Hi All! Welcome to the Reading Club for Rust’s “The Book” (“The Rust Programming Language”). This is week 1 (the beginning!!).
Have a shot at going through “the reading” and post any thoughts, confusions or insights here
“The Reading”
-
Finish up to Chapter 2: “Programming a Guessing Game”
-
The Book: https://rust-book.cs.brown.edu/title-page.html (the special Brown University version with quizzes etc)
The Twitch Stream
- @sorrybookbroke@sh.itjust.works ran a twitch stream on these chapters which is now on YouTube: https://www.youtube.com/watch?v=ou2c5J6FmsM
- You might prefer watching and listening to that rather than reading the book.
- Be sure to catch future streams (will/should be weekly: https://www.twitch.tv/deerfromsmoke)
What’s Next Week?
- Chapters 3 and 4
- Start thinking about challenges or puzzles to try as we go in order to get some applied practice!
- EG, Advent of Code
- Maybe some basic/toy web apps such as a “todo”
Even though The Book is a bit verbose in these first few sections and really only touches on the basics of the language, I enjoyed going through it! Once you’ve gone to the end of just chapter 2, you’ve touched on project management with
cargo
, compiling withrustc
orcargo
,match
statements, and some of the other syntactical details of the language. It’s definitely enough to get you started if you’re new to the language!
For me, my outstanding questions, which come from the final exercise that builds a “guessing game” (code extracted below for easy reference):
- Why are macros covered so much later in the book (ch 19)? … not for mere mortals?
- I don’t think I really know at all what an object like
Ordering
actually is and what mechanically happened when it was used in the match statement. - Why did I import/use
rand::Rng
but then writerand.thread_rng().gen_range()
. That I’m importing something not explicitly used in the code (Rng
is never used) feels off to me. I can only guess that this was a shortcut to get to use a higher level interface. But what isRng
? - This will probably come up when we cover “Ownership” … but it strikes me now that we were passing variables by reference (eg
and
. Given rust’s concern with ownership and the “borrow checker” as a means of preventing memory safety issues … why isn’t it the default behaviour that a variable is passed by reference? Why do we have to explicitly pass the reference ourselves (with the ampersand syntax
)?
use std::io; use std::cmp::Ordering; use rand::Rng; fn main() { println!("Guess the number!"); let secret_number = rand::thread_rng().gen_range(1..=100); // println!("Secret number is: {secret_number}"); loop { println!("Please input your guess."); let mut guess = String::new(); io::stdin() .read_line(&mut guess) .expect("Failed to read line"); // let guess: u32 = guess.trim().parse().expect("Please type a number!"); let guess: u32 = match guess.trim().parse() { Ok(num) => num, Err(_) => continue, }; println!("you guessed: {guess}"); match guess.cmp(&secret_number) { Ordering::Less => println!("Too small!"), Ordering::Greater => println!("Too big!"), Ordering::Equal => { println!("You win"); break; } } } }
This is just my attempt at answering (I’m learning too):
Macros are easy to use, allowing beginners to write ‘hello world’ for example, but hide a bunch of complicated stuff we’re unlikely to understand until later. (I vaguely remember from Java that just writing something to the console was a whole bunch statements strung together.)
I found it useful to document what each line was doing in the example, to get my head around the terminology.
std
is a crate,io
is a module in thestd
crate, and that module provides methods likestdin()
std
is a crate,cmp
is a module in thestd
crate, and that module provides enums likeOrdering
rand
is a crate,Rng
is a trait from therand
crate, and that trait provides methods likethread_rng()
Ordering
is a enum - a type with a list of variants with developer-friendly names. In the same way that a ‘DaysOfTheWeek’ enum would have ‘Monday’, ‘Tuesday’ …, etc,Ordering
has Less, Greater, or Equal.match
statements work with enums, in a ‘for this, do that’ kind of way.Rng
is a trait, which feature a lot in Rust, but is something I’ve had difficulty getting my head around. Where I’m at so far is they mostly provide convenience. Therand
create provides a number of structures in itsrngs
module -ThreadRng
,OsRng
,SmallRng
, andStdRng
, depending on what you want to use to generate randomness. Instead of sayingThreadRng
provides thegen_range()
method, andOsRng
provides thegen_range()
method, etc, Rust just says they implement theRng
trait. SinceRng
provides thegen_range()
method, it means that everything that implements it will provide it.thread_rng()
returns aThreadRng
structure, which providesgen_range()
via theRng
trait, but we need to bring that trait into scope with theuse
keyword in order to be able to call it.For the default behaviour of passing owned vs. borrowed variables, I guess it’s useful to explicitly state “I’m giving this variable to another function, because I don’t intend to use it anymore”, so that if you inadvertently do, the compiler can catch it and error.
Thanks!!
Rng
is a trait, which feature a lot in Rust, but is something I’ve had difficulty getting my head around. Where I’m at so far is they mostly provide convenience. Therand
create provides a number of structures in itsrngs
module -ThreadRng
,OsRng
,SmallRng
, andStdRng
, depending on what you want to use to generate randomness. Instead of sayingThreadRng
provides thegen_range()
method, andOsRng
provides thegen_range()
method, etc, Rust just says they implement theRng
trait.This makes a lot of sense actually. Thanks! It does lead to the awkward situation where you’d have to know that Rng is the underlying trait of the rest of the module and so import it. But like I said, I’m guessing that’s because this example is aiming only for easy high-level usage of
rand
. I’d guess there’s a lower level way of using therand
crate that would involve more direct imports, perhapsuse rand::ThreadRng
?
For the default behaviour of passing owned vs. borrowed variables, I guess it’s useful to explicitly state “I’m giving this variable to another function, because I don’t intend to use it anymore”, so that if you inadvertently do, the compiler can catch it and error.
Makes sense.
Sooo … is passing by value a thing in rust? Or does just about every method take only reference types as arguments?
I wondered the same, and tried
use rand::rngs::ThreadRng
but the compiler wouldn’t accept it, and suggest I use the trait instead. So it looks like the compiler can be helpful in identifying these things (and maybe traits are the first thing that Rust developers look for when reading the docs for a crate).(wrongness edited out and hopefully corrected in new comment)
and maybe traits are the first thing that Rust developers look for when reading the docs for a crate
Yep. Makes a lot of sense. Probably gotta start thinking in terms of traits at some point. I haven’t spun up any LSP yet but hopefully that can help surface these sorts of things.
Still, at the moment, it does seem like a wrinkle in the usability of the language that you import something which implicitly underlies what you actually want to use.
Also, Thanks!
Sooo … is passing by value a thing in rust? Or does just about every method take only reference types as arguments?
I think this is an occasion where a vague familiarity with other languages ended up confusing me with Rust. The ‘&’ sign doesn’t mean ‘pass by reference’ in the same way as it does in C. Anything with a size that’s fixed at compile time is typically passed by value, whereas variables who’s size might change are passed by reference. The ‘&’ in Rust isn’t about that. For variables that are passed by reference, the ‘&’ is about whether the ownership of that memory address is transferred or not.
To illustrate:
fn abc(v: String) { println!("v is {}", v); } fn main() { let mut v=String::from("ab"); v.push('c'); abc(v); // println!("v is {}", v); }
works fine as it is, but will error if you uncomment the second println! The ‘v’ variable was passed by reference, but it’s ownership was transferred, so it can’t be referred to again.
Thanks!
Seems I gotta dig into the borrow checker before thinking too much about this!
Otherwise, compiling your code snippet is a nice illustration of how helpful the compiler tries to be … lots of tips in the output there! To anyone else, just try running
rustc
on this, with the secondprintln!
uncommented and see the output, which is half error half linting.
For variables that are passed by reference, the ‘&’ is about whether the ownership of that memory address is transferred or not.
Yea. So if the second
println!
were uncommented, how could we compile this? From what you’ve said, I’d guess that&
means “borrow” (ie, not “move” ownership).So if we alter
abc
to take a&String
type and notString
, and therefore only “borrow” the variable, and then pass inand not
v
to pass in a “borrowed” variable, it should compile.fn abc(v: &String) { println!("v is {}", v); } fn main() { let mut v=String::from("ab"); v.push('c'); abc(&v); println!("v is {}", v); }
It seems to!
Of course, as the compiler suggests, we could instead just pass in
v.clone()
which presumably creates a new variable and effectively “passes by value”.
Digging in a bit more, what happens if
abc
(tries to) mutate the variable?We can add
v.push('X')
toabc
and see if we get different printouts. As the compiler would tell us, we would need to make the argumentv
mutable for this to work.fn abc(mut v: String) { v.push('X'); println!("v is {}", v); } fn main() { let mut v=String::from("ab"); v.push('c'); abc(v.clone()); println!("v is {}", v); } // OUTPUT: // v is abcX // v is abc
I’m not clear on why I don’t have to declare that the
v.clone()
is mutable in anyway though.What about trying the same with a "borrowed’ variable?
Well we need mutable borrowed variables, so necessary adjustments to the types of
abc
and its call inmain
. And adding an additional mutation ofv
inmain
afterabc
is called, and we get two different println outputs, with each mutation applying to the same variable.fn abc(v: &mut String) { v.push('X'); println!("v is {}", v); } fn main() { let mut v=String::from("ab"); v.push('c'); abc(&mut v); v.push('Y'); println!("v is {}", v); } // OUTPUT // v is abcX // v is abcXY
Seems I gotta dig into the borrow checker before thinking too much about this!
It’s covered in detail in chapter 4.
The ‘&’ sign doesn’t mean ‘pass by reference’ in the same way as it does in C. Anything with a size that’s fixed at compile time is typically passed by value, whereas variables who’s size might change are passed by reference.
Are you sure? I noticed the rust book said this:
A reference is like a pointer in that it’s an address we can follow to access the data stored at that address; that data is owned by some other variable. Unlike a pointer, a reference is guaranteed to point to a valid value of a particular type for the life of that reference.
No, I’m not sure, tbh. It’s a concept I’m struggling with, and I reliant on others to correct/question me. I was trying to answer the question of whether things get passed by value or not, and I wanted to say yeah, loads of things do (anything who’s size is known at compile-time) and to caution against thinking too much in C terms.
Here’s where my thinking is now: a variable without a & in front is passed by value. A primitive type (i.e. something who’s size is known) will be copied. So if a=4 and you pass it to a function, you can still refer to a later. A variable-length type (e.g. a String) can’t be copied, so it is moved, and referring to it later will be an error.
A variable with a & in front is indeed a reference. It’s a memory address, so it’s of fixed size. For either a primitive or a variable-length type, the address can be copied when passed to a function, so it can be referred to again later without issue.
This feels more correct to me, so hopefully it is. If not, I’m sure someone will have a better answer soon (this community is growing well!).
On the whole primitive types and references thing, I find it helps me to remember that a reference/pointer (subtle but importance difference between rust and C where rust has more guarantees around a pointer to make it a "reference) is also basically just a number like a “primitive” i32 etc. And references/pointers obviously (?) have to get passed by value or copied (in order to “survive” their original stack frame right?), so passing any primitive by value or copying it really isn’t different from passing it by reference, apart from when you’re running a borrow checker for managing memory of course.
This will make more sense once we (this community) get to the 4th week/session and properly get into ownership, but lemme try to explain anyways:
A variable without a
&
in front is moved into the [function’s] scope.Upon exiting a scope, Rust automatically drops/de-allocates any variables that were owned by/moved into said scope.
This is why you need to either
-
pass by ref / have the scope borrow a reference to the variable, or
-
have your function return the variable/object/memory handle that was moved into it
when you want a variable/some data to “out-live” being passed as argument to a function call.
Most often you will use 1), but there are some cases where it can be much nicer to move things “into” a function call and store what you “get back out” (i.e. the return value). Using a “[Type-]State” pattern/approach is a good example of such a case (here’s a better explanation than I can give in a lemmy comment).
Example:
struct Unauthenticated; struct Authenticated { name: String }; impl Unauthenticated { fn login(self, username: String) -> Authenticated { Authenticated { name: username } } } pub fn main() { let un_authed_user = Unauthenticated::new(); let authed_user = un_authed_user.login("Alice"); // `un_authed_user` has effectively been moved into `authed_user` }
Here, we as programmers don’t need to worry about making sure
un_authed_user
gets cleaned up before the program exits, and we don’t need to worry about data that could have been stored insideun_authed_user
being freed too early (and thus not being available toauthed_user
).Admittedly, this is a contrived example that doesn’t really need to worry about ownership, it’s just the bare minimum to illustrate the idea of moving data into and then back out of a function scope. I don’t know of a small enough “real-world” example to give here instead.
Thank you.
I think there’s more to it though, in that simple values aren’t moved, they’re always copied (with any
&
in front indicating whether it’s the value to copy or the address)To illustrate:
fn how_many(a: u32, fruit: String) { println!("there are {} {}", a, fruit); } fn main() { let a=4; let fruit = String::from("Apples"); how_many(a, fruit); println!("the amount was {}", a); // this works println!("the fruit was {}", fruit); // this fails }
The ‘a’ was copied, and the ‘fruit’ was moved.
-
The Enum
Ordering
provides compile-time safety. For example, if cmp() takes a string or int, the compiler can’t catch invalid inputs (“less”, “equal”, -123, …) at compile time and crash at runtime.Hmm, not sure I’m entirely with you.
If the argument to
cmp
is of an incorrect or incompatible type (where AFAIU the parent object and argument have to be the same type, eg u32), that alone will be surfaced at compile time no?If so, then Ordering is actually relatively trivial. It’s an enum, with variants for each possible outcome of a comparison on orderable variables (eg numbers).
And the output of
cmp
is an Ordering type, which is nice for match statements, as, AFAIU, it forces us to address all possible scenarios (each being a variant of the Ordering enum).But the compile time safety will come from basic type checking on the argument to
cmp
.Am I off base here?
Oh I was completely wrong.
cmp()
takes a number (not Ordering) and returns Ordering. Sorry for bothering you.Pretty sure that’s what this is all about! A safe space to work through the ideas and details without worrying about being wrong. I wouldn’t have understood this better if you didn’t “bother” me and now we (and anyone else reading this presumably) are both better off!