C++Talk.NET C++ language newsgroups   Archives   FAQ   Search   Memberlist   Usergroups   Register   Profile   Log in to check your private messages   Log in  What exactly the const keyword means in this C++ code? Goto page 1, 2  Next          C++Talk.NET Forum Index -> C++ Language (Moderated) View previous topic :: View next topic   Author Message Sergei Zyablov Guest Posted: Tue Sep 21, 2004 6:29 pm    Post subject: What exactly the const keyword means in this C++ code? ....Whether it influece the stack? What benefit it offers? struct Record; const Record func(){ Record res; // ... return res; } Thanks in advance. [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top mmm Guest Posted: Wed Sep 22, 2004 9:58 am    Post subject: Re: What exactly the const keyword means in this C++ code? keyword const means read only. const returntype func() : it means that returntype is readonly (ex: func().objMbr = 3; will failed) -> const record func(); -> read only record returned -> const record * func(); -> rw pointer on read only record returned -> record *const func(); -> read only pointer on rw record returned returntype func(const type arg,const type &arg2); -> arg & arg2 are read only -> arg is duplicated before call when making argument list in stack -> arg2 is in fact a pointer visible in the method as it was the pointed object directly (&=reference), const &arg is very usefull when type is a class or a struct and when arg is accessed only for read operations to boost call perfo since it avoids the construction of a temporary object. Personnaly, I don't use the form &arg (without const kw) for method arguments, i prefer declare io arguments by passing pointer explicitly (dcl : type *pArg) Only meaningfull for class/struct methods (no sense for global methods): returntype func() const : it means that method doesn't modify class instance (*this) You can provide a method with an implementation for const operation different from rw operation, as following: It is correct to write for the same class/struct: const returntype &func() const; returntype &func(); const is a keyword very usefull : - it permits to increase compilation step checks, - it's a usefull information on methods to understand what they do, - and in some cases, most of times for implementation of base/library objects (like array templates, string object,...) it can be used efficiently to optimize code speed by differenciating rw of read only operations. But I think too that const keyword is not fine enough in c++, and so it is relatively difficult but not impossible to use correctly. A keyword to know absolutely to success with const keyword: : mutable, at the condition to use it only when necessary, and to use as follow: struct/class blabla { mutable membertype mbr; -> it means that mbr is accessible for write operation in a const method } Another thing: const has no effect on stack since you can cast a const variable to a rw variable and so compiler reserve space for const variables. Ex: const int a = 0; int &b = *(int*)&a; b=1; Beuurkk, you say ? you have right, it works for globals, method arguments, stack var: but please do not use this, be kind with your compiler, and with eyes of others (sorry for them ). I probaly forget things ;-) [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Victor Bazarov Guest Posted: Wed Sep 22, 2004 9:59 am    Post subject: Re: What exactly the const keyword means in this C++ code? "Sergei Zyablov" <zbl (AT) gmail (DOT) ru> wrote... Quote: ...Whether it influece the stack? What benefit it offers? None, in most cases. Quote: struct Record; const Record func(){ Record res; // ... return res; } struct C { void foo(); void bar() const; }; const C foobar(); int main() { foobar().foo(); // will give an error foobar().bar(); } As you can see, if I declare the returned object 'const', I can't call a non-const member function for it (thus modifying it on the fly). That's the only "benefit" I can think of. Victor [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Zian Smith Guest Posted: Wed Sep 22, 2004 10:04 am    Post subject: Re: What exactly the const keyword means in this C++ code? [email]zbl (AT) gmail (DOT) ru[/email] (Sergei Zyablov) wrote in message news:<c76c8a0f.0409210306.23d554ad (AT) posting (DOT) google.com>... Quote: ...Whether it influece the stack? What benefit it offers? struct Record; const Record func(){ Record res; // ... return res; } if you mean the benefits of returning a const from a function, it essentially means you cannot use the function on the left hand side of your expression. So in your case something like func() = foo_record; will be flagged by the compiler. I doubt you'll come across many scenarios where you will have the function on the LHS of the expression. (Maybe if you're returning a reference to a static variable in the function, and dont want a user to modify that variable?) One common situation however is when you're overloading operaters. Ensuring that the overloaded operator returns a const means you prevent the user from doing stuff like: a + b = c + d; -Z.Smith [ 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: Wed Sep 22, 2004 6:38 pm    Post subject: Re: What exactly the const keyword means in this C++ code? In article <cir2sg$ae2$1 (AT) news-reader2 (DOT) wanadoo.fr>, mmm <mmm (AT) devbooster (DOT) com> writes Quote: Another thing: const has no effect on stack since you can cast a const variable to a rw variable and so compiler reserve space for const variables. Sorry but that is entirely false. The language allows you to cast away top level const qualification, but the consequences of doing so are entirely the responsibility of the programmer. Nowhere in the language is there any requirement for how local variables shall be implemented. Quote: Ex: const int a = 0; At this point the compiler is not required to provide any storage for a, however it can mark that if it does so it is entitled to place that storage in write protected memory. As the programmer has specified that a contains an immutable value the compiler is entitled to fold all instances of a into a single one. Even if the function is recursive, I think it is allowed to provide a single block of memory initialised to zero at load time (even though a has not been declared static. Quote: int &b = *(int*)&a; The compiler is required to believe the programmer however it is not required to make the programmer's subsequent code work as expected. Quote: b=1; Boom. Quote: Beuurkk, you say ? you have right, it works for globals, method arguments, stack var: but please do not use this, be kind with your compiler, and with eyes of others (sorry for them ). What works? You have lied to the compiler and the results are undefined whether it is a local variable or something else is quite irrelevant. Quote: I probaly forget things -- Francis Glassborow ACCU Author of 'You Can Do It!' see http://www.spellen.org/youcandoit For project ideas and contributions: http://www.spellen.org/youcandoit/projects [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top jon hanson Guest Posted: Wed Sep 22, 2004 6:51 pm    Post subject: Re: What exactly the const keyword means in this C++ code? [email]zbl (AT) gmail (DOT) ru[/email] (Sergei Zyablov) wrote in message news:<c76c8a0f.0409210306.23d554ad (AT) posting (DOT) google.com>... Quote: ...Whether it influece the stack? What benefit it offers? struct Record; const Record func(){ Record res; // ... return res; } well it stops the following from compiling :- struct Record { int x; }; const Record test1 () { Record r; r.x = 10; return r; } void useRecord (Record& r) { r.x = 20; } void test2 () { useRecord (test1 ()); } useRecord has to be changed to something like :- void useRecord (const Record& r) { // r.x read-only; } jon Quote: Thanks in advance. [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Markus Elfring Guest Posted: Wed Sep 22, 2004 6:53 pm    Post subject: Re: What exactly the const keyword means in this C++ code? Quote: const Record func(); A constant return type prevents to try to change the temporary returned value. This setting is a detail of the topic "const-correctness". [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top mmm Guest Posted: Thu Sep 23, 2004 8:00 am    Post subject: Re: What exactly the const keyword means in this C++ code? "Francis Glassborow" wrote: Quote: mmm (AT) devbooster (DOT) com> writes Another thing: const has no effect on stack since you can cast a const variable to a rw variable and so compiler reserve space for const variables. Sorry but that is entirely false. The language allows you to cast away top level const qualification, but the consequences of doing so are entirely the responsibility of the programmer. Nowhere in the language is there any requirement for how local variables shall be implemented. Sorry for this big mistake, and thanks for your sympathic response , I have wrote this a little too fast without serious check, resorry. I never recommand this "style" of programming, I "just" misconjectured the code generated by const global & local vars. It's not entirely false, it's just at most a possible error source if C cast are used without precautions, it "works" for method arguments for all compilers () (except for fastcall like calling conv, i don't try this ). You have right, "Behaviour" for this case depends of compilers implementations and target OS/processors capabilities concerning segment protection (most of them now ). After verif (on vc6), the poor code i gived crashes at exec time for globals (for the reason you give: data segment protected from write op, here the compiler should be able to see that generated code for my example can't work, ds:[xx] is directly affected). For const local vars, vc6 replaces directly when possible occurences of const local&global vars calls by their values, and by the way i have "discovered" a totally uninteresting thing(): vc6 has a funny behaviour, my sample code changes the constants(local const only) at compilation step before substitutions (affectations was before displaying values)... Ok, vc is not a reference... Neverthe'less, sincerely thanks for your explainations and corrections. [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Peter A. Kerzum Guest Posted: Thu Sep 23, 2004 8:12 am    Post subject: Re: What exactly the const keyword means in this C++ code? All said above being correct, I want to put a note here: as of primary design intention 'const' keyword (actually first named 'readonly' as of D&E) serves 2 different purposes: 1. Covered above - to restrict access to data as read-only. 2. Applied to global or static variables - to ensure that data will not be modified during runtime. As this 2 cases may look identical, they are actually different. The point here is that really const, second case data structure might be initialized in compilation stage and put into 'code' or 'initialized data' segments that are read-only in terms of CPU. Therefore C++ standard says that writing to memory initialy marked as const should be considered an 'undefined behaviour'. Eg: int main(void) { const static int i=5; int& b=const_cast<int&>(i); b=10; // this is OK to get Segmentation violation here 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 kanze@gabi-soft.fr Guest Posted: Fri Sep 24, 2004 11:23 am    Post subject: Re: What exactly the const keyword means in this C++ code? Francis Glassborow <francis (AT) robinton (DOT) demon.co.uk> wrote Quote: Ex: const int a = 0; At this point the compiler is not required to provide any storage for a, however it can mark that if it does so it is entitled to place that storage in write protected memory. As the programmer has specified that a contains an immutable value the compiler is entitled to fold all instances of a into a single one. Even if the function is recursive, I think it is allowed to provide a single block of memory initialised to zero at load time (even though a has not been declared static. If the function takes the address and does something with it, different instances in different occurances of the function are required to have different addresses. Quote: int &b = *(int*)&a; The compiler is required to believe the programmer however it is not required to make the programmer's subsequent code work as expected. b=1; Boom. Maybe. If it would reliably go boom, that would be nice. More likely, the compiler will continue to use 0 each time it sees an a, because it knows that the value cannot change. Or sometimes it will use the new value, and sometimes the old. (In a simple case, all of my compilers use the old value. But I'm sure that I could, without too much difficulty, construct cases where they mixed the values.) -- James Kanze GABI Software http://www.gabi-soft.fr Conseils en informatique orientée objet/ Beratung in objektorientierter Datenverarbeitung 9 place Sémard, 78210 St.-Cyr-l'École, France, +33 (0)1 30 23 00 34 [ 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: Fri Sep 24, 2004 11:23 am    Post subject: Re: What exactly the const keyword means in this C++ code? "mmm" <mmm (AT) devbooster (DOT) com> wrote Quote: keyword const means read only. It's a bit more complicated than that. Const is part of the type system. But I don't think that the original poster was asking about const in general. I think he was wondering about its use on a return value. After all, return values are rvalues, and for non-class types, top level const is ignored on rvalues. Quote: const returntype func() : it means that returntype is readonly (ex: func().objMbr = 3; will failed) If objMbr is not of class type, func().objMbr = 3 is illegal with or without the const. Quote: -> const record func(); -> read only record returned -> const record * func(); -> rw pointer on read only record returned -> record *const func(); -> read only pointer on rw record returned Except that return values are rvalues, so top level const is ignored on non-class types. The only time top level const on a return value plays a role is for class types with member functions -- you cannot call non-const member functions on the returned object if it is declared const. I seem to remember someone (Scott Meyers?) recommending that class type return values be systematically declared const, since the operator= is normally a non-const function, and you don't want to see it called on your return value. IMHO, this makes sense, but I can't say that I've seen this convention widely implemented. Quote: returntype func(const type arg,const type &arg2); -> arg & arg2 are read only Again, top level const is ignored on function parameters in a function declaration. It only has an effect in the body of the function. Quote: -> arg is duplicated before call when making argument list in stack -> arg2 is in fact a pointer visible in the method as it was the pointed object directly (&=reference), That is a frequent implementation. But references aren't pointers, according to the standard. Quote: const &arg is very usefull when type is a class or a struct and when arg is accessed only for read More importantly, you cannot bind a temporary to a non-const reference. [...] Quote: Another thing: const has no effect on stack since you can cast a const variable to a rw variable and so compiler reserve space for const variables. Ex: const int a = 0; int &b = *(int*)&a; b=1; Beuurkk, you say ? you have right, it works for globals, method arguments, stack var: but please do not use this, be kind with your compiler, and with eyes of others (sorry for them ). It's undefined behavior. It may work, some of the time, or it may not. Stack or not. (In practice, I know of no compiler where it works systematically.) -- James Kanze GABI Software http://www.gabi-soft.fr Conseils en informatique orientée objet/ Beratung in objektorientierter Datenverarbeitung 9 place Sémard, 78210 St.-Cyr-l'École, France, +33 (0)1 30 23 00 34 [ 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: Fri Sep 24, 2004 2:08 pm    Post subject: Re: What exactly the const keyword means in this C++ code? In article <cit6ov$1at$1 (AT) news-reader1 (DOT) wanadoo.fr>, mmm <mmm (AT) devbooster (DOT) com> writes Quote: "Francis Glassborow" wrote: [email]mmm (AT) devbooster (DOT) com[/email]> writes Another thing: const has no effect on stack since you can cast a const variable to a rw variable and so compiler reserve space for const variables. Sorry but that is entirely false. The language allows you to cast away top level const qualification, but the consequences of doing so are entirely the responsibility of the programmer. Nowhere in the language is there any requirement for how local variables shall be implemented. Sorry for this big mistake, and thanks for your sympathic response , Sorry if my writing style came across as being a personal criticism. I sometimes have to swallow hard and count to ten when someone tells me that I am entirely wrong about something, but then I move on and learn from my error (and my increased knowledge is a direct result of my willingness to risk being wrong) Quote: I have wrote this a little too fast without serious check, resorry. I never recommand this "style" of programming, I "just" misconjectured the code generated by const global & local vars. It's not entirely false, it's just at most a possible error source if C cast are used without precautions, it "works" for method arguments for all compilers () (except for fastcall like calling conv, i don't try this ). I think you may not understand the implications of undefined behaviour. The particularly nasty aspect of it is that it often manifests by doing exactly what the programmer expects. Then the code is compiled with a different set of compiler switches, or a different compiler, or a new version of the existing one and some actually nasty behaviour occurs. Any form of cast that attempts to remove const qualification from an object (note that a reference is not an object in itself but refers to an object) that has been declared as const allows the compiler to be as nasty as it likes. Of course implementors do not actually set out to be nasty (though that might be a positive feature if the implementation was for learning purposes only) but the language rules allow them certain types of latitude based on the assumption that programmers do not deliberately lie to the compiler. I believe that const_cast should never be applied to an object but only to a reference to const or pointer to const object. IOWs it is OK to sometimes remove a 'read only' qualifier but never correct to remove an immutable one. Unfortunately C++ uses the same keyword for both even though it is always clear in context as to which meaning is intended. What makes it harder is that there is no runtime mechanism to determine if an object is immutable. That lack makes most uses of const_cast<> suspect. About the only place that I would use const_cast<> is in implementing the idiom: mytype const * foo(mytype const *); mytype * foo (mytype * mt){ return const_cast<mytytpe *>foo(const_cast<mytype const *>mt); } Whose main purpose is to prevent references to non-const objects acquiring const qualification. In addition I would view most uses of C style casts and function style casts with deep suspicion. I do accept their use in a limited number of situations in informal code. The main case being code such as: enum colour { red=1, green=2, blue=4}; colour const yellow = colour(red + green); -- Francis Glassborow ACCU Author of 'You Can Do It!' see http://www.spellen.org/youcandoit For project ideas and contributions: http://www.spellen.org/youcandoit/projects [ 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: Fri Sep 24, 2004 2:09 pm    Post subject: Re: What exactly the const keyword means in this C++ code? In article <d4039ed4.0409221159.7a3c76f0 (AT) posting (DOT) google.com>, Peter A. Kerzum <kerzum (AT) mail (DOT) ru> writes Quote: All said above being correct, I want to put a note here: as of primary design intention 'const' keyword (actually first named 'readonly' as of D&E) serves 2 different purposes: 1. Covered above - to restrict access to data as read-only. 2. Applied to global or static variables - to ensure that data will not be modified during runtime. Why do you believe that the second case only applies to static (either declared as static or at namespace scope) variables? -- Francis Glassborow ACCU Author of 'You Can Do It!' see http://www.spellen.org/youcandoit For project ideas and contributions: http://www.spellen.org/youcandoit/projects [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Chris Uzdavinis Guest Posted: Fri Sep 24, 2004 10:18 pm    Post subject: Re: What exactly the const keyword means in this C++ code? [email]j0n_h4ns0n (AT) yahoo (DOT) co.uk[/email] (jon hanson) wrote in message Quote: What benefit it offers? well it stops the following from compiling :- struct Record { int x; }; const Record test1 () { Record r; r.x = 10; return r; } void useRecord (Record& r) { r.x = 20; } void test2 () { useRecord (test1 ()); } useRecord has to be changed to something like :- void useRecord (const Record& r) { // r.x read-only; } You're correct in that the above code won't compile without your suggested fix, but the "const" keyword on the return type of test1() has no bearing on that. The reason your code fails to compile is that test1() returns a temporary varaible, and you're trying to bind a reference to a non-const object to that temporary (when passing it to useRecord()). That is not allowed. However, binding a reference-to-const to an unnamed temporary variable is allowed. -- Chris [ 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: Sat Sep 25, 2004 2:50 pm    Post subject: Re: What exactly the const keyword means in this C++ code? In article <d6652001.0409230029.2520c445 (AT) posting (DOT) google.com>, [email]kanze (AT) gabi-soft (DOT) fr[/email] writes Quote: Francis Glassborow <francis (AT) robinton (DOT) demon.co.uk> wrote in message news:<1kwSXTT$NWUBFwbM (AT) robinton (DOT) demon.co.uk>... Ex: const int a = 0; At this point the compiler is not required to provide any storage for a, however it can mark that if it does so it is entitled to place that storage in write protected memory. As the programmer has specified that a contains an immutable value the compiler is entitled to fold all instances of a into a single one. Even if the function is recursive, I think it is allowed to provide a single block of memory initialised to zero at load time (even though a has not been declared static. If the function takes the address and does something with it, different instances in different occurances of the function are required to have different addresses. I am not going to say you are mistaken, but where is that requirement in the C++ Standard? And does that requirement also apply to cases where the const object is referenced? Given that it is undefined behaviour to modify a const instance, and that there is no non-trivial construction/destruction semantics for instances of fundamental types the only remaining potential is for comparison of the pointers to two instances. Even simultaneous reads in different threads running on different processors should not be problematic. In order to obtain the addresses of two instances the function has to be recursive. Any old recursive function is not enough, it has to either store that address in a non-local object or it must have a suitable parameter by which the address can be acquired. Unless there is a good counter-example, I think I would argue that a compiler should be allowed to treat physically const objects of fundamental types and types with trivial ctors/dtors as static objects. -- Francis Glassborow ACCU Author of 'You Can Do It!' see http://www.spellen.org/youcandoit For project ideas and contributions: http://www.spellen.org/youcandoit/projects [ 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 Goto page 1, 2  Next Page 1 of 2   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.