Lint double negation !count != 0

[idubrov]

What it does

I think, clippy should lint the following code:

if !count != 0 {
}

My brain melts trying to read this expression. It's not that difficult to read, just something immediately throwing me off.

Drawbacks

None.

Example

if !count != 0 {
}

Could be written as:

if count == 0 {
}

[ghost]

This should probably be an enhancement to nonminimal_bool

[sim031999]

Hello, I want to work on this ! But I'm new to open source and would need help getting started

[flip1995]

Hey @sim031999, welcome to Clippy! We have documentation in the doc directory of this repository on how Clippy works. For this issue the adding_lints documentation (especially the first three sections, Setup -> Testing) explains best how to get started with hacking on Clippy.

https://github.com/rust-lang/rust-clippy/blob/master/doc/common_tools_writing_lints.md This document explains some reoccurring patterns you'll find in Clippy with code examples.

If you have any questions, feel free to ask here, in discord, or open a WIP PR.

[sim031999]

Hi @flip1995, I have figured out that the following file ( after line 115) needs changes: https://github.com/rust-lang/rust-clippy/blob/master/clippy_lints/src/booleans.rs#L115. But the quine_mc_cluskey technique which is used in it, is not there for !=, can you suggest how do I proceed further?

[flip1995]

You can't use the quine_mc_clusky check for this. Just implement a check based on the AST/HIR structure after this line:

rust-clippy/clippy_lints/src/booleans.rs

Line 68 in bb67423

NonminimalBoolVisitor { cx }.visit_body(body)

Maybe not in the check_fn, but in the check_expr method.

[ghost]

I think I was mistaken about extending nonminimal_bool because I mistook !count != 0 for !(count != 0). I don't think any code in nonminimal_bool will help with this.

Maybe a better fit would be to add this to bit_mask module since there is a lot of bit comparison logic in there.

[flip1995]

I think the issue meant !(count != 0), which I get from the example.

Replacing !count != 0 with count == 0 would change the semantics, whereas replacing !(count != 0) with count == 0 wouldn't.

@idubrov can you clarify here?

[idubrov]

Right, in my case it was an integer variable, so I made an assumption that expression was parsed as !(count != 0) (and this definitely was implied meaning in our code). However, now I'm totally confused:

fn main() {
    let count = 0;
    // Original expression
    if !count != 0 {
        eprintln!("hello!");
    }
    // This is what I thought it means to compiler
    if !(count != 0) {
        eprintln!("hello!");
    }
    // This compiles? Why???
    if (!count) != 0 {
        eprintln!("hello!");
    }
    // This does not compile (makes sens; count is not a `bool`)
    if (!count) && true {
        eprintln!("hello!");
    }
}

https://play.rust-lang.org/?version=stable&mode=debug&edition=2018&gist=a141aa1438ec32ef629a9a68638b8739

[flip1995]

!count returns the bitwise inverse of count (Ones' complement).

I just saw that !(x != y) already gets linted by nonminimal_bool. Playground

I think there is already an issue open for suspicious uses of the Not/! operations in integer comparisons, but I can't find it.

[idubrov]

Ah, got it! The fact that ! works for bitwise operators was a surprise (though, know I remember I knew that before...).

[flip1995]

Anyway, this is a new lint. The behavior of the lint would be: Lint when there is !var inside a comparison (!=/==) of integral types, but not if it is on a literal (i.e. count != !0 is fine). The bad_mask module should be good for this lint.

[idubrov]

Yeah; now that I understand how it works, it actually turned out to be a logic error in our case!

Branch was intended to do count == 0, but effectively, it was doing if count != 0xffff_ffff_ffff_ffff_usize, which didn't make sense.

[DevAccentor]

@rustbot claim

[MorrisLaw]

@rustbot claim

[GuillaumeGomez]

Opened #12248.