Is there a design reason why beast static_string uses size_t for length?
Internally as a member, not talking about template argument. I would naively assume that performance overhead of having conversion from uint8_t/uint16_t(and uint32_t if you want to allow "insanely" large static_strings) to size_t is not big enough to justify "large"(in context of microoptimizations for space) overhead due to alignment and size requirements of size_t(on 64b machines). Recently I other context I complained that there is no way to nicely embed godbolt/source examples in here :), so here is a link with example: https://godbolt.org/z/Y4Y5zKj86 regards, Ivan
On Fri, Apr 15, 2022 at 6:43 PM Ivan Matek wrote:
Internally as a member, not talking about template argument.
Note that boost::static_string (Boost.StaticString) does not always use size_t: https://github.com/boostorg/static_string/blob/develop/include/boost/static_... // Find the smallest width integral type that can hold a value as large as N (Glen Fernandes) template<std::size_t N> using smallest_width = typename std::conditional<(N <= (std::numeric_limits<unsigned char>::max)()), unsigned char, typename std::conditional<(N <= (std::numeric_limits<unsigned short>::max)()), unsigned short, typename std::conditional<(N <= (std::numeric_limits<unsigned int>::max)()), unsigned int, typename std::conditional<(N <= (std::numeric_limits<unsigned long>::max)()), unsigned long, typename std::conditional<(N <= (std::numeric_limits<unsigned long long>::max)()), unsigned long long, std::size_t>::type>::type>::type>::type>::type; Maybe boost::beast::static_string can become an alias to boost::static_string. Glen
On Sat, Apr 16, 2022 at 3:32 AM Glen Fernandes <glen.fernandes@gmail.com> wrote:
Note that boost::static_string (Boost.StaticString) does not always use size_t:
https://github.com/boostorg/static_string/blob/develop/include/boost/static_... Maybe boost::beast::static_string can become an alias to boost::static_string.
Weird. :) I would have expected them to be the same, I guess you did the fix I wanted but it was never backported(IIRC static_string was take out of beast to become standalone, but could be wrong). In any case my problem is solved, but if Vinnie sees this I would be interested to know if difference is intentional or not.
P.S. If aliasing was not done due to lack of time I could work on it, but I fear that largest part of work will be reviewing if changes did not break anything, so I guess I will not be helping a lot/// but if there is an issue to make this happen please let me know. On Sat, Apr 16, 2022 at 12:40 PM Ivan Matek <libbooze@gmail.com> wrote:
On Sat, Apr 16, 2022 at 3:32 AM Glen Fernandes <glen.fernandes@gmail.com> wrote:
Note that boost::static_string (Boost.StaticString) does not always use size_t:
https://github.com/boostorg/static_string/blob/develop/include/boost/static_... Maybe boost::beast::static_string can become an alias to boost::static_string.
Weird. :) I would have expected them to be the same, I guess you did the fix I wanted but it was never backported(IIRC static_string was take out of beast to become standalone, but could be wrong). In any case my problem is solved, but if Vinnie sees this I would be interested to know if difference is intentional or not.
On Sat, Apr 16, 2022 at 3:44 AM Ivan Matek via Boost <boost@lists.boost.org> wrote:
if there is an issue to make this happen please let me know.
participants (3)
-
Glen Fernandes -
Ivan Matek -
Vinnie Falco