Repository navigation
Step::steps_between does not distinguish overflow and unimplemented (unstable) #48117
Description
It returns Option<size>: https://doc.rust-lang.org/std/iter/trait.Step.html#tymethod.steps_between
And the comment says it uses None for overflow: https://doc.rust-lang.org/src/core/iter/range.rs.html#29
But it also uses None for unimplemented https://doc.rust-lang.org/src/core/iter/range.rs.html#155
And as a result, Range needs to return (0, None) when it gets None: https://doc.rust-lang.org/src/core/iter/range.rs.html#235-240
It would be nice to either
- Require that the method is implemented accurately (easy for fundamental integers; harder if
Stepis expected to support graph walks, but maybe not harder thanPartialOrd) - Change the type to something else, maybe
Option<Option<usize>>or be more hint-like with(usize, Option<usize>) - Something else
cc #42168
Reacted by Felipe S. S. Schneider
Reacted by Felipe S. S. Schneider
Activity
CAD97 commentedon May 15, 2020
The links appear to have lost their original position.
Does the original concern still stand? (Does it stand with #69659?)
scottmcm commentedon Aug 4, 2020
@CAD97 Yes, I think there's still essentially the same problem -- the Invariants section still has
Nonefor both unreachable (steps_between(1, 0)) and overflow (steps_between(0_u128, 1<< 127)).So if this is used for
size_hint, then gettingNoneback fromsteps_betweenmeans you're stuck returning(0, None), when you'd rather be able to return either(0, Some(0))or(usize::MAX, None).scottmcm commentedon Jul 12, 2026
Looks like this was updated, but I realized it's unclear, at least to me, what to do if it's actually unreachable -- should that be marked distinctly? Step only requires PartialOrd, not Ord, so
aandbmight both have successors and predecessors but neither is reachable from the other.Or if
(0, None)is always a legal response, shouldn't it be a provided implementation likeIterator::step_by?