I need a layer of abstraction involving QWidget that can be QGLWidget, and I wonder if there is a way to say to the compiler, "Any time you have a doubt (ambiguities) try to use the default base I give you", of course if there is ambiguities it can't resolve with the default choice it prompts errors just like it does. My aim is not have to explicitly solve each of ambiguities one by one since I will always re-direct them to the same class.
Quick setup,
#Qt inheritance (very roughly...)
class QWidget {};
class QGLWidget : public QWidget {};
#my side
class MyAbstract : public QWidget {}; //used by a factory
class MyClass1 : public MyAbstract {};
class MyClass2 : public MyAbstract, public QGLWidget{};
I'am aware compiler can't determine by its-self witch duplicated methods to use for the MyClass2 class, since QGLwidget inherits and re-implement most of the QWidget, but can I tell to the complier to use QGLWidget first since I know that's what I want ?
Qt is just an example here.
I personally doubt that this kind of automatic disambiguation is feasible in C++ at the language level.
What is possible is, case-by-case, disambiguate by explicitly giving the class whose method should be executed, like this:
QGLWidget::ambiguous_method(...
This is not what you are asking for, I know, and I am sure you already know about it. I am saying just for completeness.
On another hand, I am not sure that this kind of automatic disambiguation would be desirable or simply helpful, because the main point about multiple inheritance being "delicate" is the replication of data inside of the derived class. If you had automatic disambiguation, you would end up using sometimes (when there is no ambiguity) the partial object corresponding to a base class, and in other cases the partial object corresponding to the other base class (because of automatic disambiguation) and you would get a mosaic of things that would not make any sense, i.e. a corrupt object...
Finally, I think that this kind of automatic disambiguation would be infeasible in case you have more complex inheritance diagrams, like, following your example:
class Nasty : QGLWidget {};
class Very_nasty : Nasty, MyClass2 {};
There would be no possibility of automatic disambiguation. Indeed, say that the classes you provided form a library and that you decided, when building the library, to use MyClass2::QGLWidget as a base for disambiguation.
Now, I take your library and define two more classes like the ones I gave. Very_nasty inherits QGLWidget from Nasty and from Class2; each one has got a QGWidget inside, and overall I have 3 of them (because Class2 already inherited it twice).
Now suppose that for me, a base class for disambiguation should be Very_Nasty::Nasty::QGLWidget, given the semantics of my class. If you say that automatic disambiguation is a way to resolve ambiguities with multiple inheritance, I should be able to specify it with each case of multiple inheritance.
What would happen if I call through Very_nasty a method inherited from MyClass2?
What would happen if I call through Very_nasty a method inherited from Nasty?
They would take two different disambiguation paths. Clash.
Good answer: don't model anything as inheritance until it's absolutely necessary.
Exact answer: use virtual base class.
Related
Musing on a Sunday...
Deriving from a class brings all names from the base class into the scope of the derived class. However, it also adds the base class non-static data members to every instance of the derived class.
Is there a way to achieve the former without the latter? I'm asking in the interest of concise notation.
Obviously, when the base class doesn't have any data members, I get what I want. There are quite a few empty classes or class templates in the standard library defined to do just that - inject names summarily into a class scope through inheritance. There's even the empty base class optimization to make this as cheap as possible.
But if I wanted to do the same with a non-empty base class, I would be tempted to employ something like:
struct Bar {
using struct Foo;
};
But, alas, that's not supported by C++. My question is, is there another way to achieve the same which I overlooked?
To provide a more complete example:
struct Foo {
enum { some_constant=42 };
// data members follow here ...
};
struct Bar {
using class Foo; // this doesn't compile
int f();
};
int Bar::f() {
return some_constant; // I want to use the constant directly, without Foo::
}
One clumsy way could be to split the definitions in Foo into two classes, one with the constants (which would be an empty class I could derive from without penalty) and the other with the data members, but that looks rather like an inelegant hack to me.
If there isn't a clean way to achieve this, maybe someone can provide a rationale for why it doesn't exist, or perhaps shouldn't exist.
Deriving from a class brings all names from the base class into the scope of the derived class.
Let me stop you there. Yes, it is true that inheriting from a base class causes the (non-private) names in the base class to be accessible from the derived class definition. However, that's not why you inherit from a base class; that's merely the mechanism by which inheritance achieves its goal.
To publicly inherit from a base class is to make a statement about the relationship between the derived and base classes. You're saying that every instance of the derived class should behave like the base class in virtually all ways. Even virtual function overriding still carries with it the expectation that the derived class implementations of these methods are conceptually doing the same job, just in a way appropriate for that derived class.
This is true even of mixin-style base classes, where the base class is used to define common functionality that is imported into a particular derived class. In such interfaces, there is little expectation of a user explicitly talking to base class definitions. But this provision of common functionality is ultimately still based on the semantic idea of a derived class being a base class. And that's very important for many of them to do their job.
Consider what is probably the most prominent mixin in the C++ standard library: std::enable_shared_from_this<T>. It has non-static data members, without which it couldn't actually provide the functionality it does (well, it could, but you would have to provide some interface in your derived class to store them, so it may as well do it).
This is true of private inheritance, though there is some modification. While to the outside world, the derived class is just a derived class, to the class definition itself, it still remains a base class. It remains wholly a base class, along with all the baggage that comes along with it.
Do not mistake the mechanism for the meaning. Mechanisms are important; don't get me wrong. But those mechanisms exist to facilitate meaning.
Having a class contain everything of some other class except the non-static data members is, semantically, nonsense. It doesn't mean anything about the relationship between the types. And you've essentially admitted that the main reason you want this is so that you don't have to scope-qualify the names defined in the "base" class.
This is a mechanical reason, not a semantic one. You shouldn't employ a semantic tool like inheritance to escape the mechanical consequences of how you have chosen to design your types.
In reference to your specific example you could make the constants you want to access static, which will allow you to access them from the second class by fully qualifying with the "base" class
I am after your opinion on how best to implement an inheritance pattern in C++. I have two base classes, say
class fooBase{
protected:
barBase* b;
};
class barBase{};
where fooBase has a barBase. I intend to put these classes in a library, so that wherever I have a fooBase it can use its barBase member.
I now intend to create a specialisation of these in a specific program
class fooSpec : public fooBase{};
class barSpec : public barBase{};
Now I want fooSpec::b to point to a barSpec instead of a barBase. I know that I can just initialise b with a new barSpec, but this would require me to cast the pointer to a barSpec whenever I wanted to use specific functions in the specialisation wouldn't it?
Is there another way that this is often acheived?
Cheers.
Create a method in your specclass to cast the b into the special version.
That way instead of casting it all the time, it looks like a getter.
On the other hand OO is about programming towards interfaces and not objects. So what you are doing here looks like programming towards objects. But the is difficult to see as this example is purely theoretical.
You may consider the template solution:
template <class T>
class fooBase{
protected:
T* b;
};
and then use it as
class fooSpec : public fooBase<barSpec>{};
while ordinarily, the base would be used as fooBase<barBase>.
Is this what you want?
Normally we create a function that has the cast and returns the pointer -- and use that instead of the member directly.
Now I want fooSpec::b to point to a barSpec instead of a barBase.
There's no such thing as fooSpec::b. b belongs to fooBase, and your new class fooSpec is a (specialization of) a fooBase. You can't change the fact that b, a fooBase member, is of type barBase. This is a property of all the instances of fooBase that you can't invalidate in the particular subset of instances concerned by your specialization.
I know that I can just initialise b with a new barSpec, but this would
require me to cast the pointer to a barSpec whenever I wanted to use
specific functions in the specialisation wouldn't it?
Yes and no. Yes, you need to do that cast; but no, you don't need to do it every time. You can encapsulated in a function of fooSpec.
Is there another way that this is often acheived?
Not that I'm aware of.
this would require me to cast the pointer to a barSpec whenever I wanted to use specific functions in the specialisation wouldn't it?
That depends on whether the method you are trying to invoke is defined in the superclass and whether it is virtual.
You need to cast the pointer before invoking a method if one of the following is true...
The method belongs to the subclass only
The superclass has an implementation of the method and the subclass's implementation does not override the implementation in the superclass. This amounts to a question of whether the function is a virtual function.
Avoid data members in non-leaf classes, use pure virtual getters instead. If you follow this simple rule, your problem solves itself automatically.
This also makes most non-leaf classes automatically abstract, which may seem like an undue burden at first, but you get used to it and eventually realize it's a Good Thing.
Like most rules, this one is not absolute and needs to be broken now and then, but in general it's a good rule to follow. Give it a try.
If it looks too extreme, you may try one of the design patterns that deal with dual hierarchies such as Stairway to Heaven.
As noted in this answer:
high reliance on dynamic_cast is often an indication your design has gone wrong.
What I'd like to know is how can I call a custom function in a derived class, for which there is no same-name function in the base class, but do so using a base class pointer and perhaps without dynamic_cast if in fact there is a better way.
If this function was a virtual function defined in both, that's easy. But this is a unique function only in the derived class.
Perhaps dynamic_cast is the best way afterall?
In order to call a function of Derived class you have to obtain a pointer to derived class. As an option (depending on situation) you may want using static_cast instead of dynamic, but as you said:
it is often an indication your design has gone wrong
Also, sometimes I think it's ok to use casts. When I was designing a GUI library for a game it has a base class Widget and lots of subclasses. An actual window layout was made in an editor and later some Loader class was inflating this layout. In order to fill widgets from the layout with actual specific for each widget data (game related) I made a method for quering widget's child from a widget. This function retuned Widget* and then I dynamic_casted it to actual type. I have not found a better design for this.
Later I also found that GUI system on Android works the same way
What I'd like to know is how can I call a custom function in a derived class ... without dynamic_cast if in fact there is a better way
As indicated in the quote, it's a design issue, not an implementation issue. There's no "better way" to call that function; the "better way" is to redesign your types so that subtypes don't need to add functionality to their parents. By doing so, your types satisfy (a common interpretation of) the Liskov Substitution Principle, and are easier to use since users don't need to know about the subtypes at all.
If it's impossible or unreasonably difficult to redesign the types in such a way, then perhaps you do need RTTI. The advice doesn't say "All use of ...", just "High reliance on ...", meaning that RTTI should be a last resort, not a default approach.
This is more like an option then a real answer, so don't stone me to death.
class Derived;
class Base
{
public:
virtual Derived * getDerived()const
{
return NULL;
}
};
class Derived : public Base
{
public:
virtual Derived * getDerived()const
{
return this;
}
};
I guess you get the picture...
P.S. Mike Seymour, thanks :-)
I'm currently in the design phase of a class library and stumbled up on a question similar to "Managing diverse classes with a central manager without RTTI" or "pattern to avoid dynamic_cast".
Imagine there is a class hierarchy with a base class Base and two classes DerivedA and DerivedB that are subclasses of Base. Somewhere in my library there will be a class that needs to hold lists of objects of both types DerivedA and DerivedB. Further suppose that this class will need to perform actions on both types depending on the type. Obviously I will use virtual functions here to implement this behavior. But what if I will need the managing class to give me all objects of type DerivedA?
Is this an indicator of a bad class design because I have the need to perform actions only on a subset of the class hierarchy?
Or does it just mean that my managing class should not use a list of Base but two lists - one for DerivedA and one for DerivedB? So in case I need to perform an action on both types I would have to iterate over two lists. In my case the probability that there will be a need to add new subclasses to the hierarchy is quite low and the current number is around 3 or 4 subclasses.
But what if I will need the managing class to give me all objects of
type DerivedA?
Is this an indicator of a bad class design because I have the need to
perform actions only on a subset of the class hierarchy?
More likely yes than no. If you often need to do this, then it makes sense to question whether the hierarchy makes sense. In that case, you should separate this into two unrelated lists.
Another possible approach is to also handle it through virtual methods, where e.g. DeriveB will have a no-op implementation for methods which don't affect that. It is hard to tell without knowing more information.
It certainly is a sign of bad design if you store (pointers to) objects together that have to be handled differently.
You could however just implement this differing behaviour as an empty function in the base class or use the visitor pattern.
You can do it in several ways.
Try to dynamic_cast to specific class (this is a bruteforce solution, but I'd use it only for interfaces, using it for classes is a kind of code smell. It'll work though.)
Do something like:
class BaseRequest {};
class DerivedASupportedRequest : public BaseRequest {};
Then modify your classes to support the method:
// (...)
void ProcessRequest(const BaseRequest & request);
Create a virtual method bool TryDoSth() in a base class; DerivedB will always return false, while DerivedA will implement the required functionality.
Alternative to above: Create method Supports(Action action), where Action is an enum defining possible actions or groups of actions; in such case calling DoSth() on class, which does not support given feature should result in thrown exception.
Base class may have a method ActionXController * GetControllerForX(); DerivedA will return the actual controller, DerivedB will return nullptr.
Similarly, base class can provide method: BaseController * GetController(Action a)
You asked, if it is a bad design. I believe, that it depends on how much functionality is common and how much is different. If you have 100 common methods and only one different, it would be weird to hold these data in separate lists. However, if count of different methods is noticeable, consider changing design of your application. This may be a general rule, but there are also exceptions. It's hard to tell without knowing the context.
I am relatively new to "design patterns" as they are referred to in a formal sense. I've not been a professional for very long, so I'm pretty new to this.
We've got a pure virtual interface base class. This interface class is obviously to provide the definition of what functionality its derived children are supposed to do. The current use and situation in the software dictates what type of derived child we want to use, so I recommended creating a wrapper that will communicate which type of derived child we want and return a Base pointer that points to a new derived object. This wrapper, to my understanding, is a factory.
Well, a colleague of mine created a static function in the Base class to act as the factory. This causes me trouble for two reasons. First, it seems to break the interface nature of the Base class. It feels wrong to me that the interface would itself need to have knowledge of the children derived from it.
Secondly, it causes more problems when I try to re-use the Base class across two different Qt projects. One project is where I am implementing the first (and probably only real implementation for this one class... though i want to use the same method for two other features that will have several different derived classes) derived class and the second is the actual application where my code will eventually be used. My colleague has created a derived class to act as a tester for the real application while I code my part. This means that I've got to add his headers and cpp files to my project, and that just seems wrong since I'm not even using his code for the project while I implement my part (but he will use mine when it is finished).
Am I correct in thinking that the factory really needs to be a wrapper around the Base class rather than the Base acting as the factory?
You do NOT want to use your interface class as the factory class. For one, if it is a true interface class, there is no implementation. Second, if the interface class does have some implementation defined (in addition to the pure virtual functions), making a static factory method now forces the base class to be recompiled every time you add a child class implementation.
The best way to implement the factory pattern is to have your interface class separate from your factory.
A very simple (and incomplete) example is below:
class MyInterface
{
public:
virtual void MyFunc() = 0;
};
class MyImplementation : public MyInterface
{
public:
virtual void MyFunc() {}
};
class MyFactory
{
public:
static MyInterface* CreateImplementation(...);
};
I'd have to agree with you. Probably one of the most important principles of object oriented programming is to have a single responsibility for the scope of a piece of code (whether it's a method, class or namespace). In your case, your base class serves the purpose of defining an interface. Adding a factory method to that class, violates that principle, opening the door to a world of shi... trouble.
Yes, a static factory method in the interface (base class) requires it to have knowledge of all possible instantiations. That way, you don't get any of the flexibility the Factory Method pattern is intended to bring.
The Factory should be an independent piece of code, used by client code to create instances. You have to decide somewhere in your program what concrete instance to create. Factory Method allows you to avoid having the same decision spread out through your client code. If later you want to change the implementation (or e.g. for testing), you have just one place to edit: this may be e.g. a simple global change, through conditional compilation (usually for tests), or even via a dependency injection configuration file.
Be careful about how client code communicates what kind of implementation it wants: that's not an uncommon way of reintroducing the dependencies factories are meant to hide.
It's not uncommon to see factory member functions in a class, but it makes my eyes bleed. Often their use have been mixed up with the functionality of the named constructor idiom. Moving the creation function(s) to a separate factory class will buy you more flexibility also to swap factories during testing.
When the interface is just for hiding the implementation details and there will be only one implementation of the Base interface ever, it could be ok to couple them. In that case, the factory function is just a new name for the constructor of the actual implementation.
However, that case is rare. Except when explicit designed having only one implementation ever, you are better off to assume that multiple implementations will exist at some point in time, if only for testing (as you discovered).
So usually it is better to split the Factory part into a separate class.