Nested Matchers in GTest - c++

Question

I would like to use some existing matchers in other matcher. I know about the MatcherInterface solution but I was wondering can I use matchers which were defined by MATCHER_P. If found this solution:
struct Foo
{
double x;
double y;
};
struct Bar
{
Foo foo;
int i;
};
MATCHER_P(EqFoo, foo, "")
{
::testing::Matcher<double> x_matcher = ::testing::DoubleNear(foo.x, 0.0001);
if (!x_matcher.MatchAndExplain(arg.x, result_listener))
{
return false;
}
::testing::Matcher<double> y_matcher = ::testing::DoubleNear(foo.y, 0.0001);
if (!y_matcher.MatchAndExplain(arg.y, result_listener))
{
return false;
}
return true;
}
MATCHER_P(EqBar, bar, "")
{
::testing::Matcher<Foo> foo_matcher = EqFooMatcherP<Foo>(bar.foo);
if (!foo_matcher.MatchAndExplain(arg.foo, result_listener))
{
return false;
}
if (bar.i != arg.i)
{
return false;
}
return true;
}
TEST_F(TestClass, BarTest)
{
Bar bar_val{{10.12, 76.43}, 78};
Bar bar_exp{{10.12, 99.99}, 78};
EXPECT_THAT(bar_val, EqBar(bar_exp));
}
I am just wondering, is there any better and nicer solution to
use my own MATCHER_P matcher in another one
use an original GTest matcher in another one.

Answer 1

The correct way is to use, as much as possible, the matchers from gtest/gmock. Only if there is no already provided matchers - use your own.
In your example - it is just as simple as this:
auto EqFoo(const Foo& expected)
{
return ::testing::AllOf(
::testing::Field(&Foo::x, ::testing::DoubleNear(expected.x, 0.0001)),
::testing::Field(&Foo::y, ::testing::DoubleNear(expected.y, 0.0001))
);
}
auto EqBar(const Bar& expected)
{
return ::testing::AllOf(
::testing::Field(&Bar::foo, EqFoo(expected.foo)),
::testing::Field(&Bar::i, expected.i)
);
}
More general approach is to use overloads:
auto MatchDouble(double expected)
{
return ::testing::DoubleNear(expected.x, 0.0001);
}
auto MatchFoo(::testing::Matcher<double> x, ::testing::Matcher<double> y)
{
return ::testing::AllOf(
::testing::Field(&Foo::x, x),
::testing::Field(&Foo::y, y)
);
}
auto MatchFoo(double x, double y)
{
return MatchFoo(MatchDouble(x), MatchDouble(y));
}
auto MatchBar(::testing::Matcher<Foo> foo, ::testing::Matcher<int> i)
{
return ::testing::AllOf(
::testing::Field(&Bar::foo, foo),
::testing::Field(&Bar::i, expected.i),
);
}
auto MatchBar(const Bar& expected)
{
return MatchBar(expected.foo, expected.i);
}
So your test:
TEST_F(TestClass, BarTest)
{
Bar bar_val{{10.12, 76.43}, 78};
Bar bar_exp{{10.12, 99.99}, 78};
EXPECT_THAT(bar_val, MatchBar(bar_exp));
// or - e.g. you can match only Bar::y if other things are irrelevant in your test
EXPECT_THAT(bar_val, MatchBar(MatchFoo(_, MatchDouble(2.001)), _);
}
Anyway - using MATCHER_P should be rather rare case, my own observation is that this macro is really overused.
In case your project is pre-C++14 - use ::testing::Matcher<T> instead of auto as return type for all of these functions.