C++Talk.NET C++ language newsgroups   Archives   FAQ   Search   Memberlist   Usergroups   Register   Profile   Log in to check your private messages   Log in  Is &* a good idea? Goto page 1, 2  Next          C++Talk.NET Forum Index -> C++ Language (Moderated) View previous topic :: View next topic   Author Message David Turner Guest Posted: Fri Jan 30, 2004 6:04 pm    Post subject: Is &* a good idea? Hi everyone I recently had reason to implement a proxy class, and got to thinking. In several places I have done something similar to this: std::auto_ptr<Foo> object = new Foo(); bar(&*object); where bar() takes a Foo*. It strikes me that &* is a very unnatural construct. I appreciate that having auto_ptr<Foo> automatically convertible to a Foo* might lead us into trouble with the "assignment is move" semantic, however, for non-copyable proxies, is automatic conversion really such a bad thing? On the pro side, it makes the code a little less ugly. On the con side, it might cause one to forget that one is not actually using an object of type Foo*... Any thoughts? Regards David Turner [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Thomas Tutone Guest Posted: Sat Jan 31, 2004 1:45 am    Post subject: Re: Is &* a good idea? "David Turner" <david (AT) firepro (DOT) co.za> wrote: Quote: Is &* a good idea? Generally, no. Quote: I recently had reason to implement a proxy class, and got to thinking. In several places I have done something similar to this: std::auto_ptr<Foo> object = new Foo(); bar(&*object); bar(object.get()); use the get() member function. That's what it's there for. Best regards, Tom __________________________________ Do you Yahoo!? Yahoo! SiteBuilder - Free web site building tool. Try it! http://webhosting.yahoo.com/ps/sb/ [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Ivan Vecerina Guest Posted: Sat Jan 31, 2004 1:46 am    Post subject: Re: Is &* a good idea? Hi, "David Turner" <david (AT) firepro (DOT) co.za> wrote Quote: I recently had reason to implement a proxy class, and got to thinking. In several places I have done something similar to this: std::auto_ptr<Foo> object = new Foo(); bar(&*object); where bar() takes a Foo*. It strikes me that &* is a very unnatural construct. I appreciate that having auto_ptr<Foo> automatically convertible to a Foo* might lead us into trouble with the "assignment is move" semantic, however, for non-copyable proxies, is automatic conversion really such a bad thing? It addition to having real dangers, implicit conversions can be confusing to ususpecting readers maintaining the code. Quote: On the pro side, it makes the code a little less ugly. On the con side, it might cause one to forget that one is not actually using an object of type Foo*... Exactly. Quote: Any thoughts? std::auto_ptr<> has a member function intended to do exactly what you want: .get() Everyone knows and understand what it means, and is able to grep for it. Furthermore: what if the auto_ptr instance contains a null pointer? &* would lead to undefined behavior -- never a good choice. .get() will return a null pointer, which might be what you want... So: bar( object.get() ); hth, Ivan -- http://ivan.vecerina.com/contact/?subject=NG_POST <- e-mail contact form [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Jonathan Turkanis Guest Posted: Sat Jan 31, 2004 1:47 am    Post subject: Re: Is &* a good idea? "David Turner" <david (AT) firepro (DOT) co.za> wrote Quote: Hi everyone I recently had reason to implement a proxy class, and got to thinking. In several places I have done something similar to this: std::auto_ptr<Foo> object = new Foo(); bar(&*object); where bar() takes a Foo*. It strikes me that &* is a very unnatural construct. <snip Definitely. The problem is that operator->() is a unary operator, but can't be called like a normal unary operator. The typical usage in this case would be bar(object.get()); (Also, you should write auto_ptr<Foo> object(new Foo); Implicit conversion doesn't work in the reverse direction, either.) Jonathan [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Philipp Bachmann Guest Posted: Sat Jan 31, 2004 1:48 am    Post subject: Re: Is &* a good idea? Hello David, Quote: std::auto_ptr<Foo> object = new Foo(); bar(&*object); You could also call "auto_ptr<>::get()": bar(object.get()); Quote: where bar() takes a Foo*. It strikes me that &* is a very unnatural construct. I appreciate that having auto_ptr<Foo> automatically convertible to a Foo* might lead us into trouble with the "assignment is move" semantic, however, for non-copyable proxies, is automatic conversion really such a bad thing? Making potentially expensive or dangerous things explicit is a good design principle of the standard library. Consider e.g. someone designs a factory method which returns not a pointer, but an auto_ptr<>: struct foo { virtual ~foo(void); virtual void a(void); }; class bar : public foo { [...] public: [...] }; std::auto_ptr< foo > createBar(void) { return new bar; } And consider, auto_ptr<> provided an automatic conversion to foo* and the user thought, "createBar()" actually returns foo* and therefore added a call to "delete", then the object created will be destructed a second time on leaving the block. With current "auto_ptr<>", your code won't even compile, which is better than a crash. Quote: On the pro side, it makes the code a little less ugly. On the con side, it might cause one to forget that one is not actually using an object of type Foo*... "auto_ptr<>" is not the only application for "&*" - I often come across the need to write "&*" with iterators. The confusion the expression "&*" may cause is due to the fact, that in C you could just erase it away, because "&" is just the inverse of "*" there. But with a language which allowes you to overload "operator*(void)" the programmers have to get used to "&*" in my opinion. Cheers, Philipp. [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Bob Bell Guest Posted: Sat Jan 31, 2004 1:55 am    Post subject: Re: Is &* a good idea? "David Turner" <david (AT) firepro (DOT) co.za> wrote Quote: Hi everyone I recently had reason to implement a proxy class, and got to thinking. In several places I have done something similar to this: std::auto_ptr<Foo> object = new Foo(); bar(&*object); where bar() takes a Foo*. It strikes me that &* is a very unnatural construct. I appreciate that having auto_ptr<Foo> automatically convertible to a Foo* might lead us into trouble with the "assignment is move" semantic, however, for non-copyable proxies, is automatic conversion really such a bad thing? On the pro side, it makes the code a little less ugly. On the con side, it might cause one to forget that one is not actually using an object of type Foo*... Any thoughts? bar(object.get()) reads better to me Bob Bell [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Sean Kelly Guest Posted: Sat Jan 31, 2004 10:43 am    Post subject: Re: Is &* a good idea? "David Turner" <david (AT) firepro (DOT) co.za> wrote Quote: the pro side, it makes the code a little less ugly. On the con side, it might cause one to forget that one is not actually using an object of type Foo*... Any thoughts? template<typename Ty> inline Ty* get_ptr( Ty* val ) { return val; } template<typename Ty> inline Ty const* get_ptr( std::auto_ptr<Ty> const& val ) { return val.get(); } Problem solved It's semantically meaningful without requiring changes to auto_ptr. Sean [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Daniel Anderson Guest Posted: Sat Jan 31, 2004 10:45 am    Post subject: Re: Is &* a good idea? "David Turner" <david (AT) firepro (DOT) co.za> wrote Quote: Hi everyone I recently had reason to implement a proxy class, and got to thinking. In several places I have done something similar to this: std::auto_ptr<Foo> object = new Foo(); bar(&*object); where bar() takes a Foo*. It strikes me that &* is a very unnatural construct. I appreciate that having auto_ptr<Foo> automatically convertible then do not use it, do: bar(object.get()); that's what I do Danderson [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Dave Harris Guest Posted: Sat Jan 31, 2004 6:25 pm    Post subject: Re: Is &* a good idea? [email]david (AT) firepro (DOT) co.za[/email] (David Turner) wrote (abridged): Quote: I recently had reason to implement a proxy class, and got to thinking. In several places I have done something similar to this: std::auto_ptr<Foo> object = new Foo(); bar(&*object); where bar() takes a Foo*. It strikes me that &* is a very unnatural construct. With auto_ptr<> you can use get(). I find it more of an issue with iterators. For example: void use( int *p ); void demo( vector<int>::iterator vi, list<int>::iterator li ) { use( &*vi ); use( &*li ); } It's not uncommon to have functions like use() which take a pointer so as to be neutral about where their data comes from (and please don't tell me to make use() a template). If the data actually comes from a standard container, I often have to convert an iterator into a pointer. &* is the best way I know to do it. -- Dave Harris, Nottingham, UK [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Peter Dimov Guest Posted: Sun Feb 01, 2004 3:26 pm    Post subject: Re: Is &* a good idea? [email]brangdon (AT) cix (DOT) co.uk[/email] (Dave Harris) wrote in message news:<memo.20040131112238.932D (AT) brangdon (DOT) m>... Quote: david (AT) firepro (DOT) co.za (David Turner) wrote (abridged): I recently had reason to implement a proxy class, and got to thinking. In several places I have done something similar to this: std::auto_ptr<Foo> object = new Foo(); bar(&*object); where bar() takes a Foo*. It strikes me that &* is a very unnatural construct. With auto_ptr<> you can use get(). I find it more of an issue with iterators. For example: void use( int *p ); void demo( vector<int>::iterator vi, list<int>::iterator li ) { use( &*vi ); use( &*li ); } It's not uncommon to have functions like use() which take a pointer so as to be neutral about where their data comes from (and please don't tell me to make use() a template). The plain &*vi is usually incorrect. You either have void use( int & r ); and pass it *vi, or you have void use( int * p ); and pass it vi == v.end()? 0: &*vi. [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top ta0kira Guest Posted: Sun Feb 01, 2004 8:09 pm    Post subject: Re: Is &* a good idea? "David Turner" <david (AT) firepro (DOT) co.za> wrote Quote: where bar() takes a Foo*. It strikes me that &* is a very unnatural construct. The only reason I can think of to use &* is in a function that you will change a pointer as well as return a value: size_t HalfArray(double &*aArray, size_t sSize) { aArray = new double[sSize / 2]; return sSize / 2; } As far as a conversion operator, you would have to make its return type &* for it to return &*, which I think would be a pretty bad idea, but if it is a conversion operator to just * then I think that is pretty safe. If you used a class with a conversion operator to * in a function that took &* I think it just creates a temp copy anyway. Not sure if all this is what you were talking about... ta0kira (Maybe using '&*const' would be safer???) [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Dave Harris Guest Posted: Mon Feb 02, 2004 5:49 pm    Post subject: Re: Is &* a good idea? [email]pdimov (AT) mmltd (DOT) net[/email] (Peter Dimov) wrote (abridged): Quote: The plain &*vi is usually incorrect. You either have void use( int & r ); and pass it *vi, You're assuming a house style about the use of references here. In some shops pointers are used even if they are not allowed to be NULL. Quote: or you have void use( int * p ); and pass it vi == v.end()? 0: &*vi. If you know the iterator is valid there's no need to test against v.end(), and indeed such a test may not be possible at the point of call. It wasn't in my example. (This is true even if use() does accept null pointers.) -- Dave Harris, Nottingham, UK [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top ta0kira Guest Posted: Tue Feb 10, 2004 8:07 pm    Post subject: Re: Is &* a good idea? Quote: With auto_ptr<> you can use get(). I find it more of an issue with iterators. For example: void use( int *p ); void demo( vector<int>::iterator vi, list<int>::iterator li ) { use( &*vi ); use( &*li ); } That makes a lot of sense, with the exception of the fact that the return type of 'operator *' is not required to be a reference: it only has to be of type 'std::iterator_traits <Iterator> ::reference' which in this case works out fine, but do not forget that a 'std::vector <bool> ::reference' is actually a 'std::_Bit_reference'. This means that you can only use '&*' with an iterator if you know for sure that the 'operator *' return type is an actual reference. Although I understand using a * instead of a reference, it might be better to do this (to allow for ALL iterator types): void demo( vector<int>::iterator vi, list<int>::iterator li ) { int Temp = *vi; use( &Temp ); *vi = Temp; Temp = *li; use( &Temp ); *li = Temp; } That makes for 4 additional copy operations (in the case that the assignment operator and copy constructor actually COPY), but it makes it more compatible with varying iterator types. After all, the purpose of an iterator is so that you don't have to have pointers and references as your public interface. Who knows, maybe the data manipulated by an iterator is not internally represented as a 'referenceable object' (i.e. it doesn't literally contain a pointer to an object of the expected type, such as 'std::vector <bool> ::iterator'; hence '::reference' being 'std::_Bit_reference' and not 'bool&'...) ta0kira [ 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: Fri Feb 13, 2004 1:52 am    Post subject: Re: Is &* a good idea? In article <b5c14083.0402100059.2b8cc8d (AT) posting (DOT) google.com>, ta0kira wrote: Quote: With auto_ptr<> you can use get(). I find it more of an issue with iterators. For example: void use( int *p ); void demo( vector<int>::iterator vi, list<int>::iterator li ) { use( &*vi ); use( &*li ); } That makes a lot of sense, with the exception of the fact that the return type of 'operator *' is not required to be a reference: it only has to be of type 'std::iterator_traits <Iterator> ::reference' Table 65 says the "reference" member of a container class has type "lvalue of T" and the way to represent lvalue-ness in the type system is to use reference types. Quote: which in this case works out fine, but do not forget that a 'std::vector <bool> ::reference' is actually a 'std::_Bit_reference'. snip Which means vector<bool> is not a container: <http://std.dkuug.dk/jtc1/sc22/wg21/docs/lwg-active.html#96>. -- Ben Hutchings Humour is the best antidote to reality. [ 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: Fri Feb 13, 2004 1:53 am    Post subject: Re: Is &* a good idea? ta0kira wrote: Quote: "David Turner" <david (AT) firepro (DOT) co.za> wrote in message news:<1075463184.630072 (AT) tanzanite (DOT) firepro.co.za>... where bar() takes a Foo*. It strikes me that &* is a very unnatural construct. The only reason I can think of to use &* is in a function that you will change a pointer as well as return a value: size_t HalfArray(double &*aArray, size_t sSize) snip T&* appears to be a pointer to a reference to T, but "[t]here shall be no...pointers to references." (8.3.2/4) You're talking about references to pointers, written as T*&. [ 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.