- Boost - lists.preview.boost.org

[container] stable_vector::back() is broken in Boost 1.53
by Klaim - Joël Lamotte 08 Apr '13

08 Apr '13

Tested on both VS2010 and VS2012, boost 1.53, with default boost no option set: ############ #include <boost/container/stable_vector.hpp> #include <iostream> int main() { boost::container::stable_vector< int > values; values.emplace_back( 42 ); int& bad_value = values.back(); // bad address int& good_value = values.front(); int& also_good_value = values.at(0); // imply values[0] std::cout<< " ADDRESS OF BAD VALUE : " << &bad_value << std::endl; std::cout<< " ADDRESS OF GOOD VALUE : " << &good_value << std::endl; std::cout<< " ADDRESS OF ALSO GOOD VALUE : " << &also_good_value << std::endl; } ############ This code displays addresses of the bad_value object very low on Debug and very high on Release. I believe that .back() is broken. However, when I test there : http://liveworkspace.org/code/3F11CK$2 it seems to work with any available compiler. So either it's LiveWorkspace which use an older boost version, or this issue arise only on VC. I made a ticket: https://svn.boost.org/trac/boost/ticket/8412 Joel Lamotte

6 10

GSoC 2013: Boost is a mentoring organization
by Boris Schaeling 08 Apr '13

08 Apr '13

Boost is a mentoring organization in this year's GSoC program. Thanks to everyone who added proposals to <https://svn.boost.org/trac/boost/wiki/SoC2013> and to Google for inviting us! According to the GSoC calendar at <https://google-melange.appspot.com/gsoc/events/google/gsoc2013> we have now two weeks where would-be student participants discuss application ideas. If you are interested in participating in the program - either as a student or as a mentor - get active now! :) 1) Have a look at our ideas list at <https://svn.boost.org/trac/boost/wiki/SoC2013>. Find something which interests you or think of a brand new project. 2) Send a mail to the mailing list at <mailto:boost@lists.boost.org> to let people know what you are interested in and what you have to offer. 3) Discuss your idea with others and try to turn it into an actual GSoC project. You need at least a project description, a goal, a mentor and a student. :) Boris

1 0

Adding some image libs to boost build
by Christian Henning 08 Apr '13

08 Apr '13

I have added jpeg, png, and tiff jam files to boost trunk. These files will be needed by gil::io tests. There are still some remaining issues which I like to discuss in this thread: 1. building these libs sometimes requires renaming header and/or source files to the correct architecture. Should we leave that up to the user or should bjam take care of it. And if so, how will we configure that. One example is jpeg: see jconfig.vc -> jconfig.h 2. Tiff has a file nmake.opt which is included in the makefile. How will we include all those options? Regards, Christian

3 5

Release milestone.
by Vicente J. Botet Escriba 08 Apr '13

08 Apr '13

Hi, I would like to merge some fixes in Boost.Thread, but I suspect that the release branche should be blocked as for https://svn.boost.org/trac/boost/wiki/ReleaseSchedule. On which stage (milestone) is the release process? Best, Vicente

2 2

[gil] numeric extension, arithmetic operators
by Harg Tholan 08 Apr '13

08 Apr '13

Hey, Do we plan to overload arithmetic operators for Pixel as wrappers of the numeric extension? or should users derive a class from Pixel adding such functionality? ATM, for diving two pixels (rgb_8 for example), I have to call the division this way: pix_r = pixel_divide_t<rgb_8_pixel_t, rgb_8_pixel_t, rgb_8_pixel_t>().operator()(pix2, pix1); The thing is that I have a template class that makes use of operator/ at some point on one of the template parameters, and I would like to be able to use it with 'Pixel' type. Thank you!

3 7

Re: [boost] future<>::then() Error returning future<void> (on trunk)
by Vicente J. Botet Escriba 07 Apr '13

07 Apr '13

Le 22/01/13 22:56, Vicente Botet a écrit : > Fernando Pelliccioni wrote >> Hi Vicente, hi all, >> >> I noticed that in the trunk there is a implementation of >> future<>::then(). >> Thank you very much for your efforts to implement it! > Hi Fernando, > > I have started with a restricted implementation. While trying to cover with > the whole requirements I didn't reach to get correct implementation. This is > way this new feature is not delivered yet. For the time been I don't think > the current implementation is ready for production code. > > >> The following example, compiles OK >> >> boost::future >> <int> >> f1 = boost::async([]() { return 123; }); >> f1.then([](boost::future >> <int> >> const& f) { >> return std::string(""); >> }); >> >> >> But this one fails to compile: >> boost::future >> <int> >> f1 = boost::async([]() { return 123; }); >> f1.then([](boost::future >> <int> >> const& f) { >> }); >> >> Here future<>::then() have to returns future >> <void> >> like ... >> >> boost::async([]() { }); >> >> Am I right? > Yes, it should. I have no access to my development environment for this > week. Could you post the compiler error? > > >> On the other hand, have you taken into account the following paper? >> http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2012/n3451.pdf > I read it and the mail exchanges related to this paper. > Currently I didn't reached to make it working at all when the future > destructor blocks, but I'm not sure this is the hidden problem as I have > also other issues when the future destructor doesn't blocks. > > I'm really sorry that the current state is not really usable :( > Hi Fernando, I have reworked the future::then function, and even if it is not yet complete (e.g. deferred is not implemented for c++98 compilers) it start to be usable for c++11 compilers, even if BOOST_THREAD_DONT_PROVIDE_FUTURE_INVALID_AFTER_GET is not defined. BTW, joining on the async future destructor works also. Please could you give it a try and let me know whatever you can find that doesn't works as you expect, Best, Vicente

2 3

C++11 and compilers not supported by vendors any more
by Antony Polukhin 06 Apr '13

06 Apr '13

Hi, I've got an almost ready patch for Boost.Any, that adds rvalues support to it via Boost.Move. But the trouble is following: Current version of patch may (and would) break compilation on ancient compilers, like MSVC6, BCC 5.5 and g++ 2.95. It can be tuned, to work even on those platforms, but: * I have no access to those antiquities, so it is impossible to run tests * It will take huge amount of time, to make it work What is the current policy in such situations (may ancient compilers support be dropped)? -- Best regards, Antony Polukhin

4 3

Re: [boost] Review managers wanted for augmented data structures
by James Hirschorn 06 Apr '13

06 Apr '13

Vadim Stadnik wrote > Hi all, > > > At present, neither Boost library nor STL provides containers using > augmented data structures. I suggest that the Boost community consider > solving the challenging problem – how to extend STL by integrating > advanced > data structures. > > ... > > Augmented trees are non-standard due to different performance guarantees. > At the same these data structures are beneficial for complex algorithms, > since they offer a wider set of more efficient operations compared to > standard containers. The typical performance improvement for one operation > is O(log N) against O(N). For less efficient operations these trees > normally provide running time O(log N) against O(1), which is good enough > for many applications. This is very similar to a project I have been working on, with the intention of submitting to boost. The new data structure I have implemented is called a skip list. They can be used in place of balanced trees, and according to the inventor, insertion and deletion are significantly faster than for balanced trees (I have not tested this yet). See, e.g. William Pugh. 1990. Skip lists: a probabilistic alternative to balanced trees. Commun. ACM 33, 6 (June 1990), 668-676. DOI=10.1145/78973.78977 http://doi.acm.org/10.1145/78973.78977 The difference in performance guarantees are much more subtle than you describe above. For example, skip lists have expected O(log n) insert, delete and search, versus O(log n). In practice, the expected O(log n) time is just as good as true O(log n), unless a performance bound is required with certainty, because the probability of the skip list having height much larger than log n is extremely low. I also implemented what I called augmented skip lists, which allow for expected O(log n) rank lookup (i.e. random access). I wasn't aware of the counter tree library, or else I would probably have just used it instead of implementing skip lists. I used skip lists to provide "drop in" STL replacements. Now I have skip_list::set and skip_list::multiset, and I could similarly make skip_list::map and skip_list::multimap. There is also augmented_skip_list::set and augmented_skip_list::multiset. For now the augmented versions have an additional member function nth_element providing E[O(log n)] random access. Hopefully, this will be helpful towards the goals you have described. > There are two libraries in the Boost review schedule: > > http://www.boost.org/community/review_schedule.html > > "STL Extensions" and "Countertree". None of them has as yet a review > manager. The order of the reviews may not be significant. It is more > important that at least one library pass the review. This would help > integrating other advanced data structures. > > > Are there Boost experts who could volunteer to take responsibility of > review manager for any of these two libraries? Data structures are very interesting to me, and I am in between jobs, so I have time to review one of these. If I take the time to review, I hope someone will be willing to review my skip list library, if I complete it and submit it to boost. -- View this message in context: http://boost.2283326.n4.nabble.com/Review-managers-wanted-for-augmented-dat… Sent from the Boost - Dev mailing list archive at Nabble.com.

3 4

[any] C++11 support
by Antony Polukhin 06 Apr '13

06 Apr '13

Hi, There is a patch for Boost.Any, that adds rvalue references support (via Boost.Move) and noexcepts (see ticket #6999 any.patch). It also includes a set of tests, and they pass well. Previous maintainers of the library are away for a long time, so if there will be no objections till the end of next weekend, I'll apply the patch. -- Best regards, Antony Polukhin

1 0

Scope Tracer or General Scope Function with parameter or variable values
by Anlin Zhang 06 Apr '13

06 Apr '13

Hi, I'm wondering if a "Scope Tracer" or a "General Scope Function" can be easily implemented with existing boost libraries now. One example use case is to log variable or parameter values or any program state when entering or leaving a scope, but it can do more than just logging, it can, for example, lock/unlock mutex, or prepare/teardown test cases, etc. For logging requirement, to give you an idea, the expected log message looks like this: >04/10/09 05:29:12:210257 entering func foo with parameter a=10, b=1.5, d.x=10, d.y=[string], global d=2, filename:test.cpp>04/10/09 05:29:12:210257 entering while() a=10, b=1.5, d.x=10>04/10/09 05:29:12:210257 leaving while() a=9, b=1.5, d.x=9>04/10/09 05:29:12:210257 entering while() a=9, b=1.5, d.x=9>04/10/09 05:29:12:210257 leaving while() a=8, b=1.5, d.x=8>...>04/10/09 05:29:12:210257 entering while() a=2, b=1.5, d.x=2>04/10/09 05:29:12:210257 leaving while() a=1, b=1.5, d.x=1>04/10/09 05:29:12:210257 entering while() a=1, b=1.5, d.x=1>04/10/09 05:29:12:210257 leaving while() a=0, b=1.5, d.x=0>04/10/09 05:29:12:210257 leaving function foo with parameter a=0, b=1.5, d.x=0, d.y=[string], global d=2, filename:test.cpp It does not have to be logging, it can do any general processing when entering/leaving a scope. I implemented a small library to do so in 2008 mainly for tracing, it worked fine for me and I never revisited it again until recently. Wondering if boost has libraries to support such a requirment. My implementation has a template class logger_class which can manage the log-level at runtime and send the log message to standard-output or other places. Application code use a macro to invoke the library, the macro can be disabled at compile time, or weakened to only log scope without variables. Example of the macro definition: #ifdef ENABLE_ONLY_SCOPE_TRACER #define SCOPE_TRACER(logger_class, logger, args, level, scope) \ scope_tracer<logger_class> tracer_obj(logger, __FUNCTION__, __FILE__, __LINE__); \ if (logger.get_level() >= level){ \ tracer_obj.init(level, scope).noMore(); \ } #else #ifdef DISABLE_SCOPE_VAR_TRACER #define SCOPE_TRACER(logger_class, logger, args, level, scope) #else #define SCOPE_TRACER(logger_class, logger, args, level, scope) \ scope_tracer<loggerClass> tracer_obj(logger, __FUNCTION__, __FILE__, __LINE__); \ if (logger.get_level() >= level){ \ tracer_obj.init(level, scope) << args; \ tracer_obj.noMore(); \ } #endif#endif Application code example: int globalVar; void MyClass::func1( A a, B&b, C* c, int d, std::string & s){ SCOPE_TRACER(MyLogger, test_logger, a<<b<<c<<d<<s<<globalVar, LOG_LEVEL_TRACE, "func1"); ... int localVar; while(x){ SCOPE_TRACER(MyLogger, test_logger, x<<a<<b<<c<<d<<s<<globalVar<<localVar, LOG_LEVEL_TRACE, "while(x)"); ... } ... } As you see, "<<" operator is being used. But it does not have to be only "<<" operator. Basically any C++ operator can be supported between the logger class and logged type. And the logger_class is not limited to just do tracing, it can virtuall do anything. I can attach the code if needed. The code was compiled with GCC on linux and VisualStudio on Windows. If the above functionality can be easily achieved using existing boot library, please advise how. Otherwise is it worth to request such a feature in boost? Thanks,Anlin

2 1

Regression tests crashing
by Tom Kent 03 Apr '13

03 Apr '13

Sometime yesterday most of the regression tests that I run (not the incremental ones?) started crashing and not sending any results. The screen output is the general error: "No such file or directory: '/home/tomkent/teeks99-02/a/results/boost/bin.v2/libs/any/test/any_test.test'" Looking at the tail of the results/bjam.log file shows this: error: at /home/tomkent/teeks99-02/a/tools_bb/kernel/modules.jam:107 error: Unable to find file or target named error: 'has_float128.cpp' error: referred to from project at error: '../libs/multiprecision/config' That was on Ubuntu 12.04, running the gcc-4.6 tests, but the others (including windows) have the same error. I don't believe anything else changed on the machines that were running the regression tests. You can still see the previous results on the regression results page. Tom Kent

1 0

[utility]result_of - erroneous result type from a C++11 lambda
by Vicente J. Botet Escriba 03 Apr '13

03 Apr '13

Hi, I've started to use C++11 lambda functions with Boost.Thread. The following works with clang-3.2 boost::future<int> f1 = boost::async( []() -> int {return 123;} ); But it fails with gcc-4.6.2, 4.7.2 and 4.8.0 ../example/lambda_future.cpp:25:91: erreur: conversion from ‘boost::future<void>’ to non-scalar type ‘boost::future<int>’ requested boost::future<int> f1 = boost::async( []() -> int {return 123;} ); The current implementation of boost::async uses boost::result_of to get the result type of the callable parameter. Is this a know bug on boost::result_of or gcc compiler? Best, Vicente

9 15

generate algorithm
by Fabian Bösch 02 Apr '13

02 Apr '13

Dear Boost community I am developing a library which extends the STL generate algorithm. It's aim is to improve performance for highly optimized (e.g. vectorized) generation mechanisms while it retains the same interface. This is especially useful when dealing with random number generation as there are several libraries (e.g. Intel MKL or AMD ACML) which provide means to generate a block of random values at once. However, the current STL generate algorithm is not able to cope with such "range generation". Is there any interest in a such a library for Boost? Cheers, Fabian Bösch

4 5

[boost.lockfree] Lockfree on Windows?
by Philip Bennefall 02 Apr '13

02 Apr '13

Hello, I am in need of a lock free fifo for my current project so I looked into lockfree. However, in the documentation I see: boost.lockfree has been tested on the following platforms: a.. g++ 4.4, 4.5 and 4.6, linux, x86 & x86_64 b.. clang++ 3.0, linux, x86 & x86_64 My question is what the status is on Windows? The documentation for boost.atomic says that Windows has been tested, so is there a reason why lockfree might have platform specific issues? When I run the example provided for the single producer/single consumer queue, I get the expected results so it does seem to work. It also states that the queue is lock free. Is it safe to use on Windows at present? Thanks in advance for any information. Kind regards, Philip Bennefall

2 3

[extension] library under construction status
by Lorenzo Caminiti 02 Apr '13

02 Apr '13

Hello all, What's the status of Jeremy Pack's library under construction Boost.Extension? P.S. I read the docs and I have a couple of comments: 1. shared_library as RAII Why isn't shared_library an RAII object? If there's a good reason for that, a RAII object should also be provided scoped_library that will open at construction and automatically close at destruction. 2. get should use function type syntax get<result_type, arg1_type, arg2_type, ...> should be changed to get<result_type (arg1_type, arg2_type, ...)> 3. type_map purpose/name The introduction of type_map seems arbitrary... some rationale needs to be provided as soon as it's introduced in Tutorial 2. Is there a better name than type_map? That name tells me little about what it is.. maybe, factory_map given that it's a map strings to factories? 4. call(type_map) lib.call(type_map) the name call is too generic, it should be changed to set or even better get_factories: type_map types; if(!lib.call(types)) ... Changed to: factory_map factories; if(!lib.get_factories(factories)) ... 5. type_map doesn't allow for multiple constructors type_map doesn't allow for derived classes to have constructors with different signatures than the base class. It also doesn't allow for classes (derived or base) to have multiple constructors. In any case, the syntax should use member function types, changing the current syntax: std::map<std::string, factory<animal, int> >... to: std::map<std::string, factory<void (*::animal)(int)> 6. I didn't understand parameter_map purpose, usage, etc :( Thanks. --Lorenzo

3 3

Building ThreadPool with boost::function in boost::lockfree::queue
by Ryan Fogarty 01 Apr '13

01 Apr '13

I am having trouble typedef-ing a lockfree::queue with boost::function as the element for a ThreadPool class I am writing. TestJobQueue1 below compiles fine while TestJobQueue2 compile fails. I am using gcc 4.7.2 and boost_1_53_0. void doSomething() { std::cout << "Called doSomething\n"; } typedef boost::function<void ()> FunctionType; class TestJobQueue1 { public: typedef std::vector<FunctionType> FunctionArray; TestJobQueue1() { std::cout << sizeof(FunctionType) << std::endl; FunctionType f(&doSomething); FunctionArray funcArray; funcArray.push_back(f); std::cout << funcArray.size() << std::endl; } }; class TestJobQueue2 { public: typedef boost::lockfree::queue<FunctionType> QueueType; TestJobQueue2() { std::cout << sizeof(FunctionType) << std::endl; FunctionType f(&doSomething); QueueType funcQueue; funcQueue.push(f); } }; >> g++ -c -I/usr/local/boost_1_53_0/include TestFunctionCompleteness.cpp produces: In file included from TestFunctionCompleteness.cpp:11:0: /usr/local/boost_1_53_0/include/boost/lockfree/queue.hpp: In instantiation of ‘class boost::lockfree::queue<boost::function<void()> >’: TestFunctionCompleteness.cpp:42:17: required from here /usr/local/boost_1_53_0/include/boost/lockfree/queue.hpp:79:1: error: invalid application of ‘sizeof’ to incomplete type ‘boost::STATIC_ASSERTION_FAILURE<false>’ /usr/local/boost_1_53_0/include/boost/lockfree/queue.hpp:83:1: error: invalid application of ‘sizeof’ to incomplete type ‘boost::STATIC_ASSERTION_FAILURE<false>’ >> g++ -c --std=c++0x -I/usr/local/boost_1_53_0/include TestFunctionCompleteness.cpp produces: In file included from TestFunctionCompleteness.cpp:11:0: /usr/local/boost_1_53_0/include/boost/lockfree/queue.hpp: In instantiation of ‘class boost::lockfree::queue<boost::function<void()> >’: TestFunctionCompleteness.cpp:42:17: required from here /usr/local/boost_1_53_0/include/boost/lockfree/queue.hpp:79:5: error: static assertion failed: (boost::has_trivial_destructor<T>::value) /usr/local/boost_1_53_0/include/boost/lockfree/queue.hpp:83:5: error: static assertion failed: (boost::has_trivial_assign<T>::value) Is boost::function somehow an incomplete type and not expected to work in boost::lockfree::queue or is this a bug in the queue implementation? Can anyone suggest a simple workaround? Thank you, Ryan Fogarty

2 3

traits classes vs. metafunctions
by Stefan Strasser 01 Apr '13

01 Apr '13

Hi everyone, are there any guidelines for boost libraries on when to use traits classes and when to use a metafunction for each member of a would-be traits class? I've seen both in boost libraries. c++11 uses traits classes (iterator_traits, pointer traits, ...), but Abrahams and Gurtovoy [1] argue that traits classes ("traits blobs") should be avoided "at all costs" because they are an unnecessary concatenation of multiple metafunctions into one metafunction with multiple return values. in practice, this means that if a single 'traits class member' aka 'metafunction result' ought to differ from the defaults, the entire traits class has to be reimplemented by the user, not only the metafunction whose result ought to differ from the default. on the other hand, asking a user to implement 5 or more metafunctions is much more tedious than implementing a traits class. possibility 1) template<...> struct property1 : mpl::bool_<...>{}; template<...> struct property2 : mpl::bool_<...>{}; template<...> struct property3 : mpl::bool_<...>{}; 2) template<...> struct ..._traits{ typedef mpl::bool_<...> property1; typedef mpl::bool_<...> property2; typedef mpl::bool_<...> property3; }; you could provide both, but in which "direction"? does a traits class refer to metafunctions, or do metafunctions refer to a traits class? a) template<...> struct ..._traits{ typedef typename property1<...>::type property1; typedef typename property2<...>::type property2; typedef typename property3<...>::type property3; }; b) template<...> struct property1 : ..._traits<...>::property1{} template<...> struct property2 : ..._traits<...>::property2{} template<...> struct property3 : ..._traits<...>::property3{} which is accessed when "calling" the metafunction? it seems to me that metafunctions are better for the "caller" and traits classes are better for the implementor of the traits/metafunctions. in my case, the caller is a library and the implementor is the user, so you might choose a traits class over metafunctions. but what if a new member is added that wasn't there before (like "property4"). the implementor of an (old) traits class could not have known about the new member, so accessing the new member (by the library) will result in an error, while a new metafunction "property4" could simply return a default. should new properties of traits classes only be accessed after making sure they exist, using SFINAE? there isn't a precedent for this in c++11. only new traits classes were introduced, but no new members of already existing traits classes. should metafunctions always be preferred? why did c++11 choose traits classes over metafunctions, even though the concept of a metafunction was introduced in c++11 (<type_traits>). Thanks [1] Abrahams; Gurtovoy: C++ Template Metaprogramming, 2.2 Metafunctions

7 14

[config] gcc-4.8 and __GXX_EXPERIMENTAL_CXX0X__
by Vicente J. Botet Escriba 01 Apr '13

01 Apr '13

Hi, it seems that gcc-4.8 doesn't defines any more __GXX_EXPERIMENTAL_CXX0X__ when the flag -std=c++11 is given. This means that all the BOOST_NO_CXX11_ are defined. Could something be done to make usable gcc-4.8 in c++11 mode? Best, Vicente

2 4

[mpi] Boost.MPI and MPI 3!
by Renato Forti 01 Apr '13

01 Apr '13

Hi All, Boost.MPI project will support MPI version 3 with that provide some support to fault tolerance! Have any work in this area? Thanks

2 2

[build & test] clang-3.2 for Win32
by Jim Bell 01 Apr '13

01 Apr '13

I took the liberty to run regressions with clang-3.2 on Win32. It's pretty unhappy, and most failures have to do with it not building shared libraries (DLLs) correctly. Not finding symbols. I found a (non-boost) thread that talks about that being a problem with it. Even so, I had a couple questions: 1. It gets labeled clang-linux-3.2, though it's under the Windows umbrella in the regression matrix. How do we fix that? user-config.jam 'using clang ;' does seem to work fine. 2. I get these warnings. They seem like clang's problem, not ours. Am I right? ..\boost/test/execution_monitor.hpp:59:7: warning: 'dllexport' attribute only applies to variables and functions [-Wignored-attributes] class BOOST_TEST_DECL translate_exception_base { ..\boost/test/detail/config.hpp:88:29: note: expanded from macro 'BOOST_TEST_DECL' # define BOOST_TEST_DECL BOOST_SYMBOL_EXPORT ..\boost/config/platform/win32.hpp:30:42: note: expanded from macro 'BOOST_SYMBOL_EXPORT' # define BOOST_SYMBOL_EXPORT __declspec(dllexport) <built-in>:154:38: note: expanded from macro '__declspec' #define __declspec(a) __attribute__((a)) 3. I tried to coerce the regression to use only static libraries (since clang does build boost static pretty well)... --bjam-options='link=static runtime-link=static' but this made some part of the regression tests very unhappy: error: link=shared together with runtime-link=static is not allowed error: such property combination is either impossible error: or too dangerious to be of any use That's understandable, as it goes against the grain. But is there a way to inform the tests somehow? It's easy to imagine wanting to explicitly test what a shared library does, but what about cases where it doesn't matter if it's shard or static? (Are they distinguishable?) If anyone's interested, please peruse the mess I made of the regression matrix. And purge them from the matrix, too, eventually.

1 0

Associative container using a vector
by Stefan Strasser 01 Apr '13

01 Apr '13

Hi everyone, I often find myself using a specific data structure and I'm wondering whether: - it has a wellknown name - there is a generic equivalent in a boost library - or there is interest to add one. The basic idea is that you have a bunch of values stored in one object of type A: struct A{ vector<Value> values; }; and then you have multiple objects of type B that each have to associate data with each value stored in A. you could use: struct B{ map<Value,Data> data; }; but since A::values already provides a contiguous index for the values you can just as well store the data in B in a vector with matching indexes, and make sure that the indexes used in A::values don't change: struct A{ vector<Value> values; vector<std::size_t> freelist; map<Value,std::size_t> index; void erase(...){ push index of unused value to freelist; } void push_back(Value){ reuse a freelist entry or values.push_back; } }; struct B{ vector<optional<Data> > data; }; keep in mind that there is one object A for many objects B. you can now access the associated data of many objects B by looking up the index once in object A, and by wasting some space in B::data for unused entries: b.data[a.index[value]] If Data is a pointer or any other type that knows an empty() state, you can also get rid of optional<> and generalize it as two template like: template<class Value> class index; template<class Mapped,class OptionalTraits=boost_optional_traits> class vector_map; Is there anything like that?

3 4

[config] BOOST_NOEXCEPT on MSVC
by Michael Marcin 01 Apr '13

01 Apr '13

Why isn't BOOST_NOEXCEPT defined to be throw() for msvc? Microsoft Visual Studio 11.0\VC\include\yvals.h defines: #define _NOEXCEPT throw () and the rest of the standard library implementation uses it ex. void swap(thread& _Other) _NOEXCEPT { // swap with _Other _STD swap(_Thr, _Other._Thr); } http://msdn.microsoft.com/en-us/library/wfa0edys.aspx states: throw() The function does not throw an exception. However, if an exception is thrown out of a function marked throw(), the Visual C++ compiler will not call unexpected (see unexpected (CRT) and unexpected (<exception>) for more information). If a function is marked with throw(), the Visual C++ compiler will assume that the function does not throw C++ exceptions and generate code accordingly. Due to code optimizations that might be performed by the C++ compiler (based on the assumption that the function does not throw any C++ exceptions) if a function does throw an exception, the program may not execute correctly. It seems a better choice here than expanding BOOST_NOEXCEPT to nothing.

2 1

Re: [boost] [fusion] What do we do with nil ?
by Enrique Nieloud 31 Mar '13

31 Mar '13

Hi!, Sorry to bother, but is is patch officially included in version 1.53.0? I'm having trouble to compile under XCode 4 and it seems related to the "nil" word. thanks, - Enrique

1 0

Re: [boost] Review managers wanted for augmented data structures
by Dave Abrahams 31 Mar '13

31 Mar '13

on Mon Mar 04 2013, Vadim Stadnik <vadimstdk-AT-gmail.com> wrote: > On Mon, Mar 4, 2013 at 7:35 PM, Stefan Strasser <strasser(a)uni-bremen.de>wrote: > >> Am 04.03.2013 08:48, schrieb Martin Knoblauch Revuelta: >> >> Do you mean that you would declare them RandomAccess iterators? >>> >>> We can expect to have containers with millions of elements. For that >>> size, O(log N) is about 20 times faster than O(log N log N). That >>> makes a big difference. >>> >> >> isn't that a reason for providing optimized specializations of some STL >> algorithms, instead of introducing another iterator category? >> >> how would a new iterator category even help? >> as long as your iterator is a forward iterator, the user is allowed to >> call std::binary_search on your iterator, which uses the same algorithm on >> any iterator category. (only functions used by binary_search like >> std::advance are specialized on the category) >> where in all this would your tree-optimized binary_search code be called? >> >> the standard doesn't require random access iterator ops to be constant >> time. >> >> > Unfortunately, there is a problem of iterator categories with augmented > trees. The C++ standard specifies computational complexity for iterator > operations, although they are not shown in tables, since all of them are > required to be constant time. In both C++03 and C++11 this is formulated in > the section 24.1.8. Are you sure these iterators don't fall under the *amortized* constant time requirements, just like those of the ordered associative containers? -- Dave Abrahams

2 3

[gil] undocumented/unintended concept requirements in color_convert
by Michael Marcin 31 Mar '13

31 Mar '13

I was trying to add other color spaces support and noticed that the rgb to cmyk converter contains: typedef typename channel_type<P2>::type T2; T2 x = channel_traits<T2>::max_value()-get_color(dst,black_t()); if (x>0.0001f) { float x1 = channel_traits<T2>::max_value()/float(x); T2 should model: http://www.boost.org/doc/libs/1_53_0/libs/gil/doc/html/g_i_l_0264.html but this code additionally requires T2 to be greater-than comparable to float and convertible to float.

3 4