I need to call a virtual method for all classes derived from a given base base class right after the construction of the derived object. But doing so in the base class constructor will result in a pure virtual method call
Here is a simplified example:
struct Loader {
int get(int index) { return 0; }
};
struct Base{
Base() {
Loader l;
load( l ); // <-- pure virtual call!
}
virtual void load( Loader & ) = 0;
};
struct Derived: public Base {
int value;
void load( Loader &l ) {
value = Loader.get(0);
}
};
I can call load at the Derived constructor, but Derived could not know how to create a Loader. Any ideas/workarounds?
The problem is that base class construction occurs before the derived class is fully constructed. You should either call "load" from the derived class, initialise throguh a different virtual member function or create a helper function to do this:
Base* CreateDerived()
{
Base* pRet = new Derived;
pRet->Load();
return pRet;
}
The C++ FAQ calls this problem DBDI, Dynamic Binding During Construction. Mainly, the problem is to avoid the Evil two-phase construction advocated in other answers here. It's sort of "my" FAQ item -- I convinced Marshall to add it.
However, Marshall's take it on it is very general (which is good for a FAQ), while I was more concerned with the particular design/coding pattern.
So, instead of sending you to the FAQ I send you to my own blog, the article "How to avoid post-construction by using Parts Factories", which links to the relevant FAQ item, but discusses in depth the pattern.
You can just skip the first two paragraphs...
I sort of rambled there. :-)
Cheers & hth.,
Use the PIMPL pattern:
template<typename T>
class Pimpl
{
public:
Pimpl()
{
// At this point the object you have created is fully constructed.
// So now you can call the virtual method on it.
object.load();
}
T* operator->()
{
// Use the pointer notation to get access to your object
// and its members.
return &object;
}
private:
T object; // Not technically a pointer
// But otherwise the pattern is the same.
// Modify to your needs.
};
int main()
{
Pimpl<Derived> x;
x->doStuff();
}
Can't you add a method getLoader() in your Base class so that DerivedClass constructor can call it on this to get a Loader ?
As DerivedClass constructor will be called after Base class constructor, that should work fine.
Its hard to give advice unless you tell us what you are trying to accomplish, rather than how. I find that its usually better to construct such objects from a factory, which will load the required data before-hand, and then pass the data into the constructor of the object.
Many known frameworks (like MFC) do this: They make a (virtual) member-function Init() or Create() and do the initialization there and then mandate in the documentation that the user call it. I know you won't like this idea, but you just can't call a virtual method from a constructor and expect it to behave polymorphically, regardless of the methods pureness...
This may come a little late after other answers, but I'll still give it a try.
You can implement this safely, and without changing derived classes. However, you will need to change use of all these classes, which might be far worse, depending on your scenario. If you are still designing, then this might be viable alternative.
Basically, you can apply the curiously recurring template pattern and inject the initialization code after the constructor gets invoked. Furthermore, if you do it as I've written it below, you can even protect load from being called twice.
struct Loader {
int get(int index) { return 0; }
};
struct Base {
virtual ~Base() {} // Note: don't forget this.
protected:
virtual void load( Loader & ) = 0;
};
struct Derived : public Base {
int value;
protected:
void load( Loader &l ) {
value = l.get(0);
}
};
template<typename T>
class Loaded : public T
{
public:
Loaded () {
Loader l; T::load(l);
}
};
int main ( int, char ** )
{
Loaded<Derived> derived;
}
Frankly, though, I would consider an alternate design if you can. Move the code from load to your constructors and provide the loader as an a reference argument defaulting as follows:
struct Derived : public Base {
Derived ( Loader& loader = Loader() ) { ... }
};
That way, you completely avoid the problem.
Summary: your choices are the following:
If you are not limited by external constraints and don't have an extensive code base depending on this, change your design for something safer.
If you want to keep load the way it is and not change your classes too much but are willing to pay the price of changing all instantiations, apply CRTP as proposed above.
If you insist on being mostly backward compatible with existing client code, you will have to change you classes to use a PIMPL as others have suggested or live with the existing problem.
There are many ways to correct this, here is 1 suggestion fitting within your provided framework
struct Loader {
int get(int index) { return 0; }
};
struct Base{
Base() {
}
Loader & getLoader( );
private:
Loader l;
};
struct Derived: public Base {
int value;
Derived( ) {
value = getLoader().get(0);
}
};
Related
I'm working on a very, very simple data access layer (DAL) featuring two classes: DataTransferObject (DTO) and DataAccessObject (DAO). Both classes are abstract base classes and need to be inherited and modified for a specific use case.
class DataTransferObject {
protected:
//protected constructor to prevent initialization
};
class DataAccessObject {
public:
virtual bool save(DataTransferObject o) = 0;
virtual DataTransferObject* load(int id) = 0;
};
in case of a House class from the business logic layer, the implementation of the DAL classes would read something along these lines:
class Dto_House : public DataTransferObject {
public:
int stories;
string address; //...which are all members of the House class...
Dto_House(House h);
};
class Dao_House : public DataAccessObject {
public:
bool save(Dto_House h) { /*...implement database access, etc...*/ }
Dto_House* load(int id) {/*...implement database access, etc...*/ }
};
EDIT: Of course, the derived classes know about the structure of the House class and the data storage.
Simple, nice, okidoke.
Now I wanted to provide a method toObject() in the DTO class in order to quickly convert the Dto_House into a House object. I then read about the automatic return type deduction in C++14 and tried:
class DataTransferObject {
public:
virtual auto toObject() = 0;
};
But I had to discover: No automatic return type deduction for virtual functions. :(
What are your ideas about implementing a "virtual function with deduced return type" for this specific case? I want a general toObject() function in my DTO "interface".
The only thing that came to my mind was something like:
template <typename T>
class DataTransferObject {
virtual T toObject() = 0;
};
class Dto_House : public DataTransferObject<House> {
public:
int stories;
string address;
House toObject() {return House(stories, address);}
};
EDIT:
A possible use case would be:
House h(3, "231 This Street");
h.doHouseStuff();
//save it
Dto_House dtoSave(h);
Dao_House dao;
dao.save(dtoSave); //even shorter: dao.save(Dto_House(h));
//now load some other house
Dto_House dtoLoad = dao.load(id 2);
h = dtoLoad.toObject();
h.doOtherHouseStuff();
But the house does not know it can be saved and loaded.
Of course, the abstract DAO class may be derived to further refine it for the use with, e.g. Sqlite, XML files or whatever... I just presented the very basic concept.
How about setting an empty abstract class - practically, an interface, then have both of your types implement it and set this as the toObject returning reference type?
class Transferable
{
virtual ~Transferable() = 0;
}
And then:
class DataTransferObject {
public:
//Return a reference of the object.
virtual Transferable& toObject() = 0;
};
Dto_House : public DataTransferObject, Transferable { /*...*/ }
House : public DataTransferObject, Transferable { /*...*/ }
The example above is to get my point.
Even better, you can use the DataTransferObject for this cause as your returning reference type, and no other abstract class:
class DataTransferObject {
public:
virtual DataTransferObject& toObject() = 0;
};
Dto_House : public DataTransferObject { /*...*/ }
House : public DataTransferObject { /*...*/ }
Update: If you want to have the classes separated apart, separating any association between data and operations by convention, you could set the name of the base class on something that represents the data i.e.: Building, Construction etc, and then use it for the reference type in toObject.
You can also have the class manipulating those operations on the API of data manipulation.
In general, you can not have a virtual function returning different types in different subclasses, as this violates the whole concept of statically typed language: if you call DataTransferObject::toObject(), the compiler does not know what type it is going to return until runtime.
And this highlight the main problem of your design: why do you need a base class at all? How are you going to use it? Calling DataTransferObject::toObject(), even if you use some magic to get it work (or use a dynamically typed language), sounds like a bad idea since you can not be sure what the return type is. You will anyway need some casts, or some ifs, etc, to get it working — or you will be using only the functionality common for all such objects (House, Road, etc.) — but then you just need a common base class for all of them.
In fact, there is one exception to the same return type rule: if you return a pointer to a class, you can use the Covariant return type concept: a subclass may override a virtual function to return a subclass of the original return type. If all your "objects" have a common base class, you may use something along the lines of
struct DataTransferObject {
virtual BaseObject* toObject() = 0;
};
struct Dto_House : public DataTransferObject {
virtual House* toObject() { /*...*/ } // assumes that House subclasses BaseObject
};
However, this will still leave the same problem: if all you have in your code is DataTransferObject, even if you (but not the compiler) know it is a Dto_House, you will need some cast, which might be unreliable.
On the other hand, you template solution seems quite good except that you will not be able to explicitly call DataTransferObject::toObject() (unless you know the type of the object), but that's a bad idea as I have explained.
So, I suggest you think on how you are going to actually use the base classes (even write some sample code), and make your choice based on that.
I'm writing a piece of generic software that will be loaded on to many different variants of the same basic hardware. They all have the same processor, but with different peripherals and their own functions that need to be carried out. The software will know which variant it should run by reading a hardware switch value.
Here's my current implementation in a nutshell:
class MyBase
{
public:
MyBase() { }
virtual run() = 0;
}
class VariantA : public MyBase
{
public:
VariantA () { }
virtual run()
{
// Run code specific to hardware Variant-A
}
}
class VariantB : public MyBase
{
public:
VariantB () { }
virtual run()
{
// Run code specific to hardware Variant-B
}
}
void main()
{
MyBase* variant;
uint_8 switchValue = readSwitchValue();
switch(switchValue)
{
case 0:
variant = new VariantA();
break;
case 1:
variant = new VariantB();
break;
}
variant->run();
}
Now this works just fine. I read the hardware value and use a switch statement to create the new corresponding class.
The problem is that there are a lot of variants I have to deal with. Currently about 15, with the potential to add another 20-30 in the near future. I have really come to despise switch statements that run for hundreds of lines, so I'm really looking for a better way to do this, probably through templates.
I want to be able to use my hardware value to look up a type and use that type to create my new object. Ideally when I add a new variant, I create the new class, add that class type to my lookup table with it's matching hardware value, and it's good to go.
Is this possible at all? What's a good solution here?
As stated, you make a factory, but not necessarily with naive switch statements. What you can do is make a template class to create the relevant object and dynamically add these to your factory.
class VariantinatorBase {
public:
VariantinatorBase() {}
virtual ~VariantinatorBase() {}
virtual std::unique_ptr<Variant> Create() = 0;
};
template< class T >
class Variantinator : public VariantinatorBase {
public:
Variantinator() {}
virtual ~Variantinator() {}
virtual std::unique_ptr<Variant> Create() { return std::make_unique<T>(); }
};
Now you have a class factory that allows you to register these.
class VariantFactory
{
public:
VariantFactory()
{
// If you want, you can do all your Register() calls in here, and even
// make the Register() function private.
}
template< uint8_t type, typename T >
void Register()
{
Register( type, std::make_unique<Variantinator<T>>() );
}
std::unique_ptr<Variant> Create( uint8_t type )
{
TSwitchToVariant::iterator it = m_switchToVariant.find( type );
if( it == m_switchToVariant.end() ) return nullptr;
return it->second->Create();
}
private:
void Register( uint8_t type, std::unique_ptr<VariantinatorBase>&& creator )
{
m_switchToVariant[type] = std::move(creator);
}
typedef std::map<uint8_t, std::unique_ptr<VariantinatorBase> > TSwitchToVariant;
TSwitchToVariant m_switchToVariant;
};
At the beginning of your program, create the factory and register your types:
VariantFactory factory;
factory.Register<0, VariantA>();
factory.Register<1, VariantB>();
factory.Register<2, VariantC>();
Then later, you want to call on it:
std::unique_ptr<Variant> thing = factory.Create( switchValue );
You are looking for a factory
http://www.oodesign.com/factory-pattern.html
A factory is a software module (a method, a class) whose sole purpose is to create the right object for the job. An example using a factory class:
class VariantFactory
{
MyBase* CreateObject(uint_8 value);
}
And the CreateObject method can be filled out to give you the type of object that you need.
In the case of a very small selection of objects with simple construction, a simple switch statement might suffice. As soon as you get a lot of objects or ones that require more detailed construction, a factory is quite useful.
I made this a comment; let's turn it into an answer:
Personally, I think a "switch/case" block to create the appropriate class is probably an optimal solution. Just put your case statement in a static "factory" method that returns a reference to the specific class. IMHO...
Here's a good example: factory method design pattern
Class Book : public Product
{
};
class Computer : public Product
{
};
class ProductFactory
{
public:
virtual Product* Make(int type)
{
switch (type)
{
case 0:
return new Book();
case 1:
return new Computer();
[...]
}
}
}
Call it like this:
ProductFactory factory = ....;
Product* p1 = factory.Make(0); // p1 is a Book*
Product* p2 = factory.Make(1); // p2 is a Computer*
// remember to delete p1 and p2
Note that in his most excellent response, smink also suggests some other design alternatives, too.
BOTTOM LINE: There's nothing inherently "wrong" with a switch/case block. Even for a switch with many case options.
IMHO...
PS:
This really isn't creating a "dynamic type". Rather, it's "creating a static type dynamically". That would be equally true if you used a template or an enum solution as well. But again - I vastly prefer the "switch/case".
Update: I am leaving my original solution here for posterity, but consider the solution provided by paddy to be superior and less error prone. With only a couple of slight improvements I think it's actually about as good as you can possibly get.
Consider this design:
class VariantA : public MyBase
{
static MyBase *CreateMachineInstance() { return new VariantA; }
};
class VariantB : public MyBase
{
static MyBase *CreateMachineInstance() { return new VariantB; }
};
Now, all you need is an std::map that uses a uint_8 as the key and maps it to a function pointer (returning MyBase). Insert the identifiers in the map (pointing each to the appropriate machine creation function) and then read the code and just use the map to find what machine you're using.
This is loosely based on a concept/pattern called a "factory" but may break slightly if your machine constructors require different arguments or you need to perform additional per-machine initialization/operations - and from what you mention it sounds like you might.
If that's the case, you can still use this pattern but you will have to make some tweaks and rearchitect things a bit but you will end up with something much cleaner and easier to augment and maintain.
#include <stdio.h>
#include <string.h>
#include <iostream>
using namespace std;
template<class T,class T1>
class HeroHonda
{
private:
T millage;
T1 *options;
public:
HeroHonda() {
puts("constructed");
options=new T1[20];
strcpy(options,"Good millage,Powerstart");
millage=110;
}
virtual T features() {
cout<<options<<"millage is"<<millage<<endl;
return 1;
}
// virtual T Extrafeatures() = 0;
~HeroHonda() {
cout<<"destructor"<<endl;
delete [] options;
}
};
int main()
{
HeroHonda <int,char> *Ptr=new HeroHonda <int,char>;
Ptr->features();
delete Ptr;
}
I'm finding it difficult to describe this problem very concisely, so I've attached the code for a demonstration program.
The general idea is that we want a set of Derived classes that are forced to implement some abstract Foo() function from a Base class. Each of the derived Foo() calls must accept a different parameter as input, but all of the parameters should also be derived from a BaseInput class.
We see two possible solutions so far, neither we're very happy with:
Remove the Foo() function from the base class and reimplement it with the correct input types in each Derived class. This, however, removes the enforcement that it be implemented in the same manner in each derived class.
Do some kind of dynamic cast inside the receiving function to verify that the type received is correct. However, this does not prevent the programmer from making an error and passing the incorrect input data type. We would like the type to be passed to the Foo() function to be compile-time correct.
Is there some sort of pattern that could enforce this kind of behaviour? Is this whole idea breaking some sort of fundamental idea underlying OOP? We'd really like to hear your input on possible solutions outside of what we've come up with.
Thanks so much!
#include <iostream>
// these inputs will be sent to our Foo function below
class BaseInput {};
class Derived1Input : public BaseInput { public: int d1Custom; };
class Derived2Input : public BaseInput { public: float d2Custom; };
class Base
{
public:
virtual void Foo(BaseInput& i) = 0;
};
class Derived1 : public Base
{
public:
// we don't know what type the input is -- do we have to try to cast to what we want
// and see if it works?
virtual void Foo(BaseInput& i) { std::cout << "I don't want to cast this..." << std::endl; }
// prefer something like this, but then it's not overriding the Base implementation
//virtual void Foo(Derived1Input& i) { std::cout << "Derived1 did something with Derived1Input..." << std::endl; }
};
class Derived2 : public Base
{
public:
// we don't know what type the input is -- do we have to try to cast to what we want
// and see if it works?
virtual void Foo(BaseInput& i) { std::cout << "I don't want to cast this..." << std::endl; }
// prefer something like this, but then it's not overriding the Base implementation
//virtual void Foo(Derived2Input& i) { std::cout << "Derived2 did something with Derived2Input..." << std::endl; }
};
int main()
{
Derived1 d1; Derived1Input d1i;
Derived2 d2; Derived2Input d2i;
// set up some dummy data
d1i.d1Custom = 1;
d2i.d2Custom = 1.f;
d1.Foo(d2i); // this compiles, but is a mistake! how can we avoid this?
// Derived1::Foo() should only accept Derived1Input, but then
// we can't declare Foo() in the Base class.
return 0;
}
Since your Derived class is-a Base class, it should never tighten the base contract preconditions: if it has to behave like a Base, it should accept BaseInput allright. This is known as the Liskov Substitution Principle.
Although you can do runtime checking of your argument, you can never achieve a fully type-safe way of doing this: your compiler may be able to match the DerivedInput when it sees a Derived object (static type), but it can not know what subtype is going to be behind a Base object...
The requirements
DerivedX should take a DerivedXInput
DerivedX::Foo should be interface-equal to DerivedY::Foo
contradict: either the Foo methods are implemented in terms of the BaseInput, and thus have identical interfaces in all derived classes, or the DerivedXInput types differ, and they cannot have the same interface.
That's, in my opinion, the problem.
This problem occured to me, too, when writing tightly coupled classes that are handled in a type-unaware framework:
class Fruit {};
class FruitTree {
virtual Fruit* pick() = 0;
};
class FruitEater {
virtual void eat( Fruit* ) = 0;
};
class Banana : public Fruit {};
class BananaTree {
virtual Banana* pick() { return new Banana; }
};
class BananaEater : public FruitEater {
void eat( Fruit* f ){
assert( dynamic_cast<Banana*>(f)!=0 );
delete f;
}
};
And a framework:
struct FruitPipeLine {
FruitTree* tree;
FruitEater* eater;
void cycle(){
eater->eat( tree->pick() );
}
};
Now this proves a design that's too easily broken: there's no part in the design that aligns the trees with the eaters:
FruitPipeLine pipe = { new BananaTree, new LemonEater }; // compiles fine
pipe.cycle(); // crash, probably.
You may improve the cohesion of the design, and remove the need for virtual dispatching, by making it a template:
template<class F> class Tree {
F* pick(); // no implementation
};
template<class F> class Eater {
void eat( F* f ){ delete f; } // default implementation is possible
};
template<class F> PipeLine {
Tree<F> tree;
Eater<F> eater;
void cycle(){ eater.eat( tree.pick() ); }
};
The implementations are really template specializations:
template<> class Tree<Banana> {
Banana* pick(){ return new Banana; }
};
...
PipeLine<Banana> pipe; // can't be wrong
pipe.cycle(); // no typechecking needed.
You might be able to use a variation of the curiously recurring template pattern.
class Base {
public:
// Stuff that don't depend on the input type.
};
template <typename Input>
class Middle : public Base {
public:
virtual void Foo(Input &i) = 0;
};
class Derived1 : public Middle<Derived1Input> {
public:
virtual void Foo(Derived1Input &i) { ... }
};
class Derived2 : public Middle<Derived2Input> {
public:
virtual void Foo(Derived2Input &i) { ... }
};
This is untested, just a shot from the hip!
If you don't mind the dynamic cast, how about this:
Class BaseInput;
class Base
{
public:
void foo(BaseInput & x) { foo_dispatch(x); };
private:
virtual void foo_dispatch(BaseInput &) = 0;
};
template <typename TInput = BaseInput> // default value to enforce nothing
class FooDistpatch : public Base
{
virtual void foo_dispatch(BaseInput & x)
{
foo_impl(dynamic_cast<TInput &>(x));
}
virtual void foo_impl(TInput &) = 0;
};
class Derived1 : public FooDispatch<Der1Input>
{
virtual void foo_impl(Der1Input & x) { /* your implementation here */ }
};
That way, you've built the dynamic type checking into the intermediate class, and your clients only ever derive from FooDispatch<DerivedInput>.
What you are talking about are covariant argument types, and that is quite an uncommon feature in a language, as it breaks your contract: You promised to accept a base_input object because you inherit from base, but you want the compiler to reject all but a small subset of base_inputs...
It is much more common for programming languages to offer the opposite: contra-variant argument types, as the derived type will not only accept everything that it is bound to accept by the contract, but also other types.
At any rate, C++ does not offer contravariance in argument types either, only covariance in the return type.
C++ has a lot of dark areas, so it's hard to say any specific thing is undoable, but going from the dark areas I do know, without a cast, this cannot be done. The virtual function specified in the base class requires the argument type to remain the same in all the children.
I am sure a cast can be used in a non-painful way though, perhaps by giving the base class an Enum 'type' member that is uniquely set by the constructor of each possible child that might possibly inherit it. Foo() can then check that 'type' and determine which type it is before doing anything, and throwing an assertion if it is surprised by something unexpected. It isn't compile time, but it's the closest a compromise I can think of, while still having the benefits of requiring a Foo() be defined.
It's certainly restricted, but you can use/simulate coviarance in constructors parameters.
There was an interesting problem in C++, but it was more about architecture.
There are many (10, 20, 40, etc) classes describing some characteristics (mix-in classes), for example:
struct Base { virtual ~Base() {} };
struct A : virtual public Base { int size; };
struct B : virtual public Base { float x, y; };
struct C : virtual public Base { bool some_bool_state; };
struct D : virtual public Base { string str; }
// ....
The primary module declares and exports a function (for simplicity just function declarations without classes):
// .h file
void operate(Base *pBase);
// .cpp file
void operate(Base *pBase)
{
// ....
}
Any other module can have code like this:
#include "mixing.h"
#include "primary.h"
class obj1_t : public A, public C, public D {};
class obj2_t : public B, public D {};
// ...
void Pass()
{
obj1_t obj1;
obj2_t obj2;
operate(&obj1);
operate(&obj2);
}
The question is how do you know what the real type of a given object in operate() is without using dynamic_cast and any type information in classes (constants, etc)? The operate() function is used with a big array of objects in small time periods and dynamic_cast is too slow for it and I don't want to include constants (enum obj_type { ... }) because this is not the OOP-way.
// module operate.cpp
void some_operate(Base *pBase)
{
processA(pBase);
processB(pBase);
}
void processA(A *pA)
{
}
void processB(B *pB)
{
}
I cannot directly pass a pBase to these functions. And it's impossible to have all possible combinations of classes, because I can add new classes just by including new header files.
One solution that came to mind, in the editor I can use a composite container:
struct CompositeObject
{
vector<Base *pBase> parts;
};
But the editor does not need time optimization and can use dynamic_cast for parts to determine the exact type. In operate() I cannot use this solution.
So, is it possible to avoid using a dynamic_cast and type information to solve this problem? Or maybe I should use another architecture?
The real problem here is about what you are trying to achieve.
Do you want something like:
void operate(A-B& ) { operateA(); operateB(); }
// OR
void operate(A-B& ) { operateAB(); }
That is, do you want to apply an operation on each subcomponent (independently), or do you wish to be able to apply operations depending on the combination of components (much harder).
I'll take the first approach here.
1. Virtual ?
class Base { public: virtual void operate() = 0; };
class A: virtual public Base { public virtual void operate() = 0; };
void A::operate() { ++size; } // yes, it's possible to define a pure virtual
class obj1_t: public A, public B
{
public:
virtual void operate() { A::operate(); B::operate(); }
};
Some more work, for sure. Notably I don't like the repetition much. But that's one call to the _vtable, so it should be one of the fastest solution!
2. Composite Pattern
That would probably be the more natural thing here.
Note that you can perfectly use a template version of the pattern in C++!
template <class T1, class T2, class T3>
class BaseT: public Base, private T1, private T2, private T3
{
public:
void operate() { T1::operate(); T2::operate(); T3::operate(); }
};
class obj1_t: public BaseT<A,B,C> {};
Advantages:
no more need to repeat yourself! write operate once and for all (baring variadic...)
only 1 virtual call, no more virtual inheritance, so even more efficient that before
A, B and C can be of arbitrary type, they should not inherit from Base at all
edit the operate method of A, B and C may be inlined now that it's not virtual
Disadvantage:
Some more work on the framework if you don't have access to variadic templates yet, but it's feasible within a couple dozen of lines.
First thing that comes to mind is asking what you really want to achieve... but then again the second thought is that you can use the visitor pattern. Runtime type information will implicitly be used to determine at what point in the hierarchy is the final overrider of the accept method, but you will not explicitly use that information (your code will not show any dynamic_cast, type_info, constants...)
Then again, my first thought comes back... since you are asking about the appropriateness of the architecture, what is it that you really want to achieve? --without knowledge of the problem you will only find generic answers as this one.
The usual object oriented way would be to have (pure) virtual functions in the base class that are called in operate() and that get overridden in the derived classes to execute code specific to that derived class.
Your problem is that you want to decide what to do based on more than one object's type. Virtual functions do this for one object (the one left of the . or ->) only. Doing so for more than one object is called multiple dispatch (for two objects it's also called double dispatch), and in C++ there's no built-in feature to deal with this.
Look at double dispatch, especially as done in the visitor pattern.
EDIT: minor fixes (virtual Print; return mpInstance) following remarks in the answers.
I am trying to create a system in which I can derive a Child class from any Base class, and its implementation should replace the implementation of the base class.
All the objects that create and use the base class objects shouldn't change the way they create or call an object, i.e. should continue calling BaseClass.Create() even when they actually create a Child class.
The Base classes know that they can be overridden, but they do not know the concrete classes that override them.
And I want the registration of all the the Child classes to be done just in one place.
Here is my implementation:
class CAbstractFactory
{
public:
virtual ~CAbstractFactory()=0;
};
template<typename Class>
class CRegisteredClassFactory: public CAbstractFactory
{
public:
~CRegisteredClassFactory(){};
Class* CreateAndGet()
{
pClass = new Class;
return pClass;
}
private:
Class* pClass;
};
// holds info about all the classes that were registered to be overridden
class CRegisteredClasses
{
public:
bool find(const string & sClassName);
CAbstractFactory* GetFactory(const string & sClassName)
{
return mRegisteredClasses[sClassName];
}
void RegisterClass(const string & sClassName, CAbstractFactory* pConcreteFactory);
private:
map<string, CAbstractFactory* > mRegisteredClasses;
};
// Here I hold the data about all the registered classes. I hold statically one object of this class.
// in this example I register a class CChildClass, which will override the implementation of CBaseClass,
// and a class CFooChildClass which will override CFooBaseClass
class RegistrationData
{
public:
void RegisterAll()
{
mRegisteredClasses.RegisterClass("CBaseClass", & mChildClassFactory);
mRegisteredClasses.RegisterClass("CFooBaseClass", & mFooChildClassFactory);
};
CRegisteredClasses* GetRegisteredClasses(){return &mRegisteredClasses;};
private:
CRegisteredClasses mRegisteredClasses;
CRegisteredClassFactory<CChildClass> mChildClassFactory;
CRegisteredClassFactory<CFooChildClass> mFooChildClassFactory;
};
static RegistrationData StaticRegistrationData;
// and here are the base class and the child class
// in the implementation of CBaseClass::Create I check, whether it should be overridden by another class.
class CBaseClass
{
public:
static CBaseClass* Create()
{
CRegisteredClasses* pRegisteredClasses = StaticRegistrationData.GetRegisteredClasses();
if (pRegisteredClasses->find("CBaseClass"))
{
CRegisteredClassFactory<CBaseClass>* pFac =
dynamic_cast<CRegisteredClassFactory<CBaseClass>* >(pRegisteredClasses->GetFactory("CBaseClass"));
mpInstance = pFac->CreateAndGet();
}
else
{
mpInstance = new CBaseClass;
}
return mpInstance;
}
virtual void Print(){cout << "Base" << endl;};
private:
static CBaseClass* mpInstance;
};
class CChildClass : public CBaseClass
{
public:
void Print(){cout << "Child" << endl;};
private:
};
Using this implementation, when I am doing this from some other class:
StaticRegistrationData.RegisterAll();
CBaseClass* b = CBaseClass::Create();
b.Print();
I expect to get "Child" in the output.
What do you think of this design? Did I complicate things too much and it can be done easier? And is it OK that I create a template that inherits from an abstract class?
I had to use dynamic_pointer (didn't compile otherwise) - is it a hint that something is wrong?
Thank you.
This sort of pattern is fairly common. I'm not a C++ expert but in Java you see this everywhere. The dynamic cast appears to be necessary because the compiler can't tell what kind of factory you've stored in the map. To my knowledge there isn't much you can do about that with the current design. It would help to know how these objects are meant to be used. Let me give you an example of how a similar task is accomplished in Java's database library (JDBC):
The system has a DriverManager which knows about JDBC drivers. The drivers have to be registered somehow (the details aren't important); once registered whenever you ask for a database connection you get a Connection object. Normally this object will be an OracleConnection or an MSSQLConnection or something similar, but the client code only sees "Connection". To get a Statement object you say connection.prepareStatement, which returns an object of type PreparedStatement; except that it's really an OraclePreparedStatement or MSSQLPreparedStatement. This is transparent to the client because the factory for Statements is in the Connection, and the factory for Connections is in the DriverManager.
If your classes are similarly related you may want to have a function that returns a specific type of class, much like DriverManager's getConnection method returns a Connection. No casting required.
The other approach you may want to consider is using a factory that has a factory-method for each specific class you need. Then you only need one factory-factory to get an instance of the Factory. Sample (sorry if this isn't proper C++):
class CClassFactory
{
public:
virtual CBaseClass* CreateBase() { return new CBaseClass(); }
virtual CFooBaseClass* CreateFoo() { return new CFooBaseClass();}
}
class CAImplClassFactory : public CClassFactory
{
public:
virtual CBaseClass* CreateBase() { return new CAImplBaseClass(); }
virtual CFooBaseClass* CreateFoo() { return new CAImplFooBaseClass();}
}
class CBImplClassFactory : public CClassFactory // only overrides one method
{
public:
virtual CBaseClass* CreateBase() { return new CBImplBaseClass(); }
}
As for the other comments criticizing the use of inheritance: in my opinion there is no difference between an interface and public inheritance; so go ahead and use classes instead of interfaces wherever it makes sense. Pure Interfaces may be more flexible in the long run but maybe not. Without more details about your class hierarchy it's impossible to say.
Usually, base class/ derived class pattern is used when you have an interface in base class, and that interface is implemented in derived class (IS-A relationship). In your case, the base class does not seem to have any connection with derived class - it may as well be void*.
If there is no connection between base class and derived class, why do you use inheritance? What is the benefit of having a factory if factory's output cannot be used in a general way? You have
class CAbstractFactory
{
public:
virtual ~CAbstractFactory()=0;
};
This is perfectly wrong. A factory has to manufacture something that can be used immediately:
class CAbstractFactory
{
public:
virtual ~CAbstractFactory(){};
public:
CBaseClass* CreateAndGet()
{
pClass = new Class;
return pClass;
}
private:
CBaseClass* pClass;
protected:
CBaseClass *create() = 0;
};
In general, you're mixing inheritance, virtual functions and templates the way they should not be mixed.
Without having read all of the code or gone into the details, it seems like you should've done the following:
make b of type CChildClass,
make CBaseClass::Print a virtual function.
Maybe I'm wrong but I didn't find any return statement in your CBaseClass::Create() method!
Personally, I think this design overuses inheritance.
"I am trying to create a system in which I can derive a Child class from any Base class, and its implementation should replace the implementation of the base class." - I don't know that IS-A relationships should be that flexible.
I wonder if you'd be better off using interfaces (pure virtual classes in C++) and mixin behavior. If I were writing it in Java I'd do this:
public interface Foo
{
void doSomething();
}
public class MixinDemo implements Foo
{
private Foo mixin;
public MixinDemo(Foo f)
{
this.mixin = f;
}
public void doSomething() { this.mixin.doSomething(); }
}
Now I can change the behavior as needed by changing the Foo implementation that I pass to the MixinDemo.