Can I fail compilation based on constexpr if?

Question

Is there a workaround to do something like this?

if constexpr (std::floating_point<T>) {}
else if constexpr (std::integral<T>) {}
...
else static_failure("Feature expansion needed");

because if I replace static_failure with static_assert, it needs the replication of all the above conditions (they are many and complicated) and it becomes ugly.

// It does not work properly: always fail
else static_assert(false, "Feature expansion needed");
// works fine, but it is ugly
else static_assert(std::floating_point<T> || std::integral<T> || ... || ... || ..., "Feature expansion needed");

I do not prefer a runtime behavior like:

throw "Feature expansion needed";

it needs the replication of all the above conditions What do you mean? — kiner_shah
– kiner_shah, Commented Apr 10, 2023 at 10:04
It doesn't need replication of all the above conditions, and you may as well use always_false<T> as workaround to static_assert issues. That's being said, you'd better put the full requirement as a requirement in the declaration of the function rather than use static_assert to notify that these conditions fail. — ALX23z
– ALX23z, Commented Apr 10, 2023 at 10:12
Here's the trick. In C++23 you can just use static_assert(false). — n. m. could be an AI
– n. m. could be an AI, Commented Apr 10, 2023 at 10:21
@n.m. Wow! The correct answer is yours! But g++ does not support it yet. open-std.org/jtc1/sc22/wg21/docs/papers/2022/p2593r0.html — Chameleon
– Chameleon, Commented Apr 10, 2023 at 10:28

Salgar · Accepted Answer · 2023-04-10 10:11:50Z

5

Yes, you just need the static assert to be dependant on T and always be false like this:

if constexpr (std::is_integral_v<T>) {}
else
    static_assert(sizeof(T) < 0, "Only works for integral T");

https://godbolt.org/z/Yhesd58rK

answered Apr 10, 2023 at 10:11

Salgar

7,8151 gold badge27 silver badges39 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

Jarod42 Over a year ago

static_assert(sizeof(T) < 0); is pedantically ill-formed NDR as static_assert(false); even if compilers fail to diagnose the former and succeed for the later.

Red.Wave · Accepted Answer · 2023-04-10 10:14:55Z

3

replace your last if constexpr with static_assert.

Instead of:

else if constexpr (last_case)
     //Do last case
else
     static_failure("prompt");

Do this:

else {
     static_assert(last_case,"prompt");
     //Do last case
};

Some proposals are made to legalize static_assert(false, "prompt");, but it is not yet part of std.

answered Apr 10, 2023 at 10:14

Red.Wave

4,77814 silver badges21 bronze badges

Comments

Jarod42 · Accepted Answer · 2023-04-10 15:27:47Z

1

In C++23, you might use static_assert(false); in non-instantiated context. Before, it would be ill-formed NDR, but most compiler do diagnostic on that.

if constexpr (std::floating_point<T>) { /*...*/ }
else if constexpr (std::integral<T>) { /*...*/ }
/*...*/
else static_assert(false, "Feature expansion needed");

Before C++23, you have to cheat, common legal way is with always_false:

template <typename T>
struct always_false : std::false_type {};

and then

if constexpr (std::floating_point<T>) { /*...*/ }
else if constexpr (std::integral<T>) { /*...*/ }
/*...*/
else static_assert(always_false{}, "Feature expansion needed");

answered Apr 10, 2023 at 15:27

Jarod42

228k15 gold badges215 silver badges355 bronze badges

1 Comment

Chameleon Over a year ago

About C++23, unfortunately static_assert(false, ...) is not supported as of g++12.2.

Collectives™ on Stack Overflow

Can I fail compilation based on constexpr if?

3 Answers 3

1 Comment

Comments

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

3 Answers 3

1 Comment

Comments

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related