December 2016 - Boost - lists.preview.boost.org

Re: [boost] Asynchronous library now in Boost Library Incubator
by Christophe Henry 07 Dec '16

07 Dec '16

>boost.asynchronous looks to have many of the same features/functionality as HPX. >Can I ask why you chose to reimplement futures/lightweight thread pools/parallel STL/etc, >rather than working on improving HPX to suit your needs? I don't think the libraries have the same goals (or at least it seems to me so). Asynchronous is first of all an architecture tool focusing on organizing a whole application into thread worlds. The libraries also do not have the same design / tradeoffs. Asynchronous can make use of futures but encourages callbacks and tries to make these as safe as possible. The goal is to be as asynchronous (meaning non-blocking) as possible. It is no coincidence that I'm also the author of Boost.MSM. The encouraged design is a whole lot of state machines, each in its own thread (possibly sharing a thread), sending tasks to threadpools, TCP connections, etc. and waiting for callbacks. Future-based designs do not mix well with state machines' run-to-completion. The reason for the non-blocking focus is that I'm using the library at work in an industrial 24/7 application. When working on the production line, blocking is a disaster. If an algorithm goes astray, or simply takes longer than usually, blocking the production line results in huge costs. Crashes due to races also. Asynchronous provides ways to avoid both. >JB >Disclaimer : I have contributed o HPX, but I'm not one of the "German HPC Guys" :) . No hard feeling. I'm grateful for participation and the chance to explain the underlying ideas of the library. Thanks, Christophe

4 25

[serialization][test] failures in serialization causing regressions
by Barend Gehrels 06 Dec '16

06 Dec '16

> The regression matrix in develop of Boost.Geometry contains many new > compilation failures due to: > > > 1) : D:\t08\run\boost_root\boost/serialization/array.hpp(25) : error C2039: 'array' : is not a member of 'std' > > 2): > ../boost/numeric/ublas/storage.hpp: In member function ???void boost::numeric::ublas::unbounded_array<T, ALLOC>::serialize(Archive&, unsigned int)???: > ../boost/numeric/ublas/storage.hpp:331:18: error: ???make_array??? is not a member of ???boost::serialization??? > ar & serialization::make_array(data_, s);and a > > > The master is still OK. > > Can this be solved? Ping! This is still the case! Robert, can you explain this? It is (of course) not only geometry failing, also ublas or (if there are) other libraries with ublas dependencies: http://www.boost.org/development/tests/develop/developer/numeric-ublas.html All yellow! Thanks, Barend

3 8

Re: [boost] [process] Formal Review starts today, 27 October
by Bjorn Reese 04 Dec '16

04 Dec '16

On 10/27/2016 08:26 AM, Antony Polukhin wrote: > - Whether you believe the library should be accepted into Boost > * Conditions for acceptance I withhold judgement for now to give the author an opportunity to respond. At the moment I am unsure whether to vote for conditional acceptance or rejection. > - Your knowledge of the problem domain Advanced on Unix. > You are strongly encouraged to also provide additional information: > - What is your evaluation of the library's: > * Design I have a couple of serious concerns. My first concern is about the use of tagged arguments, and especially the redirection operators, in constructors. Instead I propose that the library uses the Named Parameter idiom (see [1], not to be confused with what Niall labelled "tagged name parameter" idiom.) The bp::child object should not execute in the constructor, but rather via operator(). That way we can use the named parameter idiom to build up redirections and similar modifications. For example: auto compile = bp::child("c++").output(bp::null); compile("main.cpp"); // Executes My second concern is about deadlocking. The FAQ shows two ways of deadlocking. I would expect the latter (the ls example) to work without further ado because that is how any newcomer would write their code. My third concern is about the chaining of pipes between processes. The tutorial shows an example where the output of the nm command is forwarded to the c++filt command. However, the forwarding is not done directly from the nm process to the c++filt process, but goes via the parent process. This can incur a performance degradation when nm generates much output. Is it possible to chain them directly? Besides "nm myfile | c++filt" that is. I also have the following minor comments / questions. Is it possible to redirect stderr to stdout? (like 2>&1). The reference documentation only mentions this under asynchronous pipe output. The bp::starting_dir should be renamed to bp::current_path to be consistent with Boost.Filesystem. Consider renaming bp::spawn to avoid confusion with boost::asio::spawn which is how you launch coroutines with Boost.Asio. I am missing a this_process::get_id() function to return the current process identifier. > * Implementation My major concern with the implementation is that the library is only header-only, so I cannot compile it as a static library if I wish to. There are several Boost header-only libraries that allows you to do that. The reason why I find this important for Boost.Process is because I want to avoid getting platform-specific C header files included into my project when I include a Boost.Process header. The library should throw its own process_error exception which inherits from system_error, similar to Boost.Filesystem. I did not investigate the implementation in greater detail. > * Documentation The documentation is lacking an entire chapter with an overall description of the library and the concepts used. For instance: * What are processes and pipes? * What are the possible invocation mechanisms? Some of this is mentioned in the Design Rationale, but that is not the place where developers seek their information. * How to set search paths or the current working directory. * How to use environment variables. * this_process Tutorial / Synchronous I/O: What is the default behaviour when I/O redirections are not specified? Tutorial / Asynchronous I/O: Consider including an example with boost::asio::spawn as a way to obtain the yield context. Reference / child: Does not mention how the life-time of the object and the child process are related. For instance, the documentation on the destructor only mentions that it is a destructor. Reference / cmd: bp::cmd will parse the input string as a sequence of space-separated arguments. Is it spaces or whitespaces? Is it possible to have escapes for arguments that do contain spaces? > * Tests Did not investigate. > * Usefulness A process library is a very useful addition to Boost. > - Did you attempt to use the library? If so: > * Which compiler(s) > * What was the experience? Any problems? Did not try. > - How much effort did you put into your evaluation of the review? About 4-5 hours (one of which was gratuitously donated by daylight saving time.) [1] https://en.wikibooks.org/wiki/More_C%2B%2B_Idioms/Named_Parameter

11 49

API/ABI changes report for Boost
by Ponomarenko Andrey 02 Dec '16

02 Dec '16

Hello, The report on API/ABI changes in the latest releases of Boost: https://abi-laboratory.pro/tracker/timeline/boost/ The report is generated by the open-source abi-tracker tool and updated three times a week. Hope it will be helpful for users and maintainers of the library. Thank you.

1 0

[poly_collection] Request for comments: fast polymorphic collections
by Joaquin M López Muñoz 02 Dec '16

02 Dec '16

A couple of years ago I wrote two blog entries about high-performance data structures for polymorphic objects: http://bannalia.blogspot.com/2014/05/fast-polymorphic-collections.html http://bannalia.blogspot.com/2014/05/fast-polymorphic-collections-with.html that spurred some interest in the community. Now I've had the time to turn these ideas into something resembling a full-fledged library: https://github.com/joaquintides/poly_collection http://rawgit.com/joaquintides/poly_collection/website/doc/html/index.html and I'd be very grateful if people could give their opinion on usefulness, design, etc. so as to determine potential interest in an eventual official submission to Boost. Thank you, Joaquín M López Muñoz

6 21

Re: [boost] Asynchronous library now in Boost Library Incubator
by Christophe Henry 01 Dec '16

01 Dec '16

>After having a deeper dive into the documentation and examples, one >pattern I see popping up quite >frequently is this: >future<R> fut = my_promise.get_future(); >post_callback(some_work, [this](expected<R> r) { >this->my_promise.set_value(r.get()); }); >return fut; This is simply to make examples easier to run (it finishes the example). It demonstrates that the only one which blocks in an application built on Asynchronous is main(). This might be more confusing than useful and I probably should get rid of this. >Is there any performance/usability drawback from this (might have gotten >the syntax wrong here): >return post_future(pool, some_work); > >Why is the former to be preferred to the latter, or vice versa? post_future is the preferred for simple applications or examples. It is similar to std::async (without the problem of blocking futures and with different continuations). post_callback is the design the library seeks to promote for bigger applications: post a task to a threadpool, get a callback posted to the queue of the caller's thread world. >Regarding the documentation overall, >it seems to miss a proper reference section. >For example, boost::asynchronous::expected is used almost >everywhere, but not documented. My bad. Will be corrected. Thanks, Christophe

2 2

Asynchronous library now in Boost Library Incubator
by Christophe Henry 01 Dec '16

01 Dec '16

Dear Community, The Asynchronous library is now listed in the Boost Library Incubator ( http://blincubator.com/bi_library/asynchronous-2/?gform_post_id=1654). I took a lot of time since first presenting it in C++ Now 2014 in order to extend the scope and intensively test the library. The library has been in use in production code for already 1.5 years, has survived billions of calls and usage 24/7, so its core is stable and quite tested. All algorithms have not been used though so that bugs are possible there. The reason for using the Incubator is, the library is quite big and 2 weeks review will not be sufficient so I hope for comments, discussions or pre-reviews earlier on. Asynchronous is first of all an architecture tool. It allows organizing a complete application into Thread Worlds, each world having the possibility to use the same or different threadpools for long-lasting tasks. The library provides an implementation of the Active Object pattern, extended to allow many Active Objects to live in the same World. It provides several Threadpools and many parallel algorithms making use of them. And most important of all, it allows simple, blocking-free asynchronous programming based on thread-safe callbacks. It also provides very useful logging and HTML diagnostics (see https://htmlpreview.github.io/?https://raw.githubusercontent.com/henry-ch/a…) to allow better understanding of code. I look forward to an interesting and lively discussion. Christophe

6 7

Upcoming review of the Synapse library starting tomorrow
by Edward Diener 01 Dec '16

01 Dec '16

The formal review of the Synapse library by Emil Dotchevski starts tomorrow, December 2 and is scheduled to continue through December 11. I will be making an official announcement tomorrow, but I want to give the following information today for anyone wishing to acquaint themselves with the library. As explained previously by Emil Dotchevski when requesting that Synapse be put on the Boost Review list, Synapse is a signal programming library, similar to Boost Signals2 and the signals-slots system in Qt. The main difference is that Synapse is non-intrusive: the address of any object of any static type whatsoever can be passed to synapse::emit to emit a signal. This makes it possible to emit Synapse signals from objects of third-party types as well as system objects (e.g. standard FILE pointers, HWNDs, etc.) or any other object that can be converted to a pointer. The library has been formatted to fit the Boost directory and namespace structure. To get Synapse, clone 'https://github.com/zajo/boost-synapse.git' into a directory called 'synapse' under your boost/libs directory. See the tutorial at http://zajo.github.io/boost-synapse/Tutorial.html, or read full documentation at http://zajo.github.io/boost-synapse/index.html. The documentaion can also be viewed from the local clone at boost/libs/synapse/doc/index.html. I would urge anyone who may be interested in signal/slot processing in C++ to take a good look at the Synapse library. Edward Diener, review manager

1 0

Re: [boost] Asynchronous library now in Boost Library Incubator
by Christophe Henry 01 Dec '16

01 Dec '16

>I'm trying to digest the docs for this monster work. > >One things that strikes me immediately is that the doc preparation >does not use code snippets, and so code is in boring old >black'n'white rather than colored syntax. >For me, that makes it significantly harder to read (and understand) the code examples. Ok, I'll see if I can improve this. >It also means that one doesn't have the assurance that 'What You See is What Compiles'. For most features, the doc provides a link to an example, which I regularly compile. >(I've also having trouble finding things in all the plentiful 'good stuff'. >But that's a common (and largely still unsolved) >problem with all documentation.) >Paul Please let me know what is the "good stuff" for you and I'll make it more visible. Cheers, Christophe

2 1

[compute] Remove boost::chrono lib dependencie
by Viktor Sehr 01 Dec '16

01 Dec '16

Hi, Would it be possible to remove the dependency for boost/chrono is boost/compute/detail/duration.hpp. As it already supports std::chrono via a macro, couldn't the inclusion of boost/chrono.hpp be excluded? As it is now, boost compute is dependent on having libboost_chrono-vc140-mt-gd-1_62.lib built. /Viktor

4 4