String initialization notes

FromCString.swift

// The gold standard, at least at present
let utf8Buffer: UnsafeBufferPointer<CChar> = ...
var arrayVersion = Array(utf8Buffer)
arrayVersion.append(0)
let string = String.fromCString(arrayVersion)

NSString.swift

let utf8Buffer: UnsafeBufferPointer<UInt8> = ...

// Crossing the NSString bridge performs slower than you'd like
let nsString1 = NSString(bytes: utf8Buffer.baseAddress, length: utf8Buffer.count, encoding: NSUTF8StringEncoding)!
let string1 = string1 as String

// The bridged initializer does an Array() memoization, so it's even slower
let string2 = String(bytes: utf8Buffer, encoding: NSUTF8StringEncoding)!

UnicodeScalar.swift

let utf8Buffer: UnsafeBufferPointer<UInt8> = ...
let codePointCount = ... // assume this comes from UTF16.measure(_:input:repairIllFormedSequences:) and we're isAscii

// The slowest of all, unforunately, even if you make the same isAscii optimization as _StringBuffer.fromCodeUnits(_:input:repairIllFormedSequences:minimumCapacity:).
var string1 = ""
string1.reserveCapacity(codePointCount)
transcode(UTF8.self, UTF16.self, utf8Buffer.generate(), {
    string1.append($0)
}, stopOnError: true)

// alternatively, this is ridiculously faster (50% better than fromCString)
let string2 = unsafeBitCast(_StringCore(baseAddress: utf8Buffer.baseAddress, count: utf8Buffer.count, elementShift: 0, hasCocoaBuffer: false, owner: nil), String.self)