- Boost - lists.preview.boost.org

Outcome/expected/etc/etc/etc
by Gottlob Frege 22 Jun '17

22 Jun '17

I've now read ALL the emails. Here's some thoughts, in no particular order. - I wonder, Niall, if your uses of the empty state could/should be compared to how people sometimes use -1 or 0, etc - ie "special values as errors". I typically don't like that pattern (yet I've done it anyhow - performance/laziness/etc - but typically only in localized cases, not in "APIs"). So I'm wary about this empty state thing. If we do have an empty state, I like it as a separate type - code needs to deal with it separately. I like the idea of taking a possibly-empty type, checking for empty, and creating a never-empty type, and passing that along so future code need not worry. - I think the narrow and wide contracts should be in the same type, not separate types. This is C++. I suggest "access" if you want long names. ie T & val = oc.access_value(); // possible UB (I almost also suggested "deref" but "deref_value" is wrong - you are deref'ing the outcome, not the value) - I agree with Vicente's concern that functions that behave differently need to be named differently. I haven't looked at the new (or old) API carefully enough, but as a (possibly hypothetical) example: wide_type::value() // returns value, or throw if no value narrow_type::value() // returns value, or UB if no value These functions are not the same, and should not be named the same. Since you thus need different names anyhow (ie value and access_value), why not put them both on the same API? I'm not yet sure if you end up with similar examples with possibly_empty_type::something() and never_empty_type::something(). - I see value in the idea that the narrow contract takes more typing, but I actually *like* using * and -> for the narrow contract because I have 30+ years of training in looking at * and -> and recognizing it as "this may be unsafe, examine the code to see how it is guaranteed". Boost.Optional specifically chose * for that very reason. (on the other hand, I have 0 training on spotting access_value(), but on the third hand, it is easy to grep for) - exception() returns error, error() doesn't return exception, etc: would it be better/easier/clearer if we introduced another name, such as "failure" or whatever? ie exception() only returns the stored exception, error() only returns the stored error(), and then has_failure() would return true in either case, failure_exception() returns exception, generates an exception from error if necessary, etc? - I dislike "exceptions are for exceptional circumstances". Yes it is a common thing, but it is usually wrong. Using Outcome to enable the pattern of "handle errors, but totally cancel the operation on exceptions" might actually turn out to be a useful pattern, but I think this is actually slightly different from the "exceptions are for exceptional cases" camp, as they typically use that chant to choose whether their API will use exceptions or errors, not both in a single API. So I agree with Emil - for most APIs, error-codes vs exceptions" is basically trying to handle the same thing - failure (or "disappointment"). Using both error-codes and exceptions in a single API is somewhat new. And interesting. (There is lots of code that almost-unknowingly uses error-codes and exceptions in the same API, but that is because they don't realize push_back throws or they pretend it doesn't, etc - it isn't really an API design choice) - this ring buffer thing sounds interesting. I haven't looked at it at all, so maybe this makes no sense: should it use thread-local? - I think std::expected should always throw something deriving from std::exception. So if E derives from std::exception, we can throw it. If E is exception_ptr we can (re)throw it. If E is error_code/system_error/etc we can figure out what to throw. If E is user-type, then we wrap it in bad_expected or whatever. (And/or allow user E-to-exception customization somehow). This sounds complicated, but it is easily summarized/taught as "the standard always throws std::exceptions" - I HATE the dual interface of std::filesystem. Worse, there is a chance it is setting a precedent for future std APIs (dual APIs have been proposed elsewhere). Let's please kill it before a dual API strikes again. However, the other problem with trying to "single-fy" the std::filesystem (or any other) API is that users who want error codes sometimes don't want the cost of the extra data (paths, etc). The only way to handle that may be with customizable callback functions that capture the error-info however you like, or something similar to callbacks. - I really look forward to Peter's never-empty variant - and have been hoping for one ever since std::variant was finalized. Not because I dislike std::variant, but just because I think the option should be there for those who need/want it. Regardless, std::variant, in my world, is never empty, since it is only empty when a move operator throws, which should be never and/or when I can't allocate 64 bytes, in which case I'm screwed anyhow. Some days I'm like "man just accept empty, it would be a simple API", but then I think "it is stupid for variant<int, double> to be empty". Other days I think "just double buffer when necessary" (I assume that's your direction Peter?), but then I think "I don't want double-buffering variant<list, vector> to go on/off based on whether I use MS std vs libc++ etc" and also "I don't want double buffering for cases that only happen in theory, not in practice". I like the trade-offs of std::variant. I do agree that Boost should have a variant2 with different choices than std::variant - I think that actually helps the standard - we can choose a direction and instead of telling people "tough luck" when they disagree, we can say "use boost variant2 when you want those properties". - Andrzej: > Actually, now I am thinking not about Boost.Outcome but specifying these > preconditions in the Standard. If you want to std::terminate upon some > condition rather than let the program do random things, you could say. > > *Requires:* `!r` with rescue action `std::terminate()`. This sounds a bit like the contracts proposal - where you get to define what happens when contracts are broken. I'm cautiously hopeful that contracts could help with LEWG's constant struggles with wide vs narrow. - I agree with those who were saying (or thinking) that Boost was - and should again be - the place where APIs go before going into the standard. APIs (mostly) shouldn't bypass boost and go to the standard directly. Or at least an API needs a similar user-base (both in scale and diversity - ie an internal google API has scale, but not necessarily diversity). Since I know a number of LEWG members agree with me on this, don't be surprised to see some push-back from the committee for proposals without that kind of backing, and also don't be surprised if Boost gets flooded with "the committee told me to come here first". Of course I don't speak for LEWG, so maybe none of that happens, we haven't had that discussion yet. I actually probably had more to say - I should have made notes as I was reading all... those... emails. Tony

10 75

[review] Review of Nowide (Unicode) starts today
by Frédéric Bron 22 Jun '17

22 Jun '17

Hi Everyone, The formal review of Artyom Beilis' Nowide library starts today and will last until Wed. 21st of June. Your participation is encouraged, as the proposed library is uncoupled, focused and rather small. Nowadays, everyone needs to handle Unicode but this is very difficult using only the standard library in a platform independant way. Nowide offers a very simple way to handle Unicode the same way on Windows/MacOS/Linux. Key features: * work with UTF-8 in your code, Nowide converts to OS encoding * Easy to use functions for converting UTF-8 to/from UTF-16 * A class to fixing argc, argc and env main parameters to use UTF-8 * UTF-8 aware functions: - stdio.h functions (fopen, freopen, remove, rename) - stdlib.h functions (system, getenv, setenv, unsetenv, putenv) - fstream (filebuf, fstream/ofstream/ifstream) - iostream (cout, cerr, clog, cin) Documentation: http://cppcms.com/files/nowide/html GitHub: https://github.com/artyom-beilis/nowide git clone https://github.com/artyom-beilis/nowide.git Latest tarballs: - to be unzipped in boost source: https://github.com/artyom-beilis/nowide/archive/master.zip - as a standalone library: http://cppcms.com/files/nowide/nowide_standalone.zip Nowide has standard boost layout. So, just put it into the boost source tree under libs/nowide directory for building and running tests. Or alternatively - for header only part of the library add the nowide/include path to the include path of your projects. Please post your comments and review to the boost mailing list (preferably), or privately to the Review Manager (to me ;-). Here are some questions you might want to answer in your review: - 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 most importantly: - 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. For more information about Boost Formal Review Process, see: http://www.boost.org/community/reviews.html Thank you very much for your time and efforts. Frédéric

11 70

[cmake] Minimum viable cmakeification for Boost
by Niall Douglas 22 Jun '17

22 Jun '17

I've finished the mockup which can be studied at: https://github.com/ned14/boost-bmv-cmake Just Boost.System was cmakeified. Extensive comments are throughout to explain every single line of cmake and why it's there. A small sample program is in the root of the repo to show how this cmake would be used by end user programs. Readme.md follows for those who don't like clicking on links: --- cut --- This is a mock up of a **bare minimum viable** cmakeification of the Boost C++ libraries for the community to study and evaluate. It only cmakeifies Boost.System alone, its upstream dependencies are stubbed/faked. A demo program which shows how one would link against Boost.System is in this directory (`example_client_program`) and it demos linking against boost::system as a header only library, as a static library and as a shared library. # Specific design goals of this mock up: - Modern, **highly reusable by unknown third party cmake** cmake3 only throughout. - Minimum possible cmake complexity, no cmake innovating, even at the cost of extra boilerplate. - No custom macros, functions, nor custom build logic. Only vanilla cmake3. Keep the learning curve for the build system as gentle as is possible. - **Strict** separation of non-fixed configuration and build. Child CMakeLists must only EVER define build and *fixed* configuration like "I need C++ 11 minimum". *Variable* configuration like naming, directory layout for outputs, optimisation flags etc is ALWAYS defined in the **rootmost** CMakeLists. NEVER in child CMakeLists. - All custom build logic should always be placed in rootlevel cmake scripts (command line programs written in cmake, runnable using `cmake -V`, these can take args etc) or rootlevel CMakeLists, or combinations thereof. # Stuff in there we might yet remove: - This cmake mockup explicitly documents dependencies between Boost libraries in cmake, and so when I link against say Boost.System, cmake will also link in the things Boost.System depends against. Moreover this is minimal, so ONLY the stuff minimally necessary to fulfil the need for Boost.System is built. - We allow those Boost libraries support it to present header only, static and dynamic linkage editions of themselves. This does complicate the CMakeLists however. # What we don't do: - Test the boundaries of cmake or push any limits of what's possible. - Bespoke libraries of common functions dragged in wherever. - Complex custom build logic which needs to be documented and learned off by library developers. - Do anything which gets in the way of CMakeLists reuse by any arbitrary third party cmake written by any Boost user. --- cut --- The code should be fairly self explanatory. Yes there are no tests support. We are limiting the scope of discussion to build only for now. Technical questions are welcome. Niall -- ned Productions Limited Consulting http://www.nedproductions.biz/ http://ie.linkedin.com/in/nialldouglas/

8 50

[review] Beast review coming up!
by Michael Caisse 22 Jun '17

22 Jun '17

The Boost formal review of Vinnie Falco’s Beast library will take place July 1 - 10, 2017. Beast is a C++ header-only library serving as a foundation for writing interoperable networking libraries by providing low-level HTTP/1, WebSocket, and networking protocol vocabulary types and algorithms using the consistent asynchronous model of Boost.Asio. This library will be familiar to programmers that have used Boost.Asio or the Networking TS. Users who wish to use asynchronous interfaces should already know how to create concurrent network programs using callbacks or coroutines. Beast is designed for: * Symmetry: Algorithms are role-agnostic; build clients, servers, or both. * Ease of Use: Boost.Asio users will immediately understand Beast. * Flexibility: Users make the important decisions such as buffer or thread management. * Performance: Build applications handling thousands of connections or more. * Basis for Further Abstraction. Components are well-suited for building upon. Beast can be found here: * Documentation : <http://vinniefalco.github.io/beast/> * Source code : <https://github.com/vinniefalco/Beast> * The FAQ answers real questions that come up : <http://vinniefalco.github.io/beast/beast/design_choices/faq.html> A lightning talk from CppCon 2016 introducing Beast can be found here: <https://www.youtube.com/watch?v=uJZgRcvPFwI> If you find an obvious bug or problem with the library, please open a GitHub Issue <https://github.com/vinniefalco/Beast/issues>. Vinnie has a track record of resolving issues quickly. Please feel free to take an early look at the library and ask questions. Maybe even click that github Star button! The actual review will be against a non-mutating branch starting July 1st. -- Michael Caisse Ciere Consulting ciere.com

2 1

[build] Use full compiler-version in library filenames
by Deniz Bahadir 22 Jun '17

22 Jun '17

Hi list, While reading Vinnie's mail (regarding the Boost library filenames which should additionally contain the address-model) I remembered another problem existing with library filenames. That is, that Boost libraries built with newer GCC versions (since GCC 5) never add the compiler-version into the library filenames. Building with GCC 5 and newer only adds "gcc" into the filename to determine the toolset, while building with earlier versions of GCC (e.g. GCC 4.9 or GCC 4.8) additionally adds the compiler's version (e.g. "gcc49" or "gcc48"). I assume this might be due to the fact that the version-scheme of GCC changed since GCC 5 and somehow the logic in Boost.Build no longer works because of this. But in order to install boost-libraries built by different GCC versions side-by-side in the same directory the version-number is really required in the filename. Maybe someone can fix this again? A bug-report on Github which addresses this issue already exists: https://github.com/boostorg/build/issues/191 Thanks in advance, Deniz -- BENOCS GmbH Dipl.-Inform. Deniz Bahadir Winterfeldtstr. 21 10781 Berlin Germany Phone: +49 - 30 / 577 0004-22 Email: deniz.bahadir(a)benocs.com www.benocs.com Advisory Board (Chairman): Stephan Schröder Board of Management: Dr.-Ing. Oliver Holschke, Dr.-Ing. Ingmar Poese Commercial Register: Amtsgericht Bonn HRB 19378

2 1

[review] Review of Nowide has finished
by Frédéric Bron 22 Jun '17

22 Jun '17

Thanks for you interest and the discussion. Please be patient, it will take a few days for me to compile all the discussions and give the result of the review. Kind regards, Frédéric

1 0

1.65 closes for new libraries and breaking changes this Friday
by Marshall Clow 22 Jun '17

22 Jun '17

[ The calendar says today, but I forgot to send this out, so we're pushing the deadline back two days. ] Upcoming release dates: 28-Jun: Boost 1.65.0 closed for major changes 5-Jul: Boost 1.65.0 closed for all changes, except by permission of a release manager. 8-Jul: Boost 1.65.0 closed for all changes 12-Jul: Boost 1.65.0 beta; master branch re-opened 2-Aug: Release closed for all changes, except by permission of a release manager. 5-Aug: Boost 1.65.0 closed for all changes 9-Aug: Boost 1.65.0 release -- Marshall

1 0

Re: [boost] [review] Review of Nowide (Unicode)
by Yakov Galka 21 Jun '17

21 Jun '17

Dear all, Here is my review of the library. - What is your evaluation of the design? > I welcome the UTF-8 approach on Window. However, the fact that Nowide does not handle invalid UTF-16 transparently is an issue that has to be addressed. - What is your evaluation of the implementation? > Going through the code I can't spot any implementation issues. > - What is your evaluation of the documentation? > It's clear. No issues found. > - What is your evaluation of the potential usefulness of the library? > I already use this approach through a different implementation. Even though I wouldn't bring Boost just for that single library, for projects that already use Boost it might be a valuable addition. I do think that Nowide offers a superior solution to encoding issues compared to Boost.Filesystem. > - Did you try to use the library? With what compiler? Did you have any > problems? > I did try it years ago when it was first submitted to Boost. I didn't try the recent version, so NO. > - How much effort did you put into your evaluation? A glance? A quick > reading? In-depth study? > 30 minutes reading the code + mailing list discussions. > - Are you knowledgeable about the problem domain? > I rate myself as an expert. I'm a co-author of utf8everywhere.org. DISCLAIMER: I met Artyom a couple times IRL. > - 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. > Conditional acceptance, subject to transparent invalid UTF-16 resolution. -- Yakov Galka http://stannum.co.il/

4 3

[review] Last day to send your review of Nowide (Unicode)
by Frédéric Bron 21 Jun '17

21 Jun '17

Dear all, This is the last day to submit your review of Artyom Beilis' Nowide library. So far, 12 people participated to the review (not counting the author and the review manager): - Andrzej Krzemienski - Chris Glover - Degski - Edward Diener - Gavin Lambert - Niall Douglas - Norbert Wenzel - Paul Groke - Peter Dimov - Vadim Zeitlin - Yakov Galka - Zach Laine 8 reviews where received including * 6 reviews said nowide should be accepted (with conditions): - Degski, - Peter Dimov - Andrzej Krzemienski - Chris Glover - Norbert Wenzel - Yakov Galka * 1 review did not choose between accept/do not accept: - Paul Groke * 1 review said nowide should not be accepted as is: - Zach Laine Your participation is encouraged, as the proposed library is uncoupled, focused and rather small. Nowadays, everyone needs to handle Unicode but this is very difficult using only the standard library in a platform independant way. Nowide offers a very simple way to handle Unicode the same way on Windows/MacOS/Linux. Key features: * work with UTF-8 in your code, Nowide converts to OS encoding * Easy to use functions for converting UTF-8 to/from UTF-16 * A class to fixing argc, argc and env main parameters to use UTF-8 * UTF-8 aware functions: - stdio.h functions (fopen, freopen, remove, rename) - stdlib.h functions (system, getenv, setenv, unsetenv, putenv) - fstream (filebuf, fstream/ofstream/ifstream) - iostream (cout, cerr, clog, cin) Documentation: http://cppcms.com/files/nowide/html GitHub: https://github.com/artyom-beilis/nowide git clone https://github.com/artyom-beilis/nowide.git Latest tarballs: - to be unzipped in boost source: https://github.com/artyom-beilis/nowide/archive/master.zip - as a standalone library: http://cppcms.com/files/nowide/nowide_standalone.zip Nowide has standard boost layout. So, just put it into the boost source tree under libs/nowide directory for building and running tests. Or alternatively - for header only part of the library add the nowide/include path to the include path of your projects. Please post your comments and review to the boost mailing list (preferably), or privately to the Review Manager (to me ;-). Here are some questions you might want to answer in your review: - 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 most importantly: - 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. For more information about Boost Formal Review Process, see: http://www.boost.org/community/reviews.html Thank you very much for your time and efforts. Frédéric

1 0

[sort] Timsort review result
by Steven Ross 21 Jun '17

21 Jun '17

After reviewing the responses, I have rejected Timsort. Here are the reasons: 1. No author/maintainer responses to the questions on boost (except for forwarding some documentation links). We need a maintainer for all our source, and I'm not volunteering to maintain new functions as complex as Timsort. 2. There was no response on the Timsort bug reported during the review, nor any tests for it. 3. We need better tests. 4. Francisco is including a more efficient stable sort as part of his additions to the library, including handling special cases like Timsort handles better than stable_sort does. Francisco will maintain his code. I'd like to note that many people THINK they are experts on sorting because they covered it in class or may have read about it in a textbook, but most do not understand hybrid sorts well. As a specific example, I saw O(N*log(N)) mentioned as the worst-case speed limit of general-case sorts, when in fact Spreadsort beats that as described in its docs: http://www.boost.org/doc/libs/1_61_0/libs/sort/doc/html/index.html#sort.ove… Many people also don't carefully analyze for worst-case performance, which can be tricky. With regards to future sorting library reviews, I'll get at least a one-page doc describing the algorithm for the review in advance, but the boost documentation formatting is an unreasonable request for a simple sorting function, and I don't see the value it adds for these simple reviews. It is useful for a full library, but we'll be integrating the documentation into our full library after a new function is accepted. It isn't accurate to call it a mini-review (as described here: http://www.boost.org/community/reviews.html#Fast-Track) because it's reviewing completely new source we're talking about adding to the library. That said, do you still want me to a call the upcoming one for pdqsort a "mini-review"?

3 3