wrapping negation -% should work on unsigned ints

[thejoshwolfe]

This is used in C in these algorithms: http://www.pcg-random.org/posts/bounded-rands.html (search for "-range").

Semantics:

const U = u32;
const I = i32;
var u: U = foo();
assert(-%u == @bitCast(U, -%@bitCast(I, u)));
assert(-%u == ~u +% 1);

I'm trying to port this code to Zig, and the workarounds with @bitCast or ~ obscure the origin and intent of the code.

[andrewrk]

I'm not convinced. Everywhere else in Zig, operations such as negation refer to the mathematical meaning, not a bitwise interpretation. Negation of an unsigned integer does not make sense mathematically. The @bitCast code above makes it clear.

I will re-open this issue if a counter-argument is provided.

[daurnimator]

Negation of an unsigned integer does not make sense mathematically

-%u should be equivalent to 0 -% u.

[andrewrk]

OK, when you put it that way, it makes sense, and I agree.