- Boost-users - lists.boost.org

fixing graph labels
by Harry Waye 17 Apr '07

17 Apr '07

Hello, Have created a graph using erdos_renyi_generator and the vertices are labelled 1 to n. When I remove a vertex the labels are then 1 to n-1. Is there any way to stop this relabelling from happening? Heres some info: typedef adjacency_list<setS, vecS, undirectedS> Graph; typedef erdos_renyi_iterator<minstd_rand, Graph> ERGen; Graph g(ERGen(gen, n, p), ERGen(), n); Cheers, Harry

2 1

Thread safety and weak_ptr questions
by Dizzy 17 Apr '07

17 Apr '07

Hello struct A {} // executes before spawning any boost threads shared_ptr<A> sp(new A()); weak_ptr<A> wp(sp); // then in thread1 sp.reset(); // in thread thread2 shared_ptr<A> sp2(wp.lock()); Is the above code thread safe ? Is sp2 always either a default initilized shared_ptr<A> or a valid shared_ptr to the instance of A initially referenced by sp ? I couldn't find documentation on weak_ptr thread safety issues just on shared_ptr. I am unable to determine from the documentation what is the intended behaivour for the above code. I could look into the boost code and/or test it but I want to know not how it works but how it's intended to work (portably and all) :) Thank you. -- Mihai RUSU Email: dizzy(a)roedu.net GPG : http://dizzy.roedu.net/dizzy-gpg.txt WWW: http://dizzy.roedu.net "Linux is obsolete" -- AST

2 1

Re: [Boost-users] Boost Regex in a DLL
by Metasebia Moges 17 Apr '07

17 Apr '07

Dear Sir, First of all, thank you for your reply. Your replies for my requests were very helpful. >You can use Boost.Regex in a dll by defining BOOST_REGEX_DYN_LINK when >building your dll, but as long as the regex code is isolated to your dll and >doesn't spill over into it's public interface then this shouldn't be >necessary. Other than that it's hard to say what the issue might be without >seeing your code. As you said, the regex code is isolated in my dll and doesn't spill over into the public interface. Then the problem could be with my code below. Would you please inspect and correct me if I am wrong somewhere? I read a phrase from a file and search for words preceded by "W=" and if the word is different from !NULL, sil, and SIL, I store them in a list called PhraseList for further processing. boost::regex re("W=\\w+"); boost::regex sre("!NULL|sil|SIL"); string s, sbuf; string::const_iterator begin, end; size_t length; sbuf = phrase; begin = sbuf.begin(); end = sbuf.end(); boost::match_results<std::string::const_iterator>matches; boost::match_flag_type flags = boost::match_default; while(boost::regex_search(begin, end, matches, re, flags)) { if(boost::regex_search(phrase, re)) { s= matches.str(); length=matches.str().size(); string item =s.substr(2,length-1); if(boost::regex_match(item,sre)) break; else{ phraseList.addString (item.c_str()); break; }//else } //if }//while I have the following code which does the job without any problem at the expense of "not using boost". I have a line read from a file: // Parse the line len=strlen(phrase); //If the phrase is a content bearing term, put it into phrase list while(i<len){ if(phrase[i] != 'W') i++; else{ i=i+2; item=&phrase[i]; if (strncmp (item,"!NULL ",6) == 0|| strncmp(item,"sil ",4)==0 || strncmp(item,"SIL ",4)==0) break; else { memcpy(phrase, item, strlen(item)+1); phraseList.addString (phrase); break; } } } I don't like the second option in spite of the fact that it works. Moreover, I want to use boost.regex for other complex tasks and I want to test it with this simple real life problem. Your input is highly appreciated! Thank you in advance! Best regards, mmoges _______________________________________________ Boost-users mailing list Boost-users(a)lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users __________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com

2 1

regex_iterator with for_each question...
by Graham Reitz 17 Apr '07

17 Apr '07

I thought that for_each() goes while (it != end). How does for_each() 'know' what end is here? It doesn't seem to be initialized to anything. Thanks much, Graham class regex_callback { int m_sum; public: regex_callback() : m_sum(0) {} template <typename T> void operator()(const T& what) { m_sum += boost::lexical_cast<int>(what[1].str().c_str()); } int sum() const { return m_sum; } }; int main() { boost::regex("(\\d+),?"); string s = "1,1,2,3,5,8,13,21"; boost::sregex_iterator it(s.begin(), s.end(), reg); // Is end initialized to anything here? boost::sregex_iterator end; regex_callback c; // How is end valid here? int sum = for_each(it, end, c).sum(); return 0; }

3 2

[Serialization] Closing tag in XML description missing
by Ruediger Berlich 17 Apr '07

17 Apr '07

Hi there, I've got the problem that the XML description of my serialized classes seem to be missing the final closing tag. Funnily enough loading of the XML descriptions nevertheless works. The platform is OpenSUSE 10.2, g++ 4.1.2, Boost 1.33.1 . I'll try to illustrate the problem with a small program. Output further below. /************************************************************************/ // headers + namespace statements removed to save space class test { friend class boost::serialization::access; template<class Archive> void serialize(Archive & ar, const unsigned int version){ using boost::serialization::make_nvp; ar & make_nvp("x",x); } int x; public: test(int i){ x = i; } void printX(void){ cout << "x = " << x << endl; } }; main() { test Test(1); // I want a stringstream, as my applicion is in remote processing ostringstream oss; assert(oss.good()); boost::archive::xml_oarchive oa(oss); oa << BOOST_SERIALIZATION_NVP(Test); // Closing tag is missing in output, see below cout << oss.str() << flush; // But the file has the correct content, see below ofstream ofs("print.xml"); assert(ofs.good()); boost::archive::xml_oarchive oa2(ofs); oa2 << BOOST_SERIALIZATION_NVP(Test); // Try to load the incorrect XML description works !?!?? test Test2(2); istringstream iss(oss.str()); assert(iss.good()); boost::archive::xml_iarchive ia(iss); ia >> BOOST_SERIALIZATION_NVP(Test2); // Funnily enough this seems to work Test2.printX(); // correctly prints "x = 1", not "x = 2" } /************************************************************************/ So when I give the xml_oarchive an ostringstream as argument (because my application is in remote processing and I don't need file output), the result is: /************************************************************************/ <?xml version="1.0" encoding="UTF-8" standalone="yes" ?> <!DOCTYPE boost_serialization> <boost_serialization signature="serialization::archive" version="3"> <Test class_id="0" tracking_level="0" version="0"> <x>1</x> </Test> /************************************************************************/ So here the closing tag </boost_serialization> is missing. If I do the same with an ofstream (see program) the file content is o.k. . /************************************************************************/ <?xml version="1.0" encoding="UTF-8" standalone="yes" ?> <!DOCTYPE boost_serialization> <boost_serialization signature="serialization::archive" version="3"> <Test class_id="0" tracking_level="0" version="0"> <x>1</x> </Test> </boost_serialization> /************************************************************************/ Funnily enough my problem with the ostringstream does not seem to have an influence on the loading of the XML description. What am I doing wrong ? Thanks and Best Regards, Ruediger

2 2

Failed gcc link and targets
by Douglas Linford 17 Apr '07

17 Apr '07

2 Davids, The output at the top of the compile after running: bjam "-sTOOLS=gcc" install > out.txt 2>&1 shows python errors followed by: ...found 13313 targets... ...updating 250 targets... And then throughout the compile there are a lot of link errors and FileClone operation not permitted errors. The the compile completes with a lot of skipped errors ending with: ....failed updating 110 targets... ...skipped 118 targets... ...updated 22 targets... So...the following list from the Boost site gave the order for compiling which I completed in order with my appended notes: 1. Download Boost <http://boost.org/more/getting_started.html#step1>. 2. Install Boost.Jam<http://boost.org/more/getting_started.html#step2>. ( I installed bjam to: /usr/local/bin/bjam) 3. Configure your compiler toolset<http://boost.org/more/getting_started.html#step3>. (This I did not do...I didn't know what they wanted me to do) 4. Go to Boost distribution directory<http://boost.org/more/getting_started.html#step4> . 5. Build and install<http://boost.org/more/getting_started.html#step5> . Could some of this be the result of the fact that I didn't configure my compiler? Thanks guys... Douglas

3 4

[ptr_container]Visual C++ 2005 visualizers for ptr_container classes and boost::array
by Bill Buklis 16 Apr '07

16 Apr '07

Since I work with the ptr_container classes (especially ptr_vector and ptr_map) quite a bit (very nice library!), I found it a little frustrating attempting to debug container and iterator variables for these classes as compared with how well the std containers show up in the VC++ watch window. After some thought and experimentation I figured out how to add these classes to the VC++ visualizer so that now these objects will display almost as well as the std containers. For those of you unfamiliar with the built-in visualizer (no surprise as there is no official documentation for it), there is a text file called autoexp.dat that contains definitions on how to display custom objects in the watch window (including CString, std::string, and the standard containers). I found this very helpful blog article that explains this in more detail: http://www.virtualdub.org/blog/pivot/entry.php?id=120. The autoexp.dat file is usually located in C:\Program Files\Microsoft Visual Studio 8\Common7\Packages\Debugger. Place the visualizer code (included down below) at the end of the "visualizer" section of autoexp.dat (probably near the end of the file). Some notes: 1. ptr_vector expands outs the same way as std::vector<T*> would rather than std::vector<T>. I did it this way so you can easily see the memory address of each contained object. 2. ptr_map unfortunately doesn't expand out exactly the same as std::map<K,T*>. This is because ptr_map uses std::map<K,void*> internally and not map<K,T*>. I couldn't figure out any way to auto convert the sub-level appropriately. As a result the expanded entries only show the value objects and not KEY and VALUE (e.g. where VALUE is the "second" member). As a compromise I display the KEY elements in the preview section so that you can still see them. The preview section can be seen in two places: as a tooltip when you highlight the variable and on the first line next to the variable in the watch window. If anyone can think of a workaround or a better alternative let me know. 3. Also included is an entry for boost::array, so that these can be seen in the watch window in the same manner as a static array. Any questions? Enjoy, -- Bill -- Add the following text to autoexp.dat: ;--------------------------------------------------------------------------- --- ; boost::ptr_vector ;--------------------------------------------------------------------------- --- boost::ptr_vector<*,*>{ children ( #array ( expr: (($T1**)($c.c_._Myfirst))[$i], size : $c.c_._Mylast-$c.c_._Myfirst ) ) preview ( #( "[", $e.c_._Mylast - $e.c_._Myfirst , "](", #array ( expr : (($T1**)($c.c_._Myfirst))[$i], size : $c.c_._Mylast-$c.c_._Myfirst ), ")" ) ) } ;--------------------------------------------------------------------------- ; boost::ptr_vector::iterator ;--------------------------------------------------------------------------- boost::void_ptr_iterator<std::_Vector_iterator<*>,*>{ preview ( #( ($T2*)(*$c.iter_._Myptr) ) ) children ( #( ptr: ($T2*)(*$c.iter_._Myptr) ) ) } ;--------------------------------------------------------------------------- ; boost::ptr_set::iterator ;--------------------------------------------------------------------------- boost::void_ptr_iterator<*,*>{ preview ( #( ($T2*) $e.iter_._Ptr->_Myval) ) children ( #(ptr: ($T2*) $e.iter_._Ptr->_Myval) ) } ;--------------------------------------------------------------------------- ; boost::ptr_map ;--------------------------------------------------------------------------- boost::ptr_map<*,*,*>{ preview ( #( "[", $e.c_._Mysize, "](", #tree ( head : $c.c_._Myhead->_Parent, skip : $c.c_._Myhead, size : $c.c_._Mysize, left : _Left, right : _Right ) : $e._Myval.first, ")" ) ) children ( #tree ( head : $c.c_._Myhead->_Parent, skip : $c.c_._Myhead, size : $c.c_._Mysize, left : _Left, right : _Right ) : ($T2*)$e._Myval.second ) } ;--------------------------------------------------------------------------- ; boost::ptr_map ;--------------------------------------------------------------------------- boost::ptr_multimap<*,*,*>{ preview ( #( "[", $e.c_._Mysize, "](", #tree ( head : $c.c_._Myhead->_Parent, skip : $c.c_._Myhead, size : $c.c_._Mysize, left : _Left, right : _Right ) : $e._Myval.first, ")" ) ) children ( #tree ( head : $c.c_._Myhead->_Parent, skip : $c.c_._Myhead, size : $c.c_._Mysize, left : _Left, right : _Right ) : ($T2*)$e._Myval.second ) } boost::ptr_map_iterator<*,*,*>{ preview ( #( "(", $c.iter_._Ptr->_Myval.first, ",", ($T3*) $c.iter_._Ptr->_Myval.second, ")" ) ) children ( #( key: $c.iter_._Ptr->_Myval.first, value: ($T3*) $c.iter_._Ptr->_Myval.second, _ptr: $c.iter_ ) ) } ;--------------------------------------------------------------------------- ; boost::ptr_set ;--------------------------------------------------------------------------- boost::ptr_set<*,*>{ preview ( #( "[", $e.c_._Mysize , "](", #tree ( head : $c.c_._Myhead->_Parent, skip : $c.c_._Myhead, left : _Left, right : _Right, size : $c.c_._Mysize ) : ($T1*)$e._Myval, ")" ) ) children ( #tree ( head : $c.c_._Myhead->_Parent, skip : $c.c_._Myhead, left : _Left, right : _Right, size : $c.c_._Mysize ) : ($T1*)$e._Myval ) } ;--------------------------------------------------------------------------- ; boost::ptr_set ;--------------------------------------------------------------------------- boost::ptr_multiset<*,*>{ preview ( #( "[", $e.c_._Mysize , "](", #tree ( head : $c.c_._Myhead->_Parent, skip : $c.c_._Myhead, left : _Left, right : _Right, size : $c.c_._Mysize ) : ($T1*)$e._Myval, ")" ) ) children ( #tree ( head : $c.c_._Myhead->_Parent, skip : $c.c_._Myhead, left : _Left, right : _Right, size : $c.c_._Mysize ) : ($T1*)$e._Myval ) } ;--------------------------------------------------------------------------- ; boost::array ;--------------------------------------------------------------------------- boost::array<*,*>{ preview( $c.elems ) children ( #array ( expr: $c.elems[$i], size : $T2 ) ) }

1 0

[ptr_container]Internal use of void* containers
by Bill Buklis 16 Apr '07

16 Apr '07

Internally, the ptr_container classes use a void* container to do its magic. For example, ptr_vector uses a std::vector<void*> (and similarly for ptr_map and ptr_set) Or come to think of it, I believe it's actually vector<void**> and map<KEY,void*>, respectively. I'm curious as to why? This, of course, works just fine and since it's an internal detail it doesn't really matter. I'm just curious as to why you didn't use the more obvious (and perhaps wrong) choice of vector<T*>. Note: I'm not suggesting that it be changed. Functionally I have no issues with it, naturally, as it works. But, it does make debugging a little more interesting. But, even that is easy to get around. Just curious, -- Bill --

2 1

dereference an iterator
by Christian Holmquist 16 Apr '07

16 Apr '07

Sorry for beeing off-topic, but is there a utility in boost for dereferencing an iterator, such as the below (but that also works for boost::reference_wrapper) template<class Iterator> typename std::iterator_traits<Iterator>::reference deref(Iterator i) { return *i; } Thanks, Christian

4 4

Deep copy of objects coupled with shared and weak pointer
by Fabian Sturm 16 Apr '07

16 Apr '07

Hi! I used the boost shared pointers for quite a while successfully but now ran into a problem I don't have a nice solution yet. I have a class A which has a vector of B classes. I now need to get from the class A to all Bs which is easy (iterating over the vector) and I need to get from any B to A (which I handle with a weak pointer) But now I need to implement a copy method to get a "real" copy of A and not a pointer to the shared instance. How would I do that? The copy also points to copies of the Bees aka. deep copy. My current solution looks like this: class A; class B; typedef boost::shared_ptr< A > A_SP; typedef boost::weak_ptr< A > A_WP; typedef boost::shared_ptr< B > B_SP; class A : public boost::enable_shared_from_this< A > { public: int x; int y; std::vector< B_SP > bees; void addB ( B_SP b ); A_SP clone (); }; class B { public: A_WP weakA; int c; B_SP clone (); }; void A::addB ( B_SP b ) { b->weakA = shared_from_this(); bees.push_back ( b ); } A_SP A::clone () { A_SP newA ( new A ); newA->x = x; newA->y = y; for ( std::vector< B_SP >::const_iterator it = bees.begin (); it != bees.end (); ++it ) { newA->addB ( (*it)->clone () ); } return newA; } B_SP B::clone () { B_SP newB ( new B ); // note I can't initialize the weak_ptr here newB->c = c; return newB; } int main() { A_SP a ( new A ); B_SP b ( new B ); a->addB ( b ); // Now make a deep copy of a A_SP newA = a->clone(); return 0; } Any ideas how this could be done more beautiful? Or more C++ish? Every help is greatly appreciated, since I still need to get accustomed to C++. Thanks a lot, Fabian

1 0