Format API¶

The format API is similar in spirit to the C printf family of function but is safer, simpler and several times faster than common standard library implementations. The format string syntax is similar to the one used by str.format in Python:

fmt::format("The answer is {}.", 42);

The fmt::format function returns a string “The answer is 42.”. You can use fmt::memory_buffer to avoid constructing std::string:

fmt::memory_buffer out;
format_to(out, "For a moment, {} happened.", "nothing");
out.data(); // returns a pointer to the formatted data

The fmt::print function performs formatting and writes the result to a stream:

fmt::print(stderr, "System error code = {}\n", errno);

The file argument can be omitted in which case the function prints to stdout:

fmt::print("Don't {}\n", "panic");

The Format API also supports positional arguments useful for localization:

fmt::print("I'd rather be {1} than {0}.", "right", "happy");

Named arguments can be created with fmt::arg. This makes it easier to track what goes where when multiple arguments are being formatted:

fmt::print("Hello, {name}! The answer is {number}. Goodbye, {name}.",
           fmt::arg("name", "World"), fmt::arg("number", 42));

If your compiler supports C++11 user-defined literals, the suffix _a offers an alternative, slightly terser syntax for named arguments:

using namespace fmt::literals;
fmt::print("Hello, {name}! The answer is {number}. Goodbye, {name}.",
           "name"_a="World", "number"_a=42);

Safety¶

The library is fully type safe, automatic memory management prevents buffer overflow, errors in format strings are reported using exceptions or at compile time. For example, the code

fmt::format("The answer is {:d}", "forty-two");

throws a format_error exception with description “unknown format code ‘d’ for string”, because the argument "forty-two" is a string while the format code d only applies to integers, while

format(FMT_STRING("The answer is {:d}"), "forty-two");

reports a compile-time error for the same reason on compilers that support relaxed constexpr. See here for details.

The following code

fmt::format("Cyrillic letter {}", L'\x42e');

produces a compile-time error because wide character L'\x42e' cannot be formatted into a narrow string. You can use a wide format string instead:

fmt::format(L"Cyrillic letter {}", L'\x42e');

For comparison, writing a wide character to std::ostream results in its numeric value being written to the stream (i.e. 1070 instead of letter ‘ю’ which is represented by L'\x42e' if we use Unicode) which is rarely what is needed.

Compact Binary Code¶

The library is designed to produce compact per-call compiled code. For example (godbolt),

#include <fmt/core.h>

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

compiles to just

main: # @main
  sub rsp, 24
  mov qword ptr [rsp], 42
  mov rcx, rsp
  mov edi, offset .L.str
  mov esi, 17
  mov edx, 2
  call fmt::v5::vprint(fmt::v5::basic_string_view<char>, fmt::v5::format_args)
  xor eax, eax
  add rsp, 24
  ret
.L.str:
  .asciz "The answer is {}."

Portability¶

The library is highly portable and relies only on a small set of C++11 features:

• variadic templates

• type traits

• rvalue references

• decltype

• trailing return types

• deleted functions

• alias templates

These are available since GCC 4.8, Clang 3.0 and MSVC 19.0 (2015). For older compilers use {fmt} version 4.x which continues to be maintained and only requires C++98.

The output of all formatting functions is consistent across platforms. In particular, formatting a floating-point infinity always gives inf while the output of printf is platform-dependent. For example,

fmt::print("{}", std::numeric_limits<double>::infinity());

always prints inf.

Ease of Use¶

{fmt} has a small self-contained code base with the core library consisting of just three header files and no external dependencies. A permissive MIT license allows using the library both in open-source and commercial projects.