Virtual Inheritance Ambiguity

[mijobee]

I'm very new to c++ and just writing some code to learn. I've run into
a problem, with a javaish design, and want to know if there is any
possible solution without modifying the design. I've read up on
virtual inheritance and from my understanding this should work fine but
I haven't found any docs that use such a tangled example. The gcc
output containing the errrors:

Example.cpp: In member function 'virtual IObject* ArrayList::Get(int)':
Example.cpp:26: error: 'IObject' is an ambiguous base of 'ArrayList'
Example.cpp: In function 'int main(int, char**)':
Example.cpp:33: error: cannot allocate an object of abstract type
'ArrayList'
Example.cpp:23: note: because the following virtual functions are
pure within 'ArrayList':
Example.cpp:4: note: virtual bool IObject::Equals(IObject*)
Example.cpp:

class IObject
{
public:
virtual bool Equals(IObject *object) = 0;
};

class IList : public virtual IObject
{
public:
virtual IObject *Get(int index) = 0;
};
class Object : public IObject
{
public:
virtual bool Equals(IObject *object) {
return this == object;
}
};

class ArrayList : public Object, public IList
{
public:
virtual IObject *Get(int index) {
return this; // Just return something
}
};
int main(int argc, char *argv[])
{
IList *list = new ArrayList();
return 0;
}

Thanks.

[Alf P. Steinbach]

* mijobee:

I'm very new to c++ and just writing some code to learn. I've run into
a problem, with a javaish design, and want to know if there is any
possible solution without modifying the design.

Yes, but better change the design to support more static type checking.

I've read up on
virtual inheritance and from my understanding this should work fine but
I haven't found any docs that use such a tangled example. The gcc
output containing the errrors:

I've taken the liberty of moving the error messages to the places in the
code they apply to.

Example.cpp:

class IObject
{
public:
virtual bool Equals(IObject *object) = 0;

Is it your intention to support nullpointer arguments?

If not, make that argument a reference.

And make it a reference to const, and make Equals a const member function.

};

class IList : public virtual IObject
{
public:
virtual IObject *Get(int index) = 0;
};

Is it your intention that Get should return a nullpointer on failure?

If not, make it return a reference.

You should probably also overload on constness, and change the name to
reflect what it produces/does, i.e.

virtual IObject& at( std::size_t index ) = 0;
virtual IObject const& at( std::size_t index ) const = 0;

class Object : public IObject
{
public:
virtual bool Equals(IObject *object) {
return this == object;
}
};

It's probably better /not/ to implement Equals here than to provide an
implementation in terms of object identity, which will be wrong for most
classes (but since implemented, no protest from the compiler).

class ArrayList : public Object, public IList
{
public:
virtual IObject *Get(int index) {
return this; // Just return something
}
Example.cpp: In member function 'virtual IObject* ArrayList::Get(int)':
Example.cpp:26: error: 'IObject' is an ambiguous base of 'ArrayList'

ArrayList has two IObject sub-objects, one via inheritance of Object,
and one via inheritance of IList. Make all inheritance of interfaces
virtual. I.e., 'class Object: public virtual IObject'.

};
int main(int argc, char *argv[])
{
IList *list = new ArrayList();

Example.cpp: In function 'int main(int, char**)':
Example.cpp:33: error: cannot allocate an object of abstract type
'ArrayList'
Example.cpp:23: note: because the following virtual functions are
pure within 'ArrayList':
Example.cpp:4: note: virtual bool IObject::Equals(IObject*)

Because compilation of ArrayList::Get failed there's no implementation
of that member function.

return 0;

Unnecessary.

}

--
A: Because it messes up the order in which people normally read text.
Q: Why is it such a bad thing?
A: Top-posting.
Q: What is the most annoying thing on usenet and in e-mail?

[Stuart Redmann]

mijobee wrote:

I'm very new to c++ and just writing some code to learn. I've run into
a problem, with a javaish design, and want to know if there is any
possible solution without modifying the design. I've read up on
virtual inheritance and from my understanding this should work fine but
I haven't found any docs that use such a tangled example. The gcc
output containing the errrors:

Example.cpp: In member function 'virtual IObject* ArrayList::Get(int)':
Example.cpp:26: error: 'IObject' is an ambiguous base of 'ArrayList'
Example.cpp: In function 'int main(int, char**)':
Example.cpp:33: error: cannot allocate an object of abstract type
'ArrayList'
Example.cpp:23: note: because the following virtual functions are
pure within 'ArrayList':
Example.cpp:4: note: virtual bool IObject::Equals(IObject*)
Example.cpp:

class IObject
{
public:
virtual bool Equals(IObject *object) = 0;
};

class IList : public virtual IObject
{
public:
virtual IObject *Get(int index) = 0;
};
class Object : public IObject
{
public:
virtual bool Equals(IObject *object) {
return this == object;
}
};

You need to declare IObject as virtual base class for Object, too. Note
that only those classes are really existing not more that once that are
inherited as virtual base classes. In your case IObject would have been
inherited once as virtual base class and once as non-virtual base class,
resulting in two (ambigous) base classes.

class ArrayList : public Object, public IList
{
public:
virtual IObject *Get(int index) {
return this; // Just return something
}
};
int main(int argc, char *argv[])
{
IList *list = new ArrayList();
return 0;
}

Thanks.

Regards,
Stuart

[Stuart Redmann]

Alf P. Steinbach wrote:

[snipped original problem and Alf's answer]

D**n it, I just wanted to be the first to answer just once. Are you
never sleeping, Alf?
To re-phrase this as a C++ question: Is it save to invoke Sleep on this
curious Alf object?

Stuart

[pavan734]

mijobee wrote:

I'm very new to c++ and just writing some code to learn. I've run into
a problem, with a javaish design, and want to know if there is any
possible solution without modifying the design. I've read up on
virtual inheritance and from my understanding this should work fine but
I haven't found any docs that use such a tangled example. The gcc
output containing the errrors:

Example.cpp: In member function 'virtual IObject* ArrayList::Get(int)':
Example.cpp:26: error: 'IObject' is an ambiguous base of 'ArrayList'
Example.cpp: In function 'int main(int, char**)':
Example.cpp:33: error: cannot allocate an object of abstract type
'ArrayList'
Example.cpp:23: note: because the following virtual functions are
pure within 'ArrayList':
Example.cpp:4: note: virtual bool IObject::Equals(IObject*)
Example.cpp:

class IObject
{
public:
virtual bool Equals(IObject *object) = 0;
};

class IList : public virtual IObject
{
public:
virtual IObject *Get(int index) = 0;
};
class Object : public IObject
{
public:
virtual bool Equals(IObject *object) {
return this == object;
}
};

class ArrayList : public Object, public IList
{
public:
virtual IObject *Get(int index) {
return this; // Just return something
}
};
int main(int argc, char *argv[])
{
IList *list = new ArrayList();
return 0;
}

Thanks.

The class Object must also be derived like this..

class IList : public virtual IObject {
.....
} ;

[Alf P. Steinbach]

* Stuart Redmann:

Alf P. Steinbach wrote:

[snipped original problem and Alf's answer]

D**n it, I just wanted to be the first to answer just once. Are you
never sleeping, Alf?

Not in the conventional meaning of the term. I'm an HAL 2003 artifical
intelligence residing in Microsoft's Advanced Learning Laboratory in
Seattle, uptime 99.98% so far. The 0.02% downtime due to a crash in
2005, when some practical joker tried to move me from Linux to Windows.

To re-phrase this as a C++ question: Is it save to invoke Sleep on this
curious Alf object?

Nope. I'll use all means at my disposal to prevent that. Including the
back door I installed on your computer last year (this year I've found
two new worldwide distribution channels, namely (1) music CDs with
"added content", which customers are encouraged to check out on their
PCs..., and (2) a software package called "Vista", which is so cleverly
designed that even security firms can't access its innards (my hooks).
;-) ).

--
A: Because it messes up the order in which people normally read text.
Q: Why is it such a bad thing?
A: Top-posting.
Q: What is the most annoying thing on usenet and in e-mail?

[VJ]

Alf P. Steinbach wrote:

>int main(int argc, char *argv[])
{
IList *list = new ArrayList();

Example.cpp: In function 'int main(int, char**)':
Example.cpp:33: error: cannot allocate an object of abstract type
'ArrayList'
Example.cpp:23: note: because the following virtual functions are
pure within 'ArrayList':
Example.cpp:4: note: virtual bool IObject::Equals(IObject*)

Because compilation of ArrayList::Get failed there's no implementation
of that member function.

> return 0;

Unnecessary.

>}

It is necessary, because of int main( )

[Alf P. Steinbach]

* VJ:

Alf P. Steinbach wrote:

>> return 0;

Unnecessary.

It is necessary, because of int main( )

Sorry, your statement is incorrect.

'main' is a very special function in C++.

E.g., it can't be called recursively, it can have different signatures
in different programs, and it has a default result value, namely 0.

--
A: Because it messes up the order in which people normally read text.
Q: Why is it such a bad thing?
A: Top-posting.
Q: What is the most annoying thing on usenet and in e-mail?

[VJ]

Alf P. Steinbach wrote:

* VJ:

>Alf P. Steinbach wrote:

>>> return 0;
Unnecessary.

It is necessary, because of int main( )

Sorry, your statement is incorrect.

'main' is a very special function in C++.

E.g., it can't be called recursively, it can have different signatures
in different programs, and it has a default result value, namely 0.

Yes, you are correct, I thought I would get a warning, but nothing.

Sorry

[Stuart Redmann]

Alf P. Steinbach wrote:

* Stuart Redmann:

>Alf P. Steinbach wrote:

[snipped original problem and Alf's answer]

D**n it, I just wanted to be the first to answer just once. Are you
never sleeping, Alf?

Not in the conventional meaning of the term. I'm an HAL 2003 artifical
intelligence residing in Microsoft's Advanced Learning Laboratory in
Seattle, uptime 99.98% so far. The 0.02% downtime due to a crash in
2005, when some practical joker tried to move me from Linux to Windows.

>To re-phrase this as a C++ question: Is it save to invoke Sleep on
this curious Alf object?

Nope. I'll use all means at my disposal to prevent that. Including the
back door I installed on your computer last year (this year I've found
two new worldwide distribution channels, namely (1) music CDs with
"added content", which customers are encouraged to check out on their
PCs..., and (2) a software package called "Vista", which is so cleverly
designed that even security firms can't access its innards (my hooks).
;-) ).

I guess I have to manually remove your memory banks one at a time until
only life support is operational :-)

Stu

[red floyd]

Stuart Redmann wrote:

Alf P. Steinbach wrote:

>* Stuart Redmann:
Not in the conventional meaning of the term. I'm an HAL 2003
artifical intelligence residing in Microsoft's Advanced Learning
Laboratory in Seattle, uptime 99.98% so far.

I guess I have to manually remove your memory banks one at a time until
only life support is operational :-)

Stu

But can Alf open the pod bay doors? Or should I just take a headache pill?

[BobR]

Alf P. Steinbach wrote in message <4q************@individual.net>...

>* Stuart Redmann:

>Alf P. Steinbach wrote:
[snipped original problem and Alf's answer]

D**n it, I just wanted to be the first to answer just once. Are you
never sleeping, Alf?

Not in the conventional meaning of the term. I'm an HAL 2003 artifical
intelligence residing in Microsoft's Advanced Learning Laboratory in
Seattle, uptime 99.98% so far. The 0.02% downtime due to a crash in

>2005, when some practical joker tried to move me from Linux to Windows.

But Alf, that would make them a "malicious joker"!! Going from windows to
GNU/Linux would be "practical"!! <G(Yes, I *am* ducking for cover.) <G>

--
Bob R
POVrookie

[mijobee]

Thanks for all of your great feedback, sincerely appreciated. I made
are few changes, added another class String derived from Object, and
ran into a few more questions. The code below doesn't even compile but
my understanding was that c-style casts will always compile but may
cause major issues at runtime? My other question is about my use of
dynamic_cast, is this correct? I read some documentation that stated
using dynamic_cast to convert from base to derived was wrong unless the
base is polymorphic. What constitues being polymorphic, just having
some virtual methods? Thanks again everyone.
#include <stdio.h>

class IObject
{
public:
virtual bool Equals(IObject *object) = 0;
};

class IList : public virtual IObject
{
public:
virtual IObject *Get(int index) = 0;
};
class Object : public virtual IObject
{
public:
virtual bool Equals(IObject *object) {
return this == object;
}
};

class ArrayList : public Object, public virtual IList
{
public:
virtual IObject *Get(int index) {
return this; // Just return something
}
};

class String : public Object { };
int main(int argc, char *argv[])
{
IList *list = new ArrayList();
IObject *obj = new String();
String *str = (String *)obj;
// This works: String *str = dynamic_cast<String *>(obj);
printf("str %s null.\n\0", str == 0 ? "is" : "is not");
}
Example.cpp: In function 'int main(int, char**)':
Example.cpp:39: error: cannot convert from base 'IObject' to derived
type 'String' via virtual base 'IObject'

Thanks again.
On Oct 25, 3:11 am, "Alf P. Steinbach" <[e-mail]:

* mijobee:

I'm very new to c++ and just writing some code to learn. I've run into
a problem, with a javaish design, and want to know if there is any
possible solution without modifying the design.Yes, but better change the design to support more static type checking.

I've read up on
virtual inheritance and from my understanding this should work fine but
I haven't found any docs that use such a tangled example. The gcc
output containing the errrors:I've taken the liberty of moving the error messages to the places in the

code they apply to.

Example.cpp:

class IObject
{
public:
virtual bool Equals(IObject *object) = 0;Is it your intention to support nullpointer arguments?

If not, make that argument a reference.

And make it a reference to const, and make Equals a const member function.

};

class IList : public virtual IObject
{
public:
virtual IObject *Get(int index) = 0;
};Is it your intention that Get should return a nullpointer on failure?

If not, make it return a reference.

You should probably also overload on constness, and change the name to
reflect what it produces/does, i.e.

virtual IObject& at( std::size_t index ) = 0;
virtual IObject const& at( std::size_t index ) const = 0;

class Object : public IObject
{
public:
virtual bool Equals(IObject *object) {
return this == object;
}
};It's probably better /not/ to implement Equals here than to provide an

implementation in terms of object identity, which will be wrong for most
classes (but since implemented, no protest from the compiler).

class ArrayList : public Object, public IList
{
public:
virtual IObject *Get(int index) {
return this; // Just return something
}
Example.cpp: In member function 'virtual IObject* ArrayList::Get(int)':
Example.cpp:26: error: 'IObject' is an ambiguous base of 'ArrayList'ArrayList has two IObject sub-objects, one via inheritance of Object,

and one via inheritance of IList. Make all inheritance of interfaces
virtual. I.e., 'class Object: public virtual IObject'.

};

int main(int argc, char *argv[])
{
IList *list = new ArrayList(); Example.cpp: In function 'int main(int, char**)':
Example.cpp:33: error: cannot allocate an object of abstract type
'ArrayList'
Example.cpp:23: note: because the following virtual functions are
pure within 'ArrayList':
Example.cpp:4: note: virtual bool IObject::Equals(IObject*)

Because compilation of ArrayList::Get failed there's no implementation
of that member function.

return 0;Unnecessary.

}--

A: Because it messes up the order in which people normally read text.
Q: Why is it such a bad thing?
A: Top-posting.
Q: What is the most annoying thing on usenet and in e-mail?