From ae692174244c506d03d4a4d5928081f8fcf31919 Mon Sep 17 00:00:00 2001 From: Conrad Ludgate Date: Thu, 10 Feb 2022 09:47:54 +0000 Subject: [PATCH 01/13] begin rewriting chapter on subtyping --- src/subtyping.md | 40 ++++++++++++++++++++++++++++++++++++++++ 1 file changed, 40 insertions(+) diff --git a/src/subtyping.md b/src/subtyping.md index 79b29be..8c8a180 100644 --- a/src/subtyping.md +++ b/src/subtyping.md @@ -1,5 +1,45 @@ # Subtyping and Variance +Rust uses lifetimes to track the relationships between borrows and ownership. +However, a naive implementation of lifetimes would be either too restrictive, +or permit undefined behaviour. Let's see a few examples: + +```rust,ignore +fn debug<'a>(a: &'a str, b: &'a str) { + println!("a = {:?} b = {:?}", a, b) +} + +fn main() { + let a: &'static str = "hello"; + { + let b = String::from("world"); + let b = &b; // 'b has a shorter lifetime than 'static + debug(a, b); + } +} +``` + +In an overly restrictive implementation of lifetimes, since `a` and `b` have differeing lifetimes, +we might see the following error: + +```text +error[E0308]: mismatched types + --> src/main.rs:6:16 + | +6 | debug(a, b); + | ^ + | | + | expected `&'static str`, found struct `&'b str` +``` + +This is over-restrictive. In this case, what we want is to accept any type that lives "at least as long" as `<'a>`. +This is what subtyping is intended to fix. + +Let's define lifetime `'a` to be a `subtype` of lifetime `'b`, if and only if `'a` lives _at least as long_ as `'b`. +We will denote this as `'a: 'b` + +--- + Subtyping is a relationship between types that allows statically typed languages to be a bit more flexible and permissive. From 9c17e30bf9a298a8d12ffb1c923aaa260a19431a Mon Sep 17 00:00:00 2001 From: Conrad Ludgate Date: Thu, 10 Feb 2022 13:54:55 +0000 Subject: [PATCH 02/13] continue --- src/subtyping.md | 236 +++++++++++++++++------------------------------ 1 file changed, 87 insertions(+), 149 deletions(-) diff --git a/src/subtyping.md b/src/subtyping.md index 8c8a180..ce01d15 100644 --- a/src/subtyping.md +++ b/src/subtyping.md @@ -2,11 +2,41 @@ Rust uses lifetimes to track the relationships between borrows and ownership. However, a naive implementation of lifetimes would be either too restrictive, -or permit undefined behaviour. Let's see a few examples: +or permit undefined behavior. -```rust,ignore -fn debug<'a>(a: &'a str, b: &'a str) { - println!("a = {:?} b = {:?}", a, b) +In order to allow flexible usage of lifetimes +while also preventing mis-use, Rust uses a combination of **Subtyping** and **Variance**. + +## Subtyping + +Subtyping is the idea that one type can be a *subtype* of another. +Let's define that `A: B` is equivalent to saying '`A` is a subtype of `B`'. +What this is suggesting to us is that the set of *requirements* that `B` defines +are completely satisfied by `A`. `A` may then have more requirements. + +An example of simple subtyping that exists in the language are [supertraits](https://doc.rust-lang.org/stable/book/ch19-03-advanced-traits.html?highlight=supertraits#using-supertraits-to-require-one-traits-functionality-within-another-trait) + +```rust +use std::fmt; + +trait OutlinePrint: fmt::Display { + fn outline_print(&self) { + todo!() + } +} +``` + +Here, we have that `OutlinePrint: fmt::Display` (`OutlinePrint` is a *subtype* of `Display`), +because it has all the requirements of `fmt::Display`, plus the `outline_print` function. + +However, subtyping in traits is not that interesting in the case of Rust. +Here in the nomicon, we're going to focus more with how subtyping interacts with **lifetimes** + +Take this example + +```rust +fn debug(a: T, b: T) { + println!("a = {:?} b = {:?}", a, b); } fn main() { @@ -24,163 +54,73 @@ we might see the following error: ```text error[E0308]: mismatched types - --> src/main.rs:6:16 - | -6 | debug(a, b); - | ^ - | | - | expected `&'static str`, found struct `&'b str` + --> src/main.rs:10:16 + | +10 | debug(a, b); + | ^ + | | + | expected `&'static str`, found struct `&'b str` ``` -This is over-restrictive. In this case, what we want is to accept any type that lives "at least as long" as `<'a>`. -This is what subtyping is intended to fix. - -Let's define lifetime `'a` to be a `subtype` of lifetime `'b`, if and only if `'a` lives _at least as long_ as `'b`. -We will denote this as `'a: 'b` - ---- +This is over-restrictive. In this case, what we want is to accept any type that lives *at least as long* as `'b`. +Let's try using subtyping with our lifetimes. -Subtyping is a relationship between types that allows statically typed -languages to be a bit more flexible and permissive. +Let's define a lifetime to have the a simple set of requirements: `'a` defines a region of code in which a value will be alive. +Now that we have a defined set of requirements for lifetimes, we can define how they relate to each other. +`'a: 'b` if and only if `'a` defines a region of code that **completely contains** `'b`. -Subtyping in Rust is a bit different from subtyping in other languages. This -makes it harder to give simple examples, which is a problem since subtyping, -and especially variance, is already hard to understand properly. As in, -even compiler writers mess it up all the time. +`'a` may define a region larger than `'b`, but that still fits our definition. +Going back to our example above, we can say that `'static: 'b`. -To keep things simple, this section will consider a small extension to the -Rust language that adds a new and simpler subtyping relationship. After -establishing concepts and issues under this simpler system, -we will then relate it back to how subtyping actually occurs in Rust. - -So here's our simple extension, *Objective Rust*, featuring three new types: +For now, let's accept the idea that subtypes of lifetimes can be transitive (more on this in [Variance](#variance>)), +eg. `&'static str` is a subtype of `&'b str`, then we can let them coerce, and then the example above will compile ```rust -trait Animal { - fn snuggle(&self); - fn eat(&mut self); -} - -trait Cat: Animal { - fn meow(&self); +fn debug(a: T, b: T) { + println!("a = {:?} b = {:?}", a, b); } -trait Dog: Animal { - fn bark(&self); -} -``` - -But unlike normal traits, we can use them as concrete and sized types, just like structs. - -Now, say we have a very simple function that takes an Animal, like this: - - -```rust,ignore -fn love(pet: Animal) { - pet.snuggle(); +fn main() { + let a: &'static str = "hello"; + { + let b = String::from("world"); + let b = &b; // 'b has a shorter lifetime than 'static + debug(a, b); // a silently converts from `&'static str` into `&'b str` + } } ``` -By default, static types must match *exactly* for a program to compile. As such, -this code won't compile: - - -```rust,ignore -let mr_snuggles: Cat = ...; -love(mr_snuggles); // ERROR: expected Animal, found Cat -``` - -Mr. Snuggles is a Cat, and Cats aren't *exactly* Animals, so we can't love him! 😿 - -This is annoying because Cats *are* Animals. They support every operation -an Animal supports, so intuitively `love` shouldn't care if we pass it a `Cat`. -We should be able to just **forget** the non-animal parts of our `Cat`, as they -aren't necessary to love it. - -This is exactly the problem that *subtyping* is intended to fix. Because Cats are just -Animals **and more**, we say Cat is a *subtype* of Animal (because Cats are a *subset* -of all the Animals). Equivalently, we say that Animal is a *supertype* of Cat. -With subtypes, we can tweak our overly strict static type system -with a simple rule: anywhere a value of type `T` is expected, we will also -accept values that are subtypes of `T`. - -Or more concretely: anywhere an Animal is expected, a Cat or Dog will also work. - -As we will see throughout the rest of this section, subtyping is a lot more complicated -and subtle than this, but this simple rule is a very good 99% intuition. And unless you -write unsafe code, the compiler will automatically handle all the corner cases for you. - -But this is the Rustonomicon. We're writing unsafe code, so we need to understand how -this stuff really works, and how we can mess it up. - -The core problem is that this rule, naively applied, will lead to *meowing Dogs*. That is, -we can convince someone that a Dog is actually a Cat. This completely destroys the fabric -of our static type system, making it worse than useless (and leading to Undefined Behavior). - -Here's a simple example of this happening when we apply subtyping in a completely naive -"find and replace" way. +## Variance - -```rust,ignore -fn evil_feeder(pet: &mut Animal) { - let spike: Dog = ...; +Above, we glossed over the fact that `'static: 'b` implied that `&'static T: &'b T`. This uses a property known as variance. +It's not always as simple as this example though, to understand that let's try extend this example a bit - // `pet` is an Animal, and Dog is a subtype of Animal, - // so this should be fine, right..? - *pet = spike; +```rust,compile_fail +fn debug(a: &mut T, b: T) { + *a = b; } fn main() { - let mut mr_snuggles: Cat = ...; - evil_feeder(&mut mr_snuggles); // Replaces mr_snuggles with a Dog - mr_snuggles.meow(); // OH NO, MEOWING DOG! + let mut a: &'static str = "hello"; + { + let b = String::from("world"); + let b = &b; + debug(&mut a, b); + } } ``` -Clearly, we need a more robust system than "find and replace". That system is *variance*, -which is a set of rules governing how subtyping should compose. Most importantly, variance -defines situations where subtyping should be disabled. +This has a memory bug in it. -But before we get into variance, let's take a quick peek at where subtyping actually occurs in -Rust: *lifetimes*! - -> NOTE: The typed-ness of lifetimes is a fairly arbitrary construct that some -> disagree with. However it simplifies our analysis to treat lifetimes -> and types uniformly. - -Lifetimes are just regions of code, and regions can be partially ordered with the *contains* -(outlives) relationship. Subtyping on lifetimes is in terms of that relationship: -if `'big: 'small` ("big contains small" or "big outlives small"), then `'big` is a subtype -of `'small`. This is a large source of confusion, because it seems backwards -to many: the bigger region is a *subtype* of the smaller region. But it makes -sense if you consider our Animal example: Cat is an Animal *and more*, -just as `'big` is `'small` *and more*. - -Put another way, if someone wants a reference that lives for `'small`, -usually what they actually mean is that they want a reference that lives -for *at least* `'small`. They don't actually care if the lifetimes match -exactly. So it should be ok for us to **forget** that something lives for -`'big` and only remember that it lives for `'small`. - -The meowing dog problem for lifetimes will result in us being able to -store a short-lived reference in a place that expects a longer-lived one, -creating a dangling reference and letting us use-after-free. - -It will be useful to note that `'static`, the forever lifetime, is a subtype of -every lifetime because by definition it outlives everything. We will be using -this relationship in later examples to keep them as simple as possible. - -With all that said, we still have no idea how to actually *use* subtyping of lifetimes, -because nothing ever has type `'a`. Lifetimes only occur as part of some larger type -like `&'a u32` or `IterMut<'a, u32>`. To apply lifetime subtyping, we need to know -how to compose subtyping. Once again, we need *variance*. - -## Variance +If we were to expand this out, we'd see that we're trying to assign a `&'b str` into a `&'static str`, +but the problem is that as soon as `b` goes out of scope, `a` is now invalid, even though it's supposed to have a `'static` lifetime. -Variance is where things get a bit complicated. +However, the implementation of `debug` is valid. +Therefore, this must mean that `&mut &'static str` should **not** a *subtype* of `&mut &'b str`, +even if `'static` is a subtype of `'b`. -Variance is a property that *type constructors* have with respect to their -arguments. A type constructor in Rust is any generic type with unbound arguments. +Variance is the way that Rust defines the transitivity of subtypes through their *type constructor*. +A type constructor in Rust is any generic type with unbound arguments. For instance `Vec` is a type constructor that takes a type `T` and returns `Vec`. `&` and `&mut` are type constructors that take two inputs: a lifetime, and a type to point to. @@ -192,20 +132,18 @@ A type constructor F's *variance* is how the subtyping of its inputs affects the subtyping of its outputs. There are three kinds of variance in Rust. Given two types `Sub` and `Super`, where `Sub` is a subtype of `Super`: -* `F` is *covariant* if `F` is a subtype of `F` (subtyping "passes through") -* `F` is *contravariant* if `F` is a subtype of `F` (subtyping is "inverted") -* `F` is *invariant* otherwise (no subtyping relationship exists) +* F is **covariant** if `F` is a subtype of `F` (the subtype property is passed through) +* F is **contravariant** if `F` is a subtype of `F` (the subtype property is "inverted") +* F is **invariant** otherwise (no subtyping relationship exists) -If `F` has multiple type parameters, we can talk about the individual variances -by saying that, for example, `F` is covariant over `T` and invariant over `U`. +If we remember from the above examples, +it was ok for us to treat `&'a T` as a subtype of `&'b T` if `'a: 'b`, +therefore we can say that `&'a T` is *covariant* over `'a`. -It is very useful to keep in mind that covariance is, in practical terms, "the" -variance. Almost all consideration of variance is in terms of whether something -should be covariant or invariant. Actually witnessing contravariance is quite difficult -in Rust, though it does in fact exist. +Also, we saw that it was not ok for us to treat `&mut &'a T` as a subtype of `&mut &'b T`, +therefore we can say that `&mut T` is *invariant* over `T` -Here is a table of important variances which the rest of this section will be devoted -to trying to explain: +Here is a table of some other type constructors and their variances: | | | 'a | T | U | |---|-----------------|:---------:|:-----------------:|:---------:| From 0492daf82c2a3653341265c2075d90547faf051b Mon Sep 17 00:00:00 2001 From: Conrad Ludgate Date: Thu, 10 Feb 2022 14:08:59 +0000 Subject: [PATCH 03/13] clarify some points --- src/subtyping.md | 12 +++++++----- 1 file changed, 7 insertions(+), 5 deletions(-) diff --git a/src/subtyping.md b/src/subtyping.md index ce01d15..3792cbe 100644 --- a/src/subtyping.md +++ b/src/subtyping.md @@ -5,14 +5,16 @@ However, a naive implementation of lifetimes would be either too restrictive, or permit undefined behavior. In order to allow flexible usage of lifetimes -while also preventing mis-use, Rust uses a combination of **Subtyping** and **Variance**. +while also preventing their misuse, Rust uses a combination of **Subtyping** and **Variance**. ## Subtyping Subtyping is the idea that one type can be a *subtype* of another. -Let's define that `A: B` is equivalent to saying '`A` is a subtype of `B`'. -What this is suggesting to us is that the set of *requirements* that `B` defines -are completely satisfied by `A`. `A` may then have more requirements. + +Let's define that `Sub` is a subtype of `Super` (we'll be using the notation `Sub: Super` throughout this chapter) + +What this is suggesting to us is that the set of *requirements* that `Super` defines +are completely satisfied by `Sub`. `Sub` may then have more requirements. An example of simple subtyping that exists in the language are [supertraits](https://doc.rust-lang.org/stable/book/ch19-03-advanced-traits.html?highlight=supertraits#using-supertraits-to-require-one-traits-functionality-within-another-trait) @@ -72,7 +74,7 @@ Now that we have a defined set of requirements for lifetimes, we can define how `'a` may define a region larger than `'b`, but that still fits our definition. Going back to our example above, we can say that `'static: 'b`. -For now, let's accept the idea that subtypes of lifetimes can be transitive (more on this in [Variance](#variance>)), +For now, let's accept the idea that subtypes of lifetimes can be transitive (more on this in [Variance](#variance)), eg. `&'static str` is a subtype of `&'b str`, then we can let them coerce, and then the example above will compile ```rust From a43237778a577e5f162f31f9890e0b1caa78565a Mon Sep 17 00:00:00 2001 From: Conrad Ludgate Date: Thu, 10 Feb 2022 15:20:18 +0000 Subject: [PATCH 04/13] more explanations --- src/subtyping.md | 171 +++++++++++++++++++++++++---------------------- 1 file changed, 92 insertions(+), 79 deletions(-) diff --git a/src/subtyping.md b/src/subtyping.md index 3792cbe..0bda409 100644 --- a/src/subtyping.md +++ b/src/subtyping.md @@ -42,11 +42,11 @@ fn debug(a: T, b: T) { } fn main() { - let a: &'static str = "hello"; + let hello: &'static str = "hello"; { - let b = String::from("world"); - let b = &b; // 'b has a shorter lifetime than 'static - debug(a, b); + let world = String::from("world"); + let world = &world; // 'b has a shorter lifetime than 'static + debug(hello, world); } } ``` @@ -58,10 +58,10 @@ we might see the following error: error[E0308]: mismatched types --> src/main.rs:10:16 | -10 | debug(a, b); - | ^ - | | - | expected `&'static str`, found struct `&'b str` +10 | debug(hello, world); + | ^ + | | + | expected `&'static str`, found struct `&'b str` ``` This is over-restrictive. In this case, what we want is to accept any type that lives *at least as long* as `'b`. @@ -83,11 +83,11 @@ fn debug(a: T, b: T) { } fn main() { - let a: &'static str = "hello"; + let hello: &'static str = "hello"; { - let b = String::from("world"); - let b = &b; // 'b has a shorter lifetime than 'static - debug(a, b); // a silently converts from `&'static str` into `&'b str` + let world = String::from("world"); + let world = &world; // 'b has a shorter lifetime than 'static + debug(hello, world); // a silently converts from `&'static str` into `&'b str` } } ``` @@ -98,26 +98,25 @@ Above, we glossed over the fact that `'static: 'b` implied that `&'static T: &'b It's not always as simple as this example though, to understand that let's try extend this example a bit ```rust,compile_fail -fn debug(a: &mut T, b: T) { - *a = b; +fn assign(input: &mut T, val: T) { + *input = val; } fn main() { - let mut a: &'static str = "hello"; + let mut hello: &'static str = "hello"; { - let b = String::from("world"); - let b = &b; - debug(&mut a, b); + let world = String::from("world"); + assign(&mut hello, &world); } } ``` -This has a memory bug in it. +If this were to compile, this would have a memory bug. If we were to expand this out, we'd see that we're trying to assign a `&'b str` into a `&'static str`, but the problem is that as soon as `b` goes out of scope, `a` is now invalid, even though it's supposed to have a `'static` lifetime. -However, the implementation of `debug` is valid. +However, the implementation of `assign` is valid. Therefore, this must mean that `&mut &'static str` should **not** a *subtype* of `&mut &'b str`, even if `'static` is a subtype of `'b`. @@ -149,21 +148,21 @@ Here is a table of some other type constructors and their variances: | | | 'a | T | U | |---|-----------------|:---------:|:-----------------:|:---------:| -| * | `&'a T ` | covariant | covariant | | -| * | `&'a mut T` | covariant | invariant | | -| * | `Box` | | covariant | | +| | `&'a T ` | covariant | covariant | | +| | `&'a mut T` | covariant | invariant | | +| | `Box` | | covariant | | | | `Vec` | | covariant | | -| * | `UnsafeCell` | | invariant | | +| | `UnsafeCell` | | invariant | | | | `Cell` | | invariant | | -| * | `fn(T) -> U` | | **contra**variant | covariant | +| | `fn(T) -> U` | | **contra**variant | covariant | | | `*const T` | | covariant | | | | `*mut T` | | invariant | | -The types with \*'s are the ones we will be focusing on, as they are in -some sense "fundamental". All the others can be understood by analogy to the others: +Some of these can be explained simply in relation to the others: * `Vec` and all other owning pointers and collections follow the same logic as `Box` * `Cell` and all other interior mutability types follow the same logic as `UnsafeCell` +* `UnsafeCell` having interior mutability gives it the same variance properties as `&mut T` * `*const T` follows the logic of `&T` * `*mut T` follows the logic of `&mut T` (or `UnsafeCell`) @@ -177,8 +176,72 @@ For more types, see the ["Variance" section][variance-table] on the reference. > take references with specific lifetimes (as opposed to the usual "any lifetime", > which gets into higher rank lifetimes, which work independently of subtyping). -Ok, that's enough type theory! Let's try to apply the concept of variance to Rust -and look at some examples. +Now that we have some more formal understanding of variance, +let's go through some more examples in more detail. + +```rust,compile_fail +fn assign(input: &mut T, val: T) { + *input = val; +} + +fn main() { + let mut hello: &'static str = "hello"; + { + let world = String::from("world"); + assign(&mut hello, &world); + } +} +``` + +And what do we get when we run this? + +```text +error[E0597]: `world` does not live long enough + --> src/main.rs:9:28 + | +6 | let mut hello: &'static str = "hello"; + | ------------ type annotation requires that `world` is borrowed for `'static` +... +9 | assign(&mut hello, &world); + | ^^^^^^ borrowed value does not live long enough +10 | } + | - `world` dropped here while still borrowed +``` + +Good, it doesn't compile! Let's break down what's happening here in detail. + +First let's look at the `assign` function: + +```rust +fn assign(input: &mut T, val: T) { + *input = val; +} +``` + +All it does is take a mutable reference and a value and overwrite the referent with it. +What's important about this function is that it creates a type equality constraint. It +clearly says in its signature the referent and the value must be the *exact same* type. + +Meanwhile, in the caller we pass in `&mut &'static str` and `&'spike_str str`. + +Because `&mut T` is invariant over `T`, the compiler concludes it can't apply any subtyping +to the first argument, and so `T` must be exactly `&'static str`. + +This is counter to the `&T` case + +```rust +fn debug(a: T, b: T) { + println!("a = {:?} b = {:?}", a, b); +} +``` + +Where similarly `a` and `b` must have the same type `T`. +But since `&'a T` *is* covariant over `'a`, we are allowed to perform subtyping. +So the compiler decides that `&'static str` can become `&'b str` if and only if +`&'static str` is a subtype of `&'b str`, which will hold if `'static: 'b`. +This is true, so the compiler is happy to continue compiling this code. + +--- First off, let's revisit the meowing dog example: @@ -249,56 +312,6 @@ enough into the place expecting something long-lived. Here it is: -```rust,compile_fail -fn evil_feeder(input: &mut T, val: T) { - *input = val; -} - -fn main() { - let mut mr_snuggles: &'static str = "meow! :3"; // mr. snuggles forever!! - { - let spike = String::from("bark! >:V"); - let spike_str: &str = &spike; // Only lives for the block - evil_feeder(&mut mr_snuggles, spike_str); // EVIL! - } - println!("{}", mr_snuggles); // Use after free? -} -``` - -And what do we get when we run this? - -```text -error[E0597]: `spike` does not live long enough - --> src/main.rs:9:31 - | -6 | let mut mr_snuggles: &'static str = "meow! :3"; // mr. snuggles forever!! - | ------------ type annotation requires that `spike` is borrowed for `'static` -... -9 | let spike_str: &str = &spike; // Only lives for the block - | ^^^^^^ borrowed value does not live long enough -10 | evil_feeder(&mut mr_snuggles, spike_str); // EVIL! -11 | } - | - `spike` dropped here while still borrowed -``` - -Good, it doesn't compile! Let's break down what's happening here in detail. - -First let's look at the new `evil_feeder` function: - -```rust -fn evil_feeder(input: &mut T, val: T) { - *input = val; -} -``` - -All it does is take a mutable reference and a value and overwrite the referent with it. -What's important about this function is that it creates a type equality constraint. It -clearly says in its signature the referent and the value must be the *exact same* type. - -Meanwhile, in the caller we pass in `&mut &'static str` and `&'spike_str str`. - -Because `&mut T` is invariant over `T`, the compiler concludes it can't apply any subtyping -to the first argument, and so `T` must be exactly `&'static str`. The other argument is only an `&'a str`, which *is* covariant over `'a`. So the compiler adopts a constraint: `&'spike_str str` must be a subtype of `&'static str` (inclusive), From 8e129cc2a8f1e6d6edda4353b0af4f75190a6b21 Mon Sep 17 00:00:00 2001 From: Conrad Ludgate Date: Thu, 10 Feb 2022 15:47:57 +0000 Subject: [PATCH 05/13] remove the rest of animals --- src/subtyping.md | 151 +++++------------------------------------------ 1 file changed, 15 insertions(+), 136 deletions(-) diff --git a/src/subtyping.md b/src/subtyping.md index 0bda409..61bebcb 100644 --- a/src/subtyping.md +++ b/src/subtyping.md @@ -241,127 +241,9 @@ So the compiler decides that `&'static str` can become `&'b str` if and only if `&'static str` is a subtype of `&'b str`, which will hold if `'static: 'b`. This is true, so the compiler is happy to continue compiling this code. ---- - -First off, let's revisit the meowing dog example: - - -```rust,ignore -fn evil_feeder(pet: &mut Animal) { - let spike: Dog = ...; - - // `pet` is an Animal, and Dog is a subtype of Animal, - // so this should be fine, right..? - *pet = spike; -} - -fn main() { - let mut mr_snuggles: Cat = ...; - evil_feeder(&mut mr_snuggles); // Replaces mr_snuggles with a Dog - mr_snuggles.meow(); // OH NO, MEOWING DOG! -} -``` - -If we look at our table of variances, we see that `&mut T` is *invariant* over `T`. -As it turns out, this completely fixes the issue! With invariance, the fact that -Cat is a subtype of Animal doesn't matter; `&mut Cat` still won't be a subtype of -`&mut Animal`. The static type checker will then correctly stop us from passing -a Cat into `evil_feeder`. - -The soundness of subtyping is based on the idea that it's ok to forget unnecessary -details. But with references, there's always someone that remembers those details: -the value being referenced. That value expects those details to keep being true, -and may behave incorrectly if its expectations are violated. - -The problem with making `&mut T` covariant over `T` is that it gives us the power -to modify the original value *when we don't remember all of its constraints*. -And so, we can make someone have a Dog when they're certain they still have a Cat. - -With that established, we can easily see why `&T` being covariant over `T` *is* -sound: it doesn't let you modify the value, only look at it. Without any way to -mutate, there's no way for us to mess with any details. We can also see why -`UnsafeCell` and all the other interior mutability types must be invariant: they -make `&T` work like `&mut T`! - -Now what about the lifetime on references? Why is it ok for both kinds of references -to be covariant over their lifetimes? Well, here's a two-pronged argument: - -First and foremost, subtyping references based on their lifetimes is *the entire point -of subtyping in Rust*. The only reason we have subtyping is so we can pass -long-lived things where short-lived things are expected. So it better work! - -Second, and more seriously, lifetimes are only a part of the reference itself. The -type of the referent is shared knowledge, which is why adjusting that type in only -one place (the reference) can lead to issues. But if you shrink down a reference's -lifetime when you hand it to someone, that lifetime information isn't shared in -any way. There are now two independent references with independent lifetimes. -There's no way to mess with the original reference's lifetime using the other one. - -Or rather, the only way to mess with someone's lifetime is to build a meowing dog. -But as soon as you try to build a meowing dog, the lifetime should be wrapped up -in an invariant type, preventing the lifetime from being shrunk. To understand this -better, let's port the meowing dog problem over to real Rust. - -In the meowing dog problem we take a subtype (Cat), convert it into a supertype -(Animal), and then use that fact to overwrite the subtype with a value that satisfies -the constraints of the supertype but not the subtype (Dog). - -So with lifetimes, we want to take a long-lived thing, convert it into a -short-lived thing, and then use that to write something that doesn't live long -enough into the place expecting something long-lived. - -Here it is: - - -The other argument is only an `&'a str`, which *is* covariant over `'a`. So the compiler -adopts a constraint: `&'spike_str str` must be a subtype of `&'static str` (inclusive), -which in turn implies `'spike_str` must be a subtype of `'static` (inclusive). Which is to say, -`'spike_str` must contain `'static`. But only one thing contains `'static` -- `'static` itself! - -This is why we get an error when we try to assign `&spike` to `spike_str`. The -compiler has worked backwards to conclude `spike_str` must live forever, and `&spike` -simply can't live that long. - -So even though references are covariant over their lifetimes, they "inherit" invariance -whenever they're put into a context that could do something bad with that. In this case, -we inherited invariance as soon as we put our reference inside an `&mut T`. - -As it turns out, the argument for why it's ok for Box (and Vec, Hashmap, etc.) to -be covariant is pretty similar to the argument for why it's ok for -references to be covariant: as soon as you try to stuff them in something like a -mutable reference, they inherit invariance and you're prevented from doing anything -bad. - -However, Box makes it easier to focus on the by-value aspect of references that we -partially glossed over. - -Unlike a lot of languages which allow values to be freely aliased at all times, -Rust has a very strict rule: if you're allowed to mutate or move a value, you -are guaranteed to be the only one with access to it. - -Consider the following code: - - -```rust,ignore -let mr_snuggles: Box = ..; -let spike: Box = ..; - -let mut pet: Box; -pet = mr_snuggles; -pet = spike; -``` - -There is no problem at all with the fact that we have forgotten that `mr_snuggles` was a Cat, -or that we overwrote him with a Dog, because as soon as we moved mr_snuggles to a variable -that only knew he was an Animal, **we destroyed the only thing in the universe that -remembered he was a Cat**! - -In contrast to the argument about immutable references being soundly covariant because they -don't let you change anything, owned values can be covariant because they make you -change *everything*. There is no connection between old locations and new locations. -Applying by-value subtyping is an irreversible act of knowledge destruction, and -without any memory of how things used to be, no one can be tricked into acting on -that old information! +`Box` is also *covariant* over `T`. This would make sense, since it's supposed to be +usable the same as `&T`. If you try to mutate the box, you'll need a `&mut Box` and the +invariance of `&mut` will kick in here. Only one thing left to explain: function pointers. @@ -369,43 +251,40 @@ To see why `fn(T) -> U` should be covariant over `U`, consider the following sig ```rust,ignore -fn get_animal() -> Animal; +fn get_str() -> &'a str; ``` -This function claims to produce an Animal. As such, it is perfectly valid to +This function claims to produce a `str` bound by some liftime `'a`. As such, it is perfectly valid to provide a function with the following signature instead: ```rust,ignore -fn get_animal() -> Cat; +fn get_static() -> &'static str; ``` -After all, Cats are Animals, so always producing a Cat is a perfectly valid way -to produce Animals. Or to relate it back to real Rust: if we need a function -that is supposed to produce something that lives for `'short`, it's perfectly -fine for it to produce something that lives for `'long`. We don't care, we can -just forget that fact. +So when the function is called, all it's expecting is a `&str` which lives at least the lifetime of `'a`, +it doesn't matter if the value actually lives longer. However, the same logic does not apply to *arguments*. Consider trying to satisfy: ```rust,ignore -fn handle_animal(Animal); +fn store_ref(&'a str); ``` with: ```rust,ignore -fn handle_animal(Cat); +fn store_static(&'static str); ``` -The first function can accept Dogs, but the second function absolutely can't. +The first function can accept any string reference as long as it lives at least for `'a`, +but the second cannot accept a string reference that lives for any duration less than `'static`, +which would cause a conflict. Covariance doesn't work here. But if we flip it around, it actually *does* -work! If we need a function that can handle Cats, a function that can handle *any* -Animal will surely work fine. Or to relate it back to real Rust: if we need a -function that can handle anything that lives for at least `'long`, it's perfectly -fine for it to be able to handle anything that lives for at least `'short`. +work! If we need a function that can handle `&'static str`, a function that can handle *any* reference lifetime +will surely work fine. And that's why function types, unlike anything else in the language, are **contra**variant over their arguments. From 510938c8ac27b85fbe5c15d694b3caef8d7ac9f0 Mon Sep 17 00:00:00 2001 From: Conrad Ludgate Date: Thu, 10 Feb 2022 16:26:08 +0000 Subject: [PATCH 06/13] remove use of transitive --- src/subtyping.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/src/subtyping.md b/src/subtyping.md index 61bebcb..b2ceb7b 100644 --- a/src/subtyping.md +++ b/src/subtyping.md @@ -74,7 +74,7 @@ Now that we have a defined set of requirements for lifetimes, we can define how `'a` may define a region larger than `'b`, but that still fits our definition. Going back to our example above, we can say that `'static: 'b`. -For now, let's accept the idea that subtypes of lifetimes can be transitive (more on this in [Variance](#variance)), +For now, let's accept the idea that subtypes of lifetimes can be passed through references (more on this in [Variance](#variance)), eg. `&'static str` is a subtype of `&'b str`, then we can let them coerce, and then the example above will compile ```rust @@ -120,7 +120,7 @@ However, the implementation of `assign` is valid. Therefore, this must mean that `&mut &'static str` should **not** a *subtype* of `&mut &'b str`, even if `'static` is a subtype of `'b`. -Variance is the way that Rust defines the transitivity of subtypes through their *type constructor*. +Variance is the way that Rust defines the relationships of subtypes through their *type constructor*. A type constructor in Rust is any generic type with unbound arguments. For instance `Vec` is a type constructor that takes a type `T` and returns `Vec`. `&` and `&mut` are type constructors that take two inputs: a From aeb9d4c21db62db267ffe28ef27d0252afc3168a Mon Sep 17 00:00:00 2001 From: Conrad Ludgate Date: Fri, 11 Feb 2022 08:07:56 +0000 Subject: [PATCH 07/13] address some comments --- src/subtyping.md | 39 +++++++++++++++++++++++---------------- 1 file changed, 23 insertions(+), 16 deletions(-) diff --git a/src/subtyping.md b/src/subtyping.md index b2ceb7b..e59e8a0 100644 --- a/src/subtyping.md +++ b/src/subtyping.md @@ -9,7 +9,7 @@ while also preventing their misuse, Rust uses a combination of **Subtyping** and ## Subtyping -Subtyping is the idea that one type can be a *subtype* of another. +Subtyping is the idea that one type can be used in place of another. Let's define that `Sub` is a subtype of `Super` (we'll be using the notation `Sub: Super` throughout this chapter) @@ -21,15 +21,15 @@ An example of simple subtyping that exists in the language are [supertraits](htt ```rust use std::fmt; -trait OutlinePrint: fmt::Display { - fn outline_print(&self) { - todo!() - } +pub trait Error: fmt::Display { + fn source(&self) -> Option<&(dyn Error + 'static)>; + fn description(&self) -> &str; + fn cause(&self) -> Option<&dyn Error>; } ``` -Here, we have that `OutlinePrint: fmt::Display` (`OutlinePrint` is a *subtype* of `Display`), -because it has all the requirements of `fmt::Display`, plus the `outline_print` function. +Here, we have that `Error: fmt::Display` (`Error` is a *subtype* of `Display`), +because it has all the requirements of `fmt::Display`, plus the `source`/`description`/`cause` functions. However, subtyping in traits is not that interesting in the case of Rust. Here in the nomicon, we're going to focus more with how subtyping interacts with **lifetimes** @@ -51,7 +51,7 @@ fn main() { } ``` -In an overly restrictive implementation of lifetimes, since `a` and `b` have differeing lifetimes, +In a conservative implementation of lifetimes, since `a` and `b` have differeing lifetimes, we might see the following error: ```text @@ -64,10 +64,12 @@ error[E0308]: mismatched types | expected `&'static str`, found struct `&'b str` ``` -This is over-restrictive. In this case, what we want is to accept any type that lives *at least as long* as `'b`. +This would be rather unfortunate. In this case, +what we want is to accept any type that lives *at least as long* as `'b`. Let's try using subtyping with our lifetimes. -Let's define a lifetime to have the a simple set of requirements: `'a` defines a region of code in which a value will be alive. +Let's define a lifetime to have the a simple set of requirements: +`'a` defines a region of code in which a value will be alive. Now that we have a defined set of requirements for lifetimes, we can define how they relate to each other. `'a: 'b` if and only if `'a` defines a region of code that **completely contains** `'b`. @@ -108,20 +110,24 @@ fn main() { let world = String::from("world"); assign(&mut hello, &world); } + println!("{}", hello); } ``` -If this were to compile, this would have a memory bug. +In `assign`, we are setting the `hello` reference to point to `world`. +But then `world` goes out of scope, before the later use of `hello` in the println! + +This is a classic use-after-free bug! -If we were to expand this out, we'd see that we're trying to assign a `&'b str` into a `&'static str`, -but the problem is that as soon as `b` goes out of scope, `a` is now invalid, even though it's supposed to have a `'static` lifetime. +Our first instinct might be to blame the `assign` impl, but there's really nothing wrong here. +It shouldn't be surprising that we might want to assign a `T` into a `T`. -However, the implementation of `assign` is valid. -Therefore, this must mean that `&mut &'static str` should **not** a *subtype* of `&mut &'b str`, +The problem is that we cannot assume that `&mut &'static str` and `&mut &'b str` are compatible. +This must mean that `&mut &'static str` should **not** be a *subtype* of `&mut &'b str`, even if `'static` is a subtype of `'b`. Variance is the way that Rust defines the relationships of subtypes through their *type constructor*. -A type constructor in Rust is any generic type with unbound arguments. +A type constructor is any generic type with unbound arguments. For instance `Vec` is a type constructor that takes a type `T` and returns `Vec`. `&` and `&mut` are type constructors that take two inputs: a lifetime, and a type to point to. @@ -190,6 +196,7 @@ fn main() { let world = String::from("world"); assign(&mut hello, &world); } + println!("{}", hello); } ``` From 2c8ff4f66903373f97be3cb202ecb6ef60601a7f Mon Sep 17 00:00:00 2001 From: Conrad Ludgate Date: Fri, 11 Feb 2022 08:24:28 +0000 Subject: [PATCH 08/13] slight restructure --- src/subtyping.md | 78 +++++++++++++++++++++++++++--------------------- 1 file changed, 44 insertions(+), 34 deletions(-) diff --git a/src/subtyping.md b/src/subtyping.md index e59e8a0..089d53c 100644 --- a/src/subtyping.md +++ b/src/subtyping.md @@ -7,34 +7,7 @@ or permit undefined behavior. In order to allow flexible usage of lifetimes while also preventing their misuse, Rust uses a combination of **Subtyping** and **Variance**. -## Subtyping - -Subtyping is the idea that one type can be used in place of another. - -Let's define that `Sub` is a subtype of `Super` (we'll be using the notation `Sub: Super` throughout this chapter) - -What this is suggesting to us is that the set of *requirements* that `Super` defines -are completely satisfied by `Sub`. `Sub` may then have more requirements. - -An example of simple subtyping that exists in the language are [supertraits](https://doc.rust-lang.org/stable/book/ch19-03-advanced-traits.html?highlight=supertraits#using-supertraits-to-require-one-traits-functionality-within-another-trait) - -```rust -use std::fmt; - -pub trait Error: fmt::Display { - fn source(&self) -> Option<&(dyn Error + 'static)>; - fn description(&self) -> &str; - fn cause(&self) -> Option<&dyn Error>; -} -``` - -Here, we have that `Error: fmt::Display` (`Error` is a *subtype* of `Display`), -because it has all the requirements of `fmt::Display`, plus the `source`/`description`/`cause` functions. - -However, subtyping in traits is not that interesting in the case of Rust. -Here in the nomicon, we're going to focus more with how subtyping interacts with **lifetimes** - -Take this example +Let's start with a example. ```rust fn debug(a: T, b: T) { @@ -68,16 +41,53 @@ This would be rather unfortunate. In this case, what we want is to accept any type that lives *at least as long* as `'b`. Let's try using subtyping with our lifetimes. -Let's define a lifetime to have the a simple set of requirements: +## Subtyping + +Subtyping is the idea that one type can be used in place of another. + +Let's define that `Sub` is a subtype of `Super` (we'll be using the notation `Sub: Super` throughout this chapter) + +What this is suggesting to us is that the set of *requirements* that `Super` defines +are completely satisfied by `Sub`. `Sub` may then have more requirements. + +An example of simple subtyping that exists in the language are [supertraits](https://doc.rust-lang.org/stable/book/ch19-03-advanced-traits.html?highlight=supertraits#using-supertraits-to-require-one-traits-functionality-within-another-trait) + +```rust +use std::fmt; + +pub trait Error: fmt::Display { + fn source(&self) -> Option<&(dyn Error + 'static)>; + fn description(&self) -> &str; + fn cause(&self) -> Option<&dyn Error>; +} +``` + +Here, we have that `Error: fmt::Display` (`Error` is a *subtype* of `Display`), +because it has all the requirements of `fmt::Display`, plus the `source`/`description`/`cause` functions. + +However, subtyping in traits is not that interesting. +Here in the nomicon, we're going to focus more with how subtyping interacts with lifetimes + +Let's define a lifetime to be the simple requirement: `'a` defines a region of code in which a value will be alive. Now that we have a defined set of requirements for lifetimes, we can define how they relate to each other. -`'a: 'b` if and only if `'a` defines a region of code that **completely contains** `'b`. +`'long: 'short` if and only if `'long` defines a region of code that **completely contains** `'short`. -`'a` may define a region larger than `'b`, but that still fits our definition. -Going back to our example above, we can say that `'static: 'b`. +`'long` may define a region larger than `'short`, but that still fits our definition. -For now, let's accept the idea that subtypes of lifetimes can be passed through references (more on this in [Variance](#variance)), -eg. `&'static str` is a subtype of `&'b str`, then we can let them coerce, and then the example above will compile +> As we will see throughout the rest of this chapter, +subtyping is a lot more complicated and subtle than this, +but this simple rule is a very good 99% intuition. +And unless you write unsafe code, the compiler will automatically handle all the corner cases for you. + +> But this is the Rustonomicon. We're writing unsafe code, +so we need to understand how this stuff really works, and how we can mess it up. + +Going back to our example above, we can say that `'static: 'b`. +For now, let's also accept the idea that subtypes of lifetimes can be passed through references +(more on this in [Variance](#variance)), +eg. `&'static str` is a subtype of `&'b str`, then we can let them coerce, +and then the example above will compile ```rust fn debug(a: T, b: T) { From 8f88efd850c8221b5a98917caec13196a212d19f Mon Sep 17 00:00:00 2001 From: Conrad Ludgate Date: Fri, 11 Feb 2022 09:02:57 +0000 Subject: [PATCH 09/13] add demos for box and fn --- src/subtyping.md | 61 ++++++++++++++++++++++++++++++++++++++++++++---- 1 file changed, 57 insertions(+), 4 deletions(-) diff --git a/src/subtyping.md b/src/subtyping.md index 089d53c..c0a8e5f 100644 --- a/src/subtyping.md +++ b/src/subtyping.md @@ -120,7 +120,7 @@ fn main() { let world = String::from("world"); assign(&mut hello, &world); } - println!("{}", hello); + println!("{}", hello); // use after free 😿 } ``` @@ -258,9 +258,24 @@ So the compiler decides that `&'static str` can become `&'b str` if and only if `&'static str` is a subtype of `&'b str`, which will hold if `'static: 'b`. This is true, so the compiler is happy to continue compiling this code. -`Box` is also *covariant* over `T`. This would make sense, since it's supposed to be -usable the same as `&T`. If you try to mutate the box, you'll need a `&mut Box` and the -invariance of `&mut` will kick in here. +As it turns out, the argument for why it's ok for Box (and Vec, Hashmap, etc.) to be covariant is pretty similar to the argument for why it's ok for lifetimes to be covariant: as soon as you try to stuff them in something like a mutable reference, they inherit invariance and you're prevented from doing anything bad. + +However Box makes it easier to focus on by-value aspect of references that we partially glossed over. + +Unlike a lot of languages which allow values to be freely aliased at all times, Rust has a very strict rule: if you're allowed to mutate or move a value, you are guaranteed to be the only one with access to it. + +Consider the following code: + +```rust,ignore +let hello: Box<&'static str> = Box::new("hello"); + +let mut world: Box<&'b str>; +world = hello; +``` + +There is no problem at all with the fact that we have forgotten that `hello` was alive for `'static`, +because as soon as we moved `hello` to a variable that only knew he it was alive for `'b`, +**we destroyed the only thing in the universe that remembered it lived for longer**! Only one thing left to explain: function pointers. @@ -303,6 +318,44 @@ Covariance doesn't work here. But if we flip it around, it actually *does* work! If we need a function that can handle `&'static str`, a function that can handle *any* reference lifetime will surely work fine. +Let's see this in practice + +```rust,compile_fail +# use std::cell::RefCell; +thread_local! { + pub static StaticVecs: RefCell> = RefCell::new(Vec::new()); +} + +/// saves the input given into a thread local `Vec<&'static str>` +fn store(input: &'static str) { + StaticVecs.with(|v| { + v.borrow_mut().push(input); + }) +} + +/// Calls the function with it's input (must have the same lifetime!) +fn demo<'a>(input: &'a str, f: fn(&'a str)) { + f(input); +} + +fn main() { + demo("hello", store); // "hello" is 'static. Can call `store` fine + + { + let smuggle = String::from("smuggle"); + + // `&smuggle` is not static. If we were to call `store` with `&smuggle`, + // we would have pushed an invalid lifetime into the `StaticVecs`. + // Therefore, `fn(&'static str)` cannot be a subtype of `fn(&'a str)` + demo(&smuggle, store); + } + + StaticVecs.with(|v| { + println!("{:?}", v.borrow()); // use after free 😿 + }); +} +``` + And that's why function types, unlike anything else in the language, are **contra**variant over their arguments. From ea950766dd92bd4ac6e06081c86f60935449ec01 Mon Sep 17 00:00:00 2001 From: Conrad Ludgate Date: Sun, 27 Feb 2022 09:14:29 +0000 Subject: [PATCH 10/13] address some grammatical comments --- src/subtyping.md | 55 ++++++++++++++++++++++++++---------------------- 1 file changed, 30 insertions(+), 25 deletions(-) diff --git a/src/subtyping.md b/src/subtyping.md index c0a8e5f..5645326 100644 --- a/src/subtyping.md +++ b/src/subtyping.md @@ -5,12 +5,12 @@ However, a naive implementation of lifetimes would be either too restrictive, or permit undefined behavior. In order to allow flexible usage of lifetimes -while also preventing their misuse, Rust uses a combination of **Subtyping** and **Variance**. +while also preventing their misuse, Rust uses **subtyping** and **variance**. -Let's start with a example. +Let's start with an example. ```rust -fn debug(a: T, b: T) { +fn debug<'a>(a: &'a str, b: &'a str) { println!("a = {:?} b = {:?}", a, b); } @@ -18,13 +18,13 @@ fn main() { let hello: &'static str = "hello"; { let world = String::from("world"); - let world = &world; // 'b has a shorter lifetime than 'static + let world = &world; // 'world has a shorter lifetime than 'static debug(hello, world); } } ``` -In a conservative implementation of lifetimes, since `a` and `b` have differeing lifetimes, +In a conservative implementation of lifetimes, since `hello` and `world` have differing lifetimes, we might see the following error: ```text @@ -34,23 +34,25 @@ error[E0308]: mismatched types 10 | debug(hello, world); | ^ | | - | expected `&'static str`, found struct `&'b str` + | expected `&'static str`, found struct `&'world str` ``` This would be rather unfortunate. In this case, -what we want is to accept any type that lives *at least as long* as `'b`. +what we want is to accept any type that lives *at least as long* as `'world`. Let's try using subtyping with our lifetimes. ## Subtyping Subtyping is the idea that one type can be used in place of another. -Let's define that `Sub` is a subtype of `Super` (we'll be using the notation `Sub: Super` throughout this chapter) +Let's define that `Sub` is a subtype of `Super` (we'll be using the notation `Sub: Super` throughout this chapter). What this is suggesting to us is that the set of *requirements* that `Super` defines are completely satisfied by `Sub`. `Sub` may then have more requirements. -An example of simple subtyping that exists in the language are [supertraits](https://doc.rust-lang.org/stable/book/ch19-03-advanced-traits.html?highlight=supertraits#using-supertraits-to-require-one-traits-functionality-within-another-trait) +An example of simple subtyping that exists in the language is [supertraits][supertraits]: + +[supertraits]: https://doc.rust-lang.org/book/ch19-03-advanced-traits.html#using-supertraits-to-require-one-traits-functionality-within-another-trait ```rust use std::fmt; @@ -66,12 +68,15 @@ Here, we have that `Error: fmt::Display` (`Error` is a *subtype* of `Display`), because it has all the requirements of `fmt::Display`, plus the `source`/`description`/`cause` functions. However, subtyping in traits is not that interesting. -Here in the nomicon, we're going to focus more with how subtyping interacts with lifetimes +Here in the Rustonomicon, we're going to focus more with how subtyping interacts with lifetimes. Let's define a lifetime to be the simple requirement: -`'a` defines a region of code in which a value will be alive. -Now that we have a defined set of requirements for lifetimes, we can define how they relate to each other. -`'long: 'short` if and only if `'long` defines a region of code that **completely contains** `'short`. + +> `'a` defines a region of code. + +Now that we have a defined set of requirements for lifetimes, we can define how they relate to each other: + +> `'long : 'short` if and only if `'long` defines a region of code that **completely contains** `'short`. `'long` may define a region larger than `'short`, but that still fits our definition. @@ -83,11 +88,11 @@ And unless you write unsafe code, the compiler will automatically handle all the > But this is the Rustonomicon. We're writing unsafe code, so we need to understand how this stuff really works, and how we can mess it up. -Going back to our example above, we can say that `'static: 'b`. +Going back to our example above, we can say that `'static : 'world`. For now, let's also accept the idea that subtypes of lifetimes can be passed through references (more on this in [Variance](#variance)), -eg. `&'static str` is a subtype of `&'b str`, then we can let them coerce, -and then the example above will compile +_e.g._ `&'static str` is a subtype of `&'world str`, then we can let a `&'static str` "downgrade" into a `&'world str`. +With that, the example above will compile: ```rust fn debug(a: T, b: T) { @@ -98,16 +103,16 @@ fn main() { let hello: &'static str = "hello"; { let world = String::from("world"); - let world = &world; // 'b has a shorter lifetime than 'static - debug(hello, world); // a silently converts from `&'static str` into `&'b str` + let world = &world; // 'world has a shorter lifetime than 'static + debug(hello, world); // hello silently downgrades from `&'static str` into `&'world str` } } ``` ## Variance -Above, we glossed over the fact that `'static: 'b` implied that `&'static T: &'b T`. This uses a property known as variance. -It's not always as simple as this example though, to understand that let's try extend this example a bit +Above, we glossed over the fact that `'static : 'b` implied that `&'static T : &'b T`. This uses a property known as _variance_. +It's not always as simple as this example, though. To understand that, let's try to extend this example a bit: ```rust,compile_fail fn assign(input: &mut T, val: T) { @@ -133,11 +138,11 @@ Our first instinct might be to blame the `assign` impl, but there's really nothi It shouldn't be surprising that we might want to assign a `T` into a `T`. The problem is that we cannot assume that `&mut &'static str` and `&mut &'b str` are compatible. -This must mean that `&mut &'static str` should **not** be a *subtype* of `&mut &'b str`, +This means that `&mut &'static str` **cannot** be a *subtype* of `&mut &'b str`, even if `'static` is a subtype of `'b`. -Variance is the way that Rust defines the relationships of subtypes through their *type constructor*. -A type constructor is any generic type with unbound arguments. +Variance is the concept that Rust borrows to define relationships about subtypes through their *type constructor*s. +A type constructor is any generic item in Rust. For instance `Vec` is a type constructor that takes a type `T` and returns `Vec`. `&` and `&mut` are type constructors that take two inputs: a lifetime, and a type to point to. @@ -244,7 +249,7 @@ Meanwhile, in the caller we pass in `&mut &'static str` and `&'spike_str str`. Because `&mut T` is invariant over `T`, the compiler concludes it can't apply any subtyping to the first argument, and so `T` must be exactly `&'static str`. -This is counter to the `&T` case +This is counter to the `&T` case: ```rust fn debug(a: T, b: T) { @@ -252,7 +257,7 @@ fn debug(a: T, b: T) { } ``` -Where similarly `a` and `b` must have the same type `T`. +where similarly `a` and `b` must have the same type `T`. But since `&'a T` *is* covariant over `'a`, we are allowed to perform subtyping. So the compiler decides that `&'static str` can become `&'b str` if and only if `&'static str` is a subtype of `&'b str`, which will hold if `'static: 'b`. From 86b1c8759465bdc6b2d4a7edc03e7691c990a11f Mon Sep 17 00:00:00 2001 From: Timo Date: Sat, 2 Apr 2022 15:42:09 -0400 Subject: [PATCH 11/13] Copy-edit subtyping.md --- src/subtyping.md | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/src/subtyping.md b/src/subtyping.md index 5645326..a4673ea 100644 --- a/src/subtyping.md +++ b/src/subtyping.md @@ -244,7 +244,7 @@ All it does is take a mutable reference and a value and overwrite the referent w What's important about this function is that it creates a type equality constraint. It clearly says in its signature the referent and the value must be the *exact same* type. -Meanwhile, in the caller we pass in `&mut &'static str` and `&'spike_str str`. +Meanwhile, in the caller we pass in `&mut &'static str` and `&'world str`. Because `&mut T` is invariant over `T`, the compiler concludes it can't apply any subtyping to the first argument, and so `T` must be exactly `&'static str`. @@ -263,9 +263,9 @@ So the compiler decides that `&'static str` can become `&'b str` if and only if `&'static str` is a subtype of `&'b str`, which will hold if `'static: 'b`. This is true, so the compiler is happy to continue compiling this code. -As it turns out, the argument for why it's ok for Box (and Vec, Hashmap, etc.) to be covariant is pretty similar to the argument for why it's ok for lifetimes to be covariant: as soon as you try to stuff them in something like a mutable reference, they inherit invariance and you're prevented from doing anything bad. +As it turns out, the argument for why it's ok for Box (and Vec, HashMap, etc.) to be covariant is pretty similar to the argument for why it's ok for lifetimes to be covariant: as soon as you try to stuff them in something like a mutable reference, they inherit invariance and you're prevented from doing anything bad. -However Box makes it easier to focus on by-value aspect of references that we partially glossed over. +However Box makes it easier to focus on the by-value aspect of references that we partially glossed over. Unlike a lot of languages which allow values to be freely aliased at all times, Rust has a very strict rule: if you're allowed to mutate or move a value, you are guaranteed to be the only one with access to it. @@ -279,7 +279,7 @@ world = hello; ``` There is no problem at all with the fact that we have forgotten that `hello` was alive for `'static`, -because as soon as we moved `hello` to a variable that only knew he it was alive for `'b`, +because as soon as we moved `hello` to a variable that only knew it was alive for `'b`, **we destroyed the only thing in the universe that remembered it lived for longer**! Only one thing left to explain: function pointers. From 15174604f942a930583596d795f459ab5145f220 Mon Sep 17 00:00:00 2001 From: Conrad Ludgate Date: Mon, 11 Apr 2022 10:14:08 +0100 Subject: [PATCH 12/13] remove supertraits replace subtyping syntax remove type constructors --- src/subtyping.md | 77 +++++++++++++++++------------------------------- 1 file changed, 27 insertions(+), 50 deletions(-) diff --git a/src/subtyping.md b/src/subtyping.md index a4673ea..97b4ebd 100644 --- a/src/subtyping.md +++ b/src/subtyping.md @@ -10,6 +10,7 @@ while also preventing their misuse, Rust uses **subtyping** and **variance**. Let's start with an example. ```rust +// Note: debug expects two parameters with the *same* lifetime fn debug<'a>(a: &'a str, b: &'a str) { println!("a = {:?} b = {:?}", a, b); } @@ -45,38 +46,18 @@ Let's try using subtyping with our lifetimes. Subtyping is the idea that one type can be used in place of another. -Let's define that `Sub` is a subtype of `Super` (we'll be using the notation `Sub: Super` throughout this chapter). +Let's define that `Sub` is a subtype of `Super` (we'll be using the notation `Sub <: Super` throughout this chapter). What this is suggesting to us is that the set of *requirements* that `Super` defines are completely satisfied by `Sub`. `Sub` may then have more requirements. -An example of simple subtyping that exists in the language is [supertraits][supertraits]: - -[supertraits]: https://doc.rust-lang.org/book/ch19-03-advanced-traits.html#using-supertraits-to-require-one-traits-functionality-within-another-trait - -```rust -use std::fmt; - -pub trait Error: fmt::Display { - fn source(&self) -> Option<&(dyn Error + 'static)>; - fn description(&self) -> &str; - fn cause(&self) -> Option<&dyn Error>; -} -``` - -Here, we have that `Error: fmt::Display` (`Error` is a *subtype* of `Display`), -because it has all the requirements of `fmt::Display`, plus the `source`/`description`/`cause` functions. - -However, subtyping in traits is not that interesting. -Here in the Rustonomicon, we're going to focus more with how subtyping interacts with lifetimes. - -Let's define a lifetime to be the simple requirement: +Now, in order to use subtyping with lifetimes, we need to define the requirement of a lifetime: > `'a` defines a region of code. Now that we have a defined set of requirements for lifetimes, we can define how they relate to each other: -> `'long : 'short` if and only if `'long` defines a region of code that **completely contains** `'short`. +> `'long <: 'short` if and only if `'long` defines a region of code that **completely contains** `'short`. `'long` may define a region larger than `'short`, but that still fits our definition. @@ -88,7 +69,7 @@ And unless you write unsafe code, the compiler will automatically handle all the > But this is the Rustonomicon. We're writing unsafe code, so we need to understand how this stuff really works, and how we can mess it up. -Going back to our example above, we can say that `'static : 'world`. +Going back to our example above, we can say that `'static <: 'world`. For now, let's also accept the idea that subtypes of lifetimes can be passed through references (more on this in [Variance](#variance)), _e.g._ `&'static str` is a subtype of `&'world str`, then we can let a `&'static str` "downgrade" into a `&'world str`. @@ -111,7 +92,7 @@ fn main() { ## Variance -Above, we glossed over the fact that `'static : 'b` implied that `&'static T : &'b T`. This uses a property known as _variance_. +Above, we glossed over the fact that `'static <: 'b` implied that `&'static T <: &'b T`. This uses a property known as _variance_. It's not always as simple as this example, though. To understand that, let's try to extend this example a bit: ```rust,compile_fail @@ -141,43 +122,39 @@ The problem is that we cannot assume that `&mut &'static str` and `&mut &'b str` This means that `&mut &'static str` **cannot** be a *subtype* of `&mut &'b str`, even if `'static` is a subtype of `'b`. -Variance is the concept that Rust borrows to define relationships about subtypes through their *type constructor*s. -A type constructor is any generic item in Rust. -For instance `Vec` is a type constructor that takes a type `T` and returns -`Vec`. `&` and `&mut` are type constructors that take two inputs: a -lifetime, and a type to point to. +Variance is the concept that Rust borrows to define relationships about subtypes through their generic parameters. -> NOTE: For convenience we will often refer to `F` as a type constructor just so +> NOTE: For convenience we will define a generic type `F` so > that we can easily talk about `T`. Hopefully this is clear in context. -A type constructor F's *variance* is how the subtyping of its inputs affects the +The type `F`'s *variance* is how the subtyping of its inputs affects the subtyping of its outputs. There are three kinds of variance in Rust. Given two types `Sub` and `Super`, where `Sub` is a subtype of `Super`: -* F is **covariant** if `F` is a subtype of `F` (the subtype property is passed through) -* F is **contravariant** if `F` is a subtype of `F` (the subtype property is "inverted") -* F is **invariant** otherwise (no subtyping relationship exists) +* `F` is **covariant** if `F` is a subtype of `F` (the subtype property is passed through) +* `F` is **contravariant** if `F` is a subtype of `F` (the subtype property is "inverted") +* `F` is **invariant** otherwise (no subtyping relationship exists) If we remember from the above examples, -it was ok for us to treat `&'a T` as a subtype of `&'b T` if `'a: 'b`, +it was ok for us to treat `&'a T` as a subtype of `&'b T` if `'a <: 'b`, therefore we can say that `&'a T` is *covariant* over `'a`. -Also, we saw that it was not ok for us to treat `&mut &'a T` as a subtype of `&mut &'b T`, +Also, we saw that it was not ok for us to treat `&mut &'a U` as a subtype of `&mut &'b U`, therefore we can say that `&mut T` is *invariant* over `T` -Here is a table of some other type constructors and their variances: +Here is a table of some other generic types and their variances: -| | | 'a | T | U | -|---|-----------------|:---------:|:-----------------:|:---------:| -| | `&'a T ` | covariant | covariant | | -| | `&'a mut T` | covariant | invariant | | -| | `Box` | | covariant | | -| | `Vec` | | covariant | | -| | `UnsafeCell` | | invariant | | -| | `Cell` | | invariant | | -| | `fn(T) -> U` | | **contra**variant | covariant | -| | `*const T` | | covariant | | -| | `*mut T` | | invariant | | +| | 'a | T | U | +|-----------------|:---------:|:-----------------:|:---------:| +| `&'a T ` | covariant | covariant | | +| `&'a mut T` | covariant | invariant | | +| `Box` | | covariant | | +| `Vec` | | covariant | | +| `UnsafeCell` | | invariant | | +| `Cell` | | invariant | | +| `fn(T) -> U` | | **contra**variant | covariant | +| `*const T` | | covariant | | +| `*mut T` | | invariant | | Some of these can be explained simply in relation to the others: @@ -260,7 +237,7 @@ fn debug(a: T, b: T) { where similarly `a` and `b` must have the same type `T`. But since `&'a T` *is* covariant over `'a`, we are allowed to perform subtyping. So the compiler decides that `&'static str` can become `&'b str` if and only if -`&'static str` is a subtype of `&'b str`, which will hold if `'static: 'b`. +`&'static str` is a subtype of `&'b str`, which will hold if `'static <: 'b`. This is true, so the compiler is happy to continue compiling this code. As it turns out, the argument for why it's ok for Box (and Vec, HashMap, etc.) to be covariant is pretty similar to the argument for why it's ok for lifetimes to be covariant: as soon as you try to stuff them in something like a mutable reference, they inherit invariance and you're prevented from doing anything bad. From 54ca7d1a34e24ee158a524ddfbe66441198a764c Mon Sep 17 00:00:00 2001 From: Eric Huss Date: Sat, 27 May 2023 18:54:51 -0700 Subject: [PATCH 13/13] Apply some review suggestions. --- src/subtyping.md | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) diff --git a/src/subtyping.md b/src/subtyping.md index 97b4ebd..f63b532 100644 --- a/src/subtyping.md +++ b/src/subtyping.md @@ -12,7 +12,7 @@ Let's start with an example. ```rust // Note: debug expects two parameters with the *same* lifetime fn debug<'a>(a: &'a str, b: &'a str) { - println!("a = {:?} b = {:?}", a, b); + println!("a = {a:?} b = {b:?}"); } fn main() { @@ -25,7 +25,7 @@ fn main() { } ``` -In a conservative implementation of lifetimes, since `hello` and `world` have differing lifetimes, +In a conservative implementation of lifetimes, since `hello` and `world` have different lifetimes, we might see the following error: ```text @@ -72,12 +72,12 @@ so we need to understand how this stuff really works, and how we can mess it up. Going back to our example above, we can say that `'static <: 'world`. For now, let's also accept the idea that subtypes of lifetimes can be passed through references (more on this in [Variance](#variance)), -_e.g._ `&'static str` is a subtype of `&'world str`, then we can let a `&'static str` "downgrade" into a `&'world str`. +_e.g._ `&'static str` is a subtype of `&'world str`, then we can "downgrade" `&'static str` into a `&'world str`. With that, the example above will compile: ```rust -fn debug(a: T, b: T) { - println!("a = {:?} b = {:?}", a, b); +fn debug<'a>(a: &'a str, b: &'a str) { + println!("a = {a:?} b = {b:?}"); } fn main() { @@ -95,7 +95,7 @@ fn main() { Above, we glossed over the fact that `'static <: 'b` implied that `&'static T <: &'b T`. This uses a property known as _variance_. It's not always as simple as this example, though. To understand that, let's try to extend this example a bit: -```rust,compile_fail +```rust,compile_fail,E0597 fn assign(input: &mut T, val: T) { *input = val; } @@ -106,7 +106,7 @@ fn main() { let world = String::from("world"); assign(&mut hello, &world); } - println!("{}", hello); // use after free 😿 + println!("{hello}"); // use after free 😿 } ``` @@ -177,7 +177,7 @@ For more types, see the ["Variance" section][variance-table] on the reference. Now that we have some more formal understanding of variance, let's go through some more examples in more detail. -```rust,compile_fail +```rust,compile_fail,E0597 fn assign(input: &mut T, val: T) { *input = val; } @@ -188,7 +188,7 @@ fn main() { let world = String::from("world"); assign(&mut hello, &world); } - println!("{}", hello); + println!("{hello}"); } ``` @@ -230,7 +230,7 @@ This is counter to the `&T` case: ```rust fn debug(a: T, b: T) { - println!("a = {:?} b = {:?}", a, b); + println!("a = {a:?} b = {b:?}"); } ```