New Lint - Pointless mutation

[kinnison]

What it does

The lint should detect situations in which newly created values are mutated and then immediately dropped during assignments.

Advantage

Would spot situations such as the example below and warn the user that it is not acting as the author intended.

Drawbacks

Deliberate discarding of values may end up being slightly less ergonomic in some corner cases.

Example

This example is possibly a bit wordy, but is a simplified example of something encountered during a code review at work:

    struct Thingy {
        foo: Option<(u32, u32)>,
    }

    let mut thingy = Thingy { foo: None };
    thingy.foo = Some((0, 1));

    // This operation is effectively useless (no side-effects etc.)
    thingy.foo.unwrap().1 = 2;

    // What intuitively we'd expected to pass
    assert_eq!(thingy.foo.unwrap().1, 2);

The lint is clearly meant to trigger on the thingy.foo.unwrap().1 = 2; line and probably ought to suggest:

Could be written as:

    thingy.foo.as_mut().unwrap().1 = 2;

Though the exact details of the suggestion would likely tie the lint to only working on replacing the innards of Options or similar.

[kinnison]

Some extra thoughts:

If the thing being unwrapped wasn't Copy then we'd get a use-after-move error anyway, so this only seems to apply in cases where the assignment is into an lvalue which was copied.

[GuillaumeGomez]

I think it's a valid case.

[kinnison]

If someone could guide me into how I might go about building such a lint (eg. pointing me at anything which might be similar) I'm prepared to have a go at a solution myself; though it'd be my first time in the clippy codebase.

[leonardo-m]

thingy.foo.unwrap().1 = 2;

Perhaps it's rustc that should warn for code like this, instead of Clippy?

[kinnison]

thingy.foo.unwrap().1 = 2;

Perhaps it's rustc that should warn for code like this, instead of Clippy?

I'd be quite content if it were rustc itself which warned, rather than it being a Clippy extra, but I'm not really in a position to say which would be better.