Test that default constructed objects are zero

[NAThompson]

In a generic context, we often need to compare arithmetic types to zero. However, mp-units types cannot be default constructed from (say) double, so we cann write (e.g.)

if (x < T(0)) {
  // ...
}

Instead, we have to use

constexpr auto ZERO = T{};
if (x < ZERO) {
   // ...
}

This begs the question: Do we have a guarantee that ZERO==0.0? Add a unit test that verifies this behavior.

[JohelEGP]

I think these should go before

mp-units/test/unit_test/static/quantity_test.cpp

Line 122 in 7a270a6

// construction from a value

mp-units/test/unit_test/static/quantity_point_test.cpp

Line 310 in 7a270a6

// construction from a value

And test that (0 * m).numerical_value_in(m) == quantity<m, int>::zero().
The earlier tests already check that zero returns 0 or 0.0.

[JohelEGP]

I was under the impression that

mp-units/test/unit_test/static/quantity_test.cpp

Lines 58 to 63 in 7a270a6

	static_assert(std::is_trivially_default_constructible_v<quantity<isq::length[m]>>);
	static_assert(std::is_trivially_copy_constructible_v<quantity<isq::length[m]>>);
	static_assert(std::is_trivially_move_constructible_v<quantity<isq::length[m]>>);
	static_assert(std::is_trivially_copy_assignable_v<quantity<isq::length[m]>>);
	static_assert(std::is_trivially_move_assignable_v<quantity<isq::length[m]>>);
	static_assert(std::is_trivially_destructible_v<quantity<isq::length[m]>>);

mp-units/test/unit_test/static/quantity_test.cpp

Line 72 in 7a270a6

static_assert(std::is_trivially_copyable_v<quantity<isq::length[m]>>);

implied that.
But thinking about it some more, I'm not too convinced that they guarantee it.

[NAThompson]

I think these should go before

mp-units/test/unit_test/static/quantity_test.cpp

Line 122 in 7a270a6

// construction from a value

mp-units/test/unit_test/static/quantity_point_test.cpp

Line 310 in 7a270a6

// construction from a value

I of course have no objection to doing this, but then I'd have to bring in the math.h header into quantity_test.cpp. Feels like it would decrease the orthogonality of the test suite.

[JohelEGP]

quantity_test.cpp already tests that specializations of quantity model std::default_initializable.
It also tests quantity::zero.
You shouldn't need math.h to connect both in quantity_test.cpp.

[NAThompson]

It appears that it is unclear what this PR accomplishes-if anything. I think this can safely be closed.

If anyone finds compelling value in it I'll reopen.

[mpusz]

However, mp-units types cannot be default constructed from (say) double, so we cann write (e.g.)

if (x < T(0)) {
  // ...
}

You actually do not mean a default initialization but value/zero-initalization (https://en.cppreference.com/w/cpp/language/zero_initialization) here.

In generic code, you can always write:

if (x < T{}) {
  // ...
}

A more detailed discussion on this subject can be found at https://wg21.link/p2982R1#comparison-against-zero. We will discuss this in the Committee and check if there is a guidance in any direction.