- Boost-users - lists.boost.org

Variadic versions of all the vector200 map50 stuff
by Michael Steinberg 19 Sep '13

19 Sep '13

Hello, I'm currently tracking down all the slowdowns I encounter when using boost in my Visual Studio solutions. Much of the slowdowns seem to happen, because there's a wide inclusion-usage of <boost/typeof/vector.hpp> throughout boost. This seems to be a replicate of <boost/mpl/vector.hpp>, which is not as much used, but equally evil. Inclusion of any of these headers leads to a 20MB increase in Visual Studio IntelliSense database size. I tested to comment out the inclusion of all the generated <vectorXXX.hpp>. This indeed works fine and is a real relief in Visual Studio responsiveness, because although these headers seem to be included from everywhere, they dont seem to be actually used. Tracking down the inclusion paths proves difficult to me, because the Visual-Studio tool doesn't seem to be able to handle the Preprocessor-generated include directives that boost uses extensively (and I'm just a newbie). With the advent of variadic-enabled compilers (Visual Studio 2013RC just arrived), is anybody working on making this mpl-stuff variadic? That would be a major help for almost anybody using tools like IntelliSense or VisualAssist (the vector200.hpp is a wopping 2mb of totally redundant template goodness...). I think I could make the appropriate changes to my local copy, however I have no idea how to write proper boost-worthy code or even how to file it for inclusion. Also I'm not too sure why the different sized vectors have different names and if the client code relies on that, etc., but I could probably find out. Are there any "template-instantiation-coverage"-tests included, so I could check that my variadics are compatible in the use-cases? Any help would be appreciated! With Kind Regards Michael

2 1

[boost-regex] 1.47.0 regexp_search with match_partial does not return 'what'
by Frank Bergemann 18 Sep '13

18 Sep '13

2 1

Static map like array or range?
by Duane Murphy 18 Sep '13

18 Sep '13

I find myself often building lookup tables for conversion or mapping of one piece of data to another. Common conversions include * converting from one string to a different string (e.g. text translation), * converting from an enum to a descriptive string * converting from a string to an enum * converting from a string to a function or functor The key aspect of these lookup tables is that they are static. If the table is reasonably large, say 7-10 items, I ensure that the table is presorted and use lower_bounds() to search the table. Some uses have started using map_list_of() to initialize the map. My inner optimizer finds this disturbing as inherently static information is being allocated dynamically in order to due to the lookup. I believe it would be more intuitive if there was a map like interface to the lookup table. The implementation would use the static nature of the data to its advantage. In particular the implementation would not require the data to be copied, re-allocated, etc. I've been searching the boost libraries, but have not been able to find something like this. Thank you for your suggestions. …Duane

3 2

Copy fusion::vector -> {std,boost}::array
by Rodrigo Madera 17 Sep '13

17 Sep '13

Checking the archives, I see that a *lot* was discussed on [1], and there was a mention (waaay after much debate and proposed solutions) on having a copy and copy_if to do such copies. Mr. Steven Watanabe mentioned a copy impl was lying around his computer [2]. Did it actually make it into Boost? Thanks for any input, Rodrigo [1] http://lists.boost.org/boost-users/2010/10/63168.php [2] http://lists.boost.org/boost-users/2010/10/63343.php

2 2

Generic “Observable” mixin based on Boost
by Павел Фролов 17 Sep '13

17 Sep '13

I find Observer pattern a very useful tool to decrease coupling between modules of a system. Boost provides Signals2 which can be used for that purpose. But adding Signal2 objects to arbitrary classes is not very convenient, because your must manually implement RegisterEvent/FireEvent functions. What I was looking for is a reusable mixin which could add “Observable” behavior to arbitrary class with minimal hassle. Ideally, the user of mixing must only provide event table which contains event tags and handler signatures, e.g. (in pseudocode): EventTable { EventA: void(), EventB: int(bool), ... } I came up with working proof of concept which uses CRTP-based mixin and traits class for event table declaration. I'm new to template meta-programming, so I'm looking forward to comments about possible improvements. In particular, I'm not very satisfied with syntax boost::fusion::pair<Show, boost::optional<boost::signals2::signal<void()>>> (see attachement). I think, I can do better by allowing user to specify just boost::fusion::pair<Show, void()> and then generate proper type with mpl::transform and/or fusion::transform. This way is shorter and implementation details (optional, signals2) are hidden from the user. The mixin code (full code with example usage in attachment): #include <utility> #include <boost/fusion/container/map.hpp> #include <boost/fusion/sequence/intrinsic/at_key.hpp> #include <boost/fusion/sequence/intrinsic/value_at_key.hpp> #include <boost/optional.hpp> #include <boost/signals2.hpp> #include <boost/utility/in_place_factory.hpp> //***************************************************************************** // EventSource mixin //***************************************************************************** // event source traits base (concrete event source must create specialization) template<typename ConcreteEventSource> struct Events; // event source mixin (using CRTP) template<typename ConcreteEventSource> class EventSource { public: typedef ::Events<ConcreteEventSource> Events; private: typedef typename Events::EventTable EventTable; // helper type generator template<typename Event> struct GetType { // boost::optional type typedef typename boost::fusion::result_of:: value_at_key<EventTable, Event>::type SignalWrapper; // boost::signal type for particular boost::optional type typedef typename SignalWrapper::value_type Signal; // slot return type for particular boost::signal type typedef typename Signal::result_type SignalReturnType; }; public: // connect event handler template<typename Event, typename Handler> void Connect(Handler&& handler) { GetSignal<Event>()->connect(std::move(handler)); } protected: // generate nullary signal template<typename Event> typename GetType<Event>::SignalReturnType FireEvent() { return GetSignal<Event>()->operator()(); } // generate unary signal template<typename Event, typename Arg1> typename GetType<Event>::SignalReturnType FireEvent(Arg1 arg1) { return GetSignal<Event>()->operator()(arg1); } // ... more overloads (MSVC does not support variadic templates) private: // creates signal if not exists, returns reference to it template<typename Event> typename GetType<Event>::SignalWrapper& GetSignal() { auto& signal = boost::fusion::at_key<Event>(signals_); if (!signal) signal = boost::in_place(); return signal; } EventTable signals_; };

1 0

[units]
by Robert Ramey 16 Sep '13

16 Sep '13

I have an urgent and definite need for the functionality which boost.units is meant to provide. The concept is quite simple so I had expected it to be relatively easy to use. I had anticipated spending a few hours perusing the documentation and examples, adding code to my application and moving on. T'was not to be. I'm now on day 3 (or 4?) trying to figure this thing out. The main hurdle is that the documentation is indecipherable. There are a number of templated classes defined which depend upon each other. I expect to find a page for each of the templates along with a description of what the template arguments should be. (Ideally these should be backed up with concept checking classes). Ideally there would be a small example showing how one would use these templates. It seems that the library is layered from low level ideas to higher level ones. The documentation touches on this - but its not really formalized. Theres a reference section which includes scores of things like base_units types for wide variety of units - but no where is there any indication how these should be used - not even in the examples are they used. There is a concept of "system" (of dimensions?) and the idea of mixing units from different systems and the idea that the same dimension (eg length) have multiple renderings in the system (nautical vs metric). I realize I've likely got this wrong and I'm not asking for anyone to explain this to me here - my point is that what should be very simple is incredibly confusing. I still haven't figured out whether it's only the documentation which is incomplete or if it is the actual logical design of the library itself is what's flawed. I'm still working on this. I don't remember much traffic on this list regarding this library and if I google information on boost C++ units library I don't get a whole lot of hits. This could mean that that the library is so easy to use no one ever needs to ask about it or it could be that almost no one every uses it. In Boost we don't gather statistics on usage by library so I don't which of these it might be. So my question is: Has anyone had good success with this library? How much effort did it take? What did you have to do to make it work. I would be curious about any other user feed back in these questions. Robert Ramey

2 1

Re: [Boost-users] [Boost] Command Line to run a specific Boost Test Case using BOOST_AUTO_TEST_CASE_TEMPLATE
by Hemanth Choudary M V (RBEI/ETA1) 16 Sep '13

16 Sep '13

Hi Gennadiy, I have tried the same with the Boost 1.54.0 (Release Version) and it is not working. Looking forward to the next version which supports this. Mit freundlichen Grüßen / Best Regards, Hemanth Choudary MV RBEI/ETA1

1 0

Overhead in boost IPC
by nithesh salian 16 Sep '13

16 Sep '13

Hi, I am interested in knowing the overhead associated with using boost message queue and shared memory. i.e., What is the latency associated with using these mechanisms for IPC? Thanks in advance. -- Regards, Nithesh Gangadhar Salian .

1 0

Re: [Boost-users] [units]
by Kelly, Dan 16 Sep '13

16 Sep '13

I have had much the same experience as you. The library seems to hold a lot of promise and is from what I can see very sophisticated. But I had the same problems with the documentation and examples. I struggled for some time to figure out the difference between units and base units (and dimensions and base dimensions) and when you would use each. If someone could put together a decent tutorial I'm sure I would use this library quite frequently. -----Original Message----- From: Robert Ramey [ramey(a)rrsd.com<mailto:ramey@rrsd.com>] Sent: Sunday, September 15, 2013 01:14 PM Eastern Standard Time To: boost-users(a)lists.boost.org Subject: [Boost-users] [units] I have an urgent and definite need for the functionality which boost.units is meant to provide. The concept is quite simple so I had expected it to be relatively easy to use. I had anticipated spending a few hours perusing the documentation and examples, adding code to my application and moving on. T'was not to be. I'm now on day 3 (or 4?) trying to figure this thing out. The main hurdle is that the documentation is indecipherable. There are a number of templated classes defined which depend upon each other. I expect to find a page for each of the templates along with a description of what the template arguments should be. (Ideally these should be backed up with concept checking classes). Ideally there would be a small example showing how one would use these templates. It seems that the library is layered from low level ideas to higher level ones. The documentation touches on this - but its not really formalized. Theres a reference section which includes scores of things like base_units types for wide variety of units - but no where is there any indication how these should be used - not even in the examples are they used. There is a concept of "system" (of dimensions?) and the idea of mixing units from different systems and the idea that the same dimension (eg length) have multiple renderings in the system (nautical vs metric). I realize I've likely got this wrong and I'm not asking for anyone to explain this to me here - my point is that what should be very simple is incredibly confusing. I still haven't figured out whether it's only the documentation which is incomplete or if it is the actual logical design of the library itself is what's flawed. I'm still working on this. I don't remember much traffic on this list regarding this library and if I google information on boost C++ units library I don't get a whole lot of hits. This could mean that that the library is so easy to use no one ever needs to ask about it or it could be that almost no one every uses it. In Boost we don't gather statistics on usage by library so I don't which of these it might be. So my question is: Has anyone had good success with this library? How much effort did it take? What did you have to do to make it work. I would be curious about any other user feed back in these questions. Robert Ramey _______________________________________________ Boost-users mailing list Boost-users(a)lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users ________________________________ IMPORTANT: The information contained in this email and/or its attachments is confidential. If you are not the intended recipient, please notify the sender immediately by reply and immediately delete this message and all its attachments. Any review, use, reproduction, disclosure or dissemination of this message or any attachment by an unintended recipient is strictly prohibited. Neither this message nor any attachment is intended as or should be construed as an offer, solicitation or recommendation to buy or sell any security or other financial instrument. Neither the sender, his or her employer nor any of their respective affiliates makes any warranties as to the completeness or accuracy of any of the information contained herein or that this message or any of its attachments is free of viruses.

1 0

[coroutine] Resuming of coroutine from another thread
by Jason Roehm 16 Sep '13

16 Sep '13

I'm thinking up a design for a task-processing system that uses Boost.Asio and Boost.Coroutine. As part of that design, I would have a pool of worker threads executing work items. During processing of a particular work item, it's possible that the worker might need to wait on some other operation (e.g. I/O) to complete. However, while it waits, I'd like to put that thread to work doing something else useful from the queue. I was thinking of using coroutines to provide the ability to suspend the processing of one task, restarting it later when it is ready to move forward again. In my hypothetical framework, the thread that finishes the work might not be the same one that started it in the first place. I recently came across this Stack Overflow question: http://stackoverflow.com/questions/8557067/thread-safe-coroutines-with-asio which inquired as to the safety of creating a coroutine in one thread and then resuming it in another. The answers there indicate that this is not supported, and they link to Boost.Coroutine documentation that explains the underlying problems and why it isn't supported. However, that documentation was from the original Boost.Coroutine by Giovanni Deretta (which I'm not sure was ever actually accepted into Boost). The latest version by Oliver Kowalke indicates that it is a follow-up on that library, but doesn't have any specific caveats in the documentation regarding the interaction between threads and coroutines. Does the contemporary Boost.Coroutine have the same limitation, or is it safe to create a coroutine in one thread and then resume it in another? Jason

2 1