- Boost-users - lists.boost.org

serialization auto_ptr std::string
by Olivier Destrebecq 25 Apr '06

25 Apr '06

I'm including serialization in my library and am running into problems. One of my class include an std::auto_ptr <std::string> I'm including #include <boost/serialization/string.hpp> to handle the string serialization i downloaded the serialization example for an auto_ptr and modified it to allow for XML serialization (see end of email). But when i compile get the following error saying: error: 'struct std::basic_string<char, std::char_traits<char>, std::allocator<char> >' has no member named 'serialize' If i replace the std::auto_ptr <std::string> with just std::string it works, so i assume something is wrong with my auto_ptr code Feel free to ask me more question or send me beginners advice as i'm just beginning with BOOST. Thanks auto_ptrSerialization.hpp: #include <list> #include <memory> #include <fstream> #include <boost/config.hpp> // std::autoptr inteface wrong in dinkumware #include <boost/detail/workaround.hpp> #include <boost/archive/xml_iarchive.hpp> #include <boost/archive/xml_oarchive.hpp> #include <boost/detail/workaround.hpp> #include <boost/serialization/split_free.hpp> namespace boost { namespace serialization { ///////////////////////////////////////////////////////////// // implement serialization for auto_ptr<T> // note: this must be added to the boost namespace in order to // be called by the library template<class Archive, class T> inline void save( Archive & ar, const std::auto_ptr<T> &t, const unsigned int file_version ){ // only the raw pointer has to be saved // the ref count is rebuilt automatically on load T *objectPtr = t.get(); ar << BOOST_SERIALIZATION_NVP(objectPtr); } template<class Archive, class T> inline void load( Archive & ar, std::auto_ptr<T> &t, const unsigned int file_version ){ T *objectPtr; ar >> BOOST_SERIALIZATION_NVP(objectPtr); // note that the reset automagically maintains the reference count #if BOOST_WORKAROUND(BOOST_DINKUMWARE_STDLIB, == 1) t.release(); t = std::auto_ptr<T>(objectPtr); #else t.reset(objectPtr); #endif } // split non-intrusive serialization function member into separate // non intrusive save/load member functions template<class Archive, class T> inline void serialize( Archive & ar, std::auto_ptr<T> &t, const unsigned int file_version ){ split_free(ar, t, file_version); } } // namespace serialization } // namespace boost Best, Olivier Destrebecq olivierd(a)real.com

2 2

Re: [Boost-users] BGL: filtered graph question
by Grégoire Dooms 25 Apr '06

25 Apr '06

boost-users-request(a)lists.boost.org wrote: > Date: Tue, 25 Apr 2006 08:54:12 +0200 > From: Sebastian Weber <sebastian.weber(a)physik.tu-darmstadt.de> > Subject: [Boost-users] BGL: filtered graph question > > Hello everybody! > > I use the filtered_graph framework to restrict a BFS visit to a small > region on my graph. BUT, I provide the starting vertex to the BFS > routine by a vertice_descriptor of the underlying/unfiltered graph. Of > course, this vertice descriptor matches the predicate. My code compiles > fine, but I started to ask myself if this actually behaves correct? > > As I could not find a hint within the docs, does someone know? > > A quick look at the code tells that it is perfectly fine. The type of vertex_descriptor is the same as for the underlying graph and the vertex iterator is simply a filter_iterator using the provided predicate. Best, -- Grégoire Dooms

3 4

Boost::program_optiions - Defining The Config File As A Cmdline Parameter
by Adam Hartshorne 25 Apr '06

25 Apr '06

Hi All, I want to be able to do the following using the Boost::program_options library. 1) Parse a set of command line options in a standard way (I know how to achieve this). 2) Have an command line option which is "get other command line options from a config file defined as the argument to this option. The example multiple_sources is very close to the sort of thing I want, however in that example the config file is hard coded. I want the user to interactively set the config file at runtime. If I try simply putting the following lines in the set of if statements my program crashes (I know this is still hard coding the cfg file, but it was a first attempt before I made it more general), if ((vm.count("input-file")) { ifstream ifs("multiple_sources.cfg"); store(parse_config_file(ifs, config_file_options), vm); notify(vm); } Any help would be much appreciated, Adam

3 3

Boost::program_options - Problem Compiling Examples On Linux Box
by Adam Teasdale Hartshorne 25 Apr '06

25 Apr '06

Hi All, I am having some trouble compiling the examples for the program_options library on a linux box. I following is the output of my attempt to try and compile the first.cpp example. I obviously understand that error message is due to the compiler been unable to see the appropriate function in the given libraries. I therefore assume I should be including so other libraries, but I don't know which ones are needed. Interestingly I can get the simple_ls.cpp example to compile and function ok. Any help would be much appreciated, Adam g++ -c -pipe -Wall -W -O2 -pipe -m64 -fexceptions -D_REENTRANT -DQT_NO_DEBUG -DQT_THREAD_SUPPORT -DQT_SHARED -I/usr/lib64/qt-3.3/mkspecs/default -I. -Idcs/improc2/adam/boost_1_33_1/ -I/usr/lib64/qt-3.3/include -I.moc/ -o .obj/first.o first.cpp g++ -o cmdLineTest .obj/first.o -L/usr/lib64/qt-3.3/lib -L/usr/X11R6/lib64 -L/dcs/improc2/adam/boost_1_33_1/stage/lib/ -lboost_filesystem-gcc-d-1_33_1 -lboost_program_options-gcc-1_33_1 -lqt-mt -lXext -lX11 -lm -lpthread .obj/first.o(.text+0x57): In function `main': : undefined reference to `boost::program_options::options_description::options_description(std::basic_string<char, std::char_traits<char>, std::allocator<char> > const&)' .obj/first.o(.gnu.linkonce.t._ZNK5boost15program_options25basic_command_line_parserIcE3runEv+0x5): In function `boost::program_options::basic_command_line_parser<char>::run() const': : undefined reference to `boost::program_options::common_command_line_parser::run() const' .obj/first.o(.gnu.linkonce.t._ZN5boost15program_options25basic_command_line_parserIcEC1EiPPc+0x42): In function `boost::program_options::basic_command_line_parser<char>::basic_command_line_parser(int, char**)': : undefined reference to `boost::program_options::common_command_line_parser::common_command_line_parser(std::vector<std::basic_string<char, std::char_traits<char>, std::allocator<char> >, std::allocator<std::basic_string<char, std::char_traits<char>, std::allocator<char> > > > const&)' collect2: ld returned 1 exit status make: *** [cmdLineTest] Error 1

2 2

program_options: iterate variable_map
by Stephan Stiglmayr 25 Apr '06

25 Apr '06

Hello! I'm looking for a way to "iterate" through a variable_map after the command line options were stored, in order to print out all the options, or all with newly values. I know that you can test if a specific options was read with the "count()" method, but this is cumbersome. Is there a possiblity to do that? Regards, Stephan.

4 8

Re: [Boost-users] Problem with iterator, map, shared_ptr and temp lated class
by James Hughes 25 Apr '06

25 Apr '06

Spot on! Thanks for that. reading my templates book has given me a good idea of why, so hopefully I wont make the same mistake again! James > -----Original Message----- > From: Joaquín Mª López Muñoz [mailto:joaquin@tid.es] > Sent: 25 April 2006 15:17 > To: boost-users(a)lists.boost.org > Subject: Re: [Boost-users] Problem with iterator, map, shared_ptr and > templated class > > > > James Hughes ha escrito: > > > Hello all, > > > > I am a bit stumped by a compile fault.. > > > > This is a fragment of the code > > [...] > > > for (handlerMap::iterator pos = LogMap.begin(); pos > != LogMap.end(); > > I think this has to be > > for (typename handlerMap::iterator pos = > LogMap.begin(); pos != > LogMap.end(); > > HTH, > > Joaquín M López Muñoz > Telefónica, Investigación y Desarrollo > > > _______________________________________________ > Boost-users mailing list > Boost-users(a)lists.boost.org > http://lists.boost.org/mailman/listinfo.cgi/boost-users >

1 0

Why Linux May Never Be a True Desktop OS
by Andreas Sæbjørnsen 25 Apr '06

25 Apr '06

http://www.technewsworld.com/rsstory/50091.html Why Linux May Never Be a True Desktop OS E-Mail Article<http://www.technewsworld.com/perl/mailit/?id=50091> Print Version <http://www.technewsworld.com/rsstory/50091.html#> [image: Why Linux May Never Be a True Desktop OS] By Rob Enderle TechNewsWorld 04/24/06 5:00 AM PT With Linux, the customer often expects to get the product for free and wants the retail price of Windows deducted from his/her purchase price. There are no funds passed back to the vendor and, because Linux is different, customers tend to place more service calls -- at $85 a call. As a result, the vendor generally ends up losing money. *The Formula for Total Customer Experience* Nothing is more powerful than seeing real users interact with your e-Commerce application, Web site, or software. Use Morae, a single, collaborative software solution, to identify barriers to conversion, while watching real users navigate your site. Free White Paper!<http://www.technewsworld.com/rsstory/50091.html> This week I'm speaking at the Desktop Linux Summit in San Diego on why the PC OEMs don't, and probably never will, fully support Linux on the desktop. This is somewhat of a deja vu for me since a decade ago my team made a similar presentation at IBM <http://www.ibm.com/> (NYSE: IBM) [image: Latest News about IBM]<http://www.technewsworld.com/perl/search.pl?query=IBM&scope=network>on OS/2 and why it wouldn't beat Windows in its time. For the purpose of this piece I'm using the word "Linux" as an open source OS catch-all so I don't have to call out every distribution or variant. Now, as it was then, the core of the argument wasn't the technology, the strength of Microsoft <http://www.microsoft.com/> (Nasdaq: MSFT) [image: Latest News about Microsoft]<http://www.technewsworld.com/perl/search.pl?query=Microsoft&scope=network>, or anything, really, outside IBM's control. The reason was that IBM would simply not do what was necessary to take the market away from Microsoft. When I look at Microsoft these days I recall something former Hewlett-Packard <http://www.hp.com/> (NYSE: HPQ) [image: Latest News about Hewlett-Packard]<http://www.technewsworld.com/perl/search.pl?query=Hewlett-Packard&scope=net…>Chairman Lou Platt once said: "My biggest problem is finding people who will tell me the truth." I've never seen Microsoft as weak, from an execution standpoint, before, but currently the firm appears weak. Taking advantage of that weakness remains elusive for both Linux players and Apple<http://www.apple.com/>(Nasdaq: AAPL) [image: Latest News about Apple]<http://www.technewsworld.com/perl/search.pl?query=Apple&scope=network>, however. Let's take a look at the reasons why it's been difficult for them to move dramatically into the business or consumer desktop computer spaces. Classifieds Find IT Consultants<http://www.ectnews.com/adsys/link/?creative=3338&ENN_target=http://links.in…> Post Your Project for Free. Get Bids from Thousands of Pre-Screened Consultants. ------------------------------ Laptop DC power jacks parts & service<http://www.ectnews.com/adsys/link/?creative=3338&ENN_target=http://links.in…> Your laptop doesn't power up? You don't have to replace the whole motherboard, you can just replace the part. We stock DC power s... ------------------------------ Computer, PDA & iPod Accessories<http://www.ectnews.com/adsys/link/?creative=3338&ENN_target=http://links.in…> Dell, IBM, Sony, Apple, HP, Compaq, Toshiba, Palm, iPod, and more. Factory-direct Prices, Same-day Shipping, 30-day Money-back Gu... Doing What Is Necessary When it comes to competition, typically you have the folks who are willing to do what is necessary to win and the folks who aren't. In my personal experience, I've seen again and again scenarios where a team would lay out a plan for what they deemed was needed in order to be successful in a market and then they'd watch as the executive in charge cherry picked the things he/she wanted to do. As a result, the project failed, often disastrously. People like to put filters over their eyes and ignore important steps they don't want to perform on the road to achieving a goal. All of us have seen examples of when this derails the plan. With OS/2, the last major challenger to Windows, IBM had a list of things it needed to do in order for the product to be successful. This list laid out staffing levels, marketing expense, structural changes to the company and partnerships, and its foundation was one of the most extensive pieces of research that company ever did. However as these recommendations went up the chain of command they were changed or in some cases, ignored. The end result? OS/2 failed. The requirements for OS/2's success included broad third party hardware<http://www.technewsworld.com/rsstory/50091.html#> [image: Powerful Yet Simple: HP ProLiant ML110 G3. Just $688 with the Intel(r) Pentium(r) 4 Processor.] <http://www.technewsworld.com/rsstory/50091.html> and software [image: operating system]<http://www.technewsworld.com/rsstory/50091.html>support, developer support, and Windows 32-bit support. In order to get into the hardware, however, OS/2 couldn't come from IBM. This was probably the single biggest deal breaker. Compaq, HP and others wouldn't support an OS from a competitor and, much like it was with Sun and Java<http://java.sun.com/> [image: Latest News about Java]<http://www.technewsworld.com/perl/search.pl?query=Java&scope=network>and Netscape and Navigator, the biggest weakness that OS/2 had was IBM not doing what the market required of it to be successful. After reading up <http://ask.slashdot.org/article.pl?sid=06/04/18/122246> on a subject that speaks to a likely OEM desktop requirement -- strong vendor supported drivers -- I deeply doubt whether the folks working on Linux will make the difficult tradeoffs needed to assure success. Think of the iPod, for example. In order to achieve success Apple had to support Windows, as distasteful as that must have been for the firm. The OEM Cost of Linux and Windows Windows is free to the OEMs. In fact, not only is it free, but Microsoft<http://www.technewsworld.com/rsstory/50091.html#>, in effect, pays them to take it. Regardless of the cost, Windows is a logical choice, and a straight pass. Dell <http://www.dell.com/> (Nasdaq: DELL) [image: Latest News about Dell]<http://www.technewsworld.com/perl/search.pl?query=Dell&scope=network>pays about $80 for it and typically charges about $80 for it. There is rarely much of a mark up. If Microsoft were to lower its price that lowered price would be reflected in virtually all desktop hardware immediately. Microsoft provides a number of services which include development support, service support, marketing support, technicians, classes, databases and support materials, and it picks up a lot of the service load as well. In addition, it provides marketing co-op dollars, incentives for early adoption of new products, and ensures a somewhat level playing field (which could be good or bad) for the vendors. Linux, on the other hand, is often viewed as "free." Even when it comes with a fee, that fee is also passed through without any mark up. However, the hardware vendor must now provide all of the things that Microsoft has been providing, too. Particularly painful are services, as customers expect a similar experience with open source that they have with Windows and will often call the vendor to help troubleshoot. The vendor takes all of the related costs; there are typically no dependable (on-site) resources, unless the vendor supplies some of its own. With Linux, the customer often expects to get the product for free and wants the retail price of Windows deducted from his/her purchase price. There are no funds passed back to the vendor and, because Linux is different, customers tend to place more service calls -- at $85 a call. As a result, the vendor generally ends up losing money on average with Linux. This is the primary reason vendors don't want to do desktop Linux, unless there is an extremely lucrative services contract tied to it. Given services revenue often flows to the services division, the desktop hardware divisions still do their best to throw their bodies in front of these efforts. Too Many Versions An old rule of the desktop is that support costs go up by the square of the number of different products you have to support. This may actually understate the problem now that much of this support is done in remote sites like India. For each version of a product the support team has to be trained, support scripts have to be created and updated; patches have to be collected, tested and deployed; and line changes have to be tested against every operating system version shipped. This is part of the reason that IBM hardware was not competitive in the early nineties; the extra cost of supporting OS/2 was theirs and theirs alone while IBM software got the largest portion of OS/2 revenue (that division was profitable, though OS/2 wasn't). There is broad diversity among Linux buyers, and Red Hat<http://www.redhat.com/>(Nasdaq: RHAT) [image: Latest News about Red Hat]<http://www.technewsworld.com/perl/search.pl?query=Red%20Hat&scope=network>, which is the closest to a standard Linux OS, isn't yet dominant enough to claim victory. The OEMs might be able to handle one version of Linux but until there is only one they have to handle they will probably avoid doing it. Excessive Advocacy A few years ago I recall getting a note from an Apple guy who said something to the effect of, "I'm sure glad all of the crazy Apple supporters moved over to Linux." While I seriously doubt all of the crazy Apple zealots moved to Linux, I can say that they were instrumental in getting Apple out of large enterprises. Zealots, regardless of cause, have a tendency to treat very poorly those who don't share their beliefs, and, historically, they have been much more likely to misuse e-mail [image: Integrate BlackBerry with your CRM Solution]<http://www.technewsworld.com/rsstory/50091.html>or take a dispute to a public forum. They tend to be human resource nightmares and anyone wanting to create a powerful argument against either platform has only to take the decision maker to any of a number of fan Web sites to create the impression of significant risk. This can be particularly problematic now with employee blogs starting to spread and more internal disagreements being made public. Management typically does not want to create an environment where company dirty laundry is aired or where employees organize to fight executive orders. Yet zealots appear to do this as a matter of course: their beliefs come first, even though this is technology, not religion. With all this in mind, IT executives want to be able to decide to abandon a Linux or Apple trial with faith that the employees would follow those orders and not revolt. They also want to feel in control with regard to how the technology is used and not cede that to an external or internal employee advocacy group without conscious thought. In conclusion, there are a few key requirements for alternative platforms to be successful, success being measured by significant market share, in the general market. A Reliable Road Map to Linux Desktop Success The list of clear requirements includes: - Clear OEM resources to match or exceed those currently provided by Microsoft. This would include R&D support and co-marketing dollars. - Embrace existing desktop requirements (roadmaps, application support, proprietary drivers, consistent patch releases schedules and documentation, clear escalation lines for support). - Seek out reasonable advocates who will take direction from their executives and not from other advocates. - Make it profitable for the OEM. Recall that for Microsoft it took about 5 years, US$10 billion and Y2K to move the majority of the installed PC base to another kernel -- and they had to kill the old one off to do it. In the end, I look at this short list and still doubt it is reasonably achievable. I also wonder if Linux were actually modified to be able to comply with this list whether it would even be Linux anymore. That deserves a little thought. Generally speaking, you have to give something up in order to get something else. However there are times when what you sacrifice is actually more important than what you gain.

1 1

Using bind/lambda to extract graph bundled properties from overloaded operator[]
by Sean Kelly 25 Apr '06

25 Apr '06

Hi All, I'm attempting to use bind to extract bundled properties from a adjacency_list graph via the syntax below. ToyGraph g; ToyGraph::vertex_descriptor v; bind(&ToyGraph::operator[], ref(g), _1)(v); this fails to compile (I assume because opertor[] is overloaded) bind(&ToyGraph::swap, ref(g), _1)(g); works fine -- any suggestions on how to make the former work are appreciated. thanks Sean Windows XP Professional SP1, Boost 1_32_0

1 0

Problem with iterator, map, shared_ptr and templated class
by James Hughes 25 Apr '06

25 Apr '06

Hello all, I am a bit stumped by a compile fault.. This is a fragment of the code template<typename T> void LCLogSource<T>::SetLoggers(std::map<T, boost::shared_ptr<LCLogHandler<T> > > &LogMap) { typedef std::map<T, boost::shared_ptr<LCLogHandler<T> > > handlerMap ; for (handlerMap::iterator pos = LogMap.begin(); pos != LogMap.end(); ++pos) // **** compile error here { yada yada I get a compile fault on the for line, 'expected ; before pos'. Now this fault has only started happening since I decided to template this class. It appears that the combination of the template, and the complexity of the map declaration is causing things to go awry, and I don't know why. I have tried a few alternatives but always get a problem of some description. Is there some reason I cannot use iterators on a stl::map where the map key defn. depends on the template parameter? In case this is nothing to do with the shared_ptr in the map, does anyone have a list where templates programming is discussed so I can post there? James

2 1

General Boost documentation
by John Krajewski 25 Apr '06

25 Apr '06

I'm new to generic programming and finding the various concepts used in Boost quite confusing (named parameters, property maps, etc). Is there general documentation for the programming methods used in Boost? Perhaps a good book to buy?

4 4