June 2011 - Boost - lists.preview.boost.org

[fusion] Qualification of 'at' and 'deref' in fusion (Ticket #4755)
by Gabriel Redner 15 Jun '11

15 Jun '11

I have just been bitten by a simple issue in fusion. The 'at' and 'deref' functions are used unqualified, which causes conflicts when mpl::at and mpl::deref are also visible. This issue has been reported as Ticket #4755 [1], and a patch was attached eight months ago. Is there anyone I can prod to get this fix committed? I have also updated the ticket with some relevant information. Thanks, -Gabe [1] https://svn.boost.org/trac/boost/ticket/4755

3 4

[intrusive] unordered_set insert/find question
by Tim Blechmann 15 Jun '11

15 Jun '11

hi all, i am a bit confused about the behavior of boost::intrusive::unordered_set. i can insert() an object, but i am unable to find() it: my basic workflow is: // insert: map_type * elem = new map_type(); std::pair<iterator, bool> inserted = slot_resolver_map.insert(*elem); bool success = inserted.second; assert(success); // object has been inserted correctly // verify: iterator this_iterator = inserted.first; iterator found = slot_resolver_map.find(*elem); iterator end = slot_resolver_map.end(); assert(this_iterator == found); // fails assert(found != end); // fails so basically if i insert an element, i cannot find it anymore ... i have a reduced test case, which shows this behavior at [1] ... is this my fault (not using the unordered_set correctly) or is it a bug in boost.intrusive? thanks, tim

3 4

MS VC10 std::numeric_limits<float>::max_digits10 is wrong for float
by Paul A. Bristow 15 Jun '11

15 Jun '11

Inexplicably, MS for VS 10 have chosen std::numeric_limits<float>::max_digits10 == 8 for VS 10 rather than 9, when the formula max_digits10 = 2 + std::numeric_limits<FPT>::digits * 3010/10000; and values have been known for 30 years. std::numeric_limits<T>:: max_digits10; is 'new for C++0X'. Briefly, max_digits10 is the number of decimal digits that you need to output to make sure that a one least significant bit (ULP) change produces a different decimal digit string. Any more digits are just noise with no useful information. Using max_digits10 rather than the formula above may be more efficient, and is certainly very much clearer. "Executive Summary" ================ Don't use max_digits10 for float or floating-point template types for MS VS - yet! You can safely (and very usefully) use cout.precision(std::numeric_limits<double>::max_digits10); when you want to be sure that all significant decimal digits are output. GCC provides the correct values of 9 for float (and 17 for double). The Fuller Story =========== This might affect anyone planning to use VS10 and this new (and very useful numeric_limits value) with built-in *type float, or a floating-point template type*, potentially, for example, with Boost Lexical_cast, Serialization. Round tripping would be 1 bit wrong in perhaps 1/3 of cases! If max_digits10 was used, Boost.Test might detect values that are different, but are reported to be the same, for example. As far as I know, none of these utilities yet try to use max_digits10. Mercifully, this fault only affects type T = float (not the most popular double or long double), and only in new code that uses the 'new for C++0X' numeric_limits<T>:: max_digits10. Other types like double and long double (and possibly user defined types that specialize numeric_limits) are, I believe, correct. So you can safely (and very usefully) use cout.precision(std::numeric_limits<double>::max_digits10); when you want to be sure that all significant decimal digits are output. (If you want this to be 'Boost portable' you will have to use the inscrutably named BOOST_NO_NUMERIC_LIMITS_LOWEST as a proxy for "C++0x numeric_limits before using new-for-C++0X max_digits10. Unless we add yet another macro for max_digits10? Groan.). The workaround is to continue to use the well-tried and test Kahan formula max_digits10 = 2 + std::numeric_limits<FPT>::digits * 3010/10000; Reported and confirmed as https://connect.microsoft.com/VisualStudio/feedback/details/668921/std-nume… ts10-value-of-8-is-wrong-and-should-be-9 (It might help if, you could confirm you can 'reproduce' this too). I trust this mistake will be corrected pronto - but I'm not holding my breath :-( I will post again if it is resolved. Paul --- Paul A. Bristow, Prizet Farmhouse, Kendal LA8 8AB UK +44 1539 561830 07714330204 pbristow(a)hetp.u-net.com

5 13

[MSM] exit_ps stuks in case of outer state machine uses Row with event to event base class
by Richard Szabo 15 Jun '11

15 Jun '11

Hi Christophe and All after splitting our complicated several level deep contained state machines to small run-time instances. I have to use feature Event Hierarchy (http://svn.boost.org/svn/boost/trunk/libs/msm/doc/HTML/ch03s02.html#d0e1176) which is really cool. But this feature in combination with pseudo exits state, breaks the pseudo exit state functionality and the pseudo exit state is entered but never left. I have made a small example where the problem is reproduced see attachment. (the file name is misleading was just lazy to rename it something more appropriate) Here is the output of the execution : entering: Idle MS1_ leaving: Idle Action: MS1::onEvent1 entering: RunningStateMachine MS1_ entering: RunningStateMachine::PseudoEntry1 leaving: RunningStateMachine::PseudoEntry1 entering: Inner1 RunningStateMachine_ entering: InnerState11 leaving: InnerState11 leaving: Inner1 entering: RunningStateMachine::PseudoEntry1 I have tried it in our host test environment only so VC9 compiler and with boost trunk from Friday last week. Cheers Richard.

2 2

[review][assign] Formal review of Assign v2 begins 15th June
by John Bytheway 14 Jun '11

14 Jun '11

The formal review of the Assign v2 library upgrade by Erwann Rogard starts tomorrow, June 15th and is scheduled to continue until June 24th. About Assign v2 =============== Assign v2 is an upgrade of Boost.Assign (1.0) that redefines its functionality, with optional C++0x support and new features. It is a compact interface for executing the repetitive tasks of assigning or inserting container elements. Please see the change log (linked below) for more detail on the motivation for this library and pointers for navigating the documentation. Where to get it =============== The documentation is online at: <http://svn.boost.org/svn/boost/sandbox/assign_v2/libs/assign/v2/doc/html/in…> With the change log at: <http://svn.boost.org/svn/boost/sandbox/assign_v2/libs/assign/v2/doc/html/bo…> The Assign v2 source can be found in the Boost sandbox at: <http://svn.boost.org/svn/boost/sandbox/assign_v2/> or downloaded from: <http://www.boostpro.com/vault/index.php?action=downloadfile&filename=assign…> Review guidelines ================= Reviews should be submitted to the developer list (boost(a)lists.boost.org) with '[assign]' in the subject. Or if you don't wish to for some reason, you can send them privately to me. If so, please let me know whether or not you'd like your review to be forwarded to the list. All reviews are appreciated, even if you don't have time for an in-depth study. The Assign v2 library consists of several largely independent components: Reference containers and chaining, container modification and generation, and conversion of ranges to containers. If you wish, feel free to review and even vote on these separately. For your review you may wish to consider the following questions: - What is your evaluation of the design? - What is your evaluation of the implementation? - What is your evaluation of the documentation? - What is your evaluation of the potential usefulness of the library? - Did you try to use the library? With what compiler? Did you have any problems? - How much effort did you put into your evaluation? A glance? A quick reading? In-depth study? - Are you knowledgeable about the problem domain? And finally, every review should answer this question: - Do you think the library should be accepted as a Boost library? Be sure to say this explicitly so that your other comments don't obscure your overall opinion. More detail on the review process is available at: http://www.boost.org/community/reviews.html John Bytheway Review Manager

1 0

[documentation/doxygen] Trouble with partial template specializations in documentation
by Vicente Botet 14 Jun '11

14 Jun '11

Hi, I'm having some trouble with partial template specializations documentation while generating the Boost.Conversion documentation. The documentation for the specialization is not generated sometimes. I have added the file type_erasure/any.hpp that specialize the any template class and it works as expected with my Jamfile. I don't know why the specializations in conversion/assign_to.hpp are not all generated. The specialization on conversion/boost and conversion/std don't appear neither. Could someone help me? Thanks in advance, Vicente -- View this message in context: http://boost.2283326.n4.nabble.com/documentation-doxygen-Trouble-with-parti… Sent from the Boost - Dev mailing list archive at Nabble.com.

1 0

Holding char arrays in boost variants, what is the correct way?
by Michael Goldshteyn 14 Jun '11

14 Jun '11

What is the correct way of getting several character arrays into a variant: // This doesn't work typedef boost::variant<int,char[256],char[8192]> MyType; // This does, but seems like a lot of work template <int len> struct MyChar { char val_[len]; }; typedef boost::variant<int,MyChar<256>,MyChar<8192> > MyType; What is the best way to do this? Thanks, Michael Goldshteyn

4 7

[container_bptree] updates and performance measurements
by Vadim Stadnik 14 Jun '11

14 Jun '11

Hi all, The code and documentation of the submitted library container_bptree have been updated. The new version can be downloaded using this link: http://www.boostpro.com/vault/index.php?action=downloadfile&filename=contai… 1. For those who do not know this library I’d suggest before opening the documentation to spend some time to the answer the following warm-up question: Is it possible to sum up any N consecutive numbers in less than N-1 operations? 2. The library and tests compile and run correctly on two C++ compilers: MSVC++ 9/10 (tested with Visual Studio 2008/2010, including Express editions), and GCC 4.5.2 (tested with MinGW). 3. Lists of template parameters of class templates have been modified to simplify the generation of classes of STL variants of containers. 4. Performance measurements have been added to the documentation. They show that submitted B+ trees are highly competitive with Red Black trees in basic operations. 5. I would like to thank Beman Dawes for suggestion to adapt and apply map tests from his project [BTree]. Class templates of the namespace container_bptree can be used to generate three types of map containers with bidirectional and random access iterators. All of them pass successfully the adapted tests on Windows systems. At this stage it would be particularly valuable in addition to my own tests to run independent tests for associative containers with random access iterators. This is why I am wondering if there are Boost libraries, which include classes of such associative containers and test code. 6. Since I am a newcomer to Boost community I do not know how the current level of quality of the library container_bptree meets Boost standards. Anyone who tried the code and read the documentation is welcome to make comments and suggestions. I am quite happy to implement another list of improvements before the formal request for interest to this library. Regards, Vadim Stadnik

1 0

fusion concepts
by Robert Ramey 14 Jun '11

14 Jun '11

I'm trying to implement concept checking for at least parts of the boost::fusion library. My motivations are several: a) I'm interested in making another library/program based on fusion and I believe that having implemented concepts will make this job easier. b) I'm interested in the concept checking concept and this is way to pracitice. Attached is the current state of my efforts. I have a few problems with this. I think most of these are specific to fusion but maybe not. a) line 29 of test_misc.cpp invokes concept checking on a fusion vector. lines 28 & 29 of fusion/concepts.hpp are intended to verify that result_of::next<I> have been defined for the for the forward iterator associated with the fusion vector. The BOOST_CONCEPT_ASSERT fails with element 6 from the sequence. I conclude that the recurrsive invocation of concept checking on the result of next<I> is crapping out at the end of the sequence. This makes sense in a practical way - BUT it would voilate the requirements on the return type from result_of::end<vector>. Sooooo i) am I right about this? ii) if so how should it be addressed? fixing the concept implementation (how?) would "hide" the violation. Seems to be that the documentation on forward sequence/iterator should be tweaked and my concept checking code should be fixed up to match. Any thoughts on this? b) lines 34-38 illustrate a different problem. I can't figure out what distance(i,j) is supposed to return. I would think that result_of::distance(i, j) should be a signed integer of some sort. So I'm not understanding why this assertion traps. I not sure if it's a side effect of the problem cited in a) above. c) In another file not included here (test_concepts.hpp). I've attempted to implement archtypes for these types. I managed to do it, but I'm not all that confident that i did it right. After doing this, I came to understand the fusion extension mechanism. I could re-implement the archtypes with this and I'm confident that it would be "correct" BUT I'm wonder if this just wouldn't be a circular conclusion. That is would doing this really be testing the concepts or would generating archetypes in this way just guarentee that things were in some sense internally consistent but not necessarily reflection the concepts as stated in the documentation. Any comments/help would be appreciated. Robert Ramey begin 666 concepts.hpp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a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a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a)+$H^.CIT>7!E/@T*+R\@(" @(" @("DI M.PT*(" @(" @("!S86UE7W1Y<&4H:2 ]/2!J+"!B*3L-"B @(" @(" @<V%M M95]T>7!E*&D@(3T@:BP@8BD[#0H@(" @(" @('-A;65?='EP92AD97)E9BAI M*2P(a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` end begin 666 test_misc.cpp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` end

5 11

Re: [boost] [gsoc] boost.simd news from the front.
by Ivan Le Lann 14 Jun '11

14 Jun '11

----- "David A. Greene" <greened(a)obbligato.org> a écrit : > Mathias Gaunard <mathias.gaunard(a)ens-lyon.org> writes: > > > Making data parallelism simpler is the goal of NT2. And we do that > by > > removing loops and pointers entirely. > > First off, I want to apologize for sparking some emotions. That was > not > my intent. I am deeply sorry for not expressing myself well. > > NT2 sounds very interesting! Does it generate the loops given calls > into generic code? > > Let me try to clarify my thinking about boost.simd a bit. > > - There is a place for boost.simd. It's important that I emphasize > this > upfront. > > - Here is how I see a typical programmer using it, given a loop nest: > > - Programmer tries to run the compiler on it, examines code > - Code sometimes (maybe most of the time) executes poorly > - If not, done > > - Programmer restructures loop nest to expose parallelism > - Try compiler directives first, if available (tell compiler > which > loops to interchange, where to cache block, blocking factors, > which loops to collapse, etc.) > - Otherwise, hand-restructure (ouch!) > > - Programmer tries compiler again on restructured loop nest > - Code may execute poorly > - If not, done > > - Programmer adds directives to tell the compiler which loops > to vectorize, which to leave scalar, etc. > - Code may still execute poorly > - If not, done > > - Programmer uses boost.simd to write vector code at a higher level > than provided compiler intrinsics > > Does that seem like a reasonable use case? > I hope I won't disappoint you, but in my case I'd put boost.simd a few ranks up. Because, from what I've seen in slides, it's a quite easy and really clean way to do things. I've always been taught to dive into dirty things as a last resort, so I'd use boost.simd first, and eventually, if performance is still not met, try hand-written hints. Regards, Ivan

1 0