isn't as important as you seem to think. Your constant recommendations to learn assembly (from your own mid-90s-style website no-less) just aren't relevant in most cases. Learning assembly is valuable, and I'd even say it's something the majority of programmers should take the time to do eventually, but it just isn't useful to beginners and it isn't relevant to the original poster's question or any of the following discussion.

This statement shows how little you know about programming. You defined a PIXEL wrong. Alpha should be in leftmost byte (0x*AA*BBCCDD).

Actually, THIS statement shows how little YOU know about programming. You've now asserted the alpha should always be the leftmost byte of a pixel -- and that to do otherwise is incorrect -- more than once, but as has already been pointed out to you, that's simply a convention and isn't significantly more or less popular than alternatives.

"I don’t get the point of this post" - I merely demonstrated that OOP is not neccessary.

Actually, you sort of showed -- in a rather poor way that was probably meaningless to the original poster given you neglected to to say what you were doing rather than just dumping some code -- that OO can be implemented in C even though the language doesn't explicitly support the paradigm through language features. OOP is a set of guiding principles for designing code -- not just use of the "class" (or related) keyword(s) -- and like any other paradigm is useful in some situations when used correctly, but isn't universally applicable.

Sharing your opinion and contributing to the discussion are good. Acting like you're better than everyone else while spouting nonsense and spamming your (often irrelevant) website link are not. Drop the attitude...

L. Spiro: Let me show how to draw/paint/airbrush/sculpt anything

...and consider this an official moderator instruction to stop bringing up irrelevant stuff to show off every time you feel you've been challenged. This topic isn't even slightly about art or yours -- or anyone else's -- artistic abilities, and I recall this isn't the first topic in which you've done so.
But before you accuse me of abusing moderator powers or anything of the sort, note that you're welcome to challenge anything I've said as long as you stay within the site's rules -- you're only being instructed not to post things that are completely off topic.

[L. Spiro]

You think you're so right, but you don't even know what your code converts to.
Because it's faster to send less[sic] parameters, but you don't know anything about push-call sequences.

Actually I have written a C compiler, a disassembler, and an assembler. And a debugger and everything else listed here: http://memoryhacking.com/feature.php

L. Spiro: Let me show how to draw/paint/airbrush/sculpt anything: http://www.facebook....7/photos_stream

I don’t understand why you posted this completely unrelated bit, but, since we are sharing, I drew this when I was 12:

http://l-spiro.devia...gallery/4844241

uglybdavis: Forgive me for complimenting you "Assemblers are dumb and they suck. ASM is is for newbies who know nothing about how to code" - This statement shows how little you know about programming.

By the way, he didn’t attack you. It was a joke and a reference to xkcd.



It would generally be better if you didn’t try to make yourself sound amazing in a place such as this.
You will invariably find others who are much more knowledgeable and skilled than yourself. Of course, that is usually a good thing, just not when you are trying to pick fights.


L. Spiro

Edited by L. Spiro, 25 October 2012 - 08:24 AM.

[slayemin]

Back on topic....

To OP: I follow a similar pattern. I consider myself somewhat of a novice at designing game architectures, so take what I say with a grain of salt:
I have a base class which every game object inherits from (already may be a bad idea). The class is abstract and pretty much worries about assigning an incrementing ID, possibly accepting a given name, maintaining a queue of messages, and enforcing the implementation of update and init functions for classes which inherit from it, and acting as a generic object which can be used by containers (pretty much polymorphism). That's it. Any inheriting classes will extend this base class.

Here is the C# code for my base class:

public abstract class GBO
{
	 UInt64 m_ID;
	 static UInt64 m_nextID = 0;
	 protected string m_name;
	 public Queue<GameMessage> Messages = new Queue<GameMessage>(10);
	 
	 public GBO()
	 {
		  m_ID = m_nextID++;
	 }

	 public GBO(string Name)
	 {
		  m_ID = m_nextID++;
		  m_name = Name;
	 }
	 public UInt64 ID
	 {
		  get { return m_ID;}
	 }
	 public string Name
	 {
		  get { return m_name;}
		  set { m_name = value;}
	 }
	 public abstract void Update();
	 public abstract void Start();		  //I use "Start" instead of "Init" for Unity3D
}

Looking at my base object and thinking about it, it does have some weaknesses:
If I decide to create particle engine and each particle is a GBO class, do I really care about the name of a particle or any game messages it may have generated? Not really. I could slice those two variables out. The ID is mostly used as a key for dictionaries and hash tables, but would a particle ever be stored in a hash table or dictionary? Not really. So, if I slice that out too, then my base class would just have an abstract "Start()" and "Update()" method. Do I even need those? I already know that all of my game objects have to implement initialization and update functions, so enforcing it is a bit of a moot point and possibly restrictive since they don't have input parameters. I might as well have a completely blank base class to support the most flexibility... but why even have an empty class if it doesn't do anything? Do I even NEED a base class?
"Oh, what about using the base class as a generic container object? That way, you can have a single list of all your objects in the game and call their update() method!"
Well, every inheriting class would have a corresponding manager class. The manager class itself can have update called on it and we'll let the manager worry about updating its objects.
Instead of:

foreach(GBO obj in m_allObjects)
	 obj.update();

we can do this:

MageMgr.Update();
MonsterMgr.Update();
PlayerMgr.Update();
BulletsMgr.Update();

Is this "better"? One immediate disadvantage is that we explicitly have to create and call the update functions for every list of items we have in the game. This adds extra programmer overhead. But, is there an advantage to asking the manager to update its contained items? I think so. The manager can worry about the game logic in regards to the object. So, for example, your mage manager would not only call update() on all of the mages in its list, it would also manage the list of mages by removing any mages which are dead (hitpoints >= 0) or perform any other trivial object specific management.
If you really like the super simple single update, you can let your manager classes derive from an abstract manager class which has an update function. Then, you'd have a list of managers, for which you update every frame:

foreach(GameObjMgr mgr in m_managers)
	 mgr.update();

Then, you just have to worry about instantiating a manager class and inserting it into the manager list.

Anyways, I really don't know much about "good" software engineering and game architecture. I may be oversimplifying the core of game development: Managing lists of "stuff" and applying rules to them.

[uart777]

spiro: Cool drawing

jbadams: Sorry, it just seems that gamedev has changed so much since it was released. This defensive-ness is caused by the disrespect towards ASM programmers like Lamothe and myself. "Acting like you're better than everyone else" - Ultimately, no one is better than anyone else. We're all just little specks of Stardust. I apologize if this is your perception of me. "You've now asserted the alpha should always be the leftmost byte of a pixel" - Yes, that's how it was originally: AA.RR.GG.BB. Otherwise, it requires shift+and. Why change it? Why store things upside down and backwards? Why cause millions of headaches?

Back to the subject: OOP is not required. Never needed it.

Edited by uart777, 25 October 2012 - 08:40 AM.

[L. Spiro]

spiro: Cool drawing

Coming from an artist such as yourself, that is a compliment. Thank you.
Your Simba t-shirt is also top-notch quality.

jbadams: Sorry, it just seems that gamedev has changed so much since it was released. This defensive-ness is caused by the disrespect towards ASM programmers like Lamothe and myself. "Acting like you're better than everyone else" - Ultimately, no one is better than anyone else. We're all just little specks of Stardust. "You've now asserted the alpha should always be the leftmost byte of a pixel" - Yes, that's how it was originally: AA.RR.GG.BB. Otherwise, it requires shift+and. Why change it? Why store things upside down and backwards? Why cause millions of headaches?

How it was originally?
I feel that you are a prime example of what was mentioned here.
You learned early-on what compilers do internally and took it to heart.
You changed the way you coded based on what you learned from one compiler. You didn’t know that other compilers behave differently and may easily generate different code.
I can tell you for sure that there are rare cases in which my C compiler will generate horribly slow code for some switch cases.

How C++ code becomes machine-language is not strictly specified and you should understand that what my code “becomes” can vary depending on the compiler I use.

Don’t spend your time studying how some compiler created some code.
It would be better to spend your time reading the C/C++ specifications, and if you are so inclined make your own compiler. You will definitely learn a lot that way.


L. Spiro

[FLeBlanc]

jbadams: Sorry, it just seems that gamedev has changed so much since it was released.

Ah, yes, I remember the old gamedev.net. You know, back in August of 2012 when you joined. Those were the days, eh?

This defensive-ness is caused by the disrespect towards ASM programmers like Lamothe and myself.

I've heard of Lamothe. Got a couple of his books. Much respect for that dude, he's been around awhile.

Never heard of you, though.

"You've now asserted the alpha should always be the leftmost byte of a pixel" - Yes, that's how it was originally: AA.RR.GG.BB. Otherwise, it requires shift+and. Why change it? Why store things upside down and backwards? Why cause millions of headaches?

How things were originally has exactly 0 bearing on how things are now. Pixel formats now have everything to do with hardware support. Modern hardware can handle RGBA data in so many different formats, that you are basically free to pick your preferred method.

[uart777]

Spiro: My only intention is to defend ASM programmers like Andre Lamothe, Michael Abrash and Diane Gruber, the queen of graphics programming. She's hot, too Wouldn't you love to have a girl like her who does programming?
"Coming from an artist such as yourself, that is a compliment" - Thank you ;) Simba's my baby.

"I can tell you for sure that there are rare cases in which my C compiler will generate horribly slow code" - Creating a HL compiler is all about converting standard/infix expressions to RPN. You can perform optimizations in RPN format: Resolve constant subexpressions, replace mul/div with shifts by power of 2, reorder cumulative operations, double jmps/jxx, etc.

Jason: Sorry again for being way off subject, but the issue was brought up and I responded. Spiro, let's create another post about writing HL compilers

[FLeBlanc]

Why would they need defending? Did anyone here attack them? Are you a troll account?