String Types

• Rust strings are complicated.
  • Sequences of Unicode values encoded in UTF-8.
  • Not null-terminated and may contain null bytes.
• There are two kinds: &str and String.

&str

• &str is a string slice (like array slice).
• "string literals" are of type &str.¹
• &strs are statically-allocated and fixed-size.
• May not be indexed with some_str[i], as each character may be multiple bytes due to Unicode.
• Instead, iterate with chars():
  • for c in "1234".chars() { ... }
• As with all Rust references, they have an associated lifetime.

¹More specifically, they have the type &'static str.

String

#[derive(PartialOrd, Eq, Ord)]
#[stable(feature = "rust1", since = "1.0.0")]
pub struct String {
    vec: Vec<u8>,
}

• Strings are heap-allocated, and are dynamically growable.
• Cannot be indexed either.
  • You can select characters with s.nth(i).
• May be coerced into an &str by taking a reference to the String.

let s0: String = String::new();
let s1: String = "foo".to_string();
let s2: String = String::from("bar");
let and_s: &str = &s0;

str

• If &str is the second string type, what exactly is str?
• An Unsized type, meaning the size is unknown at compile time.
  • You can't have bindings to strs directly, only references.

String Concatenation

• A String and an &str may be concatenated with +:

let course_code = "CIS".to_string();
let course_name = course_code + " 198";

• Concatenating two Strings requires coercing one to &str:

let course_code = String::from("CIS");
let course_num  = String::from(" 198");
let course_name = course_code + &course_num;

• You can't concatenate two &strs.

let course_name = "CIS " + "198"; // Err!

String & &str: Why?

• Like slices for Vecs, &strs are useful for passing a view into a String.
• It's expensive to copy a String around, and lending an entire String out may be overkill.
• &str therefore allows you to pass portions of a String around, saving memory.

String & &str: Why?

• Generally, start with String unless you have string literals.
  • You can then lend out &strs easily.
• i.e. Building a String via concatenation or formatting, then pass it as a function argument.

String Conversion

• However, actually converting a String into an &str requires a dereference:

use std::net::TcpStream;
TcpStream::connect("192.168.0.1:3000"); // &str
let addr = "192.168.0.1:3000".to_string();
TcpStream::connect(&*addr);

• This doesn't automatically coerce because TcpStream doesn't take an argument of type &str, but a Trait bounded type:
  • TcpStream::connect<A: ToSocketAddr>(addr: A);

Deref Coercions

• We know Rust can automatically dereference from &T to T.
• This can support dereferencing between different types too.

pub trait Deref {
    type Target: ?Sized;
    fn deref(&self) -> &Self::Target;
}

• Since String implements Deref<Target=str>, so values of &String will automatically be dereferenced to &str when possible.

Operators

• Operators are evaluated left-to-right, in the following order:
  • Unary operators: ! - * & &mut
  • as casting
  • * / % multiplicative arithmetic
  • + - additive arithmetic
  • << >> shift arithmetic
  • & bitwise and
  • ^ bitwise xor
  • | bitwise or
  • == != < > <= >= logical comparison
  • && logical and
  • || logical or
  • = .. assignment and ranges
• Also: call(), index[]

Operator Overloading

• Okay, same old, same old. We can customize these!
• Rust defines these - surprise! - using traits, in std::ops.
  • Neg, Not, Deref, DerefMut
  • Mul, Div, Mod
  • Add, Sub
  • Shl, Shr
  • BitAnd
  • BitXor
  • BitOr
  • Eq, PartialEq, Ord, PartialOrd
  • And
  • Or
• Also: Fn, FnMut, FnOnce, Index, IndexMut, Drop

Option<T>

enum Option<T> {
    None,
    Some(T),
}

• Provides a concrete type to the concept of nothingness.
• Use this instead of returning NaN, -1, null, etc. from a function.
• No restrictions on what T may be.

Option::unwrap()

• The pattern where None values are ignored is pretty common:

// fn foo() -> Option<i32>
match foo() {
    None => None,
    Some(value) => {
        bar(value)
        // ...
    },
}

Option::unwrap()

• What if we extracted the pattern match into a separate function to simplify it?

fn unwrap<T>(&self) -> T { // 🎁!
    match *self {
        None => panic!("Called `Option::unwrap()` on a `None` value"),
        Some(value) => value,
    }
}
let x = foo().unwrap();
let y = bar(x);
// ...

• Unfortunately, panic!ing on None values makes this abstraction inflexible.
• Better: use expect(&self, msg: String) -> T instead.
  • panic!s with a custom error message if a None value is found.

Option::map()

• Let's make the pattern a little better.
• We'll take an Option, change the value if it exists, and return an Option.
  • Instead of failing on None, we'll keep it as None.

fn map<U, F>(self, f: F) -> Option<U>
        where F: FnOnce(T) -> U {
    match self {
        None => None,
        Some(x) => Some(f(x))
    }
}
// fn foo() -> Option<i32>
let x = foo().map(|x| bar(x));

Option::and_then()

• There's a similar function and_then:

fn and_then<U, F>(self, f: F) -> Option<U>
      where F: FnOnce(T) -> Option<U> {
    match self {
        Some(x) => f(x),
        None => None,
    }
}
// fn foo() -> Option<i32>
let x = foo().and_then(|x| Some(bar(x)));

• Notice the type of f changes from T -> U to T -> Some(U).

Option::unwrap_or()

• If we don't want to operate on an Option value, but it has a sensible default value, there's unwrap_or.

impl<T> Option<T> {
    fn unwrap_or<T>(&self, default: T) -> T {
      match *self {
          None => default,
          Some(value) => value,
      }
    }
}

Option::unwrap_or_else()

• If you don't have a static default value, but you can write a closure to compute one:

impl<T> Option<T> {
    fn unwrap_or_else<T>(&self, f: F) -> T
            where F: FnOnce() -> T {
        match *self {
            None => f(),
            Some(value) => value,
        }
    }
}

Other

• Some other methods provided by Option:
• fn is_some(&self) -> bool
• fn is_none(&self) -> bool
• fn map_or<U, F>(self, default: U, f: F) -> U
  • where F: FnOnce(T) -> U
  • A default value: U.
• fn map_or_else<U, D, F>(self, default: D, f: F) -> U
  • where D: FnOnce() -> U, F: FnOnce(T) -> U
  • A default-generating closure: D.

Other

• fn ok_or(self, err: E) -> Result<T, E>
• fn ok_or_else(self, default: F) -> Result<T, E>
  • where F: FnOnce() -> E
  • Similar to unwrap_or but returns a Result with a default Err or closure.
• fn and<U>(self, optb: Option<U>) -> Option<U>
  • Returns None if self is None, else optb
• fn or(self, optb: Option<T>) -> Option<T>
  • returns self if self is Some(_), else optb

Result

enum Result<T, E> {
    Ok(T),
    Err(E)
}

• Result is like Option, but it also encodes an Err type.
• Also defines unwrap() and expect() methods.
• Can be converted to an Option using ok() or err().
  • Takes either Ok or Err and discards the other as None.
• Can be operated on in almost all the same ways as Option
  • and, or, unwrap, etc.

Result

• Unlike Option, a Result should always be consumed.
  • If a function returns a Result, you should be sure to unwrap/expect it, or otherwise handle the Ok/Err in a meaningful way.
  • The compiler warns you if you don't.
  • Not using a result could result (ha) in your program unintentionally crashing!

Custom Result Aliases

• A common pattern is to define a type alias for Result which uses your libary's custom Error type.

use std::io::Error;
type Result<T> = Result<T, Error>;

• Typically a convenience alias; other than fixing E = Error, this is identical to std::Result.
• Users of this type should namespace it:

use std::io;
fn foo() -> io::Result {
    // ...
}

try!

• try! is a macro, which means it generates Rust's code at compile-time.
  • This means it can actually expand to pattern matching syntax patterns.
• The code that try! generates looks roughly like this:

macro_rules! try {
    ($e:expr) => (match $e {
        Ok(val) => val,
        Err(err) => return Err(err),
    });
}
try!(Ok(42));
match Ok(42) {
    Ok(val) => val,
    Err(err) => return Err(err);
}

try!

• try! is a concise way to implement early returns when encountering errors.

let socket1: TcpStream = try!(TcpStream::connect("127.0.0.1:8000"));
// Is equivalent to...
let maybe_socket: Result<TcpStream> =
    TcpStream::connect("127.0.0.1:8000");
let socket2: TcpStream =
    match maybe_socket {
        Ok(val) => val,
        Err(err) => { return Err(err) }
    };

• This is actually a slight simplification.
  • Actual try! has some associated trait magiclogic.
• You can only call try! in a function that returns a Result.

Collections

Vec<T>

• Nothing new here.

VecDeque<T>

• An efficient double-ended Vec.
• Implemented as a ring buffer.

LinkedList<T>

• A doubly-linked list.
• Even if you want this, you probably don't want this.
  • Seriously, did you even read any of Gankro's book?

HashMap<K,V>/BTreeMap<K,V>

• Map/dictionary types.
• HashMap<K, V> is useful when you want a basic map.
  • Requires that K: Hash + Eq.
  • Uses "linear probing with Robin Hood bucket stealing".
• BTreeMap<K, V> is a sorted map (with different performance benefits).
  • Requires that K: Ord.
  • Uses a B-tree under the hood (surprise surprise).

HashSet<T>/BTreeSet<T>

• Sets for storing unique values.
• HashSet<T> and BTreeSet<T> are literally struct wrappers for HashMap<T, ()> and BTreeMap<T, ()>.
• Same tradeoffs and requirements as their Map variants.

BinaryHeap<T>

• A priority queue implemented with a binary max-heap.

Iterator

• Iterators are types which provide fn next() -> Option<T>.
• Yields Some(val) until the end of the iterator (if it exists), then None.

Iterator Consumers

• Consumers operate on an iterator and return one or more values.
• There are like a billion of these, so let's look at a few.

collect

• collect() rolls a (lazy) iterator back into an actual collection.
• The target collection must define the FromIterator trait for the Item inside the Iterator.
• collect() sometimes needs a type hint to properly compile.
  • The output type can be practically any collection.

fn collect<B>(self) -> B where B: FromIterator<Self::Item>
let vs = vec![1,2,3,4];
// What type is this?
let set = vs.iter().collect();
// Hint to `collect` that we want a HashSet back.
// Note the lack of an explicit <i32>.
let set: HashSet<_> = vs.iter().collect();
// Alternate syntax! The "turbofish" ::<>
let set = vs.iter().collect::<HashSet<_>>();

fold

fn fold<B, F>(self, init: B, f: F) -> B
    where F: FnMut(B, Self::Item) -> B;
let vs = vec![1,2,3,4,5];
let sum = vs.iter().fold(0, |acc, &x| acc + x);
assert_eq!(sum, 15);

• fold "folds up" an iterator into a single value.
  • Sometimes called reduce or inject in other languages.
• fold takes two arguments:
  • An initial value or "accumulator" (acc above) of type B.
  • A function that takes a B and the type inside the iterator (Item) and returns a B.
• Rust doesn't do tail-recursion (yet), so fold is implemented iteratively.
  • See here for more info.

filter

fn filter<P>(self, predicate: P) -> Filter<Self, P>
    where P: FnMut(&Self::Item) -> bool;

• filter takes a predicate function P and removes anything that doesn't pass the predicate.
• filter returns a Filter<Self, P>, so you need to collect it to get a new collection.

skip

fn skip(self, n: usize) -> Skip<Self>;

• Creates an iterator that skips its first n elements.

zip

fn zip<U>(self, other: U) -> Zip<Self, U::IntoIter>
    where U: IntoIterator;

• Takes two iterators and zips them into a single iterator.
• Invoked like a.iter().zip(b.iter()).
  • Returns pairs of items like (ai, bi).
• The shorter iterator of the two wins for stopping iteration.

any & all

fn any<F>(&mut self, f: F) -> bool
    where F: FnMut(Self::Item) -> bool;
fn all<F>(&mut self, f: F) -> bool
    where F: FnMut(Self::Item) -> bool;

• any tests if any element in the iterator matches the input function
• all tests all elements in the iterator match the input function
• Logical OR vs. logical AND.

enumerate

fn enumerate(self) -> Enumerate<Self>;

• Want to iterate over a collection by item and index?
• Use enumerate!
• This iterator returns (index, value) pairs.
  • index is the usize index of value in the collection.

Iterator Adapters

• Adapters operate on an iterator and return a new iterator.
• Adapters are often lazy -- they don't evaluate unless you force them to!
• You must explicitly call some iterator consumer on an adapter or use it in a for loop to cause it to evaluate.

map

fn map<B, F>(self, f: F) -> Map<Self, F>
    where F: FnMut(Self::Item) -> B;
let vs = vec![1,2,3,4,5];
let twice_vs: Vec<_> = vs.iter().map(|x| x * 2).collect();

• map takes a function and creates an iterator that calls the function on each element
• Abstractly, it takes a Collection<A> and a function of A -> B and returns a Collection<B>
  • (Collection is not a real type)

take & take_while

fn take(self, n: usize) -> Take<Self>;
fn take_while<P>(self, predicate: P) -> TakeWhile<Self, P>
    where P: FnMut(&Self::Item) -> bool;

• take creates an iterator that yields its first n elements.
• take_while takes a closure as an argument, and iterates until the closure returns false.
• Can be used on infinite ranges to produce finite enumerations:

for i in (0..).take(5) {
    println!("{}", i); // Prints 0 1 2 3 4
}

cloned

fn cloned<'a, T>(self) -> Cloned<Self>
    where T: 'a + Clone, Self: Iterator<Item=&'a T>;

• Creates an iterator which calls clone on all of its elements.
• Abstracts the common pattern vs.iter().map(|v| v.clone()).
• Useful when you have an iterator over &T, but need one over T.

drain

• Not actually an Iterator method, but is very similar.
• Calling drain() on a collection removes and returns some or all elements.
• e.g. Vec::drain(&mut self, range: R) removes and returns a range out of a vector.

Iterators

• There are many more Iterator methods we didn't cover.
• Take a look at the docs for the rest.