Improve interior mutability slide

[djmitche]

Mutex is probably more broadly understood by people coming from other languages.

Fixes #1512.

[mgeisler]

Like @hurryabit said in the issue, I also feel it's weird to use a Mutex here: other languages typically don't let the mutex wrap the data like you do in Rust. So being familiar with std::mutex or a similar type might actually confuse people.

The mutex.lock().unwrap() pattern is strange as well and it's certainly something people ask about in the Concurrency class.

Talking about Cell feel equally clear to me: it's a storage location which you can get and set on. It is a very clear concept and it avoids the baggage of thinking about threads.

[djmitche]

I went back to using just RefCell, with the same example, but improved the text and notes a little bit. WDYT?

[mgeisler]

Thanks, I think should be good.

[mgeisler]

I went back to using just RefCell, with the same example, but improved the text and notes a little bit. WDYT?

I think it looks pretty good. I really like your concise part about "if there are no references, borrowing rules cannot be broken". I had no thought of it like this myself.

Have you considered focusing on Cell only? As a way to introduce the main concept of a type T where you can call set on a &T value? To me, that is the essence of interior mutability: the typesystem says you have an shared read-only reference, and yet you can modify something. The precise default of Cell vs RefCell is less important in my opinion.

[djmitche]

I did consider using Cell instead, but I think it's such a "thin" layer that it actually obscures the meaning -- it would be easy to take away an understanding that "interior mutability is this special Cell::set that takes &self but still modifies itself".

I think RefCell hits the sweet spot of being complex enough to illustrate the point, without being too complex.

[mgeisler]

I did consider using Cell instead, but I think it's such a "thin" layer that it actually obscures the meaning -- it would be easy to take away an understanding that "interior mutability is this special Cell::set that takes &self but still modifies itself".

I think RefCell hits the sweet spot of being complex enough to illustrate the point, without being too complex.

Alright, I'm happy with that. Let's get it in! 😄