X-Git-Url: https://git.ralfj.de/rust-101.git/blobdiff_plain/800c78a128bd22bbe700164de5019ff8270fd746..38318e759c8aae48e68e413856289eb436a15b08:/src/part00.rs diff --git a/src/part00.rs b/src/part00.rs index 308c242..00eb4d1 100644 --- a/src/part00.rs +++ b/src/part00.rs @@ -1,146 +1,116 @@ -// [index](main.html) | previous | [next](part01.html) - -use std; - -// Rust-101, Part 00: Algebraic datatypes, expressions -// =================================================== - -// As a starter, we want to write a function that computes the minimum of a list. -// First, we need to write down the signature of the function: The types of its arguments and -// of the return value. In the case of our minimum function, -// we may be inclined to say that it returns a number. But then we would be in trouble: What's -// the minimum of an empty list? The type of the function says we have to return *something*. -// We could just choose 0, but that would be kind of arbitrary. What we need -// is a type that is "a number, or nothing". Such a type (of multiple exclusive options) -// is called an "algebraic datatype", and Rust lets us define such types with the keyword `enum`. -// Coming from C(++), you can think of such a type as a `union`, together with a field that -// stores the variant of the union that's currently used. - +// Rust-101, Part 00: Algebraic datatypes +// ====================================== + +// As our first piece of Rust code, we want to write a function that computes the +// minimum of a list. + +//@ ## Getting started +//@ Let us start by thinking about the *type* of our function. Rust forces us to give the types of +//@ all arguments, and the return type, before we even start writing the body. In the case of our minimum +//@ function, we may be inclined to say that it returns a number. But then we would be in trouble: What's +//@ the minimum of an empty list? The type of the function says we have to return *something*. +//@ We could just choose 0, but that would be kind of arbitrary. What we need +//@ is a type that is "a number, or nothing". Such a type (of multiple exclusive options) +//@ is called an "algebraic datatype", and Rust lets us define such types with the keyword `enum`. +//@ Coming from C(++), you can think of such a type as a `union`, together with a field that +//@ stores the variant of the union that's currently used. + +// An `enum` for "a number or nothing" could look as follows: enum NumberOrNothing { Number(i32), Nothing } - -// Notice that `i32` is the type of (signed, 32-bit) integers. To write down the type of -// the minimum function, we need just one more ingredient: `Vec` is the type of -// (growable) arrays of numbers, and we will use that as our list type. -// Observe how in Rust, the function type comes *after* the arguments. - -fn vec_min_try1(vec: Vec) -> NumberOrNothing { - // First, we need some variable to store the minimum as computed so far. - // Since we start out with nothing computed, this will again be a - // "number or nothing". Notice that we do not have to write a type - // next to `min`, Rust can figure that out automatically (a bit like - // `auto` in C++11). Also notice the `mut`: In Rust, variables are - // immutable per default, and you need to tell Rust if you want - // to change a variable later. +//@ Notice that `i32` is the type of (signed, 32-bit) integers. To write down the type of +//@ the minimum function, we need just one more ingredient: `Vec` is the type of +//@ (growable) arrays of numbers, and we will use that as our list type. + +// Observe how in Rust, the return type comes *after* the arguments. +fn vec_min(vec: Vec) -> NumberOrNothing { + //@ In the function, we first need some variable to store the minimum as computed so far. + //@ Since we start out with nothing computed, this will again be a + //@ "number or nothing": let mut min = NumberOrNothing::Nothing; + //@ We do not have to write a type next to `min`, Rust can figure that out automatically + //@ (a bit like `auto` in C++11). Also notice the `mut`: In Rust, variables are + //@ immutable per default, and you need to tell Rust if you want + //@ to change a variable later. - // Now we want to *iterate* over the list. Rust has some nice syntax for - // iterators: + // Now we want to *iterate* over the list. Rust has some nice syntax for iterators: for el in vec { - // So `el` is al element of the list. We need to update `min` accordingly, but how do we get the current + // So `el` is an element of the list. We need to update `min` accordingly, but how do we get the current // number in there? This is what pattern matching can do: match min { + // In this case (*arm*) of the `match`, `min` is currently nothing, so let's just make it the number `el`. NumberOrNothing::Nothing => { - // `min` is currently nothing, so let's just make it the number `el`. - min = NumberOrNothing::Number(el); + min = NumberOrNothing::Number(el); /*@*/ }, + // In this arm, `min` is currently the number `n`, so let's compute the new minimum and store it. + //@ We will write the function `min_i32` just after we completed this one. NumberOrNothing::Number(n) => { - // `min` is currently the number `n`, so let's compute the new minimum and store it. - let new_min = std::cmp::min(n, el); - min = NumberOrNothing::Number(new_min); + let new_min = min_i32(n, el); /*@*/ + min = NumberOrNothing::Number(new_min); /*@*/ } } + //@ Notice that Rust makes sure you did not forget to handle any case in your `match`. We say + //@ that the pattern matching has to be *exhaustive*. } // Finally, we return the result of the computation. return min; } -// Phew. We wrote our first Rust function! But all this `NumberOrNothing::` is getting kind of -// ugly. Can't we do that nicer? Indeed, we can: The following line tells Rust to take -// the constructors of `NumberOrNothing` into the local namespace: -use self::NumberOrNothing::{Number,Nothing}; -// Try moving that above the function, and removing all the occurrences `NumberOrNothing::`. -// Things should still compile, now being much less verbose! - -// However, the code is still not "idiomatic Rust code". To understand why, it is important to -// understand that Rust is an "expression-based" language. This means that most of the -// terms you write down are not just *statements* (executing code), but *expressions* -// (returning a value). This applies even to the body of entire functions! - -// For example, consider `sqr`. Between the curly braces, we are giving the *expression* -// that computes the return value. So we can just write `i * i`, the expression that -// returns the square if `i`, and make that our return value! Note that this is -// very close to how mathematicians write down functions (but with more types). -fn sqr(i: i32) -> i32 { i * i } - -// Conditionals are also just expressions. You can compare this to the ternary `? :` operator -// from languages like C. -fn abs(i: i32) -> i32 { if i >= 0 { i } else { -i } } - -// And the same applies to case distinction with `match`: Every `arm` of the match -// gives the expression that is returned in the respective case. -fn number_or_default(n: NumberOrNothing, default: i32) -> i32 { - match n { - Nothing => default, - Number(n) => n, +// Now that we reduced the problem to computing the minimum of two integers, let's do that. +fn min_i32(a: i32, b: i32) -> i32 { + if a < b { + return a; /*@*/ + } else { + return b; /*@*/ } } -// With this fresh knowledge, let us now refactor `vec_min`. -fn vec_min(v: Vec) -> NumberOrNothing { - let mut min = Nothing; - for e in v { - // First of all, notice that all we do here is compute a new value for `min`, and that we - // will always end up calling the `Number` constructor. In Rust, the structure of the code - // can express this uniformity as follows: - min = Number(match min { - Nothing => e, - Number(n) => std::cmp::min(n, e) - }); - } - // The `return` keyword exists in Rust, but it is rarely used. Instead, we typically - // make use of the fact that the entire function body is an expression, so we can just - // write down the desired return value. - min -} +// Phew. We wrote our first Rust function! But all this `NumberOrNothing::` is getting kind of +// ugly. Can't we do that nicer? -// Now that's already much shorter! Make sure you can go over the code above and actually understand -// every step of what's going on. +// Indeed, we can: The following line tells Rust to take +// the constructors of `NumberOrNothing` into the local namespace. +// Try moving that above the function, and removing all the occurrences of `NumberOrNothing::`. +use self::NumberOrNothing::{Number,Nothing}; -// To call this function, we now just need a list! Of course, ultimately we want to ask the user for -// a list of numbers, but for now, let's just hard-code something: +// To call this function, we now just need a list. Of course, ultimately we want to ask the user for +// a list of numbers, but for now, let's just hard-code something. +//@ `vec!` is a *macro* (as indicated by `!`) that constructs a constant `Vec<_>` with the given +//@ elements. fn read_vec() -> Vec { - // `vec!` is a *macro* (as you can tell from the `!`) that constructs a constant `Vec` with the given - // elements. - vec![18,5,7,1,9,27] + vec![18,5,7,1,9,27] /*@*/ } -// Finally, let's call our functions and run the code! -// But, wait, we would like to actually see something. Of course Rust can print numbers, -// but after calling `vec_min`, we have a `NumberOrNothing`. So let's write a small helper -// function that can prints such values. +// Of course, we would also like to actually see the result of the computation, so we need to print the result. +//@ Of course Rust can print numbers, but after calling `vec_min`, we have a `NumberOrNothing`. +//@ So let's write a small helper function that prints such values. +//@ `println!` is again a macro, where the first argument is a *format string*. For +//@ now, you just need to know that `{}` is the placeholder for a value, and that Rust +//@ will check at compile-time that you supplied the right number of arguments. fn print_number_or_nothing(n: NumberOrNothing) { - match n { - Nothing => println!("The number is: "), - Number(n) => println!("The number is: {}", n), - }; + match n { /*@*/ + Nothing => println!("The number is: "), /*@*/ + Number(n) => println!("The number is: {}", n), /*@*/ + }; /*@*/ } -// So putting it all together - if you type `cargo run`, it will -// run the following code: - -pub fn part_main() { +// Putting it all together: +pub fn main() { let vec = read_vec(); let min = vec_min(vec); print_number_or_nothing(min); } -// Yay, it said "1"! That's actually the right answer. Okay, we could have -// computed that ourselves, but that's besides the point. More importantly: -// You completed the first part of the course. +//@ You can now use `cargo build` to compile your *crate*. That's Rust's name for a *compilation unit*, which in +//@ the case of Rust means an application or a library.
+// Finally, try `cargo run` on the console to run it. + +//@ Yay, it said "1"! That's actually the right answer. Okay, we could have +//@ computed that ourselves, but that's beside the point. More importantly: +//@ You completed the first part of the course. -// [index](main.html) | previous | [next](part01.html) +//@ [index](main.html) | previous | [raw source](https://www.ralfj.de/git/rust-101.git/blob_plain/HEAD:/workspace/src/part00.rs) | [next](part01.html)