u64::saturating_div is useless

[ijackson]

pub const fn saturating_div(self, rhs: u64) -> u64

This function will panic if rhs is 0.

This makes no sense. The only way that u64 / u64 could overfllow is if rhs = 0.

The same is true for the other unsigned integer types. The signed types have the excuse that they wouldn't know whether to count division by zero as saturating to -ve or +ve maximum.

[Noratrieb]

I think these functions mostly exist for consistency

[ijackson]

That's all very well, but surely this one should return u64::MAX instead of panicking, on division by zero.

[Jules-Bertholet]

"Consistency" isn't usually sufficient justification for having a method. u64::abs doesn't exist, for example.

Perhaps we should just deprecate these?

[ryanavella]

The signed types have the excuse that they wouldn't know whether to count division by zero as saturating to -ve or +ve maximum.

I don't think this was the only reason. On the PR (#87921) for the std::num::Saturating type, the original implementation guaranteed x / 0 == MAX, but that was changed when @kennytm noted it could give surprising results for 0 / 0.

It's also worth noting that if u64::saturating_div saturated to u64::MAX instead of panicking, that it is somewhat redundant with u64::checked_div (the more general of the two)

[kennytm]

The signed types have the excuse that they wouldn't know whether to count division by zero as saturating to -ve or +ve maximum.

No, for signed type saturating_div makes sense because i32::MIN / -1 overflows. It has nothing to do with divide-by-zero (which also panics).

If u64::saturating_div is useless because x / 0 panics then so is u64::wrapping_div and u64::overflowing_div and u64::strict_div which are all equivalent.

[Jules-Bertholet]

If u64::saturating_div is useless because x / 0 panics

It's not useless just because of this, it's useless because it has identical behavior to u64::div in all cases. No one is proposing deprecating or changing saturating_div for the signed integers. (See my previous analogy to abs)

[kennytm]

@Jules-Bertholet All of u64::wrapping_div and u64::strict_div are identical to u64::div and u64::overflowing_div is just (x/y, false).

How is a non-existing function like u64::abs relevant? The consistency of u64::saturating_div is with u64::saturating_{add,sub,mul,...}, not i64::saturating_div.

(speaking of which there should also have been u64::unchecked_div and i64::unchecked_div which are currently missing)

[Jules-Bertholet]

All of u64::wrapping_div and u64::strict_div are identical to u64::div, and u64::overflowing_div is just (x/y, false).

Then IMO these should be deprecated also.

[AaronKutch]

No it is not useless, consistency in integer methods is extremely useful for macro writers. The only issue is that saturating_div possibly should have used the RISC-V convention where division by zero results in iX::MIN if the dividend is negative, 0 if zero, and iX::MAX if positive.

[Noratrieb]

Yeah, having this consistent integer methods makes writing macros easier (and it makes writing the standard library implementation macros easier :3), and they don't hurt. I think this should be closed.