PureXML/Sources/Parsing/StreamingDecoderCJK.swift

148 lines
6.0 KiB
Swift

extension PureXML.Parsing.StreamingDecoder {
/// Streams one character from a multi-byte CJK encoding.
mutating func nextCJK(_ encoding: PureXML.Parsing.InputEncoding) -> Character? {
switch encoding {
case .shiftJIS: nextShiftJIS()
case .eucJP: nextEUCJP()
case .eucKR: nextEUCKR()
case .gbk: nextGBK()
case .gb18030: nextGB18030()
case .big5: nextBig5()
case .iso2022jp: nextISO2022JP()
case .eucTW: nextEUCTW()
default: nextUTF8()
}
}
/// Streams one GBK character: ASCII, `0x80` (euro), or a lead+trail.
private mutating func nextGBK() -> Character? {
guard let lead = nextByte() else {
finished = true
return nil
}
if lead <= 0x7F {
return Character(Unicode.Scalar(lead))
}
if lead == 0x80 {
return "\u{20AC}"
}
guard (0x81 ... 0xFE).contains(lead) else { return Self.replacement }
guard let trail = nextByte() else { finished = true
return Self.replacement
}
return PureXML.Parsing.ByteDecoder.GBK.twoByteScalar(lead: lead, trail: trail).map(Character.init) ?? Self.replacement
}
/// Streams one GB18030 character: ASCII, `0x80` (euro), a four-byte
/// sequence when the lead is followed by a digit, or a two-byte GBK pair.
private mutating func nextGB18030() -> Character? {
guard let lead = nextByte() else {
finished = true
return nil
}
if lead <= 0x7F {
return Character(Unicode.Scalar(lead))
}
if lead == 0x80 {
return "\u{20AC}"
}
guard (0x81 ... 0xFE).contains(lead) else { return Self.replacement }
guard let second = nextByte() else { finished = true
return Self.replacement
}
if (0x30 ... 0x39).contains(second) {
guard let third = nextByte(), let fourth = nextByte() else { finished = true
return Self.replacement
}
guard (0x81 ... 0xFE).contains(third), (0x30 ... 0x39).contains(fourth) else { return Self.replacement }
let pointer = PureXML.Parsing.ByteDecoder.GB18030.fourBytePointer(lead, second, third, fourth)
return PureXML.Parsing.ByteDecoder.GB18030.rangeScalar(pointer).map(Character.init) ?? Self.replacement
}
return PureXML.Parsing.ByteDecoder.GBK.twoByteScalar(lead: lead, trail: second).map(Character.init) ?? Self.replacement
}
/// Streams one Big5 character: ASCII, or a lead+trail through the index
/// (a two-scalar combining sequence stays one grapheme, one `Character`).
private mutating func nextBig5() -> Character? {
guard let lead = nextByte() else {
finished = true
return nil
}
if lead <= 0x7F {
return Character(Unicode.Scalar(lead))
}
guard (0x81 ... 0xFE).contains(lead) else { return Self.replacement }
guard let trail = nextByte() else { finished = true
return Self.replacement
}
guard let sequence = PureXML.Parsing.ByteDecoder.Big5.mapping(lead: lead, trail: trail) else {
return Self.replacement
}
return Character(String(String.UnicodeScalarView(sequence)))
}
/// Streams one EUC-KR character: ASCII, or a lead+trail through CP949.
private mutating func nextEUCKR() -> Character? {
guard let lead = nextByte() else {
finished = true
return nil
}
if lead <= 0x7F {
return Character(Unicode.Scalar(lead))
}
guard (0x81 ... 0xFE).contains(lead) else { return Self.replacement }
guard let trail = nextByte() else { finished = true
return Self.replacement
}
return PureXML.Parsing.ByteDecoder.EUCKR.twoByteScalar(lead: lead, trail: trail).map(Character.init) ?? Self.replacement
}
/// Streams one EUC-JP character: ASCII, `0x8E`+byte (half-width katakana),
/// `0x8F`+two bytes (JIS X 0212), or a lead+byte through JIS X 0208.
private mutating func nextEUCJP() -> Character? {
guard let lead = nextByte() else {
finished = true
return nil
}
if lead <= 0x7F {
return Character(Unicode.Scalar(lead))
}
if lead == 0x8E {
guard let byte = nextByte() else { finished = true
return Self.replacement
}
guard (0xA1 ... 0xDF).contains(byte) else { return Self.replacement }
return Character(Unicode.Scalar(0xFF61 + UInt32(byte) - 0xA1) ?? "\u{FFFD}")
}
if lead == 0x8F {
guard let mid = nextByte(), let trail = nextByte() else { finished = true
return Self.replacement
}
return PureXML.Parsing.ByteDecoder.EUCJP.twoByteScalar(lead: mid, trail: trail, useJis0212: true).map(Character.init) ?? Self.replacement
}
guard let trail = nextByte() else { finished = true
return Self.replacement
}
return PureXML.Parsing.ByteDecoder.EUCJP.twoByteScalar(lead: lead, trail: trail, useJis0212: false).map(Character.init) ?? Self.replacement
}
/// Streams one Shift-JIS character: a single byte (ASCII or half-width
/// katakana) or a lead+trail two-byte sequence resolved through JIS X 0208.
private mutating func nextShiftJIS() -> Character? {
guard let lead = nextByte() else {
finished = true
return nil
}
if let single = PureXML.Parsing.ByteDecoder.ShiftJIS.singleByteScalar(lead) {
return Character(single)
}
guard PureXML.Parsing.ByteDecoder.ShiftJIS.isLead(lead) else { return Self.replacement }
guard let trail = nextByte() else {
finished = true
return Self.replacement
}
guard let scalar = PureXML.Parsing.ByteDecoder.ShiftJIS.twoByteScalar(lead: lead, trail: trail) else { return Self.replacement }
return Character(scalar)
}
}