March 2020 - Boost - lists.preview.boost.org

Adding my sorting algorithm, ska_sort, to boost, attempt 2
by Malte Skarupke 24 Nov '21

24 Nov '21

Hi everyone, I started a discussion about this in October, but at the time the maintainer of the sort library was busy, then I was busy, so now it's January... I wrote a generalized version of radix sort a couple years ago and gave a talk about it here: https://www.youtube.com/watch?v=zqs87a_7zxw I would like to get a new version of that sorting algorithm into boost. It's faster than what I showed several years ago. It's also even more general and has a new interface that should be easier to extend. I uploaded the code here: https://github.com/skarupke/sort/tree/master The documentation is also already updated with new benchmark results. Let me know what's required to get this into boost. Would you like this in the form of a pull request? One of the big things that probably needs review is the interface, since radix sorting works differently than comparison based sorting. For most normal uses you just need to provide a sort key, but more complicated cases can get more complicated. I tried very hard to make everything easy, but I'd also be curious if others can come up with better ways of doing the interface. Thanks, Malte -- Malte Skarupke malteskarupke(a)fastmail.fm

8 10

Optional library
by anshu khare 27 May '20

27 May '20

Hi mentors I was going through the optional library of boost. I was facing some problem in understanding the code base. In lines from 94 to 97 in optional.hpp, the code is as follows: struct in_place_init_if_t { Struct init_tag{}; explicit in_place_init_if_t(init_tag) {} //not able to understand this line }; I understand that we are creating a structure in_place_init_if_t and inside that we are creating another structure init_tag and then we are calling the constructor of in_place_init_if_t. But I am not able to understand that what does it the constructor do when we pass init_tag as an argument to it. Is it same as we are creating a structure Car and then creating another structure wheels inside it and passing wheels as an argument to the constructor of car as follows: struct car { Struct wheels{}; explicit car(wheels) {} //not able to understand this line }; Could you please tell what is the significance of passing wheels as an argument in this counter example?

2 1

[Range] Transform in place
by Frederick Gotham 11 May '20

11 May '20

Earlier today I wrote the following two lines of code: auto const my_range = v | boost::adaptors::sliced(1,15) | boost::adaptors::strided(2); boost::range::transform( my_range, my_range.begin(), std::bit_not<uint8_t>() ); However it would have been nice as a one-liner: boost::range::transform( v | boost::adaptors::sliced(1,15) | boost::adaptors::strided(2), std::bit_not<uint8_t>() ); Why isn't there a form of "range::transform" that can transform in place (and therefore takes exactly two arguments)?

3 2

Embarcadero C++ and Boost libraries
by Edward Diener 01 May '20

01 May '20

Embarcardero ( formerly Codegear, formerly Borland ) has had for about five years C++ compilers based on clang, along with the venerable bcc32, which is the Borland compiler of old through a number of transformations. The clang-based Windows compilers are bcc32c/bcc32x for 32-bit development and bcc64 for 64-bit development. The only difference between bcc32c and bcc32x is that the former takes the same command line flags as their legacy bcc32 compiler while the latter, like bcc64, takes clang command line flags. The clang-based C++ compilers are based on clang-5.0, and use the Dinkumware library. The clang based compilers still predefine __BORLANDC__ and __CODEGEARC__ but also predefine __clang__. The latter is the way to distinguish in preprocessor code between the bcc32 compiler and the clang-based compilers. Despite numerous workarounds in Boost code based on the predefined macros __BORLANDC__ and __CODEGEARC__ the bcc32 compiler, not being very C++ compliant even at the C++98/C++03 level at which it operates, has not really worked for Boost since about Boost 1.39. It is useless for modern day Boost. Therefore I have made efforts, in contact with some Embarcadero C++ representatives, to integrate the clang-based compilers into Boost, especially since there is a free non-commercial offering by Embarcadero called C++ Builder Community Edition at https://www.embarcadero.com/products/cbuilder/starter which any Boost developer could use to test these compilers if they wished. I am doing this in the hopes that this may become effective for Boost 1.74.0. Doing this involves: 1) Creating an embarcadero.jam toolset, based on the clang-linux.jam toolset, which supports bcc32x and bcc64. Updating the borland.jam toolset to also support bcc32c. I have done these and the PR has been merged to 'develop' on Boost Build. 2) Updating Boost Config to support the clang-based compilers. I have done this and the PR has been merged to 'develop' on Boost Config. 3) In the update to Boost Config, I have created these macros: a) BOOST_BORLANDC = __BORLANDC__ with __clang__ not defined b) BOOST_CODEGEARC = __CODEGEARC__ with __clang__ not defined c) BOOST_EMBTC = Embarcadero C++ compilers with __clang__ defined Therefore what is really needed in Boost libraries is to change all the code which currently uses __BORLANDC__ to BOOST_BORLANDC and all the code which uses __CODEGEARC__ to BOOST_CODEGEARC when Boost config has been included. When in the rarer case that Boost config has not been included, then __BORLANDC__ logic becomes __BORLANDC__ and !__clang__ and __CODEGEARC__ logic becomes __CODEGEARC__ and !__clang__. I am workiog on doing this for the numerous Boost libraries with workarounds nased on __BORLANDC__ and __CODEGEARC__ in Boost code, and will create PRs for those libraries. I only ask that library maintainers will merge the PRs I create once they have passed their CI tests, since the PRs are almost all just simple transformations regarding the __BORLANDC__ and __CODEGEARC__ in the code. If any library maintainer wants to so this on their own in the 'develop' branch it would be appreciated. Further points: The Embarcadero C++ clang-based compilers are not as good as clang itself, which is at release 9, gcc, or vc++. But they are good enough for the majority of Boost code. The Embarcadero C++ clang-based compilers support only C++17. The documentation implies that they also support C++11 and C++14 but that is not so. But since Boost libraries written for a C++ standard level are supposed to work flawlessly at a higher level this is no impediment for C++ Builder programmers using the Embarcadero C++ compilers in current Boost. I have reported a number of bugs, based on my Boost testing, for the Embarcadero C++ compilers to Embarcadero themselves in their bug reporting system. I believe that any registered user of their free community edition can report bugs to Embarcadero C++ at their bug reporting system at https://quality.embarcadero.com. Hopefully these reports will improve the compilers. It would really be nice, once the Embarcadero C++ clang-based compilers are integrated into current Boost, if regression testers could test their compilers. I will try to do so myself if I have the time and resources. I have discussed this with some Embarcadero C++ representatives about adding the Embarcadero clang-based C++ compilers to one or more CI testing services, but currently there are not any which support them of which I know. Besides the clang-based Windows compiler Embarcadero also has clang-based C++ compilers for 32-bit and 64-bit iOS ( Apple ) and for Android ( Google ). The code in Boost Config distinguishes between these environments, if necessary, with further macros based off of the mnemonic BOOST_EMBTC. It would be nice to add these configurations to embarcadero.jam once the rest of the work is done. But I have no idea how to test any of these other compilers as they are not part of the Windows based Embarcardero C++ Builder Community Edition.

2 4

Re: [boost] [review][STLInterfaces] STLInterfaces review results (ready for merge)
by Barrett Adair 30 Apr '20

30 Apr '20

On Thu, Dec 26, 2019 at 2:31 AM Barrett Adair <barrettellisadair(a)gmail.com> wrote: > > Many thanks are owed to all who participated in the formal review of Zach Laine's STLInterfaces library, especially during this busy holiday season: > > * Hadriel Kaplan - ACCEPT > * Andreas Wass - ACCEPT > * Rainer Deyke - ACCEPT > * Arthur Gruzauskas - ACCEPT > * Krystian Stasiowski - ACCEPT > * Daniela Engert - CONDITIONALLY ACCEPT > * Gavin Lambert - Technical comments > > (snip) > > The reviews clearly demonstrate a consensus. Accordingly, I'm pleased to announce that STLInterfaces is CONDITIONALLY ACCEPTED as a Boost library. > > Conditions for acceptance: > 1. Remove optional dependency on cmcstl2 (issue #27) > 2. Add Boost license comments to all files, including appveyor.yml, .travis.yml, README.md, */CMakeLists.txt (issue #26) > 3. Support Boost.Build (issue #25) > 4. Replace googletest with Boost equivalent (issue #24) > 5. Remove usage of non-existent v2_dtl members in container_interface.hpp (issue #22) > 6. Include <tuple> for std::tie usage in C++14 mode (issue #21) > 7. Fix dangling parens in container_inteface.hpp:731 (issue #20) > 8. Make Concepts usage C++20-conformant (issue #13) > 9. Remove the destructor in container_interface (issue #12) > 10. Add tests for non-copyable value types (issue #30) > > (snip) I've confirmed that STLInterfaces (https://github.com/tzlaine/stl_interfaces) has met the conditions for acceptance. Review wizards, please reply with any objections. Merge ticket here: https://github.com/boostorg/admin/issues/174 Thanks, Barrett

4 5

MySQL ASIO library
by Ruben Perez 22 Apr '20

22 Apr '20

Hi all, I have been writing an ASIO-based client for MySQL, trying to mimic what Beast is to HTTP. It currently supports SQL queries and prepared statements. It can be viewed here: https://github.com/anarthal/mysql-asio Do you guys think this has the potential to be useful or become part of Boost long term? Any feedback is very welcome. Thanks, Ruben (Anarthal).

6 12

Re: [boost] MySQL ASIO library
by Andrey Semashev 14 Apr '20

14 Apr '20

[Sending this reply to list as opposed to personal email] On Wed, Mar 4, 2020 at 3:39 PM Richard Hodges <hodges.r(a)gmail.com> wrote: > On Wed, 4 Mar 2020 at 13:25, Andrey Semashev <andrey.semashev(a)gmail.com> wrote: >> >> On Wed, Mar 4, 2020 at 1:50 PM Richard Hodges via Boost >> <boost(a)lists.boost.org> wrote: >> > >> > > Why just MySQL? >> > >> > Getting something useful released is more important than getting something >> > perfect released. >> > >> > If Ruben has a ready-to-go solution for MySQL, why not make it available to >> > users? >> >> Making it available to users is not the same as making it part of >> Boost. Boost is known for general purpose libraries, as well as more >> domain-focused solutions, but it is not a place for wrappers around >> specific other libraries. Let alone, when the said libraries already >> have C/C++ API. > > Boost also has mpi, regex, asio::ssl and python. What are these if not wrappers around common c libraries? I'm not sure about Boost.MPI, but I thought it was not a wrapper of a single library, but of a standard API that can be implemented by different libraries. Boost.Regex is not a wrapper at all; it implements regular expressions from scratch. asio::ssl is not a library but a plugin for Boost.ASIO that provides one small piece of functionality compared to the rest of the library. Boost.Python is probably closest to an exception, although it is a binding to another language (not a library), which arguably only has one C API and implementation. Yes, there is CPython, but I don't believe it offers a C API. >> > It can always be complemented or extended with Oracle, SQLite, ODBC etc etc >> > later. >> >> If the proposed library offers a stable and flexible API that can be >> backed by multiple implementations then by all means - that would be a >> very interesting proposal indeed. But the author has to demonstrate >> that the proposed user API can in fact be supported by more than one >> backend, so at least two backends need to be presented, and preferably >> with guidelines and infrastructure for adding more. > > Mpi, regex and python would dispute this arbitrary restriction. I don't think so, as per above. > Boost suffers from a lack of contribution. Is there any value in making contribution difficult? I don't think the amount of contributions by itself is the goal. There has to be value associated with the contribution. I just don't think a C++ wrapper of a specific library has enough value.

11 28

boost [math] excessive warnings via [lexical_cast] from [date_time] (and others) for clang 3.4?
by Jeff Garland 08 Apr '20

08 Apr '20

In the regression tables the following warnings are coming out of math config. I'm trying to understand what this means given that 1) the tests compile and execute fine, 2) the deadline is up, and 3) several of the cases are compiled in c++11 mode (is it pre conformance bc it's clang3.4?). I've seen this warning in other places that include lexical_cast as well. So sure I could go and define the macro to quiet this, but seems like that should be done closest to the source. In other words maybe math should only tell us that in headers that have a real issue? Or will this soon be a real issue? ../boost/math/tools/config.hpp:42:9: warning: CAUTION: One or more C++11 features were found to be unavailable [-W#pragma-messages] ../boost/math/tools/config.hpp:43:9: warning: CAUTION: Compiling Boost.Math in pre-C++11 conformance modes is now deprecated and will be removed from March 2021. [-W#pragma-messages] ../boost/math/tools/config.hpp:44:9: warning: CAUTION: Define BOOST_MATH_DISABLE_DEPRECATED_03_WARNING to suppress this message. [-W#pragma-messages] Full examples https://www.boost.org/development/tests/develop/developer/date_time.html https://www.boost.org/development/tests/develop/output/teeks99-dkr-dc3-4-98…

5 23

[Performance of circular_buffer vs Queue][Multithreading may be broken]
by Jan Hafer 04 Apr '20

04 Apr '20

Dear Boost developer, I found the circular buffer to be slower than std::queue for multithreading except in the case of a single thread (where it is notably faster). I am using resize() to have the expected size filled as in vector. Is this behavior expected or documented anywhere? What kind of speedup is to be expected against the std::queue implementation? If there is none, then what is the purpose of circular buffer? I do also have broken code on a multithreaded application that worked fine before. I do pass the circular buffer as reference via functions from file scope, which in combination with the threads (initially starting elsewhere) somehow broke the code. Do you do regular benchmarks against other implementations? I dont see linked graphs on your homepage. Regards, Jan

7 10

[gil] Review of move assignment operator for image class
by Mateusz Loskot 02 Apr '20

02 Apr '20

Hi, We're adding move semantics to some of GIL classes [1] and we'd like to receive feedback on the move-assignment operator for the image class [2] (copied below) that we've worked out so far if there are any major blunders to correct or gaps to fill for C++11 compliant implementation: image& operator=(image&& img) { if (this == std::addressof(img)) return *this; auto const exchange_memory = [](image& lhs, image& rhs) { lhs._memory = boost::exchange(rhs._memory, nullptr); lhs._align_in_bytes = boost::exchange(rhs._align_in_bytes, 0); lhs._allocated_bytes = boost::exchange(rhs._allocated_bytes, 0); lhs._view = boost::exchange(rhs._view, image::view_t{}); }; constexpr bool pocma = std::allocator_traits<Alloc>::propagate_on_container_move_assignment::value; if (pocma) { // non-sticky allocator, can adopt the memory, fast destruct_pixels(this->_view); this->deallocate(); this->_alloc = img._alloc; exchange_memory(*this, img); } else if (_alloc == img._alloc) { // allocator stuck to the rhs, but it's equivalent of ours, we can still adopt the memory destruct_pixels(_view); this->deallocate(); exchange_memory(*this, img); } else { // cannot propagate the allocator and cannot adopt the memory if (img._memory) { allocate_and_copy(img.dimensions(), img._view); destruct_pixels(img._view); img.deallocate(); img._view = image::view_t{}; } else { destruct_pixels(this->_view); this->deallocate(); this->_view = view_t{}; } } return *this; } One thing that is still missing is the noexcept with expression evaluated based on something like this template <class _Alloc> using is_noexcept = std::conditional_t < std::allocator_traits<_Alloc>::is_always_equal::value, _Equal_allocators, typename std::allocator_traits<_Alloc>::propagate_on_container_move_assignment::type >; but we there is no is_always_equal in C++11, so there are gaps to fill. Any comments are appreciated. [1] https://github.com/boostorg/gil/pull/457/ [2] https://github.com/boostorg/gil/blob/7d8df1c105e81070e30ef68ea7abb8e39111d6… Best regards, -- Mateusz Loskot, http://mateusz.loskot.net

1 1