Exercise consistency: tests or main? and allow(unused)

[djmitche]

See #1565 for context.

Also, put all text above the code sample!

[mgeisler]

Good questions!

There are some practical reasons and constraints for some of this:

• Tests are not run in the embedded playgrounds — the playgrounds need a main function.
• On the other hand, we run tests via cargo test in CI!

So I like tests because they help keep the course code correct. I also think it's fine to show people simple unit tests from day 1: they're just small functions after all and don't look scary in my opnion.

We ask people to solve most or all exercises in https://play.rust-lang.org/ or locally using an editor with rust-analyzer. This means that they can execute tests just fine. However, having passing tests be the criteria for success takes away some of the free-form experimentation that you can do when you just have some code and a main function to call it from.

Related to this, it always bothered me how Rust doctests have to use assert_eq!(foo, 123) to express the expectations in your example. It feels dry and it means I cannot talk about things like println! or dbg! output in my example. I like the Python way much more where you show an interactive console session.

However, since we're not Python and don't have such interactive tools, I guess using small unit tests is okay. A bit boring, but probably also more professional 😄