January 2018 - Boost - lists.preview.boost.org

[atomic] (op)_and_test naming
by Andrey Semashev 30 Jan '18

30 Jan '18

Hi, I would like to ask for the community opinion on the naming of the (op)_and_test functions that appeared in Boost.Atomic 1.66. As you probably know, 1.66 included a few new atomic operations, some of which have the name (op)_and_test (e.g. add_and_test, sub_and_test, etc.) These operations perform the (op) on the atomic value and test if the result is zero, so that instead of this: if (a.fetch_sub(1) - 1 == 0) one could write this: if (a.sub_and_test(1)) Also, the generated code for these operations can potentially be more efficient. Recently I received this request: https://github.com/boostorg/atomic/issues/11 which basically asks to change the result of the functions to the opposite. I can see that the current naming could be confusing, especially given that other operations like bit_test_and_set return the previous bit value (i.e. true if the bit was 1). That would be a breaking change, but the new operations were announced as experimental in the release notes, so if there is consensus that the result should indeed be changed, I'm ready to comply. Another solution is change (op)_and_test naming to something else that is more clear. In that case I could deprecate the (op)_and_test variants and provide the functionality under the new names. I would welcome suggestions for a better naming scheme, if you feel this is the right way. My only ask is that the new names be concise, if possible. Lastly, if you think the current naming is fine or there is another way to resolve the confusion, please comment as well. I would appreciate your opinion in any case. Thanks.

5 29

[review] Outcome v2: Review Period Concluded (Fri-19-Jan through Sun-28-Jan, 2018)
by charleyb123 . 29 Jan '18

29 Jan '18

Hi, Everyone-- The Boost Review Period for the formal review of Niall Douglas' Outcome (v2) library is concluded (having run Fri-19-Jan through Sun-28-Jan, 2018). Reviews were posted from: - Vinícius dos Santos Oliveira (Fri-26-Jan-2018) - Andrzej Krzemienski (Fri-26-Jan-2018) - Bjorn Reese (Sun-28-Jan-2018) - Daniela Engart (Sun-28-Jan-2018) - John P Fletcher (Sun-28-Jan-2018) In addition, online discussions included comments, observations, and critiques from a number of other community participants. It will take some time to review the discussion, make a final determination, and prepare the Review Report. Please: - Contact me (directly or on-list) if I missed identification of your review during the review period; - Contact me (directly or on-list) if you have any concerns; or to raise any salient observations or "final thoughts" Thank you! --charley Boost.Outcome(v2) Review Manager

1 0

[gsoc18] Interested in adding statistics and machine learning functions to Boost.uBLAS and Seeking for mentors
by Kanch 29 Jan '18

29 Jan '18

Dear community: My name is Long Zhang,currently in my third year of college. As subject suggests, I'm interested in adding statistics and machine learning functions to Boost.uBLAS library. I've basic knowledge of statistics (which I have been thought in school) and Machine Learning, I learned Machine Learning by Andrew Ng on Coursera, and I have tried to implement algorithms like PCA,LDA,K-means in Python, therefore, I think it won't be a problem if I have to implements it in C++. BTW, I have use C++ since my Senior High. Except components mentioned in wiki, I think I can also implement algorithms like KNN,Logistic Regression, even SVM during the summer. I also understand the efficiency of these algorithms is what matters, I will try hard to implement it as fast as it can. I know I have do some kind of programming competency test. When should I start? I mean I haven't find a mentor, don't know who should I contact if I finish the test. Can someone mentoring me? We can then talk about the test and code I used written in C++(open source code in C++) to check if I'm qualified to this project. Thank you. Best, Long -- Best regards, *Long Zhang (Kanch) <http://akakanch.com> *＠ Chengdu University of Information Technology 2015 Computer Science Major Mail:kanchisme@gmail.com No.24 Block 1, Xuefu Road, Chengdu, Sichuan, China

2 2

[outcome] Formal review
by Bjorn Reese 29 Jan '18

29 Jan '18

I. DESIGN --------- Boost.Outcome provides a useful alternative / complement to exceptions, and its design is clean. I like naming of functions with wide and narrow contracts. I suggest that result<> is renamed to basic_result<>, which will then be the customizable template class, and result<> then becomes as specialzation (template alias) that does the same as the current default result<>. This makes result<> an uncustomizable companion to std::error_code, which is useful when defining interoperable APIs. One design decision that I disagree with is the choice of sum types rather than product types for internal storage. I would rather have a lower run-time footprint at the expense of increased compile times. II. IMPLEMENTATION ------------------ Looks readable and solid. There may be an overemphasis on constexpr; I would rather have had support for C++11. BOOST_OUTCOME_TRY_UNIQUE_NAME generates identifiers starting with __t but according to the C++ standard [lex.name/3.1] "each identifier that contains a double underscore __ [...] is reserved to the implementation for any use", where implementation means the C++ compiler. III. DOCUMENTATION ------------------ The documentation is well-written, with a tutorial that is easy to understand. The API Reference could be improved though. 1. What are the requirements for ValueOrError? Looking up ValueOrError brings us to convert::value_or_error, which only states that it "matches the ValueOrError concept." The rather convoluted enable_if uses ValueOrError<U> which uses detail::ValueOrError<U> and here the trail disappears. The convert.hpp header file does contain the missing explanations though. 2. What are the requirements for EC and EP? EC and EP are used informally throughout the documentation, but are described in the Custom Payloads tutorial. In the API Reference they are suddenly called S and P instead. Please use a consistent naming throughout the documentation; preferably using more descriptive names, such as OutcomeErrorCode and OutcomeExceptionPtr. It would furthermore be helpful if they were described as concepts outside the tutorial. 3. The descriptions of value_type_if_enabled and error_type_if_enabled are obscure, and it is still not clear to me if they have any relevance for the library user. If they do, then their descriptions should be more clear; if they do not, then they should not be included in the documentation. 4. The documentation of OUTCOME_TRYX contains the full macro implementation which is not really helpful to users. This macro could possibly be expressed in terms of other macros. 5. noexcept('hidden') is used in several places. Does this mean that the noexcept condition is unspecified? If so, then it is more customary to use "unspecified" in italic. IV. MISC -------- I spent 4-5 hours reading the documentation and looking at the code. I did not compile it. V. VERDIC --------- I vote for the conditional acceptance of Boost.Outcome. The conditions are: 1. The documentation should be integratable into the Boost infrastructure. 2. People must be able to submit patches to the Boost version of the library without having to convert those patches to the standalone version.

2 1

Fwd: Fwd: [review] Review of Outcome v2 (Fri-19-Jan to Sun-28-Jan, 2018)
by degski 28 Jan '18

28 Jan '18

---------- Forwarded message ---------- From: degski <degski(a)gmail.com> Date: 28 January 2018 at 17:11 Subject: Re: [boost] Fwd: [review] Review of Outcome v2 (Fri-19-Jan to Sun-28-Jan, 2018) To: Vinícius dos Santos Oliveira <vini.ipsmaker(a)gmail.com> On 28 January 2018 at 16:44, Vinícius dos Santos Oliveira < vini.ipsmaker(a)gmail.com> wrote: > Can you elaborate on your point? If I'm left alone I'll miss your point. > I was mostly just responding to your example (which you asked me to do), and how I would do that, without having to deal with exceptions and/or outcomes. But, the more general point is, which Spirit in this case does very well, is to deal with potential problems at the source, in an explicit (comprehensive) way, without propagation of errors or exceptions. > It's becoming confusing the more you go. > It's probably because I'm rowing against the tide here. degski

2 1

Fwd: [review] Review of Outcome v2 (Fri-19-Jan to Sun-28-Jan, 2018)
by degski 28 Jan '18

28 Jan '18

Accidentally sent privately, forwarded below: ---------- Forwarded message ---------- From: degski <degski(a)gmail.com> Date: 28 January 2018 at 15:46 Subject: Re: [boost] [review] Review of Outcome v2 (Fri-19-Jan to Sun-28-Jan, 2018) To: Vinícius dos Santos Oliveira <vini.ipsmaker(a)gmail.com> On 28 January 2018 at 14:10, Vinícius dos Santos Oliveira < vini.ipsmaker(a)gmail.com> wrote: > 2018-01-28 12:12 GMT-03:00 degski via Boost <boost(a)lists.boost.org>: > >> 1. Use of outcome is more manual and cumbersome, i.e. more error-prone. >> > > What is complex about the following? > <snipped the code> I'd just use boost::cnv::spirit and specify explicitly what happens when conversion didn't work... much shorter, very well optimized (the best) when it does work (the conversion)... > Server side code... > There's more than servers in life. Handling f.e. O-O-M-situations is not gonna help me on my lap-top... a re-think and re-write of my app is in this case in order, though... or a trip to the shop and buy a bigger stick (both the "slow path"). > If the error is exceptional or not, it depends on context, not on the > algorithm (e.g. connection failure on game client code and game server code > and both of them being backed up by the same functions). Boost.Outcome > let's you do just that by converting an error into an exception. > Yeah, that's another discussion, isn't it? errors vs. exceptions. Should exceptions be exceptional, or getting back to your conversion example, should we use exceptions to indicate failure as this is the case in Boost::Lexical_cast? I know the answer I like best, and moved on to Boost.Spirit. degski

2 3

[spirit-x3]What's use case for x3::variant<OnlyOneComponent>?
by Larry Evans 28 Jan '18

28 Jan '18

Since: http://www.boost.org/doc/libs/1_65_0/libs/spirit/doc/html/spirit/qi/quick_r… doesn't mention a unary variant: variant<A> and the comments in: https://github.com/boostorg/spirit/blob/develop/include/boost/spirit/home/q… indicates variant<T> is collapsed to T, why isn't the x3::variant here: https://github.com/boostorg/spirit/blob/develop/include/boost/spirit/home/x… declared as: template <typename L, typename R, typename ...Types> struct variant instead of: template <typename ...Types> struct variant which allows sizeof...(Types)==0 instead of the minimum size of 2? That would disallow the primary and factor using's here: https://stackoverflow.com/questions/43791079/x3-linker-error-with-separate-… which make no sense because, for example, primary_def is *not* an alternative parser. -regards. Larry

1 0

Google Summer of Code with Boost
by David Bellot 28 Jan '18

28 Jan '18

Hi everyone, still a few days before we submit our proposal to Google. Don't loose the chance to become a mentor for Boost. It shines on your CV, it is a great opportunity to work with brilliant students around the world, it makes you an official C++ world-wide expert. So far, we have a nice but little list of proposals here: https://github.com/boostorg/boost/wiki/Boost-Google-Summer-of-Code-2018 It would be good to have a bit more before I go and submit to Google. This will increase our chances to be selected again this year. For Boost it's a great funding worth several thousand of dollars every year. We just don't want to miss it. Best regards, David

2 7

[gsoc18] New Astronomy Library
by Pranam Lashkari 28 Jan '18

28 Jan '18

Boost has developed many libraries which is useful in the field of science. I would like to propose a library for "Astronomy" as a student. I have been an amateur astronomer since last 8 years and being a Computer Engineering student I will be able to develop this library with little mentoring. I have some ideas about what can be integrated into it, which I can propose if the proposal seems promising. If so are there anyone who can help me out with this and mentor me? Thank you, Pranam Lashkari

4 18

[gsoc18] ublas project 3
by Aperjump 28 Jan '18

28 Jan '18

I'm Wei Wang, a CS master student focusing on high-performance computing field. `boost::ublas` project 3 adding GPU computation interests me a lot and I'd like to help add this feature to ublas. I find this project was also on last years' list and I'm curious if anyone did this before or on which stage he/she has finished. I've already read the initial source code of `ublas` in `boost 1.29`(I also read 1.66 API, and found it add one concept `container`, which used to be `bounded_array` and `unbounded_array`). I wrote a passage describing its template parameter deduction relationships. Besides, I wrote a series blogs teaching how to use openCL efficiently with proper data partition and memory usage. This is my first time participating in GSOC, and I'm a bit of confused on following question: 1. Integrating openCL requires preparing for context, command_queue, event and other "environment objects", should they also be included in this lib? 2. Take matrix matrix multiplication A*B for example. The last stage before matrix copy assignment is in `matrix_matrix_prob` class and its evaluation requires loop through all items on both matrix. If I want to add GPU compute features, I need to launch kernel for each computation expression at this step, but it seems to be contradictory to `ublas`'s lazy evaluation rationale. How to reconcile that? 3. What should I implement in the competency matrix class? Just integer matrix or template matrix class?Should I support current `ublas` interface(those typedefs and traits)? Best regards, Wei Wang -- Sent from: http://boost.2283326.n4.nabble.com/Boost-Dev-f2600599.html

2 1