init?(_:) vs init(_:)

[oscbyspro]

BinaryInteger initializers of the form T/init(_:) are abbreviations of T.exactly(_:).unwrap(). Alternatively, I could return Optional<T>. Optional<T> force-unwrapping is only one additional character ("!") and explicit error-handling is awesome. In this hypothetical rework, init?(_:) would an abbreviation of T.exactly(_:).optional().

[oscbyspro]

The more I think about it, the more I like this approach. It's easier to test since it doesn't trap. In the case of Guarantee, it would also result in fewer methods since error-handling is hoisted to Optional<T>. If I add Optional/prune(_:) then the number of Guarantee initializers would go from 5 to 2. Depending on how well LLVM optimizes unchecked(_:) it might even be possible to remove init(unsafe:).

init(_:)         // error: nil
init(unsafe:)    // error: %%%%%%

init(_:prune:)   // error: throws
init(exactly:)   // error: nil
init(_:)         // error: precondition
init(unchecked:) // error: assert
init(unsafe:)    // error: %%%%%%

[oscbyspro]

Also, if the Guarantee protocol becomes that simple, then I can probably drop it.