C++Talk.NET C++ language newsgroups   Archives   FAQ   Search   Memberlist   Usergroups   Register   Profile   Log in to check your private messages   Log in  Re: Initialization of virtual bases          C++Talk.NET Forum Index -> C++ Language (Moderated) View previous topic :: View next topic   Author Message Attila Feher Guest Posted: Tue Aug 26, 2003 6:42 pm    Post subject: Re: Initialization of virtual bases Dave Theese wrote: Quote: Hello all, The code example below has proper behavior (of course), but I'm trying to understand how the behavior is brought about. Specifically, what happens to B's and C's initialization of A??? Is it just ignored even though it explicitly appears in the code? [SNIP] Virtual bases are initialized only once, and by the most derived class type (the one you actually instantiate). IIRC they *must* be initialized (if there is no default constructor). So (while it may not be the right wording) you may say so, that the initialization done by any base class is ignored. -- Attila aka WW [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Francis Glassborow Guest Posted: Tue Aug 26, 2003 6:47 pm    Post subject: Re: Initialization of virtual bases In article <a1A2b.8257$QT5.3304@fed1read02>, Dave Theese <cheeser_1998 (AT) yahoo (DOT) com> writes Quote: Hello all, The code example below has proper behavior (of course), but I'm trying to understand how the behavior is brought about. Specifically, what happens to B's and C's initialization of A??? Is it just ignored even though it explicitly appears in the code? The rule for initilaisation of virtual bases is simple, only the most derived class does it and all other bases both direct and indirect simply use the result. It is achieved by the implementation only calling the virtual base ctor from the most derived class (and before any other non-virtual base ctors) All other calls to a virtual base ctor are suppressed. -- Francis Glassborow ACCU 64 Southfield Rd Oxford OX4 1PA +44(0)1865 246490 All opinions are mine and do not represent those of any organisation [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Ben Hutchings Guest Posted: Tue Aug 26, 2003 7:00 pm    Post subject: Re: Initialization of virtual bases In article <a1A2b.8257$QT5.3304@fed1read02>, Dave Theese wrote: Quote: Hello all, The code example below has proper behavior (of course), but I'm trying to understand how the behavior is brought about. Specifically, what happens to B's and C's initialization of A??? Is it just ignored even though it explicitly appears in the code? snip Yes. The selection of constructor and constructor arguments for a virtual base is always controlled by the complete object's constructor (even if it implicitly selects a default constructor). [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top llewelly Guest Posted: Tue Aug 26, 2003 7:05 pm    Post subject: Re: Initialization of virtual bases "Dave Theese" <cheeser_1998 (AT) yahoo (DOT) com> writes: Quote: Hello all, The code example below has proper behavior (of course), but I'm trying to understand how the behavior is brought about. Specifically, what happens to B's and C's initialization of A??? Is it just ignored even though it explicitly appears in the code? It is ignored when you create a D. It is used when you create a B or C which is not a base sub-object of a D. Quote: Thanks, Dave #include using namespace std; class A { public: A(int d): data(d) { cout << "data initialized to " << d << endl; } private: int data; }; class B: virtual public A { public: B(int d): A(d) { } }; class C: virtual public A { public: C(int d): A(d) { } }; class D: public B, public C Since the A sub-object of both B and C is virtual, D will contain only 1 A sub-object - not two. Since there is only one A sub-object, if the constructors of B and C were allowed to initialize it, we would have (potentially dangers) double initialization. So the initializations of A in B's and C's constructors are ignored when a D is created, and only the initialization of A in D's constructor is used. That's also why you get compile-time error if you don't explicitly initialize the A sub-object from D's constructor, and don't provide a default constructor for A. This is covered in TC++PL 3 (or SE) 15.2.4[.1] . Quote: { public: D(int d): A(d), B(d + 1), C(d + 2) { } }; [snip] [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Dave Theese Guest Posted: Wed Aug 27, 2003 3:14 pm    Post subject: Re: Initialization of virtual bases Let me try to refine the discussion at this point... Given that the virtual base may only be initialized once (since only one instance of it exists, by definition), I'm curious as to why, syntactically (even if not in reality), all three of the classes derived from the virtual base must initialize it. Why force a programmer to show three initializations of the virtual base given that only one (that of the most-derived class) is actually used? Seems to make for confusing code... Perhaps eliminating the superfluous initializations of the virual base would present parsing / semantic checking difficulties. I mean, you would normally expect a derived class to explicitily initialize its base if the base has no default constructor. This is a somewhat unusual case (though not that unusual), and maybe it would be tough to implement this special case and have the ability to leave out the unused initializations... Perhaps that I see errors when I remove the superfluous initializations is an artifact of my platform (VC++ 7.1). Maybe other compilers would (correctly) not force me to show the unused initializations. But I doubt it... Quote: Hello all, The code example below has proper behavior (of course), but I'm trying to understand how the behavior is brought about. Specifically, what happens to B's and C's initialization of A??? Is it just ignored even though it explicitly appears in the code? The rule for initilaisation of virtual bases is simple, only the most derived class does it and all other bases both direct and indirect simply use the result. It is achieved by the implementation only calling the virtual base ctor from the most derived class (and before any other non-virtual base ctors) All other calls to a virtual base ctor are suppressed. #include <iostream> using namespace std; class A { public: A(int d): data(d) { cout << "data initialized to " << d << endl; } private: int data; }; class B: virtual public A { public: B(int d): A(d) { } }; class C: virtual public A { public: C(int d): A(d) { } }; class D: public B, public C { public: D(int d): A(d), B(d + 1), C(d + 2) { } }; int main(void) { // Displays "data initialized to 1" D foo(1); return 0; } [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Stephen Howe Guest Posted: Thu Aug 28, 2003 10:40 am    Post subject: Re: Initialization of virtual bases Quote: Given that the virtual base may only be initialized once (since only one instance of it exists, by definition), I'm curious as to why, syntactically (even if not in reality), all three of the classes derived from the virtual base must initialize it. Well what happens if you wanted D d(2); C c(3); B b(4); Here because you have B and C object that is not part of D, the A part still needs initialising. Stephen Howe [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top kanze@gabi-soft.fr Guest Posted: Thu Aug 28, 2003 8:34 pm    Post subject: Re: Initialization of virtual bases "Dave Theese" <cheeser_1998 (AT) yahoo (DOT) com> wrote Quote: Let me try to refine the discussion at this point... Given that the virtual base may only be initialized once (since only one instance of it exists, by definition), I'm curious as to why, syntactically (even if not in reality), all three of the classes derived from the virtual base must initialize it. Why force a programmer to show three initializations of the virtual base given that only one (that of the most-derived class) is actually used? Seems to make for confusing code... The problem is that the most derived class may change, and may not be known to the compiler. Thus, practically speaking, if the virtual base class has no default constructor, every derived class must contain an initializer for the virtual base. This is NOT what the standard says; the standard says only that the most derived class must provide the initialization. It is, however, what all compilers that I know of do. In fact, the last time I tried it (some time ago, I'll admit, and I forget with which compiler), the compiler required even an abstract derived class (which could never be the most derived class of an instance) to provide an initializer. Quote: Perhaps eliminating the superfluous initializations of the virual base would present parsing / semantic checking difficulties. No particular problems for the compiler, but it would change when errors are detected. struct VB { VB( int i ) ; } ; struct L : virtual VB {} ; struct R : virtual VB {} ; struct D : L, R {} ; The compiler cannot know that D is a most derived class (nor that L or R won't be for some instances). Current implementations generate compile time errors for all three classes. If the compiler is to forgo the error in L and R, it must also forgo it in D, since it has no way of knowing that D will be the end of the inheritance. The earliest that the compiler could generate an error for D is when we tried to instantiate it -- L, R, and D will have effectively become abstract classes. To tell the truth, I don't know whether the current behavior was more or less accidental in CFront, and then copied by all later compilers, or whether it was intentional, and the wording of the standard just doesn't reflect the intent. Or, which is also quite possible, I have missed something in the standard, although §12.6.2/8 seems quite clear: the initialization is ill formed if and only if the constructor of the most derived class does not specify a mem-initializer, and the virtual base class does not have an accessible default constructor. Anyway, as it now stands, the following program is perfectly legal according to the standard, but neither g++ (3.2.2) nor Sun CC (5.1) (nor most other compilers, I suspect) accept it: #include <iostream> #include <ostream> struct VB { VB( int i ) ; } ; struct L : virtual VB {} ; struct R : virtual VB {} ; struct D : L, R {} ; struct E : D { E() : VB( 1 ) {} } ; VB::VB( int i ) { std::cout << "VB::VB: " << i << 'n' ; } int main() { E anE ; return 0 ; } At any rate, my usual rule is to require a default constructor in all virtual bases. Most of the time, the virtual base class will be abstract, will not contain any data, and so will only have a default constructor, but in the exceptional case, if no default constructor is reasonable, I will provide one which aborts with an error message -- this allows compiling something like the above, but it does shift the error detection if I try to instantiate D to runtime. -- James Kanze GABI Software mailto:kanze (AT) gabi-soft (DOT) fr Conseils en informatique orientée objet/ http://www.gabi-soft.fr Beratung in objektorientierter Datenverarbeitung 11 rue de Rambouillet, 78460 Chevreuse, France, +33 (0)1 30 23 45 16 [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Francis Glassborow Guest Posted: Thu Aug 28, 2003 8:36 pm    Post subject: Re: Initialization of virtual bases In article <3f4cd366$0$246$ed9e5944 (AT) reading (DOT) news.pipex.net>, Stephen Howe <SPAMstephen.howeGUARD (AT) tnsofres (DOT) com> writes Quote: Given that the virtual base may only be initialized once (since only one instance of it exists, by definition), I'm curious as to why, syntactically (even if not in reality), all three of the classes derived from the virtual base must initialize it. Well what happens if you wanted D d(2); C c(3); B b(4); Here because you have B and C object that is not part of D, the A part still needs initialising. In addition, a virtual base that requires explicit initialisation seems suspect to me. Of course there are exceptions, but in general a virtual base class should only have a default and possibly a copy ctor. If it has any other ctor that should (apart from a copy ctor) be its only ctor. Virtual base classes with multiple ctors are prone to abuse. -- Francis Glassborow ACCU 64 Southfield Rd Oxford OX4 1PA +44(0)1865 246490 All opinions are mine and do not represent those of any organisation [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Attila Feher Guest Posted: Thu Aug 28, 2003 8:39 pm    Post subject: Re: Initialization of virtual bases Dave Theese wrote: [SNIP] Quote: Perhaps that I see errors when I remove the superfluous initializations is an artifact of my platform (VC++ 7.1). Maybe other compilers would (correctly) not force me to show the unused initializations. But I doubt it... Unless your base classes (up to the most derived) are all abstract, why would that initialization be useless? All those base classes _can_ be instantiated and at that time _they_ will call the virtual bases constructor. -- Attila aka WW [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Display posts from previous: All Posts1 Day7 Days2 Weeks1 Month3 Months6 Months1 Year Oldest FirstNewest First         C++Talk.NET Forum Index -> C++ Language (Moderated) All times are GMT Page 1 of 1   Jump to: Select a forum C/C++----------------C LanguageC Language (Moderated)C++ Language (Moderated)C++ language (comp.lang.c++)C++ language, library and standards Non-English C/C++ Newsgroups----------------C++ (German)C++ (French)  You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum Powered by phpBB © 2001, 2006 phpBB Group SEO toolkit © 2004-2006 webmedic.