132. abs_diff — Subtract Without Caring Which Side Is Bigger

Subtracting two unsigned integers and the smaller one comes first? Instant panic. a.abs_diff(b) returns the gap as a u* regardless of which side is bigger — no branching, no overflow.

The Problem

Unsigned subtraction in Rust panics in debug and wraps in release the moment the result would go negative. You end up writing the same branch over and over:

1
2
3
4
5
6

fn gap(a: u32, b: u32) -> u32 {
    if a > b { a - b } else { b - a }
}

assert_eq!(gap(10, 3), 7);
assert_eq!(gap(3, 10), 7);

It works, but it’s noise. And the same trick on signed integers has a sneakier bug: i32::MIN.abs_diff(i32::MAX) overflows an i32 — the gap doesn’t fit in the signed range.

After: abs_diff

Every integer type carries an abs_diff method that returns the unsigned gap directly. Signed inputs come back as the matching unsigned type, so the result always fits:

1
2
3
4
5
6

assert_eq!(10u32.abs_diff(3), 7);
assert_eq!(3u32.abs_diff(10), 7);

// Signed → unsigned, no overflow at the extremes
assert_eq!((-5i32).abs_diff(5), 10u32);
assert_eq!(i32::MIN.abs_diff(i32::MAX), u32::MAX);

No if, no checked_sub, no casting through i64 to dodge overflow. One call, one number.

Where It Earns Its Keep

Distance-style calculations are the obvious fit — anywhere “how far apart are these” is the real question and the sign is incidental:

1
2
3
4
5
6

fn manhattan(a: (i32, i32), b: (i32, i32)) -> u32 {
    a.0.abs_diff(b.0) + a.1.abs_diff(b.1)
}

assert_eq!(manhattan((1, 2), (4, 6)), 7);
assert_eq!(manhattan((-3, -3), (3, 3)), 12);

It also cleans up timestamp deltas, where one side is “now” and the other could be in the past or the future:

1
2
3
4
5

let scheduled: u64 = 1_700_000_000;
let actual:    u64 = 1_699_999_995;

let drift = scheduled.abs_diff(actual);
assert_eq!(drift, 5);

Whenever you catch yourself writing if a > b { a - b } else { b - a }, reach for abs_diff instead.