- Boost-users - lists.preview.boost.org

RE: [Boost-Users] shared_ptr and incomplete types
by bjorn.karlsson＠readsoft.com 01 May '02

01 May '02

> From: Jim Hyslop [mailto:jim.hyslop@leitch.com] > > I notice that the implementation available at boost.org does not allow > the single-argument ctor to be called on an incomplete type, e.g.: [snip] > The documentation on the single-argument ctor, though, does not state > that T must be a complete type. Is the implementation > stricter than the > documentation, or is the documentation incomplete? I believe that the implementation is correct, and that belief is confirmed by the latest documentation in the CVS - T must indeed be a complete type. Another change is the addition of a default constructor, which does not require T to be complete. > I'm writing an article on the smart pointers for the CUJ, and I'd like Cool! ;-) > to know if I can correctly state "The default behavior for > shared_ptr is > to require complete types. You can, however, manage an incomplete type > if you provide a function or function object, which will be > called when > the owned object is destroyed." Since shared_ptr requires a complete type only for a few operations, such as the constructor that you mentioned, and reset(T * p), I'd rephrase to something more explicit, like: "The instantiation of a shared_ptr typically requires a complete type. [etc]". Bjorn Karlsson [Non-text portions of this message have been removed]

2 1

shared_ptr and incomplete types
by Jim Hyslop 01 May '02

01 May '02

The documentation at http://www.boost.org/libs/smart_ptr/shared_ptr.htm states that "T may be void, but in that case, either an explicit delete function must be passed in, or the pointed-to object must have a trivial destructor." I notice that the implementation available at boost.org does not allow the single-argument ctor to be called on an incomplete type, e.g.: class simple; simple * createSimple(); shared_ptr<simple> simpleFactory() { return shared_ptr<simple>(createSimple()); } The documentation on the single-argument ctor, though, does not state that T must be a complete type. Is the implementation stricter than the documentation, or is the documentation incomplete? I'm writing an article on the smart pointers for the CUJ, and I'd like to know if I can correctly state "The default behavior for shared_ptr is to require complete types. You can, however, manage an incomplete type if you provide a function or function object, which will be called when the owned object is destroyed." -- Jim

3 2

Graph Library - AllExamples.dsw, anyone got it?
by loopylozzysdad 30 Apr '02

30 Apr '02

Hi All, Just read the excellent Boost Graph Library book. The AllExamples.dsw file is not on the accompanying CD. In the Errata web page (http://siek.info/bgl.html) it says... "On the accompanying CD, the file \BGL\VC++ Projects\AllExamples.dsw is missing. You can download the missing file from here BGLVCprojects.tgz." The file there is a .dsw file but it is NOT the missing one. Has anyone got the real thing? All the best and thanks - Richard Howells

1 0

Re: Boost.Thread (VC specific)
by Holger Thiele 30 Apr '02

30 Apr '02

>-----Ursprungliche Nachricht----- >Von: bill_kempf [mailto:williamkempf@hotmail.com] >Gesendet: Dienstag, 30. April 2002 17:54 >An: Boost-Users(a)yahoogroups.com >Betreff: [Boost-Users] Re: Boost.Thread (VC specific) > > >I don't agree. If it's a good idea, then I'd expect MS to do this by >default for IDE generated builds. Such as default warning level 3? >There's some overhead involved in >generating this, there's compile time overhead (which is important to >some) Do you have figures on that? In my experience, this overhead seems to be a small price to pay for the ability to track bugs (inside your debugger!) on released code. >and there's the fact that many users will be confused by the >presence of debug information and why "stepping" through the code >behaves so strangely and unpredictably. As always, if someone doesn't know what he's doing, there's little help besides getting educated. But, seriously, I think that most users that even get into debugging release builds would surely appreciate library code that supports this. Just my 2 cents. Cheers, and thanks for one of my favorite libraries Holger Thiele D.A.V.I.D. GmbH www.david-gmbh.de

1 0

RE: [Boost-Users] operator == in shared_ptr
by bjorn.karlsson＠readsoft.com 30 Apr '02

30 Apr '02

> From: sashan [mailto:sashang@ihug.co.nz] [snip] > Is there a better way that doesn't involve 1) having to write > a wrapper class around a basic type (or any type) and 2) > doesn't involve changing the shared_ptr::operator==. > As others have pointed out, the solution is to use a function object and std::find_if. In this case, the comparison doesn't even need a separate predicate - the (exceptionally cool) Boost.Bind library can be used instead: vector<shared_ptr<int> >::iterator it=std::find_if(vec.begin(),vec.end(),bind(std::equal_to<int>(),bind(&shared _ptr<int>::operator*, _1),42)); Bjorn Karlsson [Non-text portions of this message have been removed]

1 0

operator == in shared_ptr
by sashan 30 Apr '02

30 Apr '02

Hi Because the operator== function compares pointers and not what they're pointing to this is what I've had to do to get it to find an integer. Basically I had to wrap the basic int type within another class called Int, as show below. class Int { private: shared_ptr<int> p; public: Int(int i) {p.reset(new int); *p = i;} bool operator==(const Int& r) {return *p.get() == *r.p.get();} ~Int() {}; }; void main() { vector < Int > vec; vec.push_back(Int(1)); vec.push_back(Int(2)); vec.push_back(Int(3)); Int p(4); if (find(vec.begin(), vec.end(), p) == vec.end()) { cout << "Couldn't find it"; } } Is there a better way that doesn't involve 1) having to write a wrapper class around a basic type (or any type) and 2) doesn't involve changing the shared_ptr::operator==. [Non-text portions of this message have been removed]

3 2

boost::function::set()
by Douglas Gregor 29 Apr '02

29 Apr '02

Does anyone use the boost::function::set() methods? I'd like to remove them for Boost.Function v2 to cut down on the size of the interface. If you are using set() and would like to keep using it, please tell me. Doug

1 0

RE: [Boost-Users] iterating over a tuple
by Aleksey Gurtovoy 29 Apr '02

29 Apr '02

Toon Knapen wrote: > On Friday 26 April 2002 13:37, Aleksey Gurtovoy wrote: > > Toon Knapen wrote: > > > I have a list of heterogeneous objects, e.g. > [snip sample code] > > > > And now I want to have a function that can accumulate the > > > value sof all the objects in the tuple, e.g. (pseudocode since > > > my problem is exacly that I can't figure out how to code this) > > > > > > template < typename TypleType > > > > int accumulate_values(const TupleType& t) > > > { > > > int acc = 0; > > > for(int i = 0 ; i < t.size() ; ++i ) acc += (get< i >( > t )).value(); > > > } > > > > > > Can I do this using mpl (mpl looks great but I can't figure > > > out how to do this). > > > > It's not exactly a part of MPL (the library doesn't target > "run-time" stuff > > to that extent yet; I hope it will some day), but "tuple extension" > > components I wrote some time ago > > > (http://groups.yahoo.com/group/boost/files/tuple_ext/tuple_ext-08-nov-01.zip ) > allow one to do exactly these kinds of things: > [snip aleksey's solution] > > Thanks, I like tuple_ext a _lot_ ! I am glad you found it useful. > It would be great to have complete STL-like functionality like this. Yep. The only problem with this is that we are going to end up with 3 STL-like libraries within the same language :) (the original one ("run-time"), the MPL one (pure compile time), and something in between, - HSTL (Heterogeneous Sequences Template Library :)). All three with the same algorithms, just written in different notation :). Emily's heterogeneous lists (www.oonumerics.org/tmpw01/winch.pdf, http://groups.yahoo.com/group/boost/files/alist-feb14-02.zip) 'tuple_ext' components, and MPL's v.2 half-run-time 'for_each' are obviously hanging around that "in between" thing. > Plans on adding this tuple_ext to the CVS ? Some; I think it should be a full-fledged submission that would take the prior art and build a simple framework on it. It's not a large undertaking, though - most of the concepts and some code are already here :). Aleksey

1 0

BGL: "Supergraphs?"
by David A. Greene 29 Apr '02

29 Apr '02

Hi, I've been learning the BGL and porting some existing code over to it. First off, Kudos to Jeremy, Lie-Quan and all the BGL developers. This library is great! I find the filtered_graph potentially very useful for my application except for one case. In some (not rare) instances I need to be able to create a filtered_graph that spans two graphs. In other words, I have two separate graphs. Most of the time the information I need is entirely contained in one graph so a filtered_graph works perfectly. However, sometimes I need to create a filtered_graph from several separate graphs and link them together in a well-defined manner (i.e. I easily know where the edges should go). I need a "supergraph" that is the combination of the two filtered graphs. Right now the application handles this problem by creating a new graph from the existing graphs. But this takes time and memory. I'd like to get rid of this overhead if possible, especially since these are often short-lived graphs. The filtered_graph objects would not be induced subgraphs so a subgraph object wouldn't work (subgraph also doesn't support something that spans multiple graphs anyway). What I need is the opposite of subgraph -- a unified interface to multiple filtered_graphs. This seems very hard to do because each filtered_graph will have its own vertex and edge index mappings. An extra level of indirection would solve the problem but I'm not sure the overhead would be less than simply creating a new graph. Thoughts? Opinions? Thanks again for a great library! -Dave -- "Some little people have music in them, but Fats, he was all music, and you know how big he was." -- James P. Johnson

3 6

Newbie: problems installing under RedHat 7.2
by Benjamin Han 28 Apr '02

28 Apr '02

I'm just trying to install Boost library under RedHat 7.2. I was following the build instructions found on the Boost website, but it seemed to fail for several targets. I'm wondering if it's because RH7.2 uses its version of gcc 2.96. Since my project work needs gcc 3.0+, my question is, what do I need to do in order to use gcc3/g++3 instead of gcc/g++? The build system seems exotic and I just want to get over with the installation asap and start to use the library. (does changing /usr/bin/gcc and /usr/bin/g++ links work?) Here is the build command I tried: jam -sBOOST_ROOT=. -sTOOLS="gcc" Errors (selected): ... libs/thread/build/../src/once.cpp: In function `void boost::call_once (void (*) (), once_flag &)': boost/thread/once.hpp:38: previous declaration of `void boost::call_once (void (*) (), once_flag &)' with C++ linkage libs/thread/build/../src/once.cpp:77: conflicts with new declaration ... MkDir1 libs/python/build/bin/libboost_python_static/gcc/debug/runtime-link-dynam ic/shared-linkable-true gcc-C++-action libs/python/build/bin/libboost_python_static/gcc/debug/runtime-li nk-dynamic/shared-linkable-true/types.o In file included from boost/python/errors.hpp:13, from boost/python/detail/call_object.hpp:8, from libs/python/build/../src/types.cpp:11: boost/python/detail/wrap_python.hpp:24:24: patchlevel.h: No such file or directo ry boost/python/detail/wrap_python.hpp:100:20: Python.h: No such file or directory ... (lots of Python-related errors) ... gcc-C++-action status/bin/is_function_test/gcc/debug/runtime-link-dynamic/is_fun ction_test.o boost/type_traits/function_traits.hpp: In instantiation of `boost::is_function<i nt[]>': status/../libs/type_traits/tests/is_function_test.cpp:68: instantiated from he re boost/type_traits/function_traits.hpp:148: Internal error: Segmentation fault. Please submit a full bug report. See <URL:http://bugzilla.redhat.com/bugzilla/> for instructions. ... &boost::noncopyable::operator= (const boost::noncopyable &)' is private status/../libs/utility/noncopyable_test.cpp:35: within this context ...failed updating 43 targets... ...skipped 24 targets... ...updated 629 targets... The folllowing RPMs are installed on the machine: 1. gcc-2.96-110 2. gcc3-3.1-0.23.1 3. python-1.5.2-38 4. python2-2.2-16 5. glibc-2.2.5-32 6. boost-build-1.0-1 7. boost-jam-3.0-1 Other info: kernel 2.4.18 (manually built) Any comment would be greatly appreciated, Ben

3 3