I'm wondering if there is a way to make gtest understand a user-defined types libfmt's formatter, for the means of printing a readable error output?
I know how I can teach gtest to understand user-defined types via adding the stream insertion operator operator<<
for this very user-defined type, e.g.
std::ostream& operator<<(std::ostream& stream, CpuTimes const& anything);
std::ostream& operator<<(std::ostream& stream, CpuStats const& anything);
This is well documented here.
But I heavily use libfmt, which requires a formatting function to be implemented to produce a readable and printable output of a user-defined type. This so-called fmt::formatter
is actually a template specialization, e.g.
namespace fmt {
template <> struct formatter<CpuTimes> : basics::fmt::ParseContextEmpty {
format_context::iterator format(CpuTimes const& times, format_context& ctx);
};
template <> struct formatter<CpuStats> : basics::fmt::ParseContextEmpty {
format_context::iterator format(CpuStats const& stats, format_context& ctx);
};
template <> struct formatter<CpuLimits> : basics::fmt::ParseContextEmpty {
format_context::iterator format(CpuLimits const& limits, format_context& ctx);
};
} // namespace fmt
For gtest to understand this format, you have to write the same implementation for the operator<<
over and over for every type you want gtest to print properly, like
std::ostream& operator<<(std::ostream& stream, CpuTimes const& anything) {
return stream << fmt::format("{}", anything);
}
std::ostream& operator<<(std::ostream& stream, CpuStats const& anything) {
return stream << fmt::format("{}", anything);
}
std::ostream& operator<<(std::ostream& stream, CpuLimits const& anything) {
return stream << fmt::format("{}", anything);
}
Is there a way to spare my writing this boilerplate code?
CodePudding user response:
To avoid problems caused by ambiguity of PrintTo
or operator<<
namespace is needed.
Here is some demo when namespaces is used:
#include <fmt/format.h>
#include <gmock/gmock.h>
#include <gtest/gtest.h>
namespace me {
struct Foo {
int x = 0;
double y = 0;
};
bool operator==(const Foo& a, const Foo& b)
{
return a.x == b.x && a.y == b.y;
}
}
template <>
struct fmt::formatter<me::Foo> {
char presentation = 'f';
constexpr auto parse(format_parse_context& ctx) -> decltype(ctx.begin())
{
auto it = ctx.begin(), end = ctx.end();
if (it != end && (*it == 'f' || *it == 'e'))
presentation = *it ;
if (it != end && *it != '}')
throw format_error("invalid format");
return it;
}
template <typename FormatContext>
auto format(const me::Foo& p, FormatContext& ctx) -> decltype(ctx.out())
{
return presentation == 'f'
? format_to(ctx.out(), "({}, {:.1f})", p.x, p.y)
: format_to(ctx.out(), "({}, {:.1e})", p.x, p.y);
}
};
#if VERSION == 1
namespace me {
template <typename T>
void PrintTo(const T& value, ::std::ostream* os)
{
*os << fmt::format(FMT_STRING("{}"), value);
}
}
#elif VERSION == 2
namespace me {
template <typename T>
std::ostream& operator<<(std::ostream& out, const T& value)
{
::std::operator<<(out, fmt::format(FMT_STRING("{}"), value));
return out;
}
}
#endif
class MagicTest : public testing::Test { };
TEST_F(MagicTest, CheckFmtFormater)
{
EXPECT_EQ(fmt::format("{}", me::Foo {}), "(0, 0.0)");
}
TEST_F(MagicTest, FailOnPurpuse)
{
EXPECT_EQ(me::Foo {}, (me::Foo { 1, 0 }));
}
Dropping me
namespace cause all implementations to have ambiguity problems.
Note that namespace causes that argument-dependent lookup is used.