"Smart assert" that prints the values involved in simple cases

[eira-fransham]

It's pretty annoying for a new Rust programmer to have a pretty unhelpful error message when using the default assert! macro in tests or suchlike. They have to know that the "correct" way to do it is to use assert_eq! or assert_ne!, which do print the value of their arguments. I suggest that we extend the default assert! macro to print operands in simple cases, and then fall back to the current behaviour if an explicit message is provided or if the expression is complex.

I actually thought that Python did this already and I was going to use it as precedent, but its default assertion message is even worse than Rust's.

I have a simple implementation of this here, with the message modelled on assert_eq/assert_ne's message.

[oli-obk]

Two problems I see with this:

1. it's a breaking change for types that do not implement Debug (https://play.rust-lang.org/?gist=028f79c439b15094d9495543e80cd55e&version=stable)
2. it falls back to the old assehttps://play.rust-lang.org/?gist=028f79c439b15094d9495543e80cd55e&version=stablert if the lhs or rhs expression is not a simple tt, e.g. Foo(5) == x

The first one can probably be addressed via specialization (inside the stdlib that is fine)

The second one must be addressed by writing a proc macro, macro_rules simply isn't powerful enough.

[eira-fransham]

Yeah, 2 is unfortunate. You can see in my example that I have A + B + C, that doesn't print the nicer message. The way I see it this is an incremental improvement - better than the existing situation even if it's not perfect. A user can always fall back to assert_eq!.

[Techcable]

Groovy calls these 'power asserts. They're very handy, but I worry they'd signifigantly bloat the size of the generated code. Maybe they should only be enabled in debug mode?

[pnkfelix]

This other RFC (which recently went into final comment period) is probably relevant: #2011

(Also, I want to mention that its pretty easy to override the assert message, e.g. assert!(x > y, "How can x: {} not be greater than y: {}", x, y);; I infer from the issue description that the filer is already aware of the "explicit message" option, but I suspect not everyone is...)