std.math or builtin: a tan2(rad) = (y,x) function for tan with quadrant information preserved

[expikr]

If

$$ \text{atan2}(y,x) = \theta $$

Then

$$ \text{tan2}(\theta) = (y,x) $$

Something like

pub fn tan2(rad: T) [2]T {
    return .{ @sin(rad) , @cos(rad) };
}

[mlugg]

atan2 lives in the standard library because it is a complex algorithm and not suited for a builtin. What would the advantage be of having such a tan2 function which is just computing @sin and @cos? It's not even much longer to inline the definition:

std.math.tan2(rad)
// vs
.{ @sin(rad), @cos(rad) }

...and it's arguably much clearer.

[lifthrasiir]

There is a potential speedup possible from calculating sin and cos together, and a number of math libraries do have sincos for this reason¹. Given this precedent, any equivalent function should be named std.math.sincos.

Footnotes

1. That made me wonder why atan2 is actually not asincos for a moment though ;-) Again, atan2 is also a very popular name for that particular function. ↩

[Vexu]

Zig's compiler-rt implements sincos so maybe it should be exposed as a builtin.