C++Talk.NET C++ language newsgroups   Archives   FAQ   Search   Memberlist   Usergroups   Register   Profile   Log in to check your private messages   Log in  question on std::vector default ctor Goto page 1, 2, 3, 4  Next          C++Talk.NET Forum Index -> C++ Language (Moderated) View previous topic :: View next topic   Author Message harkness Guest Posted: Mon Dec 08, 2003 11:00 pm    Post subject: question on std::vector default ctor In _The C++ Standard Library_ by Josuttis, page 149, is the assertion that std::vector<T> v(5); calls the default ctor of T 5 times. I wrote a test case for this using gcc 3.2, and the default ctor was called only once. Which of the following is true? 1) Josuttis was mistaken 2) The standard has changed 3) This is a gcc 3.2 bug (optimization?) 4) It's due to some other reason I haven't thought of TIA (test program follows) // vectorsize.cpp -- example of sizing a vector #include <vector> using std::vector; #include <iostream> using std::cout; using std::endl; using std::ostream; // poor man's profiler I got from Dr. Cargill #define DOTRACE #include "Trace.h" class MyClass { public: MyClass(); friend ostream& operator<<(ostream& os, const MyClass& mc); private: int m_i; }; int globali(0); MyClass::MyClass() { TRACE("MyClass", "Default ctor"); cout << "assigning m_i" << endl; m_i = globali++; } ostream& operator<<(ostream& os, const MyClass& mc) { os << mc.m_i; return os; } typedef vector typedef MVector::const_iterator MCIter; int main() { srand(1); // seed MVector mv(; // create and initialize // should call default ctor 8 times int i=-1; for(MCIter iter = mv.begin(); iter != mv.end(); ++iter) { cout << "mv[" << ++i << "]=" << *iter << endl; } return 0; } Output, showing only one ctor call: ..entering:MyClass Default ctor assigning m_i ..exiting: MyClass Default ctor mv[0]=0 mv[1]=0 mv[2]=0 mv[3]=0 mv[4]=0 mv[5]=0 mv[6]=0 mv[7]=0 -- Howard [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Dag Henriksson Guest Posted: Tue Dec 09, 2003 11:03 am    Post subject: Re: question on std::vector default ctor "harkness" <harkness (AT) mail (DOT) ev1.net> skrev i meddelandet news:200312081249.AA185925764 (AT) mail (DOT) ev1.net... Quote: In _The C++ Standard Library_ by Josuttis, page 149, is the assertion that std::vector<T> v(5); calls the default ctor of T 5 times. I wrote a test case for this using gcc 3.2, and the default ctor was called only once. Which of the following is true? 1) Josuttis was mistaken Yes, I think so. The constructor used above is: explicit vector(size_type n, const T& value = T(), const Allocator& = Allocator()); It creates a temporary T which is used to copy initialize the 5 elements in the vector. The default ctor should be called once and the copy ctor five times. -- Dag Henriksson [ 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 Dec 09, 2003 5:09 pm    Post subject: Re: question on std::vector default ctor In article <v07atvoifjvuek2kqmsfa1p1vjcsaq0p3a (AT) 4ax (DOT) com>, Howard <hlh_nospam (AT) iwon (DOT) com> writes Quote: I didn't see anything in Stroustrup about this, and so far, I haven't turned anything up with a websearch. I can't imagine that I'm the only person to run across this. I guess it's time to break down and order my copy of the standard. What the Standard requires is that the ctor for vector<T> make n calls to the copy ctor for T to copy the argument passed to the second (defaulted) reference parameter of the vector ctor. The default is a value initialised instance of T (i.e. T()). It is common (if not universal) that the compiler generated copy ctor for a POD is simply a call of memcpy. Try using a T with an instrumented copy ctor and you will see what happens (though now it will no longer be using memcpy, unless you do so within the copy ctor. -- Francis Glassborow ACCU Author of 'You Can Do It!' see http://www.spellen.org/youcandoit [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Roger Orr Guest Posted: Tue Dec 09, 2003 7:59 pm    Post subject: Re: question on std::vector default ctor "Dag Henriksson" <dag.henriksson (AT) quidsoft (DOT) se> wrote Quote: "harkness" <harkness (AT) mail (DOT) ev1.net> skrev i meddelandet news:200312081249.AA185925764 (AT) mail (DOT) ev1.net... In _The C++ Standard Library_ by Josuttis, page 149, is the assertion that std::vector<T> v(5); [snip] Quote: The constructor used above is: explicit vector(size_type n, const T& value = T(), const Allocator& = Allocator()); It creates a temporary T which is used to copy initialize the 5 elements in the vector. The default ctor should be called once and the copy ctor five times. I would certain expect any sensible implementation would do this - but I can't actually find anything in the standard which mandates this. There don't seem to be any explicit complexity guarantees for this std::vector constructor. Roger Orr -- MVP in C++ at www.brainbench.com [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Randy Maddox Guest Posted: Tue Dec 09, 2003 8:08 pm    Post subject: Re: question on std::vector default ctor "harkness" <harkness (AT) mail (DOT) ev1.net> wrote Quote: In _The C++ Standard Library_ by Josuttis, page 149, is the assertion that std::vector<T> v(5); calls the default ctor of T 5 times. I wrote a test case for this using gcc 3.2, and the default ctor was called only once. Which of the following is true? 1) Josuttis was mistaken 2) The standard has changed 3) This is a gcc 3.2 bug (optimization?) 4) It's due to some other reason I haven't thought of Look at the ctor you are using: explicit vector(size_type n, const T & value = T(), const Allocator & allocator = Allocator); Since you are passing only the size_type n argument the others default. So the default ctor for T is called once and the copy ctor for T is called 5 times. Randy. [ 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: Tue Dec 09, 2003 8:09 pm    Post subject: Re: question on std::vector default ctor "harkness" <harkness (AT) mail (DOT) ev1.net> wrote Quote: In _The C++ Standard Library_ by Josuttis, page 149, is the assertion that std::vector<T> v(5); calls the default ctor of T 5 times. I wrote a test case for this using gcc 3.2, and the default ctor was called only once. Which of the following is true? 1) Josuttis was mistaken 2) The standard has changed 3) This is a gcc 3.2 bug (optimization?) 4) It's due to some other reason I haven't thought of If this is actually what Josuttis wrote, he is mistaken. (He actually wrote the book some time ago, when the library was still in a state of flux, but I don't think that this point has changed.) Your statement is the equivalent of: std::vector< T > v( 5, T() ) ; In this statement, the standard requires the default constructor to be called once, and the copy constructor to be called 5 times. -- 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 John Potter Guest Posted: Tue Dec 09, 2003 8:10 pm    Post subject: Re: question on std::vector default ctor On 8 Dec 2003 18:00:33 -0500, "harkness" <harkness (AT) mail (DOT) ev1.net> wrote: Quote: In _The C++ Standard Library_ by Josuttis, page 149, is the assertion that std::vector<T> v(5); calls the default ctor of T 5 times. I wrote a test case for this using gcc 3.2, and the default ctor was called only once. Which of the following is true? 1) Josuttis was mistaken The above uses the ctor with a defaulted T(). There is no vector ctor that takes only a size_type. std::vector<T> v(5, T()); Now table 67 states that the sequence is constructed from n copies of the supplied t. Instrument your copy ctor to see the copies. Disclaimer: This could be written as two ctors with one taking a size_type and the other taking a size_type and a T. I can't find anything which allows or forbids using n default constructions in the case of the size_type only ctor. However, there is nothing which requires n default constructions. John [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Gabriel Dos Reis Guest Posted: Tue Dec 09, 2003 10:06 pm    Post subject: Re: question on std::vector default ctor [email]kanze (AT) gabi-soft (DOT) fr[/email] writes: Quote: "harkness" <harkness (AT) mail (DOT) ev1.net> wrote in message news:<200312081249.AA185925764 (AT) mail (DOT) ev1.net>... In _The C++ Standard Library_ by Josuttis, page 149, is the assertion that std::vector<T> v(5); calls the default ctor of T 5 times. I wrote a test case for this using gcc 3.2, and the default ctor was called only once. Which of the following is true? 1) Josuttis was mistaken 2) The standard has changed 3) This is a gcc 3.2 bug (optimization?) 4) It's due to some other reason I haven't thought of If this is actually what Josuttis wrote, he is mistaken. (He actually wrote the book some time ago, when the library was still in a state of flux, but I don't think that this point has changed.) Your statement is the equivalent of: std::vector< T > v( 5, T() ) ; That form is permitted, but it is not a requirement. Quote: In this statement, the standard requires the default constructor to be called once, and the copy constructor to be called 5 times. This is wrong. The Standard does not require the default constructor to be called once and the copy constructor to be called 5 times. Rather, the Standard permits both behaviour. -- Gabriel Dos Reis [email]gdr (AT) integrable-solutions (DOT) net[/email] [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Gabriel Dos Reis Guest Posted: Tue Dec 09, 2003 10:11 pm    Post subject: Re: question on std::vector default ctor John Potter <jpotter (AT) falcon (DOT) lhup.edu> writes: Quote: On 8 Dec 2003 18:00:33 -0500, "harkness" <harkness (AT) mail (DOT) ev1.net> wrote: In _The C++ Standard Library_ by Josuttis, page 149, is the assertion that std::vector<T> v(5); calls the default ctor of T 5 times. I wrote a test case for this using gcc 3.2, and the default ctor was called only once. Which of the following is true? 1) Josuttis was mistaken The above uses the ctor with a defaulted T(). There is no vector ctor that takes only a size_type. std::vector<T> v(5, T()); Now table 67 states that the sequence is constructed from n copies of the supplied t. Instrument your copy ctor to see the copies. Disclaimer: This could be written as two ctors with one taking a size_type and the other taking a size_type and a T. I can't find anything which allows or forbids using n default constructions in the case of the size_type only ctor. However, there is nothing which requires n default constructions. The Standard says that whenever a member function is described to take default arguments, the implementation is permitted to "split" that member function into as many parts. -- Gabriel Dos Reis [email]gdr (AT) integrable-solutions (DOT) net[/email] [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top John Potter Guest Posted: Wed Dec 10, 2003 11:17 am    Post subject: Re: question on std::vector default ctor On 9 Dec 2003 17:11:50 -0500, Gabriel Dos Reis <gdr (AT) integrable-solutions (DOT) net> wrote: Quote: John Potter <jpotter (AT) falcon (DOT) lhup.edu> writes: | Disclaimer: This could be written as two ctors with one taking a | size_type and the other taking a size_type and a T. I can't find | anything which allows or forbids using n default constructions in the | case of the size_type only ctor. However, there is nothing which | requires n default constructions. The Standard says that whenever a member function is described to take default arguments, the implementation is permitted to "split" that member function into as many parts. I thought I said that. If it does, it could either default construct one item and copy construct N from it or default construct N. Or can it? It must use the allocator. How do you default construct using an allocator? The member and uninitialized_X algorithms all copy construct. Is it allowed to use placement new in allocated space? John [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Bjarne Stroustrup Guest Posted: Wed Dec 10, 2003 11:25 am    Post subject: Re: question on std::vector default ctor Francis Glassborow <francis (AT) robinton (DOT) demon.co.uk> wrote Quote: In article <v07atvoifjvuek2kqmsfa1p1vjcsaq0p3a (AT) 4ax (DOT) com>, Howard [email]hlh_nospam (AT) iwon (DOT) com[/email]> writes I didn't see anything in Stroustrup about this, and so far, I haven't turned anything up with a websearch. I can't imagine that I'm the only person to run across this. I guess it's time to break down and order my copy of the standard. What the Standard requires is that the ctor for vector<T> make n calls to the copy ctor for T to copy the argument passed to the second (defaulted) reference parameter of the vector ctor. The default is a value initialised instance of T (i.e. T()). It is common (if not universal) that the compiler generated copy ctor for a POD is simply a call of memcpy. Try using a T with an instrumented copy ctor and you will see what happens (though now it will no longer be using memcpy, unless you do so within the copy ctor. That's correct, but I'd like my 2 cents about the "I didn't see anything in Stroustrup about this". On pg 447 there is a coment "n copies of val" and a paragraph that elements are initialized by the default constructor of the element type that I suspect could easily be misread. Luckily, pg 944 I happen to use exactly the initialize-with-n-copies-of-val constructor for vector as an illustration of exception safety techniques, so you can see what is done and how. - Bjarne Stroustrup; http://www.research.att.com/~bs [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top James Dennett Guest Posted: Wed Dec 10, 2003 11:38 am    Post subject: Re: question on std::vector default ctor Gabriel Dos Reis wrote: Quote: kanze (AT) gabi-soft (DOT) fr writes: | "harkness" <harkness (AT) mail (DOT) ev1.net> wrote in message | news:<200312081249.AA185925764 (AT) mail (DOT) ev1.net>... | | > In _The C++ Standard Library_ by Josuttis, page 149, is the assertion | > that | | > std::vector<T> v(5); | | > calls the default ctor of T 5 times. I wrote a test case for this | > using gcc 3.2, and the default ctor was called only once. Which of the | > following is true? | | > 1) Josuttis was mistaken | > 2) The standard has changed | > 3) This is a gcc 3.2 bug (optimization?) | > 4) It's due to some other reason I haven't thought of | | If this is actually what Josuttis wrote, he is mistaken. (He actually | wrote the book some time ago, when the library was still in a state of | flux, but I don't think that this point has changed.) Your statement is | the equivalent of: | | std::vector< T > v( 5, T() ) ; That form is permitted, but it is not a requirement. How so? There is no single-argument constructor taking an integral value directly specified: while the library may be allowed to add one, its behaviour must surely still be as if the version with a default argument were used -- and that version cannot rely on being able to use the default constructor, as it must work with types which have no default constructor. All I see is that Quote: | In this statement, the standard requires the default constructor to be | called once, and the copy constructor to be called 5 times. This is wrong. The Standard does not require the default constructor to be called once and the copy constructor to be called 5 times. Rather, the Standard permits both behaviour. How so? Where is the freedom to do this (which is beyond the "as if" rule) granted? I do accept that you may well know of rules that I don't recall, or might not have been aware of -- if so, I would like to fill one more gap in my knowledge. I think James Kanze's expectation is that there is an equivalence by definition between the form std::vector<T> v(5); and std::vector<T> v(5, T()); and I find that expectation a reasonable one based on my reading of the standard. -- James. [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top James Dennett Guest Posted: Wed Dec 10, 2003 11:39 am    Post subject: Re: question on std::vector default ctor Gabriel Dos Reis wrote: Quote: John Potter <jpotter (AT) falcon (DOT) lhup.edu> writes: | On 8 Dec 2003 18:00:33 -0500, "harkness" <harkness (AT) mail (DOT) ev1.net> wrote: | | > In _The C++ Standard Library_ by Josuttis, page 149, is the assertion that | | > std::vector<T> v(5); | | > calls the default ctor of T 5 times. I wrote a test case for this using gcc 3.2, and the | > default ctor was called only once. Which of the following is true? | | > 1) Josuttis was mistaken | | The above uses the ctor with a defaulted T(). There is no vector ctor | that takes only a size_type. | | std::vector<T> v(5, T()); | | Now table 67 states that the sequence is constructed from n copies of | the supplied t. Instrument your copy ctor to see the copies. | | Disclaimer: This could be written as two ctors with one taking a | size_type and the other taking a size_type and a T. I can't find | anything which allows or forbids using n default constructions in the | case of the size_type only ctor. However, there is nothing which | requires n default constructions. The Standard says that whenever a member function is described to take default arguments, the implementation is permitted to "split" that member function into as many parts. So long as the effects of calling those separate functions are the same (up to the usual "as if" games) as would be achieved were a single member with default arguments used and called, no? Granted, a user can often tell the difference with regular member functions (e.g. by trying to take the address of one, cast to a suitable type to detect its presence) but I would expect that the effect of calling a substituted function should be the same as calling the version explicitly defined by the standard. -- James. [ 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: Wed Dec 10, 2003 3:59 pm    Post subject: Re: question on std::vector default ctor Gabriel Dos Reis <gdr (AT) integrable-solutions (DOT) net> wrote Quote: John Potter <jpotter (AT) falcon (DOT) lhup.edu> writes: | On 8 Dec 2003 18:00:33 -0500, "harkness" <harkness (AT) mail (DOT) ev1.net> wrote: | > In _The C++ Standard Library_ by Josuttis, page 149, is the | > assertion that std::vector<T> v(5); | > calls the default ctor of T 5 times. I wrote a test case for this | > using gcc 3.2, and the default ctor was called only once. Which | > of the following is true? | > 1) Josuttis was mistaken | The above uses the ctor with a defaulted T(). There is no vector | ctor that takes only a size_type. | std::vector<T> v(5, T()); | Now table 67 states that the sequence is constructed from n copies | of the supplied t. Instrument your copy ctor to see the copies. | Disclaimer: This could be written as two ctors with one taking a | size_type and the other taking a size_type and a T. I can't find | anything which allows or forbids using n default constructions in | the case of the size_type only ctor. However, there is nothing | which requires n default constructions. The Standard says that whenever a member function is described to take default arguments, the implementation is permitted to "split" that member function into as many parts. True, but it doesn't authorize different semantics for the two variants. If an implementation splits the two, it must still ensure that the std::vector<T>::vector( size_t ) constructor meets the requirements given in table 67 for X(n, t), in particular, that it "constructs a sequence with n copies of t". The wording could definitly be more precise, but since calling the copy constructor could have observable behavior, and there is no special wording to allow eliding it here, so I would interpret this requirement to mean that the copy constructor must be called at least n times. I'm far from sure that this is the intent, but it is definitly a reasonable interpretation of what the standard actually says. -- 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 Daniel Krügler (nee Spang Guest Posted: Wed Dec 10, 2003 4:14 pm    Post subject: Re: question on std::vector default ctor Hello, John Potter! John Potter schrieb: Quote: I thought I said that. If it does, it could either default construct one item and copy construct N from it or default construct N. Or can it? It must use the allocator. How do you default construct using an allocator? The member and uninitialized_X algorithms all copy construct. Is it allowed to use placement new in allocated space? Actually some days ago I stumbled about the same restriction of std::allocator. To my opinion its interface should be extended to allow placement new default construction (construct(T*) overload). The reason is: Also the Standard Libary Container Classes does require CopyConstructible as an property of its contents, this might have different performance and exception guarantees. Could we demand, that construct(T*) does either not exist or its call should not be a well-defined expression, in case of classes, which don't provide the DefaultConstructible requirement? (SFINAE might help here) Greetings from Bremen, Daniel [ 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, 3, 4  Next Page 1 of 4   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.