COMPGROUPS.NET | Search | Post Question | Groups | Stream | About | Register

Getting to grips with boost::shared_ptr

  • Email
  • Follow


Hi there,

I have a class called Sounds which creates a number of Sound objects
in its constructor and adds them to a local std:map variable. I also
define an enum in this class which I use as the key with respect to
the map. This class therefore acts as a look-up for all the sounds I
wish to use.

I also have a Player class which has a Play(Sound&) function. This
function stops the currently playing sound and plays the new sound
supplied by the argument. I also have a Stop() function which stops
the current sound. I use a Sound* local variable to keep track of the
current sound.

The question is, how do I know that I shouldn't be using a shared_ptr
instead of this pointer? - a design change which would also require
storing shared_pts in the map previously mentioned. For example, I can
crash my program by deleting my Sounds object before calling
Player.Stop(). Using shared_ptrs would avoid this crash. However, this
is a reckless way of producing a crash, are there more innocent ways
in which it could occur for my Player class?

For me, it feels that my Player class does not take ownership of a
sound object; it simply keeps track of it and therefore I use a simple
pointer. Is this the correct mentality with regard to whether I should
be using a shared_ptr or a simple pointer?

Thanks very much for your help,

Barry
0
Reply Barry 3/24/2010 6:32:59 PM

See related articles to this posting

* Barry:
> Hi there,
> 
> I have a class called Sounds which creates a number of Sound objects
> in its constructor and adds them to a local std:map variable. I also
> define an enum in this class which I use as the key with respect to
> the map. This class therefore acts as a look-up for all the sounds I
> wish to use.
> 
> I also have a Player class which has a Play(Sound&) function. This
> function stops the currently playing sound and plays the new sound
> supplied by the argument. I also have a Stop() function which stops
> the current sound. I use a Sound* local variable to keep track of the
> current sound.
> 
> The question is, how do I know that I shouldn't be using a shared_ptr
> instead of this pointer? - a design change which would also require
> storing shared_pts in the map previously mentioned. For example, I can
> crash my program by deleting my Sounds object before calling
> Player.Stop(). Using shared_ptrs would avoid this crash. However, this
> is a reckless way of producing a crash, are there more innocent ways
> in which it could occur for my Player class?
> 
> For me, it feels that my Player class does not take ownership of a
> sound object; it simply keeps track of it and therefore I use a simple
> pointer. Is this the correct mentality with regard to whether I should
> be using a shared_ptr or a simple pointer?

It's OK.

Consider that the same problems that you envision exist with nearly all kinds of 
objects in C++. To avoid the /possibility/ of dangling pointers every object 
would have to be allocated dynamically and managed through smart pointers and/or 
garbage collection. If that is what one wants then some other language, e.g. 
Java or C#, is a more practical answer  --  it's not the C++ way.


Cheers,

- Alf
0
Reply Alf 3/25/2010 9:36:48 AM 

Barry <bg_ie@yahoo.com> wrote in news:efac0ea9-d335-42da-8b92-
cf773ad3718e@q23g2000yqd.googlegroups.com:

> Hi there,
> 
> I have a class called Sounds which creates a number of Sound objects
> in its constructor and adds them to a local std:map variable. I also
> define an enum in this class which I use as the key with respect to
> the map. This class therefore acts as a look-up for all the sounds I
> wish to use.
> 
> I also have a Player class which has a Play(Sound&) function. This
> function stops the currently playing sound and plays the new sound
> supplied by the argument. I also have a Stop() function which stops
> the current sound. I use a Sound* local variable to keep track of the
> current sound.
> 
> The question is, how do I know that I shouldn't be using a shared_ptr
> instead of this pointer? - a design change which would also require
> storing shared_pts in the map previously mentioned. For example, I can
> crash my program by deleting my Sounds object before calling
> Player.Stop(). Using shared_ptrs would avoid this crash. However, this
> is a reckless way of producing a crash, are there more innocent ways
> in which it could occur for my Player class?
> 
> For me, it feels that my Player class does not take ownership of a
> sound object; it simply keeps track of it and therefore I use a simple
> pointer. Is this the correct mentality with regard to whether I should
> be using a shared_ptr or a simple pointer?

Yes, the only thing shared_ptr does is to keep the object alive until it is 
not needed any more. If you are not deleting any Sound anyway before the 
end of the program, then you do not need shared_ptr. OTOH if you foresee 
that you might want sometimes to delete a Sound in the middle of the 
program, for example by replacing it with an updated version loaded from 
disk, then shared_ptr would probably be the simplest way to achieve this 
feature.

cheers
Paavo
0
Reply Paavo 3/26/2010 7:54:30 AM
comp.lang.c++ 49706 articles. 25 followers.

2 Replies
150 Views

Similar Articles

12/9/2013 4:02:42 PM
[PageSpeed]

Reply:

Similar Artilces:

boost
hi all, anyone out there, who uses boost with vxworks 5.5 ann/or vxworks 6.4! We use a Pentium plattform, but I think it should not be plattform dependent ... Please let me know your opinions ... best regards Heiko Elger Heiko Elger wrote: > hi all, > > anyone out there, who uses boost with vxworks 5.5 ann/or vxworks 6.4! > > We use a Pentium plattform, but I think it should not be plattform > dependent ... > > Please let me know your opinions ... > > best regards > > Heiko Elger > I have used small pieces of it. You'll be limited to those features that come only in the header files - fortunately, that's most of boost. Someone will have to do considerable work to get the pre-built boost libraries to run under VxWorks. I also only use the Gnu cross compiler, 2.96, which had the best template support for its day, but now lags substantially behind the standard. You'll find that if you have to use this compiler, you'll have some frustration getting the right compiler settings in boost, and some of the features will either require the boost workarounds or just flat won't compile. Bottom line

Best practises for boost:signal
What is the best practice for using boost::signal? Should the signal be a public field? Should an accessor method be used? Should there be some special connection methods for every signal, like the following: private: signal<void()> x; public: void connectX(const signal<void()>::slot_type& slot); void disconnectX(const signal<void()>::slot_type& slot); If you use boost:signal, how do you do it? (Okay, actually connectX should return a connection object and disconnectX should take the it as its parameter, but it's about the general idea) -- Memfis Memfis wrote: > What is the best practice for using boost::signal? Should the signal be > a public field? Should an accessor method be used? Should there be some > special connection methods for every signal, like the following: > > private: > signal<void()> x; > public: > void connectX(const signal<void()>::slot_type& slot); > void disconnectX(const signal<void()>::slot_type& slot); > > If you use boost:signal, how do you do it? > > (Okay, actually connectX should return a connection object and > disconnectX should take

a specialization in boost::result_of
Hi, I was looking at boost::result_of, came across these template specializations in namspace 'detail'. template<typename F, typename FArgs, bool HasResultType> struct get_result_of; // This one is for backward compatibility with existing functors template<typename F, typename FArgs> struct get_result_of<F, FArgs, true> { typedef typename F::result_type type; }; //This one calls the the template function result of the function object template<typename F, typename FArgs> struct get_result_of<F, FArgs, false> { typedef typename F::template result<...; #include <boost/utility/result_of.hpp> struct square { template<typename T> struct result { typedef T type; }; template<typename T> T operator()(T& x) const { return x * x; } }; struct simple { typedef float result_type; }; typedef double (*FUNC)(char); typedef int (simple::* MEMFUNC) (int, double); int main() { std::cout << typeid(boost::result_of<FUNC()>::type).name() << std::endl; //1 std::cout << typeid(boost::result_of<MEMFUNC()>::type).name() << std::endl;//2 std::cout << typeid

error in usin list with boost
i have created a "cpp" file which has the followin code and when i try to compile it..it does not know what "list" is..it says its undefined..i also tried using #include <boost/python/list.hpp> but no differe...can someone please help. thanks static list retrieval_as_list(SplitText &self, int position, int retrieveLength) { list result; int *characters = new int[retrieveLength]; int lenRet = self.RetrieveUTF32(position, characters, retrieveLength); for (int i=0;i<lenRet;i... function and not on the heap, unless boost does some mumbo jumbo that I kno= w=20 nothing of like starting the list with refcount=3D2? I don't think that's w= hat=20 you want... But for your boost usage, I have absolutely no idea! Why not write the Pyth= on=20 wrapper directly in C? Just as easy, probably. And from what my experience = is=20 with using SWIG and SIP: much less errorprone... =2D-=20 =2D-- Heiko. see you at: http://www.stud.mh-hannover.de/~hwundram/wordpress/ --nextPart1297845.xF4j8lmosl Content-Type: application/pgp-signature -----BEGIN PGP SIGNATURE----- Version

Job for Boost filtered iterators ???
NMergedChannelEvents = ???? // 3 int NMergedAIlEvents = ????; // 4 I have done these but the code lacks elegance. For 3, I think I can use Boost's filtered iterators where the predicate indicates whether the current iterator can be merged with the existing state (that is if the Boost's filtered iterators allow state). But 4 seems impossible. It seems I want to combine Boost's filtered iterators with C++0x's any_of() algorithm, and count how many times any_of() returns true. DOing this within the scope of C++ 2003 standard seems not possible. For 3, I have

problem Boost libraries
{ Mod note: I'm accepting this article because the issues raised here, essentially the quality of specific parts of the Boost library, seem to be of general interest to the C++ community. -mod/aps } A few months ago, in comp.lang.c++.moderated, Wil Evers pointed out some problems in a few Boost libraries (below). Since we are considering using all of the libraries mentioned (and also Boost.Threads and Boost.uBlas), I wonder if these problems have been fixed. I've looked at Boost-developers archives, where this message was forwarded. It seems that only the protability of Boost.Test lead to some discussion. BTW, what's a header race? The order of #inclusion or linking affecting program behavior? ========================================================== Wil Evers wrote: >> [...] For some of the Boost libraries, we've actually >> decided >> to phase out their use at the first available opportunity. These >> are: >> filesystem >> program_options >> serialization >> test [...] I'll be specific (all targeted at the official boost-1.33.1 release) : 1) filesystem: - Major problems interfacing

are the boost ptr containers exception safe?
boost::ptr_vector<animal> vec; vec.push_back( new animal ); What does happen if push_back() fails -- do I have a leak? Peter wrote: > boost::ptr_vector<animal> vec; > vec.push_back( new animal ); > > What does happen if push_back() fails -- do I have a leak? Did you know that Boost has its own discussion forums? "Peter" wrote: > boost::ptr_vector<animal> vec; > vec.push_back( new animal ); > > What does happen if push_back() fails -- do I have a leak? > No leak. All Boost libraries are designed exception safe (modulo bugs). /Pavel On Tue, 12 Dec 2006 19:19:48 +0100, "Pavel Vozenilek" wrote: >"Peter" wrote: > >> boost::ptr_vector<animal> vec; >> vec.push_back( new animal ); >> >> What does happen if push_back() fails -- do I have a leak? > >No leak. > >All Boost libraries are designed exception safe (modulo bugs). To be more precise, wrt to exception-safety (as defined in http://www.boost.org/more/generic_exception_safety.html) the function provides the basic guarantee but not the strong guarantee. Best regards, Roland Pibinger

GDB Script for Containers in Boost Library
Hi All , Require GDB Script for Containers in Boost Library in C++. Kindly provide me pointers for them. Thanking in Advance. Thanks Pallav singh

How to detect NULL for boost:shared_ptr
Hi, I am playing with boost pointer and try to wrap the following codes A* func(){ ... if(condition 1 ){ return a; } else return NULL; } Now I wrap a as shared_ptr(new A()) and change funcion as share_ptr<A> func(); The problem is that shared_ptr will not accept NULL. How should I do it in this case? Thanks Chris tradevol@yahoo.com wrote: > I am playing with boost pointer and try to wrap the following codes > > A* func(){ > ... > if(condition 1 ){ > return a; > } > else > return NULL; > > } > > Now I wrap a as shared_ptr(new A()) and change funcion as share_ptr<A> > func(); > The problem is that shared_ptr will not accept NULL. What does it mean that it "will not accept" it? Do you get a compiler error? Do you get a run-time exception? If the latter, catch it. If the former, deal with it (see FAQ 5.8). > > How should I do it in this case? Read the FAQ. V -- Please remove capital 'A's when replying by e-mail I do not respond to top-posted replies, please don't ask tradevol@yahoo.com schrieb: > Hi, > > I am playing with boost

JFileChooser potential speed boost for XP users
one potential solution for XP users who are finding JFileChooser slow to load... [Taken from another forum regarding stopping XP viewing contents of ZIP files] 1) Run regsvr32 /u %windir%\system32\zipfldr.dll 2) restart machine 3) open winzip 4) respond �yes' to question about making full association this decreased my 60 second time to open a folder with 3 small text files down to sub-second time.

Privacy Policy | All Times Are GMT(UTC)