Swift 里字符串（三）small String



small string, 只有两个 UInt64 的字，这里面存储了所有的信息。
内存布局如下：


第二个 UInt64

存储了标记位和长度信息，以及部分字符串的值

  // Get an integer equivalent to the _StringObject.discriminatedObjectRawBits
  // computed property.
  @inlinable @inline(__always)
  internal var rawDiscriminatedObject: UInt64 {
    // Reverse the bytes on big-endian systems.
    return _storage.1.littleEndian
  }

rawbit.1	值	含义
b63	1	是不可变的
b62	0/1	是否是ASCII
b61	1	是 small string
b60	0	可以获取联系utf8 code point
b59-b56	0000~1111	已经使用的长度
b55~b0		存储utf8 code point

第一个UInt64

存储的都是字符串的值

初始化 small string

最基本的初始化

  @inlinable @inline(__always)
  internal init(leading: UInt64, trailing: UInt64, count: Int) {
    _internalInvariant(count <= _SmallString.capacity)

    let isASCII = (leading | trailing) & 0x8080_8080_8080_8080 == 0
    let discriminator = _StringObject.Nibbles
      .small(withCount: count, isASCII: isASCII)
      .littleEndian // reversed byte order on big-endian platforms
    _internalInvariant(trailing & discriminator == 0)

    self.init(raw: (leading, trailing | discriminator))
    _internalInvariant(self.count == count)
  }

这是最基本的初始化方法。先根据是否是ASCII和长度值，生成一个discriminator。然后把discriminator和trailing结合，作为第二个UInt64。

根据缓存区初始化

  // Direct from UTF-8
  @inlinable @inline(__always)
  internal init?(_ input: UnsafeBufferPointer<UInt8>) {
    if input.isEmpty {
      self.init()
      return
    }

    let count = input.count
    guard count <= _SmallString.capacity else { return nil }

    // TODO(SIMD): The below can be replaced with just be a masked unaligned
    // vector load
    let ptr = input.baseAddress._unsafelyUnwrappedUnchecked
    let leading = _bytesToUInt64(ptr, Swift.min(input.count, 8))
    let trailing = count > 8 ? _bytesToUInt64(ptr + 8, count &- 8) : 0

    self.init(leading: leading, trailing: trailing, count: count)
  }

即先判断长度是否超过上限，如果超过，返回nil。
如果没有超过上限，再调用init(leading: UInt64, trailing: UInt64, count: Int) 方法。

small string 创建过程