I have a list of Parts and some of them need a pointer to an Engine, lets call them EngineParts. What I want is to find these EngineParts using RTTI and then give them the Engine.
The problem is how to design the EnginePart. I have two options here, described below, and I don't know which one to choose.
Option 1 is faster because it does not have a virtual function.
Option 2 is easier if I want to Clone() the object because without data it does not need a Clone() function.
Any thoughts? Maybe there is a third option?
Option 1:
class Part;
class EnginePart : public Part {
protected: Engine *engine
public: void SetEngine(Engine *e) {engine = e}
};
class Clutch : public EnginePart {
// code that uses this->engine
}
Option 2:
class Part;
class EnginePart : public Part {
public: virtual void SetEngine(Engine *e)=0;
};
class Clutch : public EnginePart {
private: Engine *engine;
public: void SetEngine(Engine *e) { engine = e; }
// code that uses this->engine
}
(Note that the actual situation is a bit more involved, I can't use a simple solution like creating a separate list for EngineParts)
Thanks
Virtual functions in modern compilers (from about the last 10 years) are very fast, especially for desktop machine targets, and that speed should not affect your design.
You still need a clone method regardless, if you want to copy from a pointer-/reference-to-base, as you must allow for (unknown at this time) derived classes to copy themselves, including implementation details like vtable pointers. (Though if you stick to one compiler/implementation, you can take shortcuts based on it, and just re-evaluate those every time you want to use another compiler or want to upgrade your compiler.)
That gets rid of all the criteria you've listed, so you're back to not knowing how to choose. But that's easy: choose the one that's simplest for you to do. (Which that is, I can't say based of this made-up example, but I suspect it's the first.)
Too bad that the reply stating that 'a part cannot hold the engine' is deleted because that was actually the solution.
Since not the complete Engine is needed, I found a third way:
class Part;
class EngineSettings {
private:
Engine *engine
friend class Engine;
void SetEngine(Engine *e) {engine = e}
public:
Value* GetSomeValue(params) { return engine->GetSomeValue(params); }
};
class Clutch : public Part, public EngineSettings {
// code that uses GetSomeValue(params) instead of engine->GetSomeValue(params)
}
Because GetSomeValue() needs a few params which Engine cannot know, there is no way it could "inject" this value like the engine pointer was injected in option 1 and 2. (Well.. unless I also provide a virtual GetParams()).
This hides the engine from the Clutch and gives me pretty much only one way to code it.
Related
i'm creating particle system and i want to have possibility to choose what kind of object will be showing on the screen (like simply pixels, or circle shapes). I have one class in which all parameters are stored (ParticleSettings), but without those entities that stores points, or circle shapes, etc. I thought that i may create pure virtual class (ParticlesInterface) as a base class, and its derived classes like ParticlesVertex, or ParticlesCircles for storing those drawable objects. It is something like that:
class ParticlesInterface
{
protected:
std::vector<ParticleSettings> m_particleAttributes;
public:
ParticlesInterface(long int amount = 100, sf::Vector2f position = { 0.0,0.0 });
const std::vector<ParticleSettings>& getParticleAttributes() { return m_particleAttributes; }
...
}
and :
class ParticlesVertex : public ParticlesInterface
{
private:
std::vector<sf::Vertex> m_particleVertex;
public:
ParticlesVertex(long int amount = 100, sf::Vector2f position = { 0.0,0.0 });
std::vector<sf::Vertex>& getParticleVertex() { return m_particleVertex; }
...
}
So... I know that i do not have access to getParticleVertex() method by using polimorphism. And I really want to have that access. I want to ask if there is any better solution for that. I have really bad times with decide how to connect all that together. I mean i was thinking also about using template classes but i need it to be dynamic binding not static. I thought that this idea of polimorphism will be okay, but i'm really need to have access to that method in that option. Can you please help me how it should be done? I want to know what is the best approach here, and also if there is any good answer to that problem i have if i decide to make that this way that i show you above.
From the sounds of it, the ParticlesInterface abstract class doesn't just have a virtual getParticleVertex because that doesn't make sense in general, only for the specific type ParticlesVertex, or maybe a group of related types.
The recommended approach here is: Any time you need code that does different things depending on the actual concrete type, make those "different things" a virtual function in the interface.
So starting from:
void GraphicsDriver::drawUpdate(ParticlesInterface &particles) {
if (auto* vparticles = dynamic_cast<ParticlesVertex*>(&particles)) {
for (sf::Vertex v : vparticles->getParticleVertex()) {
draw_one_vertex(v, getCanvas());
}
} else if (auto* cparticles = dynamic_cast<ParticlesCircle*>(&particles)) {
for (CircleWidget& c : cparticles->getParticleCircles()) {
draw_one_circle(c, getCanvas());
}
}
// else ... ?
}
(CircleWidget is made up. I'm not familiar with sf, but that's not the point here.)
Since getParticleVertex doesn't make sense for every kind of ParticleInterface, any code that would use it from the interface will necessarily have some sort of if-like check, and a dynamic_cast to get the actual data. The drawUpdate above also isn't extensible if more types are ever needed. Even if there's a generic else which "should" handle everything else, the fact one type needed something custom hints that some other future type or a change to an existing type might want its own custom behavior at that point too. Instead, change from a thing code does with the interface to a thing the interface can be asked to do:
class ParticlesInterface {
// ...
public:
virtual void drawUpdate(CanvasWidget& canvas) = 0;
// ...
};
class ParticlesVertex {
// ...
void drawUpdate(CanvasWidget& canvas) override;
// ...
};
class ParticlesCircle {
// ...
void drawUpdate(CanvasWidget& canvas) override;
// ...
};
Now the particles classes are more "alive" - they actively do things, rather than just being acted on.
For another example, say you find ParticlesCircle, but not ParticlesVertex, needs to make some member data updates whenever the coordinates are changed. You could add a virtual void coordChangeCB() {} to ParticlesInterface and call it after each motion model tick or whenever. With the {} empty definition in the interface class, any class like ParticlesVertex that doesn't care about that callback doesn't need to override it.
Do try to keep the interface's virtual functions simple in intent, following the Single Responsibility Principle. If you can't write in a sentence or two what the purpose or expected behavior of the function is in general, it might be too complicated, and maybe it could more easily be thought of in smaller steps. Or if you find the virtual overrides in multiple classes have similar patterns, maybe some smaller pieces within those implementations could be meaningful virtual functions; and the larger function might or might not stay virtual, depending on whether what remains can be considered really universal for the interface.
(Programming best practices are advice, backed by good reasons, but not absolute laws: I'm not going to say "NEVER use dynamic_cast". Sometimes for various reasons it can make sense to break the rules.)
I'm having trouble finding best practice information about what I believe should be a fairly common problem pattern.
I will start with a specific (software update related) example, because it makes the discussion more concrete, but the issue should be fairly generic.
Say that I have a software updater interface:
struct Software_updater {
virtual ~Software_updater() = default;
virtual void action1(const Input1& input1) = 0;
virtual void action2() = 0;
virtual bool action3(const Input2& input2) = 0;
virtual Data1 info1() = 0;
virtual Data2 info2() = 0;
// etc.
};
For my first implementation A, I am lucky, everything is straightforward.
class A_software_updater : public Software_updater {
// ...
};
A B_software_updater, however, is more complicated. Like in the A-case, it is connected to the target to update in a non-trivial manner and maintains a target connection state. But more importantly, it can update two images: the application image, and the boot loader image.
Liking what I have so far, I see no real reason to go for a refactoring, so I assume I can just build upon it. I come up with the following solution:
class B_software_updater {
public:
Software_updater& application_updater() { return application_updater_; }
Software_updater& boot_loader_updater() { return boot_loader_updater_; }
private:
class Application_updater : public Software_updater {
// ...
} application_updater_;
class Boot_loader_updater : public Software_updater {
// ...
} boot_loader_updater_;
};
I.e. I am returning non-const references to "interfaces to" member variables. Note that they cannot be const, since they mute state.
Request 1: I think the solution above is a clean one, but I would be happy to get some confirmation.
In fact, I have recently faced the issue of having to optionally provide an interface in a class, based on compile-time selection of a feature, and I believe the pattern above is a solution for that problem too:
struct Optional_interface {
virtual ~Optional_interface() = default;
virtual void action1(const Input1& input1) = 0;
virtual void action2() = 0;
virtual bool action3(const Input2& input2) = 0;
virtual Data1 info1() = 0;
virtual Data2 info2() = 0;
// etc.
};
class A_implementation {
public:
#ifdef OPTIONAL_FEATURE
Optional_interface& optional_interface() { return optional_implementation_; }
#endif
// ...
private:
#ifdef OPTIONAL_FEATURE
class Optional_implementation : public Optional_interface {
// ...
} optional_implementation_;
#endif
// ...
};
Request 2: I could not find a simple (as in: not unnecessarily complicated template-based) and clean way to express a compile-time optional inheritance at the A_implementation-level. Can you?
Better solution
Based on a comment from #ALX23z about invalidation of member variable reference upon move, I am now rejecting my initial solution (original post). That invalidation problem would not be an issue for my case, but I am in search of a generic pattern.
As usual, the solution is obvious once one has found it.
First a summary of my initial problem.
Say that I have a software updater interface (or any interface, this is just an example):
struct Software_updater {
virtual ~Software_updater() = default;
virtual void action1(const Input1& input1) = 0;
virtual void action2() = 0;
virtual bool action3(const Input2& input2) = 0;
virtual Data1 info1() = 0;
virtual Data2 info2() = 0;
// etc.
};
A B_software_updater can update two images: an application image, and a boot loader image. Therefore, it wants to provide two instances of the Software_updater interface.
A solution that is better than the one in my original post is to declare a B_application_updater and a B_boot_loader_updater, constructed from a B_software_updater&, outside of B_software_updater, and instantiated by client code.
class B_application_updater : public Software_updater {
B_application_updater(B_software_updater&);
// ...
};
class B_boot_loader_updater : public Software_updater {
B_application_updater(B_boot_loader_updater&);
// ...
};
It does have the drawback of forcing the client code to create three objects instead of only one, but I think that the cleanliness outweighs that drawback.
This will work for the optional interface too (see original post):
class A_optional_implementation : public Optional_interface {
A_optional_implementation(A_implementation&);
};
A_optional_implementation will be declared outside of A_implementation.
Applications that do not need that interface will simply not instantiate A_optional_implementation.
Additional thoughts
This is an application of the adapter design pattern!
Basically, what this answer comes down to:
An Interface class.
An Implementation class that does the job, but does not really care about the interface. It does not inherit Interface. The point of this is that Implementation could "do the job" corresponding to several interfaces, without the complexity and drawbacks of multiple inheritance (name conflicts, etc.). It could also do the job corresponding to several instances of the same interface (my case above).
An Interface_adapter class that takes an Implementation& parameter in its constructor. It inherits Interface, i.e. it effectively implements it, and that is its only purpose.
Taking a step back, I realize that this is simply an application of the adapter pattern (although Implementationin this case does not necessarily need to implement any externally defined interface - its interface is just its public member functions)!
An intermediate solution: leave the adapter classes inside the implementation class
In the solution above, I specify that the adapter classes are declared outside of the implementation classes. While this seems logical for the traditional adapter pattern case, for my case, I could just as well declare them inside the implementation class (like I did in the original post) and make them public. The client code would still have to create the implementation and adapter objects, but the adapter classes would belong to the implementation namespace, which would look nicer.
I use game state manager (intro, main menu, gameplay etc) from here. However there is one problem. A very minimalistic example:
class cApp //manages the states and gives them access to window
{
public:
cApp (RenderWindow & ref) : window(ref) {}
void changeState(cState *); //these function realy doesn't matter
void update();
void draw();
RenderWindow & window; //the same as in the article, this class not only manages state but gives them access to window etc
private:
std::vector <cState *> states;
}
The state:
class cState
{
public:
cState(cApp * ptr) : app(ptr) {}
virtual void update() = 0;
virtual void draw() = 0;
protected:
cApp * app;
}
So far everything is good. The problem is this is the part of basic framework. So the cApp is only very basic and gives access only to window. However there may be the case where the user wants to use networking in his game. Network engine isn't a part of single state so it must be at the more global (that means, cApp) level.
So the user does:
class cNetworkedApp : public cApp
{
public:
cNetworkedApp(RenderWindow & ref1, NetworkEngine & ref2)
: networking(ref2), cApp(ref1)
NetworkEngine & networking; //initialized in cNetworkedApp constructor
}
class CharacterCreationState : public cState
{
CharacterCreationState(cApp * ptr) : cState(ptr) {}
//implement pure virtual functions
void draw()
{}
void update()
{
//THE PROBLEM
//the state needs to access the network engine so casting is required
cNetworkedApp * ptr = static_cast<cNetworkedApp*>(app))
ptr->networking.sendSomething();
}
}
The only obvious solution is to include everything what may be possible in cApp, however as I said this is a framework. Of course some engines like physics engine or sound engine are things which you put into a state so that's not problem, but things like networking system must be the one object available for all states. And not every app uses it.
Do I need to redesign this code or is it okay?
Your cApp may keep with it a named list of a polymorphic type Engine, ie map<string,Engine*>, then, your user may ask the cApp if it has a given engine.
NetworkEngine would be a subclass of the pure abstract Engine.
Update
When dealing with a pointer that I you are sure that it is of the given specialized type, you should use static_cast, when you wanna query if the pointer can be casted to a type you should use dynamic_cast.
I, myself, have a safer approach for the first case, I use an assertion to guarantee that the type can be casted and use the static_cast in normal code:
Engine* fetchedEngine = cApp.fetch("network");
assert( dynamic_cast<NetworkEngine*>(fetchedEngine) != NULL );
NetworkEngine* network = static_cast<NetWorkEngine*>(fetchedEngine);
Only a object of type NetworkEngine should be putted on the "network" name, but maybe someone mistakenly put something else, the assert will make us safer without needing to worry about the overhead.
I have a C background and am a newb on C++. I have a basic design question. I have a class (I'll call it "chef" b/c the problem I have seems very analogous to this, both in terms of complexity and issues) that basically works like this
class chef
{
public:
void prep();
void cook();
void plate();
private:
char name;
char dish_responsible_for;
int shift_working;
etc...
}
in pseudo code, this gets implemented along the lines of:
int main{
chef my_chef;
kitchen_class kitchen;
for (day=0; day < 365; day++)
{
kitchen.opens();
....
my_chef.prep();
my_chef.cook();
my_chef.plate();
....
kitchen.closes();
}
}
The chef class here seems to be a monster class, and has the potential of becoming one. chef also seems to violate the single responsibility principle, so instead we should have something like:
class employee
{
protected:
char name;
int shift_working;
}
class kitchen_worker : employee
{
protected:
dish_responsible_for;
}
class cook_food : kitchen_worker
{
public:
void cook();
etc...
}
class prep_food : kitchen_worker
{
public:
void prep();
etc...
}
and
class plater : kitchen_worker
{
public:
void plate();
}
etc...
I'm admittedly still struggling with how to implement it at run time so that, if for example plater (or "chef in his capacity as plater") decides to go home midway through dinner service, then the chef has to work a new shift.
This seems to be related to a broader question I have that if the same person invariably does the prepping, cooking and plating in this example, what is the real practical advantage of having this hierarchy of classes to model what a single chef does? I guess that runs into the "fear of adding classes" thing, but at the same time, right now or in the foreseeable future I don't think maintaining the chef class in its entirety is terribly cumbersome. I also think that it's in a very real sense easier for a naive reader of the code to see the three different methods in the chef object and move on.
I understand it might threaten to become unwieldy when/if we add methods like "cut_onions()", "cut_carrots()", etc..., perhaps each with their own data, but it seems those can be dealt with by having making the prep() function, say, more modular. Moreover, it seems that the SRP taken to its logical conclusion would create a class "onion_cutters" "carrot_cutters" etc... and I still have a hard time seeing the value of that, given that somehow the program has to make sure that the same employee cuts the onions and the carrots which helps with keeping the state variable the same across methods (e.g., if the employee cuts his finger cutting onions he is no longer eligible to cut carrots), whereas in the monster object chef class it seems that all that gets taken care of.
Of course, I understand that this then becomes less about having a meaningful "object oriented design", but it seems to me that if we have to have separate objects for each of the chef's tasks (which seems unnatural, given that the same person is doing all three function) then that seems to prioritize software design over the conceptual model. I feel an object oriented design is helpful here if we want to have, say, "meat_chef" "sous_chef" "three_star_chef" that are likely different people. Moreover, related to the runtime problem is that there is an overhead in complexity it seems, under the strict application of the single responsibility principle, that has to make sure the underlying data that make up the base class employee get changed and that this change is reflected in subsequent time steps.
I'm therefore rather tempted to leave it more or less as is. If somebody could clarify why this would be a bad idea (and if you have suggestions on how best to proceed) I'd be most obliged.
To avoid abusing class heirarchies now and in future, you should really only use it when an is relationship is present. As yourself, "is cook_food a kitchen_worker". It obviously doesn't make sense in real life, and doesn't in code either. "cook_food" is an action, so it might make sense to create an action class, and subclass that instead.
Having a new class just to add new methods like cook() and prep() isn't really an improvement on the original problem anyway - since all you've done is wrapped the method inside a class. What you really wanted was to make an abstraction to do any of these actions - so back to the action class.
class action {
public:
virtual void perform_action()=0;
}
class cook_food : public action {
public:
virtual void perform_action() {
//do cooking;
}
}
A chef can then be given a list of actions to perform in the order you specify. Say for example, a queue.
class chef {
...
perform_actions(queue<action>& actions) {
for (action &a : actions) {
a.perform_action();
}
}
...
}
This is more commonly known as the Strategy Pattern. It promotes the open/closed principle, by allowing you to add new actions without modifying your existing classes.
An alternative approach you could use is a Template Method, where you specify a sequence of abstract steps, and use subclasses to implement the specific behaviour for each one.
class dish_maker {
protected:
virtual void prep() = 0;
virtual void cook() = 0;
virtual void plate() = 0;
public:
void make_dish() {
prep();
cook();
plate();
}
}
class onion_soup_dish_maker : public dish_maker {
protected:
virtual void prep() { ... }
virtual void cook() { ... }
virtual void plate() { ... }
}
Another closely related pattern which might be suitable for this is the Builder Pattern
These patterns can also reduce of the Sequential Coupling anti-pattern, as it's all too easy to forget to call some methods, or call them in the right order, particularly if you're doing it multiple times. You could also consider putting your kitchen.opens() and closes() into a similar template method, than you don't need to worry about closes() being called.
On creating individual classes for onion_cutter and carrot_cutter, this isn't really the logical conclusion of the SRP, but in fact a violation of it - because you're making classes which are responsible for cutting, and holding some information about what they're cutting. Both cutting onions and carrots can be abstracted into a single cutting action - and you can specify which object to cut, and add a redirection to each individual class if you need specific code for each object.
One step would be to create an abstraction to say something is cuttable. The is relationship for subclassing is candidate, since a carrot is cuttable.
class cuttable {
public:
virtual void cut()=0;
}
class carrot : public cuttable {
public:
virtual void cut() {
//specific code for cutting a carrot;
}
}
The cutting action can take a cuttable object and perform any common cutting action that's applicable to all cuttables, and can also apply the specific cut behaviour of each object.
class cutting_action : public action {
private:
cuttable* object;
public:
cutting_action(cuttable* obj) : object(obj) { }
virtual void perform_action() {
//common cutting code
object->cut(); //specific cutting code
}
}
Okay, so you have a load of methods sprinkled around your system's main class. So you do the right thing and refactor by creating a new class and perform move method(s) into a new class. The new class has a single responsibility and all is right with the world again:
class Feature
{
public:
Feature(){};
void doSomething();
void doSomething1();
void doSomething2();
};
So now your original class has a member variable of type object:
Feature _feature;
Which you will call in the main class. Now if you do this many times, you will have many member-objects in your main class.
Now these features may or not be required based on configuration so in a way it's costly having all these objects that may or not be needed.
Can anyone suggest a way of improving this?
EDIT: Based on suggestion to use The Null Object Design Pattern I've come up with this:
An Abstract Class Defining the Interface of the Feature:
class IFeature
{
public:
virtual void doSomething()=0;
virtual void doSomething1()=0;
virtual void doSomething2()=0;
virtual ~IFeature(){}
};
I then define two classes which implement the interface, one real implementation and one Null Object:
class RealFeature:public IFeature
{
public:
RealFeature(){};
void doSomething(){std::cout<<"RealFeature doSomething()"<<std::endl;}
void doSomething1(){std::cout<<"RealFeature doSomething()"<<std::endl;}
void doSomething2(){std::cout<<"RealFeature doSomething()"<<std::endl;}
};
class NullFeature:public IFeature
{
public:
NullFeature(){};
void doSomething(){std::cout<<"NULL doSomething()"<<std::endl;};
void doSomething1(){std::cout<<"NULL doSomething1()"<<std::endl;};
void doSomething2(){std::cout<<"NULL doSomething2()"<<std::endl;};
};
I then define a Proxy class which will delegate to either the real object or the null object depending on configuration:
class Feature:public IFeature
{
public:
Feature();
~Feature();
void doSomething();
void doSomething1();
void doSomething2();
private:
std::auto_ptr<IFeature> _feature;
};
Implementation:
Feature::Feature()
{
std::cout<<"Feature() CTOR"<<std::endl;
if(configuration::isEnabled() )
{
_feature = auto_ptr<IFeature>( new RealFeature() );
}
else
{
_feature = auto_ptr<IFeature>( new NullFeature() );
}
}
void Feature::doSomething()
{
_feature->doSomething();
}
//And so one for each of the implementation methods
I then use the proxy class in my main class (or wherever it's required):
Feature _feature;
_feature.doSomething();
If a feature is missing and the correct thing to do is ignore that fact and do nothing, you can get rid of your checks by using the Null Object pattern:
class MainThing {
IFeature _feature;
void DoStuff() {
_feature.Method1();
_feature.Method2();
}
interface IFeature {
void Method1();
void Method2();
}
class SomeFeature { /* ... */ }
class NullFeature {
void Method1() { /* do nothing */ }
void Method2() { /* do nothing */ }
}
Now, in MainThing, if the optional feature isn't there, you give it a reference to a NullFeature instead of an actual null reference. That way, MainThing can always safely assume that _feature isn't null.
An auto_ptr by itself won't buy you much. But having a pointer to an object that you lazily load only when and if you need it might. Something like:
class Foo {
private:
Feature* _feature;
public:
Foo() : _feature(NULL) {}
Feature* getFeature() {
if (! _feature) {
_feature = new Feature();
}
return _feature;
}
};
Now you can wrap that Feature* in a smart pointer if you want help with the memory management. But the key isn't in the memory management, it's the lazy creation. The advantage to this instead of selectively configuring what you want to go create during startup is that you don't have to configure – you simply pay as you go. Sometimes that's all you need.
Note that a downside to this particular implementation is that the creation now takes place the first time the client invokes what they think is just a getter. If creation of the object is time-consuming, this could be a bit of a shock to, or even a problem for, to your client. It also makes the getter non-const, which could also be a problem. Finally, it assumes you have everything you need to create the object on demand, which could be a problem for objects that are tricky to construct.
There is one moment in your problem description, that actually would lead to failure. You shouldn't "just return" if your feature is unavailable, you should check the availability of your feature before calling it!
Try designing that main class using different approach. Think of having some abstract descriptor of your class called FeatureMap or something like that, which actually stores available features for current class.
When you implement your FeatureMap everything goes plain and simple. Just ensure (before calling), that your class has this feature and only then call it. If you face a situation when an unsupported feature is being called, throw an exception.
Also to mention, this feature-lookup routine should be fast (I guess so) and won't impact your performance.
I'm not sure if I'm answering directly to your question (because I don't have any ideas about your problem domain and, well, better solutions are always domain-specific), but hope this will make you think in the right way.
Regarding your edit on the Null Object Pattern: If you already have a public interface / private implementation for a feature, it makes no sense to also create a null implementation, as the public interface can be your null implementation with no problems whatsoever).
Concretely, you can have:
class FeatureImpl
{
public:
void doSomething() { /*real work here*/ }
};
class Feature
{
class FeatureImpl * _impl;
public:
Feature() : _impl(0) {}
void doSomething()
{
if(_impl)
_impl->doSomething();
// else case ... here's your null object implementation :)
}
// code to (optionally) initialize the implementation left out due to laziness
};
This code only benefits from a NULL implementation if it is performance-critical (and even then, the cost of an if(_impl) is in most cases negligible).