yeah, it seems like the message should really be something like "expectation failed: ..."

Why? I at least usually want the panic to propagate, it’s kind of the whole point, to have actors that propagate the failure out to “infected” actors. The key is recovering :)

Actually I kind of dislike “or panic with”— the convention I find most readable is when people say eg foo.lock().expect(“not poisoned”) which conveys to me WHY it is ok. I think “or panic with” would push you to the opposite framing and I find that harder to think about.

I agree “or panic” is better but my preference continues to be something like “assert_some”. I find that conveys the time when it is appropriate to use and connects it to the assert! macro (which can also panic…).

Sadly no, not really. I’d love to see what you’ve done though!

…hello prompt injection attacks :P

I used to think that but it doesn’t work with the closure treatment. The whole point in some sense is to not require distinguishing “closure creation” and “closure invocation”. So if you just say it once, we need to automatic remove the other call… and then we can do it elsewhere.

Follow-up:

smallcultfollowing.com/babysteps/bl...

I don't see it that way at all. I think Handle has value *particularly* when you make it explicit, because otherwise `x.clone()` is quite ambiguous in terms of what it does. You really have to know the type of `x`.

You would be calling `handle` directly though...? Or at least some folks would. I agree with Carl's critique but I couldn't up with a suggestion I liked *more*, and I felt there was definite value in introducing the term "handle".

People need to be able to review but the onboarding costs are much lower. I can understand French surprisingly well but speaking it is much harder.