C++Talk.NET C++ language newsgroups   Archives   FAQ   Search   Memberlist   Usergroups   Register   Profile   Log in to check your private messages   Log in  Placement new and destructor Goto page 1, 2  Next          C++Talk.NET Forum Index -> C++ Language (Moderated) View previous topic :: View next topic   Author Message dragoncoder Guest Posted: Wed May 10, 2006 10:21 pm    Post subject: Placement new and destructor Hello all, I am reading C++ Primer 3rd edition by Lippman & Lajoie and come across the following piece of code as an example demostrating the use of placement new. Section 8.4.5 (Page 417-418) #include <iostream> #include <new> const int chunk = 16; class Foo { public: int val() { return _val; } Foo() { _val = 0; } private: int _val; }; char *buf = new char[sizeof (Foo) * chunk]; int main() { Foo *pb = new (buf) Foo; if ( pb.val() == 0 ) cout << "new expression worked!" << endl; delete[] buf; return 0; } Looking at the above code, I feel it is incorrect at 2 places. 1. Since pb is a pointer to a Foo object, the expression inside the if condition should have been pb->val() instead of pb.val(), okay that may be a printing mistake. 2. The second error is not a typographical error. I see at no place in the code the destructor of the Foo object is called. Instead delete[] buf calls the destructors of 16 char objects which is fine but I feel there should be a call to Foo's destructor before the delete[] buf. Shouldn't there be a call to pb->~Foo() before delete[] ? Please tell me, if I am missing something as this book is rated highly recommended at accu.org. [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top peter koch larsen Guest Posted: Thu May 11, 2006 1:21 pm    Post subject: Re: Placement new and destructor dragoncoder skrev: Quote: Hello all, I am reading C++ Primer 3rd edition by Lippman & Lajoie and come across the following piece of code as an example demostrating the use of placement new. I believe that book was one of the first C++ books I read (after Stroustrups), and I remember it as a very good book. But it does have flaws! Quote: Section 8.4.5 (Page 417-418) #include <iostream #include <new const int chunk = 16; class Foo { public: int val() { return _val; } Foo() { _val = 0; } private: int _val; }; char *buf = new char[sizeof (Foo) * chunk]; int main() { Foo *pb = new (buf) Foo; if ( pb.val() == 0 ) cout << "new expression worked!" << endl; delete[] buf; return 0; } Looking at the above code, I feel it is incorrect at 2 places. 1. Since pb is a pointer to a Foo object, the expression inside the if condition should have been pb->val() instead of pb.val(), okay that may be a printing mistake. You are correct. 2. The second error is not a typographical error. I see at no place in the code the destructor of the Foo object is called. Instead delete[] buf calls the destructors of 16 char objects which is fine but I feel there should be a call to Foo's destructor before the delete[] buf. Shouldn't there be a call to pb->~Foo() before delete[] ? I agree once again. This time it is not a direct error - you don't have to call destructors. And there is no leak - the buffer is deleted all right. So perhaps we should forgive Lippman and Lajoie considering that they aim to demonstrate placement new. Quote: Please tell me, if I am missing something as this book is rated highly recommended at accu.org. Then probably it is a very good book. No book is without errors and Lippman and Lajoie are both very knowledgeable - and they are fine writers. I have a feeling there might be a third error, however. The character vector is allocated via new char[]. Is new char guaranteed to allocate memory that is properly aligned for any datatype? So far as I understand it, the buffer is only guaranteed to be properly aligned for characters. /Peter [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top johnchx2@yahoo.com Guest Posted: Thu May 11, 2006 1:21 pm    Post subject: Re: Placement new and destructor dragoncoder wrote: Quote: 2. The second error is not a typographical error. I see at no place in the code the destructor of the Foo object is called. Instead delete[] buf calls the destructors of 16 char objects which is fine but I feel there should be a call to Foo's destructor before the delete[] buf. Shouldn't there be a call to pb->~Foo() before delete[] ? Since Foo has a trivial destructor, omitting the destructor call does no harm. If Foo's destructor were non-trivial and had side effects upon which the behavior of the program depends, omitting the destructor call would lead to undefined behavior (3.8.4). [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Eugene Alterman Guest Posted: Fri May 12, 2006 12:21 am    Post subject: Re: Placement new and destructor "peter koch larsen" <peter.koch.larsen (AT) gmail (DOT) com> wrote in message news:1147299859.239506.136800 (AT) i40g2000cwc (DOT) googlegroups.com... Quote: I have a feeling there might be a third error, however. The character vector is allocated via new char[]. Is new char guaranteed to allocate memory that is properly aligned for any datatype? So far as I understand it, the buffer is only guaranteed to be properly aligned for characters. Storage for objects created using new-expression is allocated by an allocation function (global operator new[] in this case) that is generally the same regardless of the type of an object being created, and it is thus aligned to satisfy the strictest alignment requirement. (3.7.3.1/2) [ 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 May 12, 2006 4:21 am    Post subject: Re: Placement new and destructor In article <1147272822.470577.270840 (AT) j33g2000cwa (DOT) googlegroups.com>, dragoncoder <pktiwary (AT) gmail (DOT) com> writes Quote: Hello all, I am reading C++ Primer 3rd edition by Lippman & Lajoie and come across the following piece of code as an example demostrating the use of placement new. A pity you did not choose to read the 4th edition. The third was good but the complete rewrite that Barbara Moo did in producing the 4th edition has turned a good book into an excellent one. Barbara is not only a fluent writer but has a deep knowledge of both programming and C++ (and has the good fortune to be married to another of the World's great programmers and writers on programming -- Andy Koenig whose influence can be seen between the words) -- Francis Glassborow ACCU Author of 'You Can Do It!'& 'You Can Program in C++' 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 kanze Guest Posted: Sat May 13, 2006 1:21 am    Post subject: Re: Placement new and destructor dan2online wrote: Quote: dragoncoder wrote: 2. The second error is not a typographical error. I see at no place in the code the destructor of the Foo object is called. Instead delete[] buf calls the destructors of 16 char objects which is fine but I feel there should be a call to Foo's destructor before the delete[] buf. Shouldn't there be a call to pb->~Foo() before delete[] ? Foo *pb = new (buf) Foo means that Foo object reuses the storage that ends the lifetime of buf. To delete 'buf' is to destroy the object at the same time. No it doesn't. delete[]buf will call the destructors of the type pointed to by buf -- here, char. Quote: So don't need to delete pb here. You're not allowed to delete pb, since pb wasn't allocated with the standard allocator. It would be cleaner, however, if they called pb->~Foo() before deleting buf. -- James Kanze GABI Software 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 Guest Posted: Sat May 13, 2006 1:21 am    Post subject: Re: Placement new and destructor peter koch larsen wrote: Quote: I have a feeling there might be a third error, however. The character vector is allocated via new char[]. Is new char guaranteed to allocate memory that is properly aligned for any datatype? So far as I understand it, the buffer is only guaranteed to be properly aligned for characters. There's a special exception when new'ing an array of character types. The results of new char[] must be sufficiently aligned for any type. (Never the less, I still prefer ::operator new(n) and ::operator delete(p) when dealing with raw memory.) -- James Kanze GABI Software 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 Eugene Alterman Guest Posted: Sat May 13, 2006 9:21 pm    Post subject: Re: Placement new and destructor "Eugene Alterman" <eugalt (AT) verizon (DOT) net> wrote in message news:hhJ8g.75476$H71.51854 (AT) newssvr13 (DOT) news.prodigy.com... Quote: "peter koch larsen" <peter.koch.larsen (AT) gmail (DOT) com> wrote in message news:1147299859.239506.136800 (AT) i40g2000cwc (DOT) googlegroups.com... I have a feeling there might be a third error, however. The character vector is allocated via new char[]. Is new char guaranteed to allocate memory that is properly aligned for any datatype? So far as I understand it, the buffer is only guaranteed to be properly aligned for characters. Storage for objects created using new-expression is allocated by an allocation function (global operator new[] in this case) that is generally the same regardless of the type of an object being created, and it is thus aligned to satisfy the strictest alignment requirement. (3.7.3.1/2) And this is not sufficient because there may be more space allocated with the array at some offset from the beginning of the allocated storage. But there is a special rule for char arrays that require them to be sufficiently aligned for any object that could be placed into such an array. (5.3.4/10) [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Tomás Guest Posted: Sat May 13, 2006 9:21 pm    Post subject: Re: Placement new and destructor kanze posted: Quote: There's a special exception when new'ing an array of character types. The results of new char[] must be sufficiently aligned for any type. (Never the less, I still prefer ::operator new(n) and ::operator delete(p) when dealing with raw memory.) If I'm dealing with raw memory, or allocating memory for an array of POD's for example, I like to use malloc -- I feel that it's more honest, so to speak. Malloc is always my first choice... it's only if I'm dealing with a fancy class type (or writing a template) that I resort to "new" and "delete". -Tomás [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Eugene Alterman Guest Posted: Sat May 13, 2006 9:21 pm    Post subject: Re: Placement new and destructor "kanze" <kanze@gabi-soft.fr> wrote in message news:1147437953.261536.79720 (AT) d71g2000cwd (DOT) googlegroups.com... Quote: peter koch larsen wrote: I have a feeling there might be a third error, however. The character vector is allocated via new char[]. Is new char guaranteed to allocate memory that is properly aligned for any datatype? So far as I understand it, the buffer is only guaranteed to be properly aligned for characters. There's a special exception when new'ing an array of character types. The results of new char[] must be sufficiently aligned for any type. That could fit into that array, that is. There would be no alignment requirements for new char[1], for example. (Never the less, I still prefer ::operator new(n) Quote: and ::operator delete(p) when dealing with raw memory.) That would make sense if new char[] has memory overhead. I wonder what could be the purpose of such an overhead? [ 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: Sun May 14, 2006 12:21 pm    Post subject: Re: Placement new and destructor In article <dXk9g.9116$j7.305540 (AT) news (DOT) indigo.ie>, Tomás <NULL (AT) NULL (DOT) NULL> writes Quote: kanze posted: There's a special exception when new'ing an array of character types. The results of new char[] must be sufficiently aligned for any type. (Never the less, I still prefer ::operator new(n) and ::operator delete(p) when dealing with raw memory.) If I'm dealing with raw memory, or allocating memory for an array of POD's for example, I like to use malloc -- I feel that it's more honest, so to speak. Malloc is always my first choice... it's only if I'm dealing with a fancy class type (or writing a template) that I resort to "new" and "delete". Then I think you are mistaken because operator new was designed to provide raw memory in C++. In many implementations the global version may just be a call to malloc() but it does not have to be. It is not a matter of honesty but a matter of making use of the language. If you use operator new you can benefit from replacement versions designed for such things as tracking memory leaks. -- 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 Tomás Guest Posted: Sun May 14, 2006 9:21 pm    Post subject: Re: Placement new and destructor Francis Glassborow posted: Quote: Then I think you are mistaken because operator new was designed to provide raw memory in C++. In many implementations the global version may just be a call to malloc() but it does not have to be. It is not a matter of honesty but a matter of making use of the language. If you use operator new you can benefit from replacement versions designed for such things as tracking memory leaks. I regularly play with references to get exactly what I want. For example, if I wanted an array of five thousand int's, I wouldn't write: int * const p = std::malloc( 5000 * sizeof(*p) ); I'd write: unsigned long const len = 5000; int (&array)[len] = *static_cast<int (*)[len]>( std::malloc( len * sizeof(*p) ) ); Then when I want to delete the memory, I can simply write: free(array); because I know that array will implicitly convert to "int*", which will implicitly convert to "void*" (which is what "free" takes as an argument). However, it gets messy if I want to use "delete", because then I have to make sure that I get the types EXACTLY right. Plus, I don't know how legal the following line is: int (&array)[len] = *static_cast<int (*)[len]>(new int[len]); I could use the following in its place: int (&array)[len] = *new int[len][0]; But then I would have to worry about passing a multi-dimensional pointer type to "delete"... (Actually it just occurred to me know that I could probably write the following:) int (&array)[len] = *new int[len][0]; delete &array; I'll have to check if that's okay... -Tomás -Tomás [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top kanze Guest Posted: Mon May 15, 2006 1:21 pm    Post subject: Re: Placement new and destructor Eugene Alterman wrote: Quote: "kanze" <kanze@gabi-soft.fr> wrote in message news:1147437953.261536.79720 (AT) d71g2000cwd (DOT) googlegroups.com... peter koch larsen wrote: I have a feeling there might be a third error, however. The character vector is allocated via new char[]. Is new char guaranteed to allocate memory that is properly aligned for any datatype? So far as I understand it, the buffer is only guaranteed to be properly aligned for characters. There's a special exception when new'ing an array of character types. The results of new char[] must be sufficiently aligned for any type. That could fit into that array, that is. There would be no alignment requirements for new char[1], for example. I think that the alignment requirement still stands (but the issue isn't important enough for me to bother looking it up). Quote: (Never the less, I still prefer ::operator new(n) and ::operator delete(p) when dealing with raw memory.) That would make sense if new char[] has memory overhead. I wonder what could be the purpose of such an overhead? Coherence with array new of types with a destructor? But that's not my motive. My motive is communicating with the reader: new X[n] means allocating and "constructing" n new objects of type X. The fact that the constructor is a no-op is in my mind irrelevant -- new char[n] tells the reader I'm allocating char's. The operator new() function tells the reader I'm allocating raw memory, which I will "construct" later. -- James Kanze GABI Software 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 James Kanze Guest Posted: Tue May 23, 2006 11:21 am    Post subject: Re: Placement new and destructor Eugene Alterman wrote: Quote: "kanze" <kanze@gabi-soft.fr> wrote in message news:1147684409.380016.226040 (AT) v46g2000cwv (DOT) googlegroups.com... Eugene Alterman wrote: "kanze" <kanze@gabi-soft.fr> wrote in message news:1147437953.261536.79720 (AT) d71g2000cwd (DOT) googlegroups.com... There's a special exception when new'ing an array of character types. The results of new char[] must be sufficiently aligned for any type. That could fit into that array, that is. There would be no alignment requirements for new char[1], for example. I think that the alignment requirement still stands (but the issue isn't important enough for me to bother looking it up). I did. :-) 5.3.4/10 quote For arrays of char and unsigned char, the difference between the result of the new-expression and the address returned by the allocation function shall be an integral multiple of the most stringent alignment requirement (3.9) of any object type whose size is no greater than the size of the array being created. /quote Does it mean that the above does not apply to signed char, BTW? That's what it says. If it were meant to apply to signed char as well, they would have written "For arrays of character type". (Or would that require it to apply to wchar_t as well?) Generally, I think that there is an intent in C++ (or parts of it), that char can be used for raw memory, as well as unsigned char. This is a change with respect to C (where char can explicitly have trapping values, so looking at an object as an array of char may not work). -- James Kanze kanze.james (AT) neuf (DOT) 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 Eugene Alterman Guest Posted: Wed May 24, 2006 4:21 pm    Post subject: Re: Placement new and destructor "James Kanze" <kanze.james (AT) neuf (DOT) fr> wrote in message news:e4mnhr$pq2$1 (AT) emma (DOT) aioe.org... Quote: Generally, I think that there is an intent in C++ (or parts of it), that char can be used for raw memory, as well as unsigned char. This is a change with respect to C (where char can explicitly have trapping values, What is that? Quote: so looking at an object as an array of char may not work). [ 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.