So Far So Rust#
As of July 13, 2022, I’ve been using Rust as my primary programming language for 145 days now. In this article, I’m going to discuss my journey so far.
While I’ve talked about this at length to my friends and even at meetups, I haven’t written about Rust as much as I’d like to. I came into this language the same way as I did to Python: sheer gut feeling. While I was considering Ruby instead of Python, here my alternative was Golang. Back then, I wasn’t educated enough to make a decision so I left it to a friend who recommended I choose Python. Today, that choice oddly involved the same friend who’s been using Rust at work. However, it also stemmed from a gut feeling that I wanted to learn Rust and not Golang for my side-projects.
An Aside
A funny thing about my journey with Rust and Ruby is that Steve Klabnik is involved in both of them. I chose Python over Ruby, and Steve is one of the most prolific contributors to Ruby on Rails. And I chose Rust over Golang, and was mildly surprised to find that Steve is involved with Rust and is one of the co-authors of the defacto book on the topic.
Another fun fact is that Armin Ronacher, the creator of Flask, my favourite Web Framework, also shifted to Rust. In fact, Armin’s talk at PyCon India 2018 convinced me that my gut feeling about Rust was right.
How did I go about learning Rust? As someone who’s only coded in Python and a little bit of Javascript, I didn’t have much of an understanding of low-level programming constructs such as memory management to really grok stuff at the beginning. The advice most newbies get is “read The Book”, and while that is great advice, I kept faltering when trying to read it. I wanted to get my hands dirty ASAP and I wanted to build something that I could use.
Note
Rust has the reputation of being a Systems Programming Language. While the language did originally target itself as being an accessible systems language, the home page now says this instead.
Don’t let the systems programming language tag scare you. Rust is designed to be used everywhere.
The opportunity came at work when I needed to build a health-check endpoint atop of our data services at Merkle Science. The original requirements needed me to hit our internal APIs and measure latency in realtime, and the prototype had been written in Python. It was introducing latency of its own and clearly that was an effect of the way it had been written. I could have rewritten it so that it stayed in Python, but that felt like a perfect reason to use Rust. I needed to write a webservice, so I sought the framework that looked the most like Flask, my preferred web-framework of choice: Rocket. In retrospect I was thrown off the more popular choice in Rust, actix-web, because I thought I had to use the actor-framework model that actix came with. Surprisingly, I learnt a long time later that actix-web doesn’t use actix under the hood anymore. The names are an unfortunate left-over from when the projects stemmed together. However, Rocket was easy to understand and I liked what I saw.
The funny thing about how I started Rust is this. I’d been trying to learn it for months and I came upon a book called Rust Web Programming by Maxwell Filton and began reading it. I didn’t complete the book, but the first two chapters gave me more than enough to try writing code on my own. Another resource I found helpful earlier was A Half-Hour to Learn Rust, by Amos Wenger.
Note
These resources worked for me, but again, read The Book, which is the definitive resource for newbie Rustaceans. I am working my way through the book now, and it teaches me a lot that I wish I knew earlier.
I had seen the trait system at work, and wanted to use it for my task, since it involved hitting similar APIs which all had their own nuances, while returning a synonymous result. I wanted to create a trait that all the checks would implement, so that it would be extensible later down the line.
After trying to write a little rust for a few days, I decided to tackle the Rustlings course, which provide interactive exercises for you to follow along. I solved all of them over a period of 3 days and I had to google for the answers to some of them. In particular, I recommend the exercises on strings to really grind some of the basics into your brain.
While trying to grok how strings work in Rust, my friend recommended the Let’s Get Rusty YouTube Channel and, in particular, the video on Strings.
After going through these resources, I began building my service and encountered a few gotchas that seem trivial today.
I couldn’t return a string slice from within a function because it was created inside the function and essentially points to something that no longer exists.
Async functions could be called from within a synchronous functions as long as I wrap them inside a blocking runtime, such as the one offered by the futures crate or the tokio crate
To implement something I’d take for granted in Python, such as a retrying mechanism, I’d have to understand how to write a function that returns a generic response. This is something I’ve documented here.
There are a few more gotchas that I ran into, especially when it came to
something as simple as the return keyword.
When I first started learning Rust, the fact that a function could be written like below, stumped me.
1fn add_ten(x: i32) -> i32 {
2 x + 10
3}
Where’s the return statement? Turns out Rust blocks all operate this way. A
block can evaluate to some value, and you could also write code like this.
1fn add_ten(x: i32) -> i32 {
2 let result: i32 = {
3 println!("I'm going to add 10 to {x}");
4 x + 10
5 };
6 result
7}
The block that contains the println! statement and its subsequent line
evaluates to x+10. Take this in for a second. Every single statement you
will see eventually runs on this simple assertion. Match statements, if-let
statements, or even a simple block like above that’s only written this way to
show you what you could do, all of them can evaluate to a result.
And the type annotations can coerce functions that use generics to return appropriate values.
Note
Rust is the first strongly typed language I have used, so forgive me if I sound like a fanboy when it comes to this. I have serious beef with Python’s type hints, and how it’s just a hint and nothing more. I love how Rust handles types and how constraining it feels, just so that I can use these constraints to not only be a better developer, but also to control my program from going haywire.
One of the things I’ve learnt writing Rust as opposed to Python is that my code fails, frequently.
I wrote a crawler to get data out of a paginated API, and my code had to account for edgecases pretty early on. The structs I wrote to deserialize payloads would fail at the first hint of a mismatch, especially when it came to types. I needed to know what could be a null value and what wouldn’t be. I needed to know the exact format for a timestamp and how to parse it.
Other things about Rust seemed to be harder to grasp. The import system was something else that I really struggled with.
1use std::env;
2
3fn main() {
4 let value = env::var("SOME_ENVIRONMENT_VARIABLE");
5 match value {
6 Ok(v) => println!("SOME_ENVIRONMENT_VARIABLE = `{}`"),
7 Err(e) => eprintln!("Error: {e:?}");
8 };
9}
In this above block, I assumed the use std::env line was an import
statement. I was blissfully unaware that I didn’t *need to import something
like this. I could, instead, have written it like follows.
1fn main() {
2 let value = std::env::var("SOME_ENVIRONMENT_VARIABLE");
3 match value {
4 Ok(v) => println!("SOME_ENVIRONMENT_VARIABLE = `{}`"),
5 Err(e) => eprintln!("Error: {e:?}");
6 };
7}
The use keyword wasn’t synonymous with Pythons import keyword. It was
more akin to import requests.client as client, bringing something into my namespace so
that I didn’t need to constantly refer to the entire import resolution path.
When I didn’t know this, using the env_logger crate was unnerving.
I didn’t know if Rust was injecting env_logger into my namespace somehow. It
made me really uncomfortable because I hate having variables in my namespace
that I didn’t control.
However, Rust was doing something else. Every crate was available to use, if I
didn’t want to use it, I could have just written out the entire import path.
Those are not all of the gotchas, though. I continue to run into them every day, but you get the idea.
Today, I’m using Rust in two projects at Merkle Science. First is the health check application I mentioned, while the second is a data-ops CLI that I wrote to help us load data into our datastores.
Choosing Rust - A Retrospective#
This article has been about how I learnt Rust, and trust me, there’s a lot more to say about that. My advice is pick a problem you’d choose any language for and choose Rust to solve it. Remember that it will take you longer, not just at first, than it would in a language like Python or Javascript. I don’t even profess to know that you’ll get guaranteed benefits, but what I can tell you is that you will learn new things; new paradigms.
I chose Rust because I felt like I wanted to code in Rust. I don’t claim to have a performance or memory-safety related answer to you. I’ve been falling slightly out of love with Python, largely because of the state of its virtual environments ecosystem or the “ship your development machine” gimic that containers seem to inculcate into every developer.
Rust is fun. I’ve had so much fun coding in Rust that it reminds me of how much fun Python used to be. I am glad I can use it at work, and I am glad I can use it for my side projects. It’s a language that feels fresh, and coding in it has taught me so much already.
I have learnt how to think about the mutability of my variables.
I think a lot more about the types of my variables and the scopes they live in.
I have learnt a lot about accounting for errors in my code.
Resultis amazing.Enumsare mind-blowingly amazing.Pattern matching is just chef’s kiss.
All of this being said and done, I don’t know if Rust is for you. Perhaps
Golang is. There’s no reason to try both. I did, and I still don’t have a
problem with Go. I prefer Rust, that’s all. If you’re trying to look for a job,
I don’t recommend Rust. There are so few jobs out there for you. If you’re
trying to rewrite your codebase for performance, let me tell you that you might
be able to do it in the language of your choice. Look into distributed
programming patterns, or even see if you can learn to profile and DRY your code.
Learn to use caching and memoization. Improve your algorithms and
datastructures. At times, your code might not even be the bottle-neck; for most
web-applications, it’s your database queries. Don’t use an ORM blindly. Learn
how SQL’s EXPLAIN statement works, you will build more efficient web
applications with ease.
Sometimes, a shell-script does everything you need a fancy CLI to do. So do that instead.
However, should you choose Rust, great job. You will struggle for a while, I
know I do even today, but you will also be able to build an amazing application
because you’re closer to the metal. You will enjoy not having to install
build-essential, or other build-time dependencies on your production
machine. You don’t need to install libgit or those pesky openssl
libraries either. Your compiled application has everything it needs.
Rust is fun. Get used to the compiler shouting at you, because it’s a good teacher.
Resources#
Note
I’ll update this list as and when I get a chance or come across something that helps me learn a newer concept.
While I used some pretty nondescript resources to learn Rust, I don’t recommend you do the same. At the time I began, a lot of the below resources didn’t exist, so I didn’t have an opportunity to use them. However, I recommend you go through these instead of following my path.
Books#
Podcasts#
New Rustcean (discontinued)