Introduction – Rust for Python Programmers

https://microsoft.github.io/RustTraining/python-book/

by linhns • 2 months ago

71 27 2 months ago

microsoft.github.io

Tiberium• 2 months ago

The whole "book" seems to be AI-generated, or at least very heavily AI-edited. Would at this point it not be easier to just tell developers to use their LLM of choice to achieve the same (or, likely, better) result?

Random chapter so you can judge the quality for yourself: https://microsoft.github.io/RustTraining/python-book/ch09-er...

And the non-stop bullet list slop just looks horrible: https://microsoft.github.io/RustTraining/python-book/ch01-in...

Seems like this isn't limited to the Python book though, and others have the same issues: https://github.com/microsoft/RustTraining/issues/14

onlyrealcuzzo6• 2 months ago

I can't recommend Rust enough. It has such a bad reputation, but it isn't that hard. I truly think it's easier than many languages with much less-intimidating reputations.

That said, one of the places Rust loses people pretty early on is an example they have early in this intro:

```rust

let parts: Vec<&str> = "a,b,c".split(',').collect(); // Vec<&str>

```

I never understood why Rust didn't / couldn't make functions able to return different outputs depending on context. If you chain `.split()` to something else that can take an iterator, you want to pass an iterator. If you don't, ~99% of people would probably rather have a collected array. And if you want an `it`, you could just do `.it()` or this is when type inference could be overloaded and you could do:

```rust

let it: Split<'_, char> = "a,b,c".split(',');

```

I think Rust should've put more effort into making the thing newbs want to do the default, and easy ways to get the most efficient thing for experts.

```rust

let parts = "a,b,c".split(); // Gives an Array/Vec

let count = "a,b,c".split().count(); // Optimized stream, no array allocation

```

It could work like that, and I think almost everyone would be happy. But it doesn't.

Instead, they've created a language that I think could have been nearly as easy as a scripting language, but isn't.

It obviously isn't only collection iterators this applies to. There's dozens of very small places that add up and make what - I believe - is an otherwise relatively easy and sensible language feel too far out of reach for too many people.

`Option<T>`, `Result<T, E>`, `Future<T>` all impact linguistically how you can interact with a Type. I think the impacts of this don't make sense to people who've never encountered this before. `Arc<T>`, `Rc<T>`, `Mutex<T>`, and `RwLock<T>`, etc also have similar consequences.

Not only do people just not get it. But also, the type system quickly becomes "scary". To do pretty basic concurrency, you need to build a pretty "scary" looking type if you come from Python.

Which is why I'm a psychopath and attempting to create a language where it defaults to the things most people want, and it's very easy for experts to override.

mightyham• 2 months ago

Having the result type of a function change based on context sounds like a horrible idea because it would introduce tons of unnecessary ambiguity. If it's an issue that you have to chain one extra collect call, just write a helper function for splitting that returns a vec.

onlyrealcuzzo6• 2 months ago

Having the result type of a function change based on context sounds like a horrible idea because it would introduce tons of unnecessary ambiguity.

I think you're assuming the language won't warn you if you're doing something cost ineffective, and that there aren't modes to compilation which will make you make things explicit whenever its ambiguous.

For a person to get started, they should be able to compile in `easy` mode and do things that make sense to them, and the compiler should only bitch at you to be pedantic when you ask for that.

Especially because an LLM can probably do that pedantic part for you.

You: write code almost as high level as a scripting language, it works, turn on strict mode, most of the time you get auto-fix solutions/options from the compiler.

That's my opinion anyhow. I assume most Rust people won't like it. That's fine. You already have your language!

I'm not trying to make a slightly different Rust for Rust people...

I'm trying to make Rust more accessible to everyone else.

embedding-shape2• 2 months ago

I never understood why Rust didn't / couldn't make functions able to return different outputs depending on context

Referential transparency probably being the first reason I could see. Having the behaviour of a callee sounds horrible and something we usually try to actively work against, you want to be able to look at the function in isolation and be able to understand how it works and what values it gives back, without jumping around and seeing where it's being called.

And yes, I'd agree with your last part, you do sound a bit like a psychopath ;) With that said, the world needs those too, so I hope your experiment is fun and brings you lots of learnings, enjoy! :)

sgeisenh• 2 months ago

This closely resembles implicit conversion from C++ and in many serious codebases it is considered poor practice because it leads to a lot of hidden control flow.

I think Rust strikes a nice balance but there’s enough magic that some people still get frustrated. Following traits can get tedious at times.

onlyrealcuzzo6• 2 months ago

This closely resembles implicit conversion from C++ and in many serious codebases it is considered poor practice because it leads to a lot of hidden control flow.

Yes, which is why there's progressive modes of compilation to not allow anything ambiguous if that's what you want (i.e. an enterprise codebase).

But, a junior can still try things out in weaker modes of compilation, and then once they've got something working, it is typically very easy to do the pedantry to remove ambiguous behavior.

pie_flavor4• 2 months ago

You may notice no language on the planet does this. That is because it is bad. Type guesstimation is a great way of ensuring random problems crop up in random places where they aren't expected and of making the typechecker much slower and more prone to unresolvability (see Swift's multi-minute compiles). All to save you from having to learn what an iterator is, in case you come from a language where lists are more common than iterators; the experience of being scared by a type, and then discovering that the type is not scary in chapter 13.2, is not actually worth making the simple type system instead staggeringly complicated.

onlyrealcuzzo6• 2 months ago

You may notice no language on the planet does this. That is because it is bad.

Perhaps I've explained this poorly, but C#, Java, Perl, & Haskell (and I'm sure others as well) do versions of this already...

You even seem to imply that Swift does, though I have almost no experience directly with Swift.

The vast majority of times it is NOT ambiguous. The compiler can flag it, IFF you want it to.

If you're coming from Python and you want to ease your learning experience, you probably don't want to hit several brick walls before you can do anything...

If you have an enterprise codebase, you probably don't want to allow anything to be compiled that could be ambiguous, so you can force that mode of compilation (and likely should)...

I don't know of any major language which have progressive modes of compilation like this. I think people will find it useful.

Maybe it'll be a disaster. Time will tell. The whole point is to intelligent design the modes such that you can't ever get a MASSIVE surprise "upgrading" from one mode to the next, any error that is too hard to resolve basically automatically / through selecting options needs to be dealt with at the easiest mode.

iloveoof• 2 months ago

This tutorial is very bad, and the time estimates are pretty absurd.

The explanations are extremely short and I imagine a new Rust dev would not understand what is going on.

The Brown tutorial is far better, compare its section on mutables and ownership to this.

And yes, this entirely thing is AI generated. Why was this created?

embedding-shape2• 2 months ago

and the time estimates are pretty absurd.

I wonder when LLMs will catch up with the new timelines, they frequently cite days/weeks of worth, then you say "Ok, implement that" and 30 minutes later everything been implemented. But seems they themselves is stuck not realizing they're not estimating for human timeframes anymore.

tclancy• 2 months ago

I actually asked Claude about that this week and it was fairly interesting: the estimates are based on historical data, so I think that suggests in the current set up there will always be wild gaps. As best I could tell, the estimates you get are probably for the model one or two whole versions ago?

onlyrealcuzzo6• 2 months ago

Link to the Brown tutorial: https://rust-book.cs.brown.edu/

pie_flavor4• 2 months ago

In my experience working with newbies, the alterations Brown makes to the chapters on borrowing are strictly worse. It is entirely focused on ramming the difference between the stack and the heap down your throat, which has nothing to do with ownership and borrowing, and newbies will frequently say that they are extremely confused by the chapter, then sigh in relief upon reading the original version. Just use the official guide, nobody has improved upon it yet. https://doc.rust-lang.org/book

Quothling• 2 months ago

Cargo vs pip/Poetry

I know this section is really just a comparrison of pyproject.toml and cargo.toml, but who on earth would use pip instead of UV as a drop-in replacement in 2026? Though calling it a comparrison is a bit of a stretch considering there is no text.

On top of that, I imagine that a lot of Python programmers who actually do use pip would also use requirements.txt and not pyproject.toml

nickpsecurity• 2 months ago

People learning Python or searching for it will run into endless answers using PIP. Then, lots of advice on how to work around PIP's problems. Then, multiple alternatives they have to consider. I only recently started using UV after going through all that.

Packaging, concurrency, and type errors had me strongly considering switching to Go or Rust recently. These are such long-solved problems in other languages that I question why we should put up with it in Python. Then, I remember it was the ecosytem, including job market and AI performance, that made me use Python.

So, maybe a Python/Rust combo... There's the extensions the OP article mentioned and a Python interpreter written in Rust.

poulpy123• 2 months ago

Because it's rust for python dev, not rust for python dev who use uv. I would understand your comment if you mentioned poetry, but pip has been the standard for years

kermatt• 2 months ago

Because Astral is now owned by a profit driven org, and that often becomes problematic.

Also because Pip ain't perfect, but it is stable in a sense.

muadddib• 2 months ago

What pathetic garbage AI slop. I apologize to everyone who clicked and wasted their time.

dartharva• 2 months ago

Jeez, what a hostile crowd. I personally am just glad companies like Microsoft, that infamously used to charge for every material they put out, have been providing free learning resources to newcomers like this, quality notwithstanding.

I am just extremely grateful in general for anyone who takes the effort to put something out there for public education for free. If you are one of those, irrespective of how your content is appreciated, please take my heartfelt thanks. You guys are at the forefront of restoring my faith in humanity.

paultopia• 2 months ago

I feel like this is a lost cause, but why do people who publish these online free book-like-substances never make the files available as a single download? Do I really need to run a scraper if I want to read this as an epub on my kindle?

frumiousirc• 2 months ago

"microsoft", okay, off to a great start. First click in the ToC that looks interesting and something I'd actually like to know as a RUST outsider:

    Common Python Pain Points That Rust Addresses

But then number one:

    1. Runtime Type Errors
    
    The most common Python production bug: passing the wrong type to a function. 
    Type hints help, but they aren’t enforced.

Uh, okay. This rarely if ever has been a problem for me and I don't usually even use type hints.

Then comes calling out the existence of None, the GIL and packaging as common "pain points". None of these have posed any problem to me essentially ever. Packaging used to be honestly annoying but since uv hit the scene, not at all.

I should have known better and stopped after reading "microsoft".

faangguyindia• 2 months ago

My question is why will Python programmer, use Rust when they can use Go?

As someone who went from C++ to java to Ruby to Javascript to Python, to Go.

I find Go enviornment much superior. Not only your code can keep working for decade and compile fine with new compilers, you get more than enough speed you need.

And also, you don't use random libraries, you precisely use to part you need which imho makes you software more reliable, cleaner and smaller.

Using libraries and other people's module will always bite you just like llm bites you because you believe you understand it but you don't know how it's actually implemented and those who actually go through library and check all that, can implement parts they need from scratch without adding 100s of things u don't need.

afdbcreid• 2 months ago

One possible reason: because Rust can interface with Python much more easily.

Introduction – Rust for Python Programmers

Related Stories