> operation will just panic at runtime if the array is too short
Exactly, rust doesn't provide a good solution to this problem at all. Panics in rust are an escape hatch used to ensure the language stays "safe" in situations where the compiler can not prove a given behaviour at compile time, but where it would have made the language too ugly if you had to wire through Result types for all the trivial operations like adding two numbers.
In my experience, panics in rust have been a major source of pain. In contrast to an exception, which you can catch, a panic behaves more like an abort. At least it has been that way in the past. Now, with a lot of libraries using panics to signalize runtime errors, coding in rust has at some times felt like I was using a bunch of badly written C libraries that internally call "abort()" and kill the process when something goes wrong that would have been totally handle-able without killing the whole process. That's the benefit of using a safe language, right?
I think lately the rust "community" has become aware of this issue and IMO the way things are going is that that panics, as they are designed, should basically not be used. But, without proper exceptions, that brings you back to the situation where an operation as trivial as adding two numbers either produces a return code that must be explicitly checked or may silently fail and produce an "undefined" result in some cases.
A panic in Rust nicely captures the category of errors which the programmer asserts should be impossible, which generally cannot be recovered well from. A panic is kind of catchable, but it only works in the sense of killing only part of the process rather than the entire process (like abort does).
If errors are expected, then you should Result instead of a panic. If people are using panics instead of Result for these kinds of errors, then the library is wrong. I'm curious what examples you have where this is happening.
Exactly, rust doesn't provide a good solution to this problem at all. Panics in rust are an escape hatch used to ensure the language stays "safe" in situations where the compiler can not prove a given behaviour at compile time, but where it would have made the language too ugly if you had to wire through Result types for all the trivial operations like adding two numbers.
In my experience, panics in rust have been a major source of pain. In contrast to an exception, which you can catch, a panic behaves more like an abort. At least it has been that way in the past. Now, with a lot of libraries using panics to signalize runtime errors, coding in rust has at some times felt like I was using a bunch of badly written C libraries that internally call "abort()" and kill the process when something goes wrong that would have been totally handle-able without killing the whole process. That's the benefit of using a safe language, right?
I think lately the rust "community" has become aware of this issue and IMO the way things are going is that that panics, as they are designed, should basically not be used. But, without proper exceptions, that brings you back to the situation where an operation as trivial as adding two numbers either produces a return code that must be explicitly checked or may silently fail and produce an "undefined" result in some cases.