April 2009 - Boost - lists.boost.org

Shorter link to getting started
by Vladimir Prus 08 Apr '09

08 Apr '09

I would like to modify the configure script to mention getting started, so that user knows where to look for information, but unfortunately, the current URL is a bit long: http://www.boost.org/doc/libs/1_38_0/more/getting_started/unix-variants.html It is 77 characters, so it does not leave any space for any other text. Can the website administrators arrange for some shorter redirect to be available? Say http://www.boost.org/getting_started/unix-variants.html ? Thanks, Volodya

4 3

[data_binding] Lets bind data
by Kasra 08 Apr '09

08 Apr '09

Hi, I am looking for interest in data_binding techniques in C++. The idea is to have a class that could bind to a list of data and when ever any of them is changed the value of the binder is changed too: data_bindee<mpz_class> x; data_bindee<mpz_class> y; typedef tuple<mpz_class,mpz_class> my_bind_tuple_type; struct my_bind_pred_type { template<typename BinderT> void operator(BinderT& me, my_bind_tuple_typeT t) { me.x = get<0>(t); me.y = 2 * get<0>(t); } }; data_binder<struct { mpz_class x,y; }, my_bind_tuple_type, my_bind_pred_type> point(x, y); // when ever data of x and y is changed the value of point // is updated too. Any suggestions?? Kasra With Best Regards -- Kasra Nassiri

1 0

Preliminary progress on Boost.Python py3k support
by Haoyu Bai 08 Apr '09

08 Apr '09

Hi, I did some preliminary work on BPL py3k support in these days. With Troy's help we have a py3k SVN tree in sandbox now, which is branched from trunk. I also tried to compile BPL with Python 3, and by following the error report of the compiler, I did some patch - now we have 14 .cpp files out of 28 in total can be compiled. The patch is committed, you can see it here: [2]. Please point me out if the code has something don't conform to Boost coding policy. There's also something interesting, maybe a bug I caught, for example: (Search the changeset to see details) 39 // XXX(bhy) Potentional memory leak here since PyObject_GetAttrString returns a new reference 40 // const char *mod = PyString_AsString(PyObject_GetAttrString( self_, const_cast<char*>("__module__"))); 41 PyObject *mod = PyObject_GetAttrString( self_, "__module__"); Also I found some potential problem with Py_ssize_t, I will take this as a chance to fix them. The biggest problem I see still is the "Text Vs. Data Instead Of Unicode Vs. 8-bit" in py3k. As we discussed and the suggestion by Niall [3], we have a clear solution for this and I will work towarding this. The other changes in py3k such as PyObject_HEAD change, int type removal and unbound method removal also affected Boost.Python's code, but they don't require logical change of BPL code and easy to handle. Hopefully in the end of this summer, we will have a Boost.Python library with Python 3 support! Thanks again for many people's help! References: [1] https://svn.boost.org/trac/boost/browser/sandbox-branches [2] https://svn.boost.org/trac/boost/changeset/52118 [3] “Nabble - Python - c++-sig - Some thoughts on py3k support,” http://www.nabble.com/Some-thoughts-on-py3k-support-td22564813.html. -- Haoyu Bai

2 1

[interprocess] Assign to shared_lock from scoped_lock
by Chard 08 Apr '09

08 Apr '09

If I have (template args removed for brevity): shared_lock sh_lock(mutex); scoped_lock sc_lock(move(sh_lock), try_to_lock); // Try for exclusive if (sc_lock) { ... sh_lock = move(sc_lock); // Drop back to shared *** not available } Should that assignment operator be available? Currently I'm working around this with: sh_lock.swap(shared_lock(move(sc_lock))); [or, alternatively: sh_lock = move(shared_lock(move(sc_lock))); ]

2 1

Re: [boost] [Any] Reference counting
by Michael Howell 08 Apr '09

08 Apr '09

Steven Watanabe <watanabesj(a)gmail.com> wrote: > Have you read http://www.gotw.ca/gotw/045.htm? I haven't read it, and have looked into atomic numbers, but couldn't find the "Boost Way" to do it (except perhaps poking around with shared_ptr innards). I wanted to use intrusive counts, since we have control over the object we are sharing. Thorsten Ottosen <thorsten.ottosen(a)dezide.com> wrote: >If you want to improve the performance of any, the thing to do is to add > a small buffer optimization such that all buildin types (or equally > small UDT) are created on the stack. I'll look into that. Mathias Gaunard <mathias.gaunard(a)ens-lyon.org> wrote: > Also, any is a value type. We do not want it to have reference > semantics. So you'd need to use COW. Actually, copy on write/implicit sharing is what I was proposing, if you had looked at the code. > Hence this is fairly useless, especially since we're getting move > semantics soon. "Soon" is relative. The trunk GCC may support it instantly (may already), but Debian Stable won't, nor will MSVC (both which will be used for a good while). Also, you usually copy the value when you store it, which may or may not involve writing. For example: #include <boost/any.hpp> using namespace boost; struct example { any variable; }; int main() { example object; example object2; any value = "I Love Boost!"; object.variable = value; object2.variable = value; return 0; } Thanks for replying, everyone. -- Michael Howell mhowell123(a)gmail.com

4 5

"[gsoc] Thread scheduler support for boost" [Re: for vicente.botet]
by Indradip Ghosh 08 Apr '09

08 Apr '09

Hi Vicente, Sorry for having missed your question earlier on my intended GSOC proposal. I actually intended to design the fine-grained scheduler for threadpool class itself. My detailed application is listed here in the googlesites: http://sites.google.com/site/threadpoolschedulergsoc2009/ Please let me know if you have any questions/ comments regarding my proposal. Thanks and Regards, Indradip. > Message: 7 > Date: Mon, 30 Mar 2009 08:36:36 +0200 > From: "vicente.botet" <vicente.botet(a)wanadoo.fr> > Subject: Re: [boost] "[gsoc] Thread scheduler support for boost" > To: <boost(a)lists.boost.org> > Message-ID: <C1B76DB53B1544F9B2C5618DB7A708EF@viboes1> > Content-Type: text/plain; charset="iso-8859-1" > > Hi, > > I'm wondering this shouldn't be considered as an extension of the Boost.Thread library more that the Boost.ThreadPool library. The ThreadPool library can be improved once these extension are in place, but also other application or libraries using threads. > > Let me know if I'm missing someething? > > Best, > Vicente

1 0

Performance measurement
by Govindan, Nandagopal (GE Healthcare) 07 Apr '09

07 Apr '09

Hi, I would like to know how performance measurement( time & memory ) done in boost. Do boost has some following APIs to measure performance?? Profiler.start() Code to be measured Profiler.stop() Printstatistics() Regards Nandagopal

3 3

[gsoc] improving any and variant
by Mathias Gaunard 07 Apr '09

07 Apr '09

A few little ideas that could eventually be joined to be a single GSoC project: - Allow visitation for Any. This can actually be generalized to any object that exposes a type_info, such as polymorphic objects. Techniques, including non-portable ones based on name mangling, could be experimented to get more efficient dispatch. - Add SBO to any. - Improve variant so that the functor doesn't require to expose a result_type. The result_type can be automatically deduced by "joining" all the possible return types obtained by boost::result_of. - Allow the easy construction of a visitor from lambda expressions, with a syntax like variant<int, std::string> a = "foo"; int b = match(a).with( args<int>(_1), args<std::string>(size(_1)) ); instead of the potentially more verbose struct my_visitor : static_visitor<int> { int operator()(int a) { return a; } int operator()(const std::string& b) { return b.size(); } }; variant<int, std::string> a = "foo";; int b = apply_visitor(my_visitor(), a);

5 5

Intrusive list/ilist and rbtree marshalling support
by Brian Wood 07 Apr '09

07 Apr '09

I'm happy to announce that version 1.8 of the C++ Middleware Writer is available on line -- http://webEbenezer.net/cgi-bin/samb.cgi<http://webebenezer.net/cgi-bin/samb.cgi>. This version adds support for the following: * Exclusive marshalling support of boost::intrusive::list and boost::intrusive::rbtree. Neither Boost Intrusive nor Boost Serialization offer serialization support of the Intrusive containers. Rather than having two types called 'list,' we refer to boost::intrusive::list as an ilist. An example of this support is here -- http://webEbenezer.net/rbtree_marshalling.html<http://webebenezer.net/rbtree_marshalling.html>. * Several new options on the user interface including: + An "Error Handling" option that indicates whether the code should be written using return codes or exceptions. The exception support is new in this release. This option affects the signatures of the generated marshalling functions. + A "Code Structure" option that indicates whether the code should be written in header-only format or separated interface and implementation files. The separated interface and implementation support is new in this release. This option also affects the signatures of the generated marshalling functions. + A choice between unsigned short or unsigned int as the type for constants written by the C++ Middleware Writer. Constants that identify messages and user defined types are written by the software. Previously the constants were always unsigned ints. * A mechanism to turn off automatic code generation for a type. Adding a // comment within a class declaration with "hand_written_marshalling_code" in the text of the comment will disable default code generation for that type. For example, class bag { // hand_written_marshalling_code ... }; As the comment indicates, you will provide the implementation of the marshalling functions for this class. This option is not available with /* */ comments. Additionally, I would like to say that I find the Intrusive library to be sound and am using it internally in the C++ Middleware Writer. Thanks to the authors: Olaf Krzikalla and Ion Gaztañaga. Brian Wood Ebenezer Enterprises www.webEbenezer.net <http://www.webebenezer.net/> "Then Samuel took a rock and set it up between Mizpah and Shen. He named it Ebenezer [Rock of Help] and said, 'Until now the L-RD has helped us.'"

1 0

[program_options] multitoken arguments
by Carlo Medas 06 Apr '09

06 Apr '09

Dears, we've been using boost::program_options and it works great. We just found a minor thing, which could be an issue, and I'm reporting it to you: when specifying and using a multitoken argument, it actually works only if put as last parameter of command line; in any other case, it eats all remaining parameters. My specification is as follows: desc.add_options() ("output,o", bpo::value<std::vector<std::string> >()->multitoken(), "use output setup for discovery/scan"); If then I want to use it like follows, it works: > fing -p1 -p2 -p3 -o 1 2 3 But if the user inverts the order, it does not work anymore, meaning that p3 is eaten by -o option. > fing -p1 -p2 -o 1 2 3 -p3 Is this a known limitation? Or maybe my specification above is not correct? Thanks in advance, Carlo Medas

2 4