Victor Zverovich
Victor Zverovich
We should be able to pull stuff from `(_._)` lazily using coroutines.
This PR adds a terminating NUL to the output buffer in the double-conversion benchmark for consistency with other benchmarks.
https://fmt.dev/latest/syntax.html#range-format-specifications
Per discussion in #3081 it might be good to have something similar to `format_as` for ostream before removing `FMT_DEPRECATED_OSTREAM`.
{fmt} doesn't support printing green day at the moment: ```c++ #include #include int main() { fmt::print(fg(fmt::color::green), "{}", std::chrono::day()); } ``` Error: ``` include/fmt/core.h:1594:63: error: implicit instantiation of undefined template 'fmt::detail::type_is_unformattable_for'...
itoa-benchmark is better than int-benchmark because it does separate measurements per digit count. However it misses side effects which gives unrealistic results for some methods. Add side effects to itoa-benchmark...
Similar to the one for gcc: https://github.com/fmtlib/format-benchmark/blob/935d0e5fd280d4e3e64b0843e4001d41210820e1/CMakeLists.txt#L39
The default FP presentation (formatting of the significand and exponent after they are produced by Dragonbox) can be slightly optimized, possibly using some ideas from `jkj::dragonbox::detail::to_chars`: https://github.com/fmtlib/fmt/issues/3675#issuecomment-3172953795.
premake4 (released around 2011) is no longer officially supported or maintained. Would you accept a PR to replace it with CMake which is a more widely used build system? I...
I just learned about your implementation of Schubfach and noticed the following discrepancy. [The paper](https://fmt.dev/papers/Schubfach4.pdf#page=21) defines $g$ (power of 10 significands) as strict overestimates: $g = \lfloor 2^{-r} 10^{-k} \rfloor...