Standardizing on the shoulders of giants

Fall C++ standards meeting trip report at Rackspace Atlanta.

Transcript

1. Standardizing on the shoulders of giants Zhihao Yuan <zhihao.yuan@rackspace.com>
2. Where std::optional<T> goes? • Voted into C++14 working draft in Bristol • Voted out in Chicago
3. Background of std::optional • Originally boost::optional • Vocabulary type[1] of a generic Nullable concept[2] ◦ Empty function std::function<...>{} ◦ Empty pointer nullptr • Literal type, in place initialization, etc.
4. Challenges • Relational operators ◦ Previously broken in Bristol ◦ New design decided within 30mins • Mixed type assignments ◦ Still exploring
5. What we learned Any attempt to standardize an existing practice should go to a Technıcal Specıfıcatıon first.
6. Prior success TR1 • std::tuple • std::function • std::shared_ptr • ...
7. From now on File System TS • boost::filesystem V3 Library Fundamentals TS • std::experimental::optional • std::experimental::any • std::experimental::string_view (string_ref)
8. However, sometimes giant is strong What is the biggest source of “existing practice” to C++?
9. , and problematic. The C Standard.
10. War of std::rand • The paper[3]: add global_urng() etc., deprecate std::rand families and random_shuffle • LEWG: No time for addition, just kill them. • LWG: +1! • The committee: No replacement; go away.
11. War of std::rand • The paper[3]: add global_urng() etc., deprecate std::rand families and random_shuffle • LEWG: No time for addition, just kill them. • LWG: +1! • The committee: No replacement
12. The procedure of depreciation • Show the problem • Find replacement • Deprecate • Deprecated means “identified as a candidate for removal from future revisions”[4]; still normative, may not get removed.
13. “Show the problem” • “XXX’s implementation of rand provides crypto-strength randomness” • “For a beginner who wants to roll dice, rand is good enough”
14. “Find replacement” • Accessibility of the global random engine • Thread-safety • Seeding • Bikeshedding ◦ Bjarne: naming debates, like syntax debates, often cover for something deeper
15. So... • Strengthen the note advising against rand • Find replacement…[5]
16. Prior success C++11 • std::unique_ptr auto_ptr • std::function unary_function, binary_function • std::bind bind1st, bind2nd • std::mem_fn mem_fun
17. From now on /* I just selfishly listed the things I know */ • std::rand, with std::randint • setiosflags..., with std::experimental::putf • vector<bool>, with ??::bitvector
18. Enough technical stuff Here is something useful.
19. What C++14 have? • Read the “Adopted” papers http://www.open- std. org/JTC1/SC22/WG21/docs/papers/2013/ • cppreference.com, items marked (C++14)
20. isocpp.org • Blog feeds • Videos • Google Groups ◦ where committee members haunting
21. Questions?
22. References [1] Pablo Halpern. “Polymorphic Memory Resources”. http://www.open-std. org/JTC1/SC22/WG21/docs/papers/2013/n3726.pdf [2] F. Cacciola, A. Krzemieński. “A proposal to add a utility class to represent optional objects (Revision 4)”. http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2013/n3672.html [3] Walter E. Brown. “Three <random>-related Proposals, v2”. http://www.open-std. org/JTC1/SC22/WG21/docs/papers/2013/n3742.pdf [4] “Compatibility features”, Annex D of Workıng Draft, Standard for Programmıng Language C++. http://isocpp.org/files/papers/N3797.pdf [5] Zhihao Yuan. “std::rand replacement” http://students.cs.niu.edu/~z1565938/randint.html