fmtlib / fmt

Reduced the library size. For example, on macOS a stripped test binary statically linked with {fmt} shrank from ~368k to less than 100k.

Added a simpler and more efficient format string compilation API:

#include <fmt/compile.h>

// Converts 42 into std::string using the most efficient method and no
// runtime format string processing.
std::string s = fmt::format(FMT_COMPILE("{}"), 42);

The old fmt::compile API is now deprecated.

Thanks @ldionne (Louis Dionne) for the suggestion and @hanickadot (Hana Dusíková) for the initial prototype.

Optimized integer formatting: format_to with format string compilation and a stack-allocated buffer is now faster than to_chars on both libc++ and libstdc++.

Optimized handling of small format strings. For example,

fmt::format("Result: {}: ({},{},{},{})", str1, str2, str3, str4, str5)

is now ~40% faster (#1685).

Applied extern templates to improve compile times when using the core API and fmt/format.h (#1452). For example, on macOS with clang the compile time of a test translation unit dropped from 2.3s to 0.3s with -O2 and from 0.6s to 0.3s with the default settings (-O0).

Before (-O2):

% time c++ -c test.cc -I include -std=c++17 -O2
c++ -c test.cc -I include -std=c++17 -O2  2.22s user 0.08s system 99% cpu 2.311 total

After (-O2):

% time c++ -c test.cc -I include -std=c++17 -O2
c++ -c test.cc -I include -std=c++17 -O2  0.26s user 0.04s system 98% cpu 0.303 total

Before (default):

% time c++ -c test.cc -I include -std=c++17
c++ -c test.cc -I include -std=c++17  0.53s user 0.06s system 98% cpu 0.601 total

After (default):

% time c++ -c test.cc -I include -std=c++17
c++ -c test.cc -I include -std=c++17  0.24s user 0.06s system 98% cpu 0.301 total

It is still recommended to use fmt/core.h instead of fmt/format.h but the compile time difference is now smaller. Thanks @alex3d for the suggestion.

Named arguments are now stored on stack (no dynamic memory allocations) and the compiled code is more compact and efficient. For example

#include <fmt/core.h>

int main() {
  fmt::print("The answer is {answer}\n", fmt::arg("answer", 42));
}

compiles to just (godbolt)

.LC0:
        .string "answer"
.LC1:
        .string "The answer is {answer}\n"
main:
        sub     rsp, 56
        mov     edi, OFFSET FLAT:.LC1
        mov     esi, 23
        movabs  rdx, 4611686018427387905
        lea     rax, [rsp+32]
        lea     rcx, [rsp+16]
        mov     QWORD PTR [rsp+8], 1
        mov     QWORD PTR [rsp], rax
        mov     DWORD PTR [rsp+16], 42
        mov     QWORD PTR [rsp+32], OFFSET FLAT:.LC0
        mov     DWORD PTR [rsp+40], 0
        call    fmt::v6::vprint(fmt::v6::basic_string_view<char>,
                                fmt::v6::format_args)
        xor     eax, eax
        add     rsp, 56
        ret

    .L.str.1:
            .asciz  "answer"

Implemented compile-time checks for dynamic width and precision (#1614):

#include <fmt/format.h>

int main() {
  fmt::print(FMT_STRING("{0:{1}}"), 42);
}

now gives a compilation error because argument 1 doesn't exist:

In file included from test.cc:1:
include/fmt/format.h:2726:27: error: constexpr variable 'invalid_format' must be
initialized by a constant expression
  FMT_CONSTEXPR_DECL bool invalid_format =
                          ^
...
include/fmt/core.h:569:26: note: in call to
'&checker(s, {}).context_->on_error(&"argument not found"[0])'
    if (id >= num_args_) on_error("argument not found");
                        ^

Added sentinel support to fmt::join (#1689)

struct zstring_sentinel {};
bool operator==(const char* p, zstring_sentinel) { return *p == '\0'; }
bool operator!=(const char* p, zstring_sentinel) { return *p != '\0'; }

struct zstring {
  const char* p;
  const char* begin() const { return p; }
  zstring_sentinel end() const { return {}; }
};

auto s = fmt::format("{}", fmt::join(zstring{"hello"}, "_"));
// s == "h_e_l_l_o"

Thanks @BRevzin (Barry Revzin).

Added support for named args, clear and reserve to dynamic_format_arg_store (#1655, #1663, #1674, #1677). Thanks @vsolontsov-ll (Vladimir Solontsov).

Added support for the 'c' format specifier to integral types for compatibility with std::format (#1652).

Replaced the 'n' format specifier with 'L' for compatibility with std::format (#1624). The 'n' specifier can be enabled via the FMT_DEPRECATED_N_SPECIFIER macro.

The '=' format specifier is now disabled by default for compatibility with std::format. It can be enabled via the FMT_DEPRECATED_NUMERIC_ALIGN macro.

Removed the following deprecated APIs:

• FMT_STRING_ALIAS and fmt macros - replaced by FMT_STRING

• fmt::basic_string_view::char_type - replaced by fmt::basic_string_view::value_type

• convert_to_int

• format_arg_store::types

• *parse_context - replaced by *format_parse_context

• FMT_DEPRECATED_INCLUDE_OS

• FMT_DEPRECATED_PERCENT - incompatible with std::format

• *writer - replaced by compiled format API

Renamed the internal namespace to detail (#1538). The former is still provided as an alias if the FMT_USE_INTERNAL macro is defined.

Improved compatibility between fmt::printf with the standard specs (#1595, #1682, #1683, #1687, #1699). Thanks @rimathia.

Fixed handling of operator<< overloads that use copyfmt (#1666).

Added the FMT_OS CMake option to control inclusion of OS-specific APIs in the fmt target. This can be useful for embedded platforms (#1654, #1656). Thanks @kwesolowski (Krzysztof Wesolowski).

Replaced FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION with the FMT_FUZZ macro to prevent interferring with fuzzing of projects using {fmt} (#1650). Thanks @asraa (Asra Ali).

Fixed compatibility with emscripten (#1636, #1637). Thanks @ArthurSonzogni (Arthur Sonzogni).

Improved documentation (#704, #1643, #1660, #1681, #1691, #1706, #1714, #1721, #1739, #1740, #1741, #1751). Thanks @senior7515 (Alexander Gallego), @lsr0 (Lindsay Roberts), @puetzk (Kevin Puetz), @fpelliccioni (Fernando Pelliccioni), Alexey Kuzmenko, @jelly (jelle van der Waa), @claremacrae (Clare Macrae), @jiapengwen (文佳鹏), @gsjaardema (Greg Sjaardema), @alexey-milovidov.

Implemented various build configuration fixes and improvements (#1603, #1657, #1702, #1728). Thanks @scramsby (Scott Ramsby), @jtojnar (Jan Tojnar), @orivej (Orivej Desh), @flagarde.

Fixed various warnings and compilation issues (#1616, #1620, #1622, #1625, #1627, #1628, #1629, #1631, #1633, #1649, #1658, #1661, #1667, #1668, #1669, #1692, #1696, #1697, #1707, #1712, #1716, #1722, #1724, #1729, #1738, #1742, #1743, #1744, #1747, #1750). Thanks @gsjaardema (Greg Sjaardema), @gabime (Gabi Melman), @johnor (Johan), @gabime (Dmitry Kurkin), @invexed (James Beach), @peterbell10, @daixtrose (Markus Werle), @petrutlucian94 (Lucian Petrut), @Neargye (Daniil Goncharov), @ambitslix (Attila M. Szilagyi), @gabime (Gabi Melman), @erthink (Leonid Yuriev), @tohammer (Tobias Hammer), @0x8000-0000 (Florin Iucha).

Fixed ostream support in sprintf (#1631).

Fixed type detection when using implicit conversion to string_view and ostream operator<< inconsistently (#1662).

Improved error reporting when trying to format an object of a non-formattable type:

fmt::format("{}", S());

now gives:

include/fmt/core.h:1015:5: error: static_assert failed due to requirement
'formattable' "Cannot format argument. To make type T formattable provide a
formatter<T> specialization:
https://fmt.dev/latest/api.html#formatting-user-defined-types"
    static_assert(
    ^
...
note: in instantiation of function template specialization
'fmt::v6::format<char [3], S, char>' requested here
  fmt::format("{}", S());
       ^

if S is not formattable.

Reduced library size by ~10%.

Always print decimal point if # is specified (#1476, #1498):

fmt::print("{:#.0f}", 42.0);

now prints 42.

Implemented the 'L' specifier for locale-specific numeric formatting to improve compatibility with std::format. The 'n' specifier is now deprecated and will be removed in the next major release.

Moved OS-specific APIs such as windows_error from fmt/format.h to fmt/os.h. You can define FMT_DEPRECATED_INCLUDE_OS to automatically include fmt/os.h from fmt/format.h for compatibility but this will be disabled in the next major release.

Added precision overflow detection in floating-point formatting.

Implemented detection of invalid use of fmt::arg.

Used type_identity to block unnecessary template argument deduction. Thanks Tim Song.

Improved UTF-8 handling (#1109):

fmt::print("┌{0:─^{2}}┐\n"
           "│{1: ^{2}}│\n"
           "└{0:─^{2}}┘\n", "", "Привет, мир!", 20);

now prints:

┌────────────────────┐
│    Привет, мир!    │
└────────────────────┘

on systems that support Unicode.

Added experimental dynamic argument storage (#1170, #1584):

fmt::dynamic_format_arg_store<fmt::format_context> store;
store.push_back("answer");
store.push_back(42);
fmt::vprint("The {} is {}.\n", store);

prints:

The answer is 42.

Thanks @vsolontsov-ll (Vladimir Solontsov).

Made fmt::join accept initializer_list (#1591). Thanks @Rapotkinnik (Nikolay Rapotkin).

Fixed handling of empty tuples (#1588).

Fixed handling of output iterators in format_to_n (#1506).

Fixed formatting of std::chrono::duration types to wide output (#1533). Thanks @zeffy (pilao).

Added const begin and end overload to buffers (#1553). Thanks @dominicpoeschko.

Added the ability to disable floating-point formatting via FMT_USE_FLOAT, FMT_USE_DOUBLE and FMT_USE_LONG_DOUBLE macros for extremely memory-constrained embedded system (#1590). Thanks @albaguirre (Alberto Aguirre).

Made FMT_STRING work with constexpr string_view (#1589). Thanks @scramsby (Scott Ramsby).

Implemented a minor optimization in the format string parser (#1560). Thanks @IkarusDeveloper.

Improved attribute detection (#1469, #1475, #1576). Thanks @federico-busato (Federico), @chronoxor (Ivan Shynkarenka), @refnum.

Improved documentation (#1481, #1523). Thanks @JackBoosY (Jack·Boos·Yu), @imba-tjd (谭九鼎).

Fixed symbol visibility on Linux when compiling with -fvisibility=hidden (#1535). Thanks @milianw (Milian Wolff).

Implemented various build configuration fixes and improvements (#1264, #1460, #1534, #1536, #1545, #1546, #1566, #1582, #1597, #1598). Thanks @ambitslix (Attila M. Szilagyi), @jwillikers (Jordan Williams), @stac47 (Laurent Stacul).

Fixed various warnings and compilation issues (#1433, #1461, #1470, #1480, #1485, #1492, #1493, #1504, #1505, #1512, #1515, #1516, #1518, #1519, #1520, #1521, #1522, #1524, #1530, #1531, #1532, #1539, #1547, #1548, #1554, #1567, #1568, #1569, #1571, #1573, #1575, #1581, #1583, #1586, #1587, #1594, #1596, #1604, #1606, #1607, #1609). Thanks @marti4d (Chris Martin), @iPherian, @parkertomatoes, @gsjaardema (Greg Sjaardema), @chronoxor (Ivan Shynkarenka), @DanielaE (Daniela Engert), @torsten48, @tohammer (Tobias Hammer), @lefticus (Jason Turner), @ryusakki (Haise), @adnsv (Alex Denisov), @fghzxm, @refnum, @pramodk (Pramod Kumbhar), @Spirrwell, @scramsby (Scott Ramsby).

Fixed ABI compatibility with libfmt.so.6.0.0 (#1471).

Fixed handling types convertible to std::string_view (#1451). Thanks @denizevrenci (Deniz Evrenci).

Made CUDA test an opt-in enabled via the FMT_CUDA_TEST CMake option.

Fixed sign conversion warnings (#1440). Thanks @0x8000-0000 (Florin Iucha).

Fixed shared library build on Windows (#1443, #1445, #1446, #1450). Thanks @egorpugin (Egor Pugin), @bbolli (Beat Bolli).

Added a missing decimal point in exponent notation with trailing zeros.

Removed deprecated format_arg_store::TYPES.

{fmt} now formats IEEE 754 float and double using the shortest decimal representation with correct rounding by default:

#include <cmath>
#include <fmt/core.h>

int main() {
  fmt::print("{}", M_PI);
}

prints 3.141592653589793.

Made the fast binary to decimal floating-point formatter the default, simplified it and improved performance. {fmt} is now 15 times faster than libc++'s std::ostringstream, 11 times faster than printf and 10% faster than double-conversion on dtoa-benchmark:

{fmt} no longer converts float arguments to double. In particular this improves the default (shortest) representation of floats and makes fmt::format consistent with std::format specs (#1336, #1353, #1360, #1361):

fmt::print("{}", 0.1f);

prints 0.1 instead of 0.10000000149011612. Thanks @orivej (Orivej Desh).

Made floating-point formatting output consistent with printf/iostreams (#1376, #1417).

Added support for 128-bit integers (#1287):

fmt::print("{}", std::numeric_limits<__int128_t>::max());

prints 170141183460469231731687303715884105727. Thanks @denizevrenci (Deniz Evrenci).

The overload of print that takes text_style is now atomic, i.e. the output from different threads doesn't interleave (#1351). Thanks @tankiJong (Tanki Zhang).

Made compile time in the header-only mode ~20% faster by reducing the number of template instantiations. wchar_t overload of vprint was moved from fmt/core.h to fmt/format.h.

Added an overload of fmt::join that works with tuples (#1322, #1330):

#include <tuple>
#include <fmt/ranges.h>

int main() {
  std::tuple<char, int, float> t{'a', 1, 2.0f};
  fmt::print("{}", t);
}

prints ('a', 1, 2.0). Thanks @jeremyong (Jeremy Ong).

Changed formatting of octal zero with prefix from "00" to "0":

fmt::print("{:#o}", 0);

prints 0.

The locale is now passed to ostream insertion (<<) operators (#1406):

#include <fmt/locale.h>
#include <fmt/ostream.h>

struct S {
  double value;
};

std::ostream& operator<<(std::ostream& os, S s) {
  return os << s.value;
}

int main() {
  auto s = fmt::format(std::locale("fr_FR.UTF-8"), "{}", S{0.42});
  // s == "0,42"
}

Thanks @dlaugt (Daniel Laügt).

Locale-specific number formatting now uses grouping (#1393 #1394). Thanks @skrdaniel.

Fixed handling of types with deleted implicit rvalue conversion to const char** (#1421):

struct mystring {
  operator const char*() const&;
  operator const char*() &;
  operator const char*() const&& = delete;
  operator const char*() && = delete;
};
mystring str;
fmt::print("{}", str); // now compiles

Enums are now mapped to correct underlying types instead of int (#1286). Thanks @agmt (Egor Seredin).

Enum classes are no longer implicitly converted to int (#1424).

Added basic_format_parse_context for consistency with C++20 std::format and deprecated basic_parse_context.

Fixed handling of UTF-8 in precision (#1389, #1390). Thanks @tajtiattila (Attila Tajti).

{fmt} can now be installed on Linux, macOS and Windows with Conda using its conda-forge package (#1410):

conda install -c conda-forge fmt

Thanks @tdegeus (Tom de Geus).

Added a CUDA test (#1285, #1317). Thanks @luncliff (Park DongHa) and @risa2000.

Improved documentation (#1276, #1291, #1296, #1315, #1332, #1337, #1395 #1418). Thanks @waywardmonkeys (Bruce Mitchener), @pauldreik (Paul Dreik), @jackoalan (Jack Andersen).

Various code improvements (#1358, #1407). Thanks @orivej (Orivej Desh), @dpacbach (David P. Sicilia),

Fixed compile-time format string checks for user-defined types (#1292).

Worked around a false positive in unsigned-integer-overflow sanitizer (#1377).

Fixed various warnings and compilation issues (#1273, #1278, #1280, #1281, #1288, #1290, #1301, #1305, #1306, #1309, #1312, #1313, #1316, #1319, #1320, #1326, #1328, #1344, #1345, #1347, #1349, #1354, #1362, #1366, #1364, #1370, #1371, #1385, #1388, #1397, #1414, #1416, #1422 #1427, #1431, #1433). Thanks @hhb, @gsjaardema (Greg Sjaardema), @gabime (Gabi Melman), @neheb (Rosen Penev), @vedranmiletic (Vedran Miletić), @dkavolis (Daumantas Kavolis), @mwinterb, @orivej (Orivej Desh), @denizevrenci (Deniz Evrenci) @leonklingele, @chronoxor (Ivan Shynkarenka), @kent-tri, @0x8000-0000 (Florin Iucha), @marti4d (Chris Martin).

Switched to the MIT license with an optional exception that allows distributing binary code without attribution.

Floating-point formatting is now locale-independent by default:

#include <locale>
#include <fmt/core.h>

int main() {
  std::locale::global(std::locale("ru_RU.UTF-8"));
  fmt::print("value = {}", 4.2);
}

prints "value = 4.2" regardless of the locale. For locale-specific formatting use the n specifier:

std::locale::global(std::locale("ru_RU.UTF-8"));
fmt::print("value = {:n}", 4.2);

prints "value = 4,2".

Added an experimental Grisu floating-point formatting algorithm implementation (disabled by default). To enable it compile with the FMT_USE_GRISU macro defined to 1:

#define FMT_USE_GRISU 1
#include <fmt/format.h>

auto s = fmt::format("{}", 4.2); // formats 4.2 using Grisu

With Grisu enabled, {fmt} is 13x faster than std::ostringstream (libc++) and 10x faster than sprintf on dtoa-benchmark (full results):

Separated formatting and parsing contexts for consistency with C++20 std::format, removing the undocumented basic_format_context::parse_context() function.

Added oss-fuzz support (#1199). Thanks @pauldreik (Paul Dreik).

formatter specializations now always take precedence over operator<< (#952):

#include <iostream>
#include <fmt/ostream.h>

struct S {};

std::ostream& operator<<(std::ostream& os, S) {
  return os << 1;
}

template <>
struct fmt::formatter<S> : fmt::formatter<int> {
  auto format(S, format_context& ctx) {
    return formatter<int>::format(2, ctx);
  }
};

int main() {
  std::cout << S() << "\n"; // prints 1 using operator<<
  fmt::print("{}\n", S());  // prints 2 using formatter
}

Introduced the experimental fmt::compile function that does format string compilation (#618, #1169, #1171):

#include <fmt/compile.h>

auto f = fmt::compile<int>("{}");
std::string s = fmt::format(f, 42); // can be called multiple times to format
                                    // different values
// s == "42"

It moves the cost of parsing a format string outside of the format function which can be beneficial when identically formatting many objects of the same types. Thanks @stryku (Mateusz Janek).

Added experimental support for the % format specifier that formats floating-point values as percentages (#1060, #1069, #1071):

auto s = fmt::format("{:.1%}", 0.42); // s == "42.0%"

Thanks @gawain-bolton (Gawain Bolton).

Implemented precision for floating-point durations (#1004, #1012):

auto s = fmt::format("{:.1}", std::chrono::duration<double>(1.234));
// s == 1.2s

Thanks @DanielaE (Daniela Engert).

Implemented chrono format specifiers %Q and %q that give the value and the unit respectively (#1019):

auto value = fmt::format("{:%Q}", 42s); // value == "42"
auto unit  = fmt::format("{:%q}", 42s); // unit == "s"

Thanks @DanielaE (Daniela Engert).

Fixed handling of dynamic width in chrono formatter:

auto s = fmt::format("{0:{1}%H:%M:%S}", std::chrono::seconds(12345), 12);
//                        ^ width argument index                     ^ width
// s == "03:25:45    "

Thanks Howard Hinnant.

Removed deprecated fmt/time.h. Use fmt/chrono.h instead.

Added fmt::format and fmt::vformat overloads that take text_style (#993, #994):

#include <fmt/color.h>

std::string message = fmt::format(fmt::emphasis::bold | fg(fmt::color::red),
                                  "The answer is {}.", 42);

Thanks @Naios (Denis Blank).

Removed the deprecated color API (print_colored). Use the new API, namely print overloads that take text_style instead.

Made std::unique_ptr and std::shared_ptr formattable as pointers via fmt::ptr (#1121):

std::unique_ptr<int> p = ...;
fmt::print("{}", fmt::ptr(p)); // prints p as a pointer

Thanks @sighingnow (Tao He).

Made print and vprint report I/O errors (#1098, #1099). Thanks @BillyDonahue (Billy Donahue).

Marked deprecated APIs with the [[deprecated]] attribute and removed internal uses of deprecated APIs (#1022). Thanks @eliaskosunen (Elias Kosunen).

Modernized the codebase using more C++11 features and removing workarounds. Most importantly, buffer_context is now an alias template, so use buffer_context<T> instead of buffer_context<T>::type. These features require GCC 4.8 or later.

formatter specializations now always take precedence over implicit conversions to int and the undocumented convert_to_int trait is now deprecated.

Moved the undocumented basic_writer, writer, and wwriter types to the internal namespace.

Removed deprecated basic_format_context::begin(). Use out() instead.

Disallowed passing the result of join as an lvalue to prevent misuse.

Refactored the undocumented structs that represent parsed format specifiers to simplify the API and allow multibyte fill.

Moved SFINAE to template parameters to reduce symbol sizes.

Switched to fputws for writing wide strings so that it's no longer required to call _setmode on Windows (#1229, #1243). Thanks @jackoalan (Jack Andersen).

Improved literal-based API (#1254). Thanks @sylveon (Charles Milette).

Added support for exotic platforms without uintptr_t such as IBM i (AS/400) which has 128-bit pointers and only 64-bit integers (#1059).

Added Sublime Text syntax highlighting config (#1037). Thanks @Kronuz (Germán Méndez Bravo).

Added the FMT_ENFORCE_COMPILE_STRING macro to enforce the use of compile-time format strings (#1231). Thanks @jackoalan (Jack Andersen).

Stopped setting CMAKE_BUILD_TYPE if {fmt} is a subproject (#1081).

Various build improvements (#1039, #1078, #1091, #1103, #1177). Thanks @luncliff (Park DongHa), @jasonszang (Jason Shuo Zang), @olafhering (Olaf Hering), @Lecetem, @pauldreik (Paul Dreik).

Improved documentation (#1049, #1051, #1083, #1113, #1114, #1146, #1180, #1250, #1252, #1265). Thanks @mikelui (Michael Lui), @foonathan (Jonathan Müller), @BillyDonahue (Billy Donahue), @jwakely (Jonathan Wakely), @kaisbe (Kais Ben Salah), @sdebionne (Samuel Debionne).

Fixed ambiguous formatter specialization in fmt/ranges.h (#1123).

Fixed formatting of a non-empty std::filesystem::path which is an infinitely deep range of its components (#1268).

Fixed handling of general output iterators when formatting characters (#1056, #1058). Thanks @abolz (Alexander Bolz).

Fixed handling of output iterators in formatter specialization for ranges (#1064).

Fixed handling of exotic character types (#1188).

Made chrono formatting work with exceptions disabled (#1062).

Fixed DLL visibility issues (#1134, #1147). Thanks @denchat.

Disabled the use of UDL template extension on GCC 9 (#1148).

Removed misplaced format compile-time checks from printf (#1173).

Fixed issues in the experimental floating-point formatter (#1072, #1129, #1153, #1155, #1210, #1222). Thanks @alabuzhev (Alex Alabuzhev).

Fixed bugs discovered by fuzzing or during fuzzing integration (#1124, #1127, #1132, #1135, #1136, #1141, #1142, #1178, #1179, #1194). Thanks @pauldreik (Paul Dreik).

Fixed building tests on FreeBSD and Hurd (#1043). Thanks @jackyf (Eugene V. Lyubimkin).

Fixed various warnings and compilation issues (#998, #1006, #1008, #1011, #1025, #1027, #1028, #1029, #1030, #1031, #1054, #1063, #1068, #1074, #1075, #1079, #1086, #1088, #1089, #1094, #1101, #1102, #1105, #1107, #1115, #1117, #1118, #1120, #1123, #1139, #1140, #1143, #1144, #1150, #1151, #1152, #1154, #1156, #1159, #1175, #1181, #1186, #1187, #1191, #1197, #1200, #1203, #1205, #1206, #1213, #1214, #1217, #1228, #1230, #1232, #1235, #1236, #1240). Thanks @DanielaE (Daniela Engert), @mwinterb, @eliaskosunen (Elias Kosunen), @morinmorin, @ricco19 (Brian Ricciardelli), @waywardmonkeys (Bruce Mitchener), @chronoxor (Ivan Shynkarenka), @remyabel, @pauldreik (Paul Dreik), @gsjaardema (Greg Sjaardema), @rcane (Ronny Krüger), @mocabe, @denchat, @cjdb (Christopher Di Bella), @HazardyKnusperkeks (Björn Schäpers), @vedranmiletic (Vedran Miletić), @jackoalan (Jack Andersen), @DaanDeMeyer (Daan De Meyer), @starkmapper (Mark Stapper).

Introduced experimental chrono formatting support:

#include <fmt/chrono.h>

int main() {
  using namespace std::literals::chrono_literals;
  fmt::print("Default format: {} {}\n", 42s, 100ms);
  fmt::print("strftime-like format: {:%H:%M:%S}\n", 3h + 15min + 30s);
}

prints:

Default format: 42s 100ms
strftime-like format: 03:15:30

Added experimental support for emphasis (bold, italic, underline, strikethrough), colored output to a file stream, and improved colored formatting API (#961, #967, #973):

#include <fmt/color.h>

int main() {
  print(fg(fmt::color::crimson) | fmt::emphasis::bold,
        "Hello, {}!\n", "world");
  print(fg(fmt::color::floral_white) | bg(fmt::color::slate_gray) |
        fmt::emphasis::underline, "Hello, {}!\n", "мир");
  print(fg(fmt::color::steel_blue) | fmt::emphasis::italic,
        "Hello, {}!\n", "世界");
}

prints the following on modern terminals with RGB color support:

Thanks @Rakete1111 (Nicolas).

Added support for 4-bit terminal colors (#968, #974)

#include <fmt/color.h>

int main() {
  print(fg(fmt::terminal_color::red), "stop\n");
}

Note that these colors vary by terminal:Thanks @Rakete1111 (Nicolas).

Parameterized formatting functions on the type of the format string (#880, #881, #883, #885, #897, #920). Any object of type S that has an overloaded to_string_view(const S&) returning fmt::string_view can be used as a format string:

namespace my_ns {
inline string_view to_string_view(const my_string& s) {
  return {s.data(), s.length()};
}
}

std::string message = fmt::format(my_string("The answer is {}."), 42);

Thanks @DanielaE (Daniela Engert).

Made std::string_view work as a format string (#898):

auto message = fmt::format(std::string_view("The answer is {}."), 42);

Thanks @DanielaE (Daniela Engert).

Added wide string support to compile-time format string checks (#924):

print(fmt(L"{:f}"), 42); // compile-time error: invalid type specifier

Thanks @XZiar.

Made colored print functions work with wide strings (#867):

#include <fmt/color.h>

int main() {
  print(fg(fmt::color::red), L"{}\n", 42);
}

Thanks @DanielaE (Daniela Engert).

Introduced experimental Unicode support (#628, #891):

using namespace fmt::literals;
auto s = fmt::format("{:*^5}"_u, "🤡"_u); // s == "**🤡**"_u

Improved locale support:

#include <fmt/locale.h>

struct numpunct : std::numpunct<char> {
 protected:
  char do_thousands_sep() const override { return '~'; }
};

std::locale loc;
auto s = fmt::format(std::locale(loc, new numpunct()), "{:n}", 1234567);
// s == "1~234~567"

Constrained formatting functions on proper iterator types (#921). Thanks @DanielaE (Daniela Engert).

Added make_printf_args and make_wprintf_args functions (#934). Thanks @tnovotny.

Deprecated fmt::visit, parse_context, and wparse_context. Use fmt::visit_format_arg, format_parse_context, and wformat_parse_context instead.

Removed undocumented basic_fixed_buffer which has been superseded by the iterator-based API (#873, #902). Thanks @superfunc (hollywood programmer).

Disallowed repeated leading zeros in an argument ID:

fmt::print("{000}", 42); // error

Reintroduced support for gcc 4.4.

Fixed compilation on platforms with exotic double (#878).

Improved documentation (#164, #877, #901, #906, #979). Thanks @kookjr (Mathew Cucuzella), @DarkDimius (Dmitry Petrashko), @HecticSerenity.

Added pkgconfig support which makes it easier to consume the library from meson and other build systems (#916). Thanks @colemickens (Cole Mickens).

Various build improvements (#909, #926, #937, #953, #959). Thanks @tchaikov (Kefu Chai), @luncliff (Park DongHa), @AndreasSchoenle (Andreas Schönle), @hotwatermorning, @Zefz (JohanJansen).

Improved string_view construction performance (#914). Thanks @gabime (Gabi Melman).

Fixed non-matching char types (#895). Thanks @DanielaE (Daniela Engert).

Fixed format_to_n with std::back_insert_iterator (#913). Thanks @DanielaE (Daniela Engert).

Fixed locale-dependent formatting (#905).

Fixed various compiler warnings and errors (#882, #886, #933, #941, #931, #943, #954, #956, #962, #965, #977, #983, #989). Thanks @Luthaf (Guillaume Fraux), @stevenhoving (Steven Hoving), @christinaa (Kristina Brooks), @lgritz (Larry Gritz), @DanielaE (Daniela Engert), @0x8000-0000 (Sign Bit), @liuping1997.

Fixed visit lookup issues on gcc 7 & 8 (#870). Thanks @medithe.

Fixed linkage errors on older gcc.

Prevented fmt/range.h from specializing fmt::basic_string_view (#865, #868). Thanks @hhggit (dual).

Improved error message when formatting unknown types (#872). Thanks @foonathan (Jonathan Müller),

Disabled templated user-defined literals when compiled under nvcc (#875). Thanks @CandyGumdrop (Candy Gumdrop),

Fixed format_to formatting to wmemory_buffer (#874).