CODEWITHSUNDEEP
c++fmt
sadig
sadig

Reputation: 521

Printing using {fmt} library

Can I print an object of the C++ class using fmt library?

fmt::print("The object is {}.", obj);

Upvotes: 9

Views: 12002

Answers (2)

vitaut
vitaut

Reputation: 55605

Yes. You can do this by providing a formatter specialization for your type as described in Formatting User-defined Types:

#include <fmt/format.h>

struct point { double x, y; };

template <> struct fmt::formatter<point> {
  constexpr auto parse(format_parse_context &ctx) { return ctx.begin(); }

  template <typename FormatContext>
  auto format(const point &p, FormatContext &ctx) const {
    return format_to(ctx.out(), "({:.1f}, {:.1f})", p.x, p.y);
  }
};

You can also reuse existing formatters via composition or inheritance in which case you might only need to implement the format function.

Upvotes: 12

Ton van den Heuvel
Ton van den Heuvel

Reputation: 10528

Yes, it is possible. As suggested in the comments, fmt provides support for custom types directly: Formatting user defined types.

I normally prefer an alternative approach using std::ostream. When you implement operator<< for std::ostream and your custom type, fmt will be able to format your custom type provided that you include <fmt/ostream.h> as well. For example:

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

struct A {};

std::ostream& operator<<(std::ostream& os, const A& a)
{
  return os << "A!";
}

int main()
{
  fmt::print("{}\n", A{});
  return 0;
}

Keep in mind that this approach will likely be much slower than the initial suggestion of going through fmt directly.

Update: To support the claim that using <fmt/ostream.h> is slower than going through fmt directly, you can use the following benchmark (using Google Benchmark):

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

#include <benchmark/benchmark.h>

struct A {};

std::ostream& operator<<(std::ostream& os, const A& a)
{
  return os << "A!";
}

struct B {};

template<>
struct fmt::formatter<B>
{
  template<typename ParseContext>
  constexpr auto parse(ParseContext& ctx)
  {
    return ctx.begin();
  }

  template<typename FormatContext>
  auto format(const B& b, FormatContext& ctx)
  {
    return format_to(ctx.out(), "B!");
  }
};

static void BM_fmt_ostream(benchmark::State& state)
{
  for (auto _ : state)
  {
    benchmark::DoNotOptimize(fmt::format("{}", A{}));
  }
}

static void BM_fmt_direct(benchmark::State& state)
{
  for (auto _ : state)
  {
    benchmark::DoNotOptimize(fmt::format("{}", B{}));
  }
}

BENCHMARK(BM_fmt_direct);
BENCHMARK(BM_fmt_ostream);

int main(int argc, char** argv)
{
  benchmark::Initialize(&argc, argv);
  benchmark::RunSpecifiedBenchmarks();
  return 0;
}

Output on my machine:

2019-10-29 12:15:57
Running ./fmt
Run on (4 X 3200 MHz CPU s)
CPU Caches:
  L1 Data 32K (x2)
  L1 Instruction 32K (x2)
  L2 Unified 256K (x2)
  L3 Unified 4096K (x1)
Load Average: 0.53, 0.50, 0.60
***WARNING*** CPU scaling is enabled, the benchmark real time measurements may be noisy and will incur extra overhead.
------------------------------------------------------
Benchmark               Time           CPU Iterations
------------------------------------------------------
BM_fmt_direct          42 ns         42 ns   16756571
BM_fmt_ostream        213 ns        213 ns    3327194

Upvotes: 15

Related Questions