Exact casts from float to int?

[BartMassey]

Unless I'm missing something, casts from float to int are currently missing. This is too bad, as these are some of the most problematic casts as provides. (Edit: I was missing something. See below.)
Suggested semantics are to only convert from floats that are exact integers (only zero bits after the decimal point) that will fit in the target int.

No prize for guessing what Rust does with (0.0f32 / 0.0) as u32. 🙂

[dhardy]

They're not missing, they're just written differently: x.cast_nearest(), x.cast_floor() etc.

Yes, as fails to fail and converts not-a-number into a number...

[BartMassey]

Ah yes, I'd forgotten. Apologies.

Could we add .cast_exact() or somesuch to the possibilities, that only casts if the "float" is already an exact integer? Or add this case to plain ol' .cast()? Maybe it's not worth the work, but I can think of some algorithms that would be surprised if their floating calculations didn't yield an integer. Ofc given floating rounding errors I suppose the next thing would be a delta…

Maybe .cast_delta(delta: float). Or maybe just make me check ranges before calling .cast_nearest() and call it a day.

Feel free to close this as wontfix, which is probably the right answer. Again, apologies.

[BartMassey]

Yes, as fails to fail and converts not-a-number into a number...

But what number? 🙂 I had a guess that turned out to be right, since there's only one "reasonable" choice; the Rust reference confirms it as canon. Ugh. I mean, I guess the Romans thought it was not a number, which should be good enough for me.

Off-topic, but I find the whole thing as funny as it is sad. Very grateful for your work on this crate, which I plan to use as my go-to tool for numeric conversions henceforth.

[dhardy]

that only casts if the "float" is already an exact integer

It's a bit too easy to end up with floats that are not quite ints though IMO, e.g.:

fn main() {
    let x = (1.0 - 0.9) * 10.0;
    println!("{}", x);
    println!("x == 1: {}", x == 1.0);
}

But what number? slightly_smiling_face

Any number? Define f(x) = nx, then f(x)/x == n for all x != 0. If you try extending that to 0 you get 0/0 == n.

[dhardy]

So...

Could we add .cast_exact() or somesuch to the possibilities, that only casts if the "float" is already an exact integer?

Technically, yes. Practically, I'm not sure if it's sufficiently useful. Perhaps we'll see if there are further requests / justifications.

Maybe .cast_delta(delta: float)

This isn't an approximate-equality library so my gut feeling is that it doesn't belong here, though that's not a big thing to add.