to_signed()

constexpr auto Chusov::to_signed ( T  val )

constexpr

Attempts to converts an integral to its signed form without losing information and throws an exception, if that is not possible.

Template Parameters

T	A type of the value to convert.

Parameters

val	A value to convert to the signed form.

Returns

A signed value equal to val.

The function template is only defined for integral T, otherwise there is substitution failure.

If T is already signed, the function performs no action.

If T is unsigned, and the signed arithmetic on the target machine is implemented as two's complement, and bit sizes of signed and unsigned forms of T are equal, the function makes a static conversion and is marked noexcept.

Otherwise, the function tries to fit the value into the destination type at runtime as follows.

• If \(\forall x: 0 \le x \le M\), where M = std::numeric_limits<std::make_signed<T>>::max() > 0 is the maximal value that can be represented with the signed type associated with the unsigned T, then signed and unsigned representations of \(x\) are equivalent. That is, if val <= M, the function yields the signed value equal to val.
• \(\forall x: n + m \le x < n\), where m = std::numeric_limits<std::make_signed<T>>::min() < 0 is the minimal value that can be represented with the signed type, and n = std::numeric_limits<T>::max() + 1 is a number of elements within the additive group \(\mathbb{Z}_n\) represented by the unsigned type T, the value produced by the function is congruent to \((x - (n + m)) + m\) in the additive group \(\{m,\dots, M\}\). Therefore, if val >= n + m, the function yields the signed value (val - (n + m)) + m.
• In all other cases, i.e. \(\forall x: M < x < n + m\) the function fails throwing ArithmeticOverflowException.