So, I'm writing code for a class that will go into a library that will be used by others. This class will intercept and process incoming messages (details are not important but it's using the activemq-cpp library). The outline of this consumer class is
class MessageConsumer {
...
public:
void runConsumer();
virtual void onMessage(const Message* message);
}
where runConsumer() sets up the connection and starts listening and onMessage() is called when a message is received.
My questions is this: People who'll use this code will each have their own way of processing the different messages. How can I keep MessageConsumer generic but offer this flexibility, while keeping their code simple?
Two options:
Should they inherit a new class from MessageConsumer and write their own onMessage()?
Should they pass a pointer to a message handling function to MessageConsumer?
What do you think, which option is better and why?
Thanks!
In one approach, clients are allowed to register a callback and then the MessageConsumer invokes the registered callback. This is something like an observer/broadcast design pattern.
The second approach, where clients have to inherit and override MessageConsumer would be something like Strategy design pattern.
Basic design goals suggest to use the weakest relationship to promote loose coupling. Since inhertiance is a stronger relationship as compared to a simple association, everything else being the same Approach 1 is preferred.
From Herb's article
"Inheritance is often overused, even
by experienced developers. Always
minimize coupling: If a class
relationship can be expressed in more
than one way, use the weakest
relationship that's practical. Given
that inheritance is nearly the
strongest relationship you can express
in C++ (second only to friendship),
it's only really appropriate when
there is no equivalent weaker
alternative."
But as James points out, it is tough to comment unless the overall design constraints are known clearly.
Inheritance will make your library more OO-friendly and may improve readability. But really, the choices are about the same since the compiler will check that the user has supplied the function (assuming you declare a pure virtual handler in the base class), and the underlying mechanism will be accomplished via a pointer anyway (virtual table in the case of inheritance).
Pure virtual functions allow the compiler to check that the client code implements the handler. Virtual dispatch is active immediately after an object is constructed, and someone looking at the derived class can reason accurately about its handling. Data needed for the handling can be conveniently and clearly grouped into the derived class. Factories can still select a particular derived class to instantiate.
Function pointers are run-time state, so there's a little more care needed to initialise them in a timely fashion, for optional run-time checks on their being set and error handling, and to reason about which set is in effect during program execution. With that comes more freedom to vary them within the lifetime of the object.
A third alternative is a template policy class, or the Curiously Recurring Template Pattern, to lock in the behaviours at compile time. This potentially allows inlining of callbacks, dead-code elimination and other optimisations.
virtual function or tepmlated functor are the way to go. These approaches give greater flexibility ad looser coupling than function pointer one.
To illustrate that - function pointer approach can be wrapped with first two, but not vice-versa.
void cbFunction();
class Interface {
virtual void act() =0 ;
};
class CbFuctionWrapper:public Interface {
public:
virtual void act() {cbFunction()};
};
class AnotherImpl: public Interface {
Context _c; // You can't pass additional context with usual function without downcasting, but OO is all about that.
public:
virtual void act() {...}
}
Related
I'm writing a Window class which propagates different types of events, listed in
enum Event {WINDOW_ClOSE=0x1, WINDOW_REDRAW=0x2, MOUSE_MOVE=0x4, ...};
to objects which have registered for notification with the window. For each type of event, I have an abstract class which any object must extend in order to allow notification. To react to, say, a MOUSE_MOVE event, my object would inherit from MouseMoveListener, which has a process_mouse_move_event() method which is called by Window. Listening to many events can be combined by extending multiple of these classes, which all inherit from the EventListener base class. To register an object, I would call
void Window::register(EventListener* object, int EventTypes)
{
if(EventTypes&WINDOW_CLOSE)
/* dynamic_cast object to WindowCloseListener*, add to internal list of all
WindowCloseListeners if the cast works, else raise error */
if(EventTypes&MOUSE_MOVE)
/* dynamic_cast object to MouseMoveListener*, add to internal list of all
MouseMoveListeners if the cast works, else raise error */
...
}
This works fine, but my gripe is that EventListener is completely empty and that seems code smelly to me. I know I could avoid this by removing EventListener altogether and having a separate Window::register for each type of event, but I feel that this would blow up my interface needlessly (especially since methods other than register might crop up with the same problem). So I guess I am looking for answers that either say:
"You can keep doing it the way you do, because ..." or
"Introduce the separate Window::register methods anyway, because ..." or of course
"You are doing it all wrong, you should ...".
EDIT:
From the link in Igors comment: What I do above only works if there is at least one virtual member in EventListener for example a virtual destructor, so the class is not technically completely empty.
EDIT 2:
I prematurely accepted n.m.'s solution as one of the type "I'm doing it all wrong". However, it is of the second type. Even if I can call EventListener->register(Window&) polymorphically, Window needs to implement a highly redundant interface (in terms of declared methods) that allows EventListeners to register for selective notification. This is equivalent to my alternative solution described above, only with the additional introduction of the EventListener class for no good reason. In conclusion, the canonical answer seems to be:
Don't do dynamic_cast + empty base class just to avoid declaring many similar functions, it will hurt you when maintaining the code later. Write the many functions.
EDIT 3:
I found a solution (using templates) which is satisfactory for me. It does not use an empty base class any more and it does not exhibit the maintenance problem pointed out by n.m.
object->registerWindow (this, EventTypes);
Of course you need to implement registerWindow for all EventListener heirs. Let them check for event types which are relevant to them.
UPDATE
If this means you need to redesign your code, then you need to redesign your code. Why is it so? Because dynamic_cast is not a proper way to do switch-on-types. It is not a proper way because every time you add a class in your hierarchy, you need to go over and possibly update all switches-by-dynamic-cast in your old code. This becomes very messy and unmaintainable very quickly, and this is exactly the reason why virtual functions were invented.
If you do your switch-on-types with virtual functions, every time you change your hierarchy you have to do... nothing. The virtual call mechanism will take care of your changes.
This is what I ended up doing:
template <int EventType> void register_(EventListener<EventType> Listener)
{
// do stuff with Listener, using more templates
};
It turned out that static polymorphism was better suited for my needs - I just wanted to avoid writing
register_mouse_motion_event(...)
register_keyboard_event(...)
and so on. This approach also nicely eliminates the need for an empty base class.
My team has written several C++ classes which implement event handling via pure virtual callbacks - for example, when a message is received from another process, the base class which handles IPC messaging calls its own pure virtual function, and a derived class handles the event in an override of that function. The base class knows the event has occurred; the derived class knows what to do with it.
I now want to combine the features provided by these base classes in a higher-level class, so for example when a message arrives from another process, my new class can then forward it on over its network connection using a similar event-driven networking class. It looks like I have two options:
(1) composition: derive classes from each of the event-handling base classes and add objects of those derived classes to my new class as members, or:
(2) multiple inheritance: make my new class a derived class of all of the event-handling base classes.
I've tried both (1) and (2), and I'm not satisfied with my implementation of either.
There's an extra complication: some of the base classes have been written using initialisation and shutdown methods instead of using constructors and destructors, and of course these methods have the same names in each class. So multiple inheritance causes function name ambiguity. Solvable with using declarations and/or explicit scoping, but not the most maintainable-looking thing I've ever seen.
Even without that problem, using multiple inheritance and overriding every pure virtual function from each of several base classes is going to make my new class very big, bordering on "God Object"-ness. As requirements change (read: "as requirements are added") this isn't going to scale well.
On the other hand, using separate derived classes and adding them as members of my new class means I have to write lots of methods on each derived class to exchange information between them. This feels very much like "getters and setters" - not quite as bad, but there's a lot of "get this information from that class and hand it to this one", which has an inefficient feel to it - lots of extra methods, lots of extra reads and writes, and the classes have to know a lot about each other's logic, which feels wrong. I think a full-blown publish-and-subscribe model would be overkill, but I haven't yet found a simple alternative.
There's also a lot of duplication of data if I use composition. For example, if my class's state depends on whether its network connection is up and running, I have to either have a state flag in every class affected by this, or have every class query the networking class for its state every time a decision needs to be made. If I had just one multiply-inherited class, I could just use a flag which any code in my class could access.
So, multiple inheritance, composition, or perhaps something else entirely? Is there a general rule-of-thumb on how best to approach this kind of thing?
From your description I think you've gone for a "template method" style approach where the base does work and then calls a pure virtual that the derived class implements rather than a "callback interface" approach which is pretty much the same except that the pure virtual method is on a completely separate interface that's passed in to the "base" as a parameter to the constructor. I personally prefer the later as I find it considerably more flexible when the time comes to plug objects together and build higher level objects.
I tend to go for composition with the composing class implementing the callback interfaces that the composed objects require and then potentially composing again in a similar style at a higher level.
You can then decide if it's appropriate to compose by having the composing object implement the callback interfaces and pass them in to the "composed" objects in their constructors OR you can implement the callback interface in its own object possibly with a simpler and more precise callback interface that your composing object implements, and compose both the "base object" and the "callback implementation object"...
Personally I wouldn't go with an "abstract event handling" interface as I prefer my code to be explicit and clear even if that leads to it being slightly less generic.
I'm not totally clear on what your new class is trying to achieve, but it sounds like you're effectively having to provide a new implementation somewhere for all of these abstract event classes.
Personally I would plump for composition. Multiple inheritance quickly becomes a nightmare, especially when things have to change, and composition keeps the existing separation of concerns.
You state that each derived object will have to communicate with the network class, but can you try and reduce this to the minimum. For instance, each derived event object is purely responsible for packaging up the event info into some kind of generic packet, and then that packet is passed to the network class to do the guts of sending?
Without knowing exactly what your new class is doing it's hard to comment, or suggest better patterns, but the more I code, the more I am learning to agree with the old adage "favour composition over inheritance"
I'm making a Gui Api for games. The user can always use inheritance on the widget and override, but I want callbacks. I want to use a templated callback system:
so if they want to have one for the mouse they inherit from a version of the templated callback base with mouseargs:
So the base would look like this:
template <typename T>
class AguiEventCallback {
public:
virtual void callback(AguiWidget* sender, T arg) = 0;
};
Is it a good idea to mix templates with polymorphism like this? Would I be better off creating callbacks for each of the types I need (mouse, keyboard, gamepad, etc)?
Thanks
Have a look at boost::function and boost::bind. Accept a function object with a defined parameter list for particular events, and callers can do what they want.
This gives callback implementations lots of flexibility and the object generating the events requires even less knowledge of the callback implementation.
For example:
typedef boost::function<void (AguiWidget* sender)> CallbackFunc;
void register_callback(CallbackFunc const& f);
And the client:
class Caller {
void do_register() { register_callback(bind(&Caller::event, this, 123, _1)); }
void event(int arg, AguiWidget* sender) { ... }
};
Just showing function/bind, many other issues ignored; eg. memory management, object lifetime.
Using a template the way you have is fine sometimes. There are issues due to the fact that you must give your template a virtual destructor and that
If you inline your virtual destructor (as you do with most template functions) some compilers find it hard to stick to the One Definition Rule, particularly if the library is used across libraries.
If you do not inline your virtual destructor you have to instantiate every type you are going to use with that template. This is my own preferred approach.
For a callback, you do have the option of using boost::function. This avoids having to derive classes from your template, create them with new and probably stick them into a shared_ptr somewhere. The downside of boost::function as a callback, I have found, is that it is harder to debug into if something goes wrong. Beware of that issue.
Momentarily accepting your virtual dispatch solution, what your templated approach guarantees is a uniformity in the callback function name and arguments. Sadly, that will force a lot of other code to disambiguate which callback is being invoked / overridden, probably causing more trouble than good.
That said, as janm said other options exist. Functors are more powerful (you can change them on an existing object at run-time, you can have lists of observers) but also have to be initialised at the right time (pure virtual functions effectively remind the programmer to supply them at compile time), and introduce a bigger variety of run-time states to reason about and understand.
You might also be able to use a template policy or the Curiously Recurring Template Pattern to supply behaviours at compile time, allowing inlining, dead code elimination, type-specific behaviours and other optimisations.
In addition to other answers here you might have a look to the boost::signal library. It implements a signal/slot mechanism which is indeed useful for GUIs. Performance is not as good as you would expect (costs more than a call to a virtual method) but for a GUI it's just fine.
The boost::signal library can also be used together with boost::bind and this combo is very powerful.
I don't like using inheritance very much for callbacks. It spawns a lot of classes with just 1 method most of the time. It's C++ not Java. you have functions, use them :)
I have an interface class similar to:
class IInterface
{
public:
virtual ~IInterface() {}
virtual methodA() = 0;
virtual methodB() = 0;
};
I then implement the interface:
class AImplementation : public IInterface
{
// etc... implementation here
}
When I use the interface in an application is it better to create an instance of the concrete class AImplementation. Eg.
int main()
{
AImplementation* ai = new AIImplementation();
}
Or is it better to put a factory "create" member function in the Interface like the following:
class IInterface
{
public:
virtual ~IInterface() {}
static std::tr1::shared_ptr<IInterface> create(); // implementation in .cpp
virtual methodA() = 0;
virtual methodB() = 0;
};
Then I would be able to use the interface in main like so:
int main()
{
std::tr1::shared_ptr<IInterface> test(IInterface::create());
}
The 1st option seems to be common practice (not to say its right). However, the 2nd option was sourced from "Effective C++".
One of the most common reasons for using an interface is so that you can "program against an abstraction" rather then a concrete implementation.
The biggest benefit of this is that it allows changing of parts of your code while minimising the change on the remaining code.
Therefore although we don't know the full background of what you're building, I would go for the Interface / factory approach.
Having said this, in smaller applications or prototypes I often start with concrete classes until I get a feel for where/if an interface would be desirable. Interfaces can introduce a level of indirection that may just not be necessary for the scale of app you're building.
As a result in smaller apps, I find I don't actually need my own custom interfaces. Like so many things, you need to weigh up the costs and benefits specific to your situation.
There is yet another alternative which you haven't mentioned:
int main(int argc, char* argv[])
{
//...
boost::shared_ptr<IInterface> test(new AImplementation);
//...
return 0;
}
In other words, one can use a smart pointer without using a static "create" function. I prefer this method, because a "create" function adds nothing but code bloat, while the benefits of smart pointers are obvious.
There are two separate issues in your question:
1. How to manage the storage of the created object.
2. How to create the object.
Part 1 is simple - you should use a smart pointer like std::tr1::shared_ptr to prevent memory leaks that otherwise require fancy try/catch logic.
Part 2 is more complicated.
You can't just write create() in main() like you want to - you'd have to write IInterface::create(), because otherwise the compiler will be looking for a global function called create, which isn't what you want. It might seem like having the 'std::tr1::shared_ptr test' initialized with the value returned by create() might seem like it'd do what you want, but that's not how C++ compilers work.
As to whether using a factory method on the interface is a better way to do this than just using new AImplementation(), it's possible it'd be helpful in your situation, but beware of speculative complexity - if you're writing the interface so that it always creates an AImplementation and never a BImplementation or a CImplementation, it's hard to see what the extra complexity buys you.
"Better" in what sense?
The factory method doesn't buy you much if you only plan to have, say, one concrete class. (But then again, if you only plan to have one concrete class, do you really need the interface class at all? Maybe yes, if you're using COM.) In any case, if you can forsee a small, fixed limit on the number of concrete classes, then the simpler implementation may be the "better" one, on the whole.
But if there may be many concrete classes, and if you don't want to have the base class be tightly coupled to them, then the factory pattern may be useful.
And yes, this can help reduce coupling -- if the base class provides some means for the derived classes to register themselves with the base class. This would allow the factory to know which derived classes exist, and how to create them, without needing compile-time information about them.
Use the 1st method. Your factory method in the 2nd option would have to be implemented per-concrete class and this is not possible to do in the interface. I.e., IInterface::create() has no idea exactly which concrete class you actually wish to instantiate.
A static method cannot be virtual, and implementing a non-static create() method in your concrete classes has not really won you anything in this case.
Factory methods are certainly useful, but this is not the correct use.
Which item in Effective C++ recommends the 2nd option? I don't see it in mine (though I don't also have the second book). That may clear up a mis-understanding.
I would go with the first option just because it's more common and more understandable. It's really up to you, but if your working on a commercial app then I would ask what my peers what they use.
I do have a very simple question there:
Are you sure you want to use a pointer ?
This question might seem unlogical but people coming from a Java background use new much often than required. In your example, creating the variable on the stack would be amply sufficient.
I have a base class "Foo" that has an Update() function, which I want to be called once per frame for every instance of that class. Given an object instance of this class called "foo", then once per frame I will call foo->Update().
I have a class "Bar" derived from my base class, that also needs to update every frame.
I could give the derived class an Update() function, but then I would have to remember to call its base::Update() function - nothing enforces my requirement that the base::Update() function is called because I have overriden it, and could easily just forget to (or choose not to) call the base:Update function.
So as an alternative I could give the base class a protected OnUpdate() function, which could be made overrideable, and call it from the base::Update() function. This removes the onus on me to remember to call base::Update() from the derived update function because I'm no longer overriding it. A Bar instance called "bar" will have bar->Update() called on it; this will first call the base class' Update() method, which will in turn call the overriden OnUpdate() function, performing the derived class' necessary updates.
Which solves everything. Except. What if I want to derive yet another updatable class, this time from the "Bar" class.
Baz (which derives from Bar) also has update requirements. If I put them in Baz's OnUpdate() function, I'm back to the original problem in that I'd have to remember to tell Baz's OnUpdate() function to call Bar's OnUpdate() function, otherwise Bar's OnUpdate() function wouldn't get called.
So really, I'd want Bar's OnUpdate() function to be non-overridable, and instead for it to call an overridable function after it has done whatever it needed to do, perhaps called OnUpdate2()...
And if I wanted to derive yet another class? OnUpdate3? OnUpdate4? AfterUpdate?
Is there a Better Way?
Further Info:
My specific problem domain is a 3d world. I've decided my base class is a "Locator" (an object with a location and orientation).
My first derived class is a "PhysicsObject" - a Locator that also has mass, velocity, collision information, etc.
My next derived class is a "Camera" - which derives from PhysicsObject. As well as position, and velocity, it also has information about the viewport, depth of field, etc.
MattK suggests simplifying the hierarchy - if a Locator is never referred to, incorporate it into PhysicsObject.
I'm also thinking about how I would go about turning the layout upside down and using composition instead of inheritance.
Perhaps a Camera HAS physics properties.
Perhaps a PhysicsObject HAS a location.
I'll have to think some more about this problem.
I like Uri's approach: "Observe the contract." Here's the rule - please follow it. Uri is right in that whatever kind of safeguards I try to put in, anyone could circumvent them, so perhaps in this case, the simplest solution is best. All my update() functions are going to have the requirement of calling their base::update() function.
Thanks for the help everyone!
Sounds like you want composition instead of inheritance. What if there was an interface IUpdateable, and Foo held a collection of IUpdateable objects, and called an Update method on each one every tick? Then Bar and Baz could just implement Update; your only worry would be how best to register them with Foo.
Based on your further info: You might want to consider your main object being analagous to your PhysicsObject, and using composition to include objects that implement specific behaviors, such as those of the Camera object.
That's a great question, I've encountered it many many times.
Unfortunately, there are at present no language mechanisms that I am familiar with for mainstream languages like C++ to do that, though I expect (at least in the future) for Java to have something with annotations.
I've used a variety of techniques including what you've suggested, each with pros and cons. Convulted approaches are not always worth the cost.
My view today is that if you really need to use inheritance (rather than composition), as it sounds here, then it is better to go for the elegant solution and observe the written contract. I admit, it's dangerous, but there are risks to the other approaches as well.
Generally speaking people are more careful reading the documentation of an overridden method than they are of a method they are using. In other words, while you would want to avoid "surprising" the user of your class, and can't count on him reading docs, you can count a little more on that in the case of inheritance, especially if you are the only user.
If you are presenting an API function and you expect many other individuals to override your subclass, you could put all kinds of sanity checks to ensure that the method was called, but in the end, you have to rely on the contract, just as so many standard library classes do.
I think that what you want is not easily doable with a class hierarchy.
One possible solution is to use a library that handle signal/slots (I've use sigslot http://sigslot.sourceforge.net/).
In the base class you declare a signal.
class Base : has_slots<> {
public:
Base() { SignalUpdate.connect(this, &Base::OnUpdate); }
void Update() { SignalUpdate.emit(); }
void OnUpdate() { cout << "Base::OnUpdate" << endl; }
private:
signal0<> SignalUpdate;
};
Now on each "derived" class you connect such signal with you own method
class Derived : public Base {
public:
Derived() { SignalUpdate.connect(this, &Derived::OnDerivedUpdate); }
void OnDerivedUpdate() { cout << "Derived::OnDerivedUpdate" << endl; }
};
(Note that this class no longer need to be a derivated from Base).
Now each time Update is called all methods that are connected will be called.
There are other framework that implement a similar behavior: boost signals, qt slots, libsigc++. You should try to take a look at these an see if they fit your needs.