Hi friends!
I am wondering if this should be considered a bug:
#[derive(Clone, Copy, Debug)]
#[repr(u8)]
enum Digit {
Zero = 0,
One,
Two,
Three,
Four,
Five,
Six,
Seven,
Eight,
Nine,
}
impl Digit {
fn make_digits_from(n: u64) -> Vec<Self> {
n.to_string()
.bytes()
.map(|b| unsafe { std::mem::transmute(b - b'0') })
.collect()
}
}
This generates this warning:
warning: multiple variants are never constructed
--> say/src/lib.rs:45:5
|
43 | enum Digit {
| ----- variants in this enum
44 | Zero = 0,
45 | One,
| ^^^
46 | Two,
| ^^^
47 | Three,
| ^^^^^
48 | Four,
| ^^^^
49 | Five,
| ^^^^
50 | Six,
| ^^^
51 | Seven,
| ^^^^^
52 | Eight,
| ^^^^^
53 | Nine,
| ^^^^
|
Yes, I know Digit is not public, it is being used/constructed in another function in order to create an English representation of the number (1234 -> "one thousand two hundred thirty-four"). However, it is clearly using and constructing each digit (it does successfully generate the English representation from Vec<Digit>). I think the issue lies with my use of std::mem::transmute, perhaps because the compiler doesn't check for the use of transmute resulting in all possible variants.
I see a potential issue with fixing this, which is that if one uses transmute, one has to manually check if all variants are actually covered. But this is already the case as it is, because you have to add #[allow(dead_code)] to get rid of the warnings anyway. So is there really a benefit to the compiler saying variants are "not constructed" in this case?
I appreciate any clarification the community might have on why this is the case, and if it is likely to be fixed.
I'm using rust 1.80.0 on NixOS x86_64 with rustup.