# Confusing @ Pattern matching **URL:** \ **Category:** help\ **Created:** [July 21, 2023, 7:02am UTC](https://users.rust-lang.org/t/confusing-pattern-matching/97271 "2023-07-21T07:02:58Z")\ **Posts on this page:** 8\ **Page:** 1
**Author:** ![TOETOE55](https://sea1.discourse-cdn.com/flex019/user_avatar/users.rust-lang.org/toetoe55/32/35203_2.png) [@TOETOE55](https://users.rust-lang.org/u/TOETOE55)\ **Post date:** [July 21, 2023, 7:02am UTC](https://users.rust-lang.org/t/confusing-pattern-matching/97271/1 "2023-07-21T07:02:58Z")
Why does the following code compile, but not after removing the comment. ```rust fn main() { let msg = Message::World { id: 123 }; match msg { x @ Message::World { id, .. } => println!("{x:?} {id}"), x => println!("{x:?}"), } } #[derive(Debug)] enum Message { // Hello { a: i32 }, World { id: i32 }, } ``` > **[Rust Playground](https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=dca3afd98e3f9f0b6eee0bdb190276f6)** > > A browser interface to the Rust compiler to experiment with the language ---
**Author:** ![hax10](https://sea1.discourse-cdn.com/flex019/user_avatar/users.rust-lang.org/hax10/32/33598_2.png) [@hax10](https://users.rust-lang.org/u/hax10)\ **Post date:** [July 21, 2023, 7:39am UTC](https://users.rust-lang.org/t/confusing-pattern-matching/97271/2 "2023-07-21T07:39:05Z")
According to the [Rust docs](https://doc.rust-lang.org/std/keyword.ref.html): > By default, [`match`](https://doc.rust-lang.org/std/keyword.match.html) statements consume all they can, which can sometimes be a problem, when you don’t really need the value to be moved and owned. i.e. in this piece of code ```rust match msg { x @ Message::World { id, .. } => println!("{x:?} {id}"), x => println!("{x:?}"), } ``` ownership of the enum `msg: Message` is moved into the first arm and `msg` will be dropped after you destructure it and extract your `id`. This means `msg` is no longer available in other match arms. This doesn't cause a problem when your enum only has one variant, because the second match arm is unreachable, but when you add the `Hello` variant to the `Message` enum, the compiler will complain about a value being used after it's moved. One possible solution is to match `msg` by reference rather than take ownership, i.e. ```rust match msg { ref x @ Message::World { id, .. } => println!("{x:?} {id}"), ref x @ Message::Hello { a, .. } => println!("{x:?} {a}"), } ``` Another would be derive `Copy` and `Clone` on your enum, which would allow Rust to make a copy of `msg` and take ownership of the copy. ---
**Author:** ![TOETOE55](https://sea1.discourse-cdn.com/flex019/user_avatar/users.rust-lang.org/toetoe55/32/35203_2.png) [@TOETOE55](https://users.rust-lang.org/u/TOETOE55)\ **Post date:** [July 21, 2023, 8:59am UTC](https://users.rust-lang.org/t/confusing-pattern-matching/97271/3 "2023-07-21T08:59:10Z")
> [@hax10](#): > > > which can sometimes be a problem, when you don’t really need the value to be moved and owned. Nop, borrow checks between different branches in pattern matching should not affect each other. ---
**Author:** ![Heliozoa](https://sea1.discourse-cdn.com/flex019/user_avatar/users.rust-lang.org/heliozoa/32/16524_2.png) [@Heliozoa](https://users.rust-lang.org/u/Heliozoa)\ **Post date:** [July 21, 2023, 9:18am UTC](https://users.rust-lang.org/t/confusing-pattern-matching/97271/4 "2023-07-21T09:18:58Z")
It seems that with an irrefutable pattern, the compiler does a partial move, but with a refutable pattern it doesn't. Not sure why, maybe someone with more compiler knowledge can explain. With a partial move, only the `id` is copied out and there's no problem accessing the enum afterwards, but without a partial move it's a bit like both sides of the @ are trying to take ownership of the whole struct and that's not allowed. ```rust fn main() { if let e @ E::A(s) = { E::A(String::new()) } {} } enum E { A(String), B, } ``` ```rust error[E0382]: use of moved value --> src/main.rs:2:21 | 2 | if let e @ E::A(s) = { E::A(String::new()) } {} | - ^ ----------------------- move occurs because value has type `E`, which does not implement the `Copy` trait | | | | | value used here after move | value moved here ``` vs ```rust fn main() { if let e @ E::A(s) = { E::A(String::new()) } {} } enum E { A(String), } ``` ```rust error[E0382]: use of partially moved value --> src/main.rs:2:12 | 2 | if let e @ E::A(s) = { E::A(String::new()) } {} | ^ - value partially moved here | | | value used here after partial move ``` ---
**Author:** ![TOETOE55](https://sea1.discourse-cdn.com/flex019/user_avatar/users.rust-lang.org/toetoe55/32/35203_2.png) [@TOETOE55](https://users.rust-lang.org/u/TOETOE55)\ **Post date:** [July 21, 2023, 11:44am UTC](https://users.rust-lang.org/t/confusing-pattern-matching/97271/6 "2023-07-21T11:44:03Z")
I seem that "the matching order" was different . For irrefutable @ pattern `x @ World { id: y, .. }`, it will be matched as `World { id, .. }` first, copy it `id` to `y`, then bind(move) it to `x`. But for normal pattern matching, it will firstly bind it to `x`, then match `World { id: y, .. }`(match after move). ---
**Author:** ![quinedot](https://sea1.discourse-cdn.com/flex019/user_avatar/users.rust-lang.org/quinedot/32/17693_2.png) [@quinedot](https://users.rust-lang.org/u/quinedot)\ **Post date:** [July 21, 2023, 6:57pm UTC](https://users.rust-lang.org/t/confusing-pattern-matching/97271/7 "2023-07-21T18:57:49Z")
Looks to be the issue below. Note that the title is inaccurate: it's a Rust 1.56+ feature, not an edition specific feature. [How the feature should work is underspecified IMO.](https://github.com/rust-lang/rust/issues/65490#issuecomment-841999009) > > > I tried the new "binding to sub-patterns" feature in the 2021 edition and it did… not work as I expected, or consistently. \[Here's a playground link\](https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=4a14b694497d8cece3bacb75fcdba5bf). > > First, I tried: > > \`\`\`rust > // NOT copy! > \#\[derive(Debug)\] > enum One { > A, > B(bool), > } > > fn main() { > let a = One::B(true); > > match a { > One::A =\> println!("A"), > other @ One::B(success) =\> { > println!("{:?}", other); > println!("{}", success); > } > } > } > \`\`\` > > I expected that even though the wrapping enum is not \`Copy\`, because the nested \`bool\` value \*is\*, it would be copied into \`success\`. However, I get the error: > > \`\`\`text > error\[E0382\]: use of moved value: \`a\` > --\> src/main.rs:13:24 > | > 9 | let a = One::B(true); > | - move occurs because \`a\` has type \`One\`, which does not implement the \`Copy\` trait > ... > 13 | other @ One::B(success) =\> { > | ---------------^^^^^^^- > | | | > | | value used here after move > | value moved here > > For more information about this error, try \`rustc --explain E0382\`. > error: could not compile \`playground\` due to previous error > \`\`\` > > Comparing this to the \[example in the release notes\](https://blog.rust-lang.org/2021/10/21/Rust-1.56.0.html#new-bindings-in-binding--pattern), where there's a non-\`Copy\` container (\`Matrix\`) with a \`Copy\`-able field (\`row\_len\`), I don't see why it shouldn't work the same way. > > Strangely (I think), if I \[remove the \`A\` variant\](https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=a77830e71e9a27ac9a3ee24062959f79), it works. So infallible patterns allow this. > > It will also work if I do \`ref other @ One::B(success)\`, but then I can't take ownership of \`other\`. > > Rust version: \`rustc 1.56.0 (09c42c458 2021-10-18)\` > Platform: Ubuntu 21.04 > Installed via Rustup ---
**Author:** ![vague](https://sea1.discourse-cdn.com/flex019/user_avatar/users.rust-lang.org/vague/32/21401_2.png) [@vague](https://users.rust-lang.org/u/vague)\ **Post date:** [July 22, 2023, 5:27am UTC](https://users.rust-lang.org/t/confusing-pattern-matching/97271/8 "2023-07-22T05:27:08Z")
> [@quinedot](#): > > [How the feature should work is underspecified IMO.](https://github.com/rust-lang/rust/issues/65490#issuecomment-841999009) > as a result, this feature (and the borrow checking rules around it) is very [well tested](https://github.com/rust-lang/rust/tree/master/src/test/ui/pattern/bindings-after-at). For the linked comment\[1\], I would rather say the feature is not well tested (or very little specified). And I raised [a type inconsistency problem on `@ binding`](https://github.com/rust-lang/rust/issues/101510) which was closed as not a bug. [Rust Playground](https://play.rust-lang.org/?version=nightly&mode=debug&edition=2021&gist=6fdf5d61b5081c9c7e2df62df05317a7) * * * 1. and a bit annoying to see a dead link there ☹ ---
**Author:** ![system](https://sea1.discourse-cdn.com/flex019/user_avatar/users.rust-lang.org/system/32/51372_2.png) [@system](https://users.rust-lang.org/u/system)\ **Post date:** [October 20, 2023, 5:27am UTC](https://users.rust-lang.org/t/confusing-pattern-matching/97271/9 "2023-10-20T05:27:41Z")
Closed on October 20, 2023, 5:27am UTC This topic was automatically closed 90 days after the last reply. We invite you to open a new topic if you have further questions or comments.