swift-mirror/stdlib/public/core/Character.swift

//===----------------------------------------------------------------------===//
//
// This source file is part of the Swift.org open source project
//
// Copyright (c) 2014 - 2016 Apple Inc. and the Swift project authors
// Licensed under Apache License v2.0 with Runtime Library Exception
//
// See http://swift.org/LICENSE.txt for license information
// See http://swift.org/CONTRIBUTORS.txt for the list of Swift project authors
//
//===----------------------------------------------------------------------===//

/// `Character` represents some Unicode grapheme cluster as
/// defined by a canonical, localized, or otherwise tailored
/// segmentation algorithm.
public struct Character :
  _BuiltinExtendedGraphemeClusterLiteralConvertible,
  ExtendedGraphemeClusterLiteralConvertible, Equatable, Hashable, Comparable {

  // Fundamentally, it is just a String, but it is optimized for the
  // common case where the UTF-8 representation fits in 63 bits.  The
  // remaining bit is used to discriminate between small and large
  // representations.  In the small representation, the unused bytes
  // are filled with 0xFF.
  //
  // If the grapheme cluster can be represented as `.small`, it
  // should be represented as such.
  @_versioned
  internal enum Representation {
    // A _StringBuffer whose first grapheme cluster is self.
    // NOTE: may be more than 1 Character long.
    case large(_StringBuffer._Storage)
    case small(Builtin.Int63)
  }

  /// Construct a `Character` containing just the given `scalar`.
  public init(_ scalar: UnicodeScalar) {
    var asInt: UInt64 = 0
    var shift: UInt64 = 0

    let output: (UTF8.CodeUnit) -> Void = {
      asInt |= UInt64($0) << shift
      shift += 8
    }

    UTF8.encode(scalar, sendingOutputTo: output)
    asInt |= (~0) << shift
    _representation = .small(Builtin.trunc_Int64_Int63(asInt._value))
  }

  @effects(readonly)
  public init(_builtinUnicodeScalarLiteral value: Builtin.Int32) {
    self = Character(
      String._fromWellFormedCodeUnitSequence(
        UTF32.self, input: CollectionOfOne(UInt32(value))))
  }

  /// Create an instance initialized to `value`.
  public init(unicodeScalarLiteral value: Character) {
    self = value
  }

  @effects(readonly)
  public init(
    _builtinExtendedGraphemeClusterLiteral start: Builtin.RawPointer,
    utf8CodeUnitCount: Builtin.Word,
    isASCII: Builtin.Int1
  ) {
    self = Character(
      String(
        _builtinExtendedGraphemeClusterLiteral: start,
        utf8CodeUnitCount: utf8CodeUnitCount,
        isASCII: isASCII))
  }

  /// Create an instance initialized to `value`.
  public init(extendedGraphemeClusterLiteral value: Character) {
    self = value
  }

  /// Create an instance from a single-character `String`.
  ///
  /// - Precondition: `s` contains exactly one extended grapheme cluster.
  public init(_ s: String) {
    // The small representation can accept up to 8 code units as long
    // as the last one is a continuation.  Since the high bit of the
    // last byte is used for the enum's discriminator, we have to
    // reconstruct it.  As a result, we can't store 0x7f in the final
    // byte, because we wouldn't be able to distinguish it from an
    // unused 0xFF byte.  Rather than trying to squeeze in other
    // one-byte code points there, we simplify decoding by banning
    // starting a code point in the last byte, and assuming that its
    // high bit is 1.
    _precondition(
      s._core.count != 0, "Can't form a Character from an empty String")
    _precondition(
      s.index(after: s.startIndex) == s.endIndex,
      "Can't form a Character from a String containing more than one extended grapheme cluster")

    let (count, initialUTF8) = s._core._encodeSomeUTF8(from: 0)
    // Notice that the result of sizeof() is a small non-zero number and can't
    // overflow when multiplied by 8.
    let bits = sizeofValue(initialUTF8) &* 8 &- 1
    if _fastPath(
      count == s._core.count && (initialUTF8 & (1 << numericCast(bits))) != 0) {
      _representation = .small(Builtin.trunc_Int64_Int63(initialUTF8._value))
    }
    else {
      if let native = s._core.nativeBuffer
              where native.start == UnsafeMutablePointer(s._core._baseAddress!){
        _representation = .large(native._storage)
        return
      }
      var nativeString = ""
      nativeString.append(s)
      _representation = .large(nativeString._core.nativeBuffer!._storage)
    }
  }

  /// Returns the index of the lowest byte that is 0xFF, or 8 if
  /// there is none.
  @warn_unused_result
  static func _smallSize(_ value: UInt64) -> Int {
    var mask: UInt64 = 0xFF
    for i in 0..<8 {
      if (value & mask) == mask {
        return i
      }
      mask <<= 8
    }
    return 8
  }

  @warn_unused_result
  static func _smallValue(_ value: Builtin.Int63) -> UInt64 {
    return UInt64(Builtin.zext_Int63_Int64(value)) | (1<<63)
  }

  internal struct _SmallUTF8 : RandomAccessCollection {
    typealias Indices = CountableRange<Int>
    
    var indices: CountableRange<Int> {
      return startIndex..<endIndex
    }

    init(_ u8: UInt64) {
      let utf8Count = Character._smallSize(u8)
      _sanityCheck(utf8Count <= 8, "Character with more than 8 UTF-8 code units")
      self.count = UInt16(utf8Count)
      self.data = u8
    }

    /// The position of the first element in a non-empty collection.
    ///
    /// In an empty collection, `startIndex == endIndex`.
    var startIndex: Int {
      return 0
    }

    /// The collection's "past the end" position.
    ///
    /// `endIndex` is not a valid argument to `subscript`, and is always
    /// reachable from `startIndex` by zero or more applications of
    /// `index(after:)`.
    var endIndex: Int {
      return Int(count)
    }

    /// Access the code unit at `position`.
    ///
    /// - Precondition: `position` is a valid position in `self` and
    ///   `position != endIndex`.
    subscript(position: Int) -> UTF8.CodeUnit {
      _sanityCheck(position >= 0)
      _sanityCheck(position < Int(count))
      // Note: using unchecked arithmetic because overflow cannot happen if the
      // above sanity checks hold.
      return UTF8.CodeUnit(
        truncatingBitPattern: data >> (UInt64(position) &* 8))
    }

    internal struct Iterator : IteratorProtocol {
      init(_ data: UInt64) {
        self._data = data
      }

      internal mutating func next() -> UInt8? {
        let result = UInt8(truncatingBitPattern: _data)
        if result == 0xFF {
          return nil
        }
        _data = (_data >> 8) | 0xFF00_0000_0000_0000
        return result
      }

      internal var _data: UInt64
    }

    internal func makeIterator() -> Iterator {
      return Iterator(data)
    }

    var count: UInt16
    var data: UInt64
  }

  struct _SmallUTF16 : RandomAccessCollection {
    typealias Indices = CountableRange<Int>
    
    init(_ u8: UInt64) {
      let count = UTF16.transcodedLength(
        of: _SmallUTF8(u8).makeIterator(),
        decodedAs: UTF8.self,
        repairingIllFormedSequences: true)!.0
      _sanityCheck(count <= 4, "Character with more than 4 UTF-16 code units")
      self.count = UInt16(count)
      var u16: UInt64 = 0
      let output: (UTF16.CodeUnit) -> Void = {
        u16 = u16 << 16
        u16 = u16 | UInt64($0)
      }
      transcode(
        _SmallUTF8(u8).makeIterator(),
        from: UTF8.self, to: UTF16.self,
        stoppingOnError: false,
        sendingOutputTo: output)
      self.data = u16
    }

    /// The position of the first element in a non-empty collection.
    ///
    /// In an empty collection, `startIndex == endIndex`.
    var startIndex : Int {
      return 0
    }

    /// The collection's "past the end" position.
    ///
    /// `endIndex` is not a valid argument to `subscript`, and is always
    /// reachable from `startIndex` by zero or more applications of
    /// `successor()`.
    var endIndex : Int {
      return Int(count)
    }

    /// Access the code unit at `position`.
    ///
    /// - Precondition: `position` is a valid position in `self` and
    ///   `position != endIndex`.
    subscript(position: Int) -> UTF16.CodeUnit {
      _sanityCheck(position >= 0)
      _sanityCheck(position < Int(count))
      // Note: using unchecked arithmetic because overflow cannot happen if the
      // above sanity checks hold.
      return UTF16.CodeUnit(truncatingBitPattern:
        data >> ((UInt64(count) &- UInt64(position) &- 1) &* 16))
    }

    var count: UInt16
    var data: UInt64
  }

  /// The hash value.
  ///
  /// **Axiom:** `x == y` implies `x.hashValue == y.hashValue`.
  ///
  /// - Note: The hash value is not guaranteed to be stable across
  ///   different invocations of the same program.  Do not persist the
  ///   hash value across program runs.
  public var hashValue: Int {
    // FIXME(performance): constructing a temporary string is extremely
    // wasteful and inefficient.
    return String(self).hashValue
  }

  typealias UTF16View = String.UTF16View
  var utf16: UTF16View {
    return String(self).utf16
  }

  @_versioned
  internal var _representation: Representation
}

extension Character : CustomDebugStringConvertible {
  /// A textual representation of `self`, suitable for debugging.
  public var debugDescription: String {
    return String(self).debugDescription
  }
}

extension String {
  /// Construct an instance containing just the given `Character`.
  public init(_ c: Character) {
    switch c._representation {
    case let .small(_63bits):
      let value = Character._smallValue(_63bits)
      let smallUTF8 = Character._SmallUTF8(value)
      self = String._fromWellFormedCodeUnitSequence(
        UTF8.self, input: smallUTF8)
    case let .large(value):
      let buf = String(_StringCore(_StringBuffer(value)))
      self = buf[buf.startIndex..<buf.index(after: buf.startIndex)]
    }
  }
}

/// `.small` characters are stored in an Int63 with their UTF-8 representation,
/// with any unused bytes set to 0xFF. ASCII characters will have all bytes set
/// to 0xFF except for the lowest byte, which will store the ASCII value. Since
/// 0x7FFFFFFFFFFFFF80 or greater is an invalid UTF-8 sequence, we know if a
/// value is ASCII by checking if it is greater than or equal to
/// 0x7FFFFFFFFFFFFF00.
internal var _minASCIICharReprBuiltin: Builtin.Int63 {
  @inline(__always) get {
    let x: Int64 = 0x7FFFFFFFFFFFFF00
    return Builtin.truncOrBitCast_Int64_Int63(x._value)
  }
}

@warn_unused_result
public func ==(lhs: Character, rhs: Character) -> Bool {
  switch (lhs._representation, rhs._representation) {
  case let (.small(lbits), .small(rbits)) where
    Bool(Builtin.cmp_uge_Int63(lbits, _minASCIICharReprBuiltin))
    && Bool(Builtin.cmp_uge_Int63(rbits, _minASCIICharReprBuiltin)):
    return Bool(Builtin.cmp_eq_Int63(lbits, rbits))
  default:
    // FIXME(performance): constructing two temporary strings is extremely
    // wasteful and inefficient.
    return String(lhs) == String(rhs)
  }
}

@warn_unused_result
public func <(lhs: Character, rhs: Character) -> Bool {
  switch (lhs._representation, rhs._representation) {
  case let (.small(lbits), .small(rbits)) where
    // Note: This is consistent with Foundation but unicode incorrect.
    // See String._compareASCII.
    Bool(Builtin.cmp_uge_Int63(lbits, _minASCIICharReprBuiltin))
    && Bool(Builtin.cmp_uge_Int63(rbits, _minASCIICharReprBuiltin)):
    return Bool(Builtin.cmp_ult_Int63(lbits, rbits))
  default:
    // FIXME(performance): constructing two temporary strings is extremely
    // wasteful and inefficient.
    return String(lhs) < String(rhs)
  }
}