|
| 1 | +using System.Text; |
| 2 | +using DotPython.Language.Text; |
| 3 | + |
| 4 | +namespace DotPython.Runtime.Managed.Execution; |
| 5 | + |
| 6 | +/// <summary> |
| 7 | +/// Decodes the runtime's qualified byte codecs without CLR fallback grouping. |
| 8 | +/// Error boundaries and BOM behavior follow CPython 3.14.7 unicodeobject.c. |
| 9 | +/// </summary> |
| 10 | +internal static class PythonBytesDecoding |
| 11 | +{ |
| 12 | + private const int MaximumTextLength = 10_000_000; |
| 13 | + |
| 14 | + internal static string Decode( |
| 15 | + byte[] bytes, |
| 16 | + int codePage, |
| 17 | + bool detectByteOrderMark, |
| 18 | + string errors, |
| 19 | + TextSpan span |
| 20 | + ) |
| 21 | + { |
| 22 | + UserObjectProtocols.Dispatcher?.CheckIterationWork(span); |
| 23 | + return new Decoder(bytes, codePage, detectByteOrderMark, errors, span).Decode(); |
| 24 | + } |
| 25 | + |
| 26 | + private sealed class Decoder( |
| 27 | + byte[] bytes, |
| 28 | + int codePage, |
| 29 | + bool detectByteOrderMark, |
| 30 | + string errors, |
| 31 | + TextSpan span |
| 32 | + ) |
| 33 | + { |
| 34 | + private readonly StringBuilder _text = new(Math.Min(bytes.Length, MaximumTextLength)); |
| 35 | + private int _position; |
| 36 | + private int _nextWork; |
| 37 | + private string _encoding = string.Empty; |
| 38 | + private bool _bigEndian; |
| 39 | + |
| 40 | + internal string Decode() |
| 41 | + { |
| 42 | + _encoding = codePage switch |
| 43 | + { |
| 44 | + 65001 => "utf-8", |
| 45 | + 20127 => "ascii", |
| 46 | + 28591 => "iso8859-1", |
| 47 | + 1200 => detectByteOrderMark && !BitConverter.IsLittleEndian |
| 48 | + ? "utf-16-be" |
| 49 | + : "utf-16-le", |
| 50 | + 1201 => "utf-16-be", |
| 51 | + _ => throw new ArgumentOutOfRangeException(nameof(codePage)), |
| 52 | + }; |
| 53 | + _bigEndian = |
| 54 | + codePage == 1201 |
| 55 | + || codePage == 1200 && detectByteOrderMark && !BitConverter.IsLittleEndian; |
| 56 | + if (codePage == 1200 && detectByteOrderMark && bytes.Length >= 2) |
| 57 | + { |
| 58 | + if (bytes[0] == 0xff && bytes[1] == 0xfe) |
| 59 | + { |
| 60 | + _bigEndian = false; |
| 61 | + _position = 2; |
| 62 | + } |
| 63 | + else if (bytes[0] == 0xfe && bytes[1] == 0xff) |
| 64 | + { |
| 65 | + _bigEndian = true; |
| 66 | + _position = 2; |
| 67 | + } |
| 68 | + _encoding = _bigEndian ? "utf-16-be" : "utf-16-le"; |
| 69 | + } |
| 70 | + |
| 71 | + while (_position < bytes.Length) |
| 72 | + { |
| 73 | + CheckProgress(); |
| 74 | + switch (codePage) |
| 75 | + { |
| 76 | + case 65001: |
| 77 | + DecodeUtf8(); |
| 78 | + break; |
| 79 | + case 1200: |
| 80 | + case 1201: |
| 81 | + DecodeUtf16(); |
| 82 | + break; |
| 83 | + default: |
| 84 | + if (codePage == 20127 && bytes[_position] >= 128) |
| 85 | + HandleError(_position + 1, "ordinal not in range(128)"); |
| 86 | + else |
| 87 | + _text.Append((char)bytes[_position++]); |
| 88 | + break; |
| 89 | + } |
| 90 | + CheckLength(); |
| 91 | + } |
| 92 | + UserObjectProtocols.Dispatcher?.CheckIterationWork(span); |
| 93 | + return _text.ToString(); |
| 94 | + } |
| 95 | + |
| 96 | + private void CheckLength() |
| 97 | + { |
| 98 | + if (_text.Length > MaximumTextLength) |
| 99 | + throw ManagedObjectProtocols.Fault( |
| 100 | + "DPY4003", |
| 101 | + "The decoded text exceeds the supported size.", |
| 102 | + span, |
| 103 | + "OverflowError" |
| 104 | + ); |
| 105 | + } |
| 106 | + |
| 107 | + private void CheckProgress() |
| 108 | + { |
| 109 | + if (_position < _nextWork) |
| 110 | + return; |
| 111 | + UserObjectProtocols.Dispatcher?.CheckIterationWork(span); |
| 112 | + // A scalar or malformed prefix consumes at most four bytes. Leave |
| 113 | + // room for it so checkpoints are never more than 256 bytes apart. |
| 114 | + _nextWork = _position > int.MaxValue - 252 ? int.MaxValue : _position + 252; |
| 115 | + } |
| 116 | + |
| 117 | + private void DecodeUtf8() |
| 118 | + { |
| 119 | + var first = bytes[_position]; |
| 120 | + if (first < 128) |
| 121 | + { |
| 122 | + _text.Append((char)first); |
| 123 | + ++_position; |
| 124 | + return; |
| 125 | + } |
| 126 | + var length = first switch |
| 127 | + { |
| 128 | + >= 0xc2 and <= 0xdf => 2, |
| 129 | + >= 0xe0 and <= 0xef => 3, |
| 130 | + >= 0xf0 and <= 0xf4 => 4, |
| 131 | + _ => 0, |
| 132 | + }; |
| 133 | + if (length == 0) |
| 134 | + { |
| 135 | + HandleError(_position + 1, "invalid start byte"); |
| 136 | + return; |
| 137 | + } |
| 138 | + |
| 139 | + var scalar = first & (0x7f >> length); |
| 140 | + for (var offset = 1; offset < length; ++offset) |
| 141 | + { |
| 142 | + if (offset >= bytes.Length - _position) |
| 143 | + { |
| 144 | + HandleError(bytes.Length, "unexpected end of data"); |
| 145 | + return; |
| 146 | + } |
| 147 | + var next = bytes[_position + offset]; |
| 148 | + if ( |
| 149 | + next is < 0x80 or > 0xbf |
| 150 | + || offset == 1 |
| 151 | + && ( |
| 152 | + first == 0xe0 && next < 0xa0 |
| 153 | + || first == 0xed && next >= 0xa0 |
| 154 | + || first == 0xf0 && next < 0x90 |
| 155 | + || first == 0xf4 && next >= 0x90 |
| 156 | + ) |
| 157 | + ) |
| 158 | + { |
| 159 | + HandleError(_position + offset, "invalid continuation byte"); |
| 160 | + return; |
| 161 | + } |
| 162 | + scalar = (scalar << 6) | (next & 0x3f); |
| 163 | + } |
| 164 | + _text.Append(char.ConvertFromUtf32(scalar)); |
| 165 | + _position += length; |
| 166 | + } |
| 167 | + |
| 168 | + private void DecodeUtf16() |
| 169 | + { |
| 170 | + if (bytes.Length - _position < 2) |
| 171 | + { |
| 172 | + HandleError(bytes.Length, "truncated data"); |
| 173 | + return; |
| 174 | + } |
| 175 | + var first = ReadUtf16(_position); |
| 176 | + if (first is < 0xd800 or > 0xdfff) |
| 177 | + { |
| 178 | + _text.Append((char)first); |
| 179 | + _position += 2; |
| 180 | + return; |
| 181 | + } |
| 182 | + if (first >= 0xdc00) |
| 183 | + { |
| 184 | + HandleError(_position + 2, "illegal encoding"); |
| 185 | + return; |
| 186 | + } |
| 187 | + if (bytes.Length - _position < 4) |
| 188 | + { |
| 189 | + HandleError(bytes.Length, "unexpected end of data"); |
| 190 | + return; |
| 191 | + } |
| 192 | + var second = ReadUtf16(_position + 2); |
| 193 | + if (second is < 0xdc00 or > 0xdfff) |
| 194 | + { |
| 195 | + HandleError(_position + 2, "illegal UTF-16 surrogate"); |
| 196 | + return; |
| 197 | + } |
| 198 | + _text.Append((char)first); |
| 199 | + _text.Append((char)second); |
| 200 | + _position += 4; |
| 201 | + } |
| 202 | + |
| 203 | + private int ReadUtf16(int index) => |
| 204 | + _bigEndian |
| 205 | + ? (bytes[index] << 8) | bytes[index + 1] |
| 206 | + : bytes[index] | (bytes[index + 1] << 8); |
| 207 | + |
| 208 | + private void HandleError(int end, string reason) |
| 209 | + { |
| 210 | + switch (errors) |
| 211 | + { |
| 212 | + case "ignore": |
| 213 | + break; |
| 214 | + case "replace": |
| 215 | + _text.Append('\ufffd'); |
| 216 | + break; |
| 217 | + case "backslashreplace": |
| 218 | + for (var index = _position; index < end; ++index) |
| 219 | + { |
| 220 | + _text.Append("\\x"); |
| 221 | + _text.Append( |
| 222 | + bytes[index] |
| 223 | + .ToString("x2", System.Globalization.CultureInfo.InvariantCulture) |
| 224 | + ); |
| 225 | + CheckLength(); |
| 226 | + } |
| 227 | + break; |
| 228 | + case "strict": |
| 229 | + var location = |
| 230 | + end == _position + 1 |
| 231 | + ? $"byte 0x{bytes[_position]:x2} in position {_position}" |
| 232 | + : $"bytes in position {_position}-{end - 1}"; |
| 233 | + throw ManagedObjectProtocols.Fault( |
| 234 | + "DPY4003", |
| 235 | + $"'{_encoding}' codec can't decode {location}: {reason}", |
| 236 | + span, |
| 237 | + "UnicodeDecodeError" |
| 238 | + ); |
| 239 | + case "xmlcharrefreplace": |
| 240 | + case "namereplace": |
| 241 | + throw ManagedObjectProtocols.Fault( |
| 242 | + "DPY4003", |
| 243 | + "don't know how to handle UnicodeDecodeError in error callback", |
| 244 | + span, |
| 245 | + "TypeError" |
| 246 | + ); |
| 247 | + default: |
| 248 | + // CPython looks handlers up only when malformed input actually |
| 249 | + // needs one. Valid data accepts even an unregistered name. |
| 250 | + throw ManagedObjectProtocols.Fault( |
| 251 | + "DPY4003", |
| 252 | + $"unknown error handler name '{errors}'", |
| 253 | + span, |
| 254 | + "LookupError" |
| 255 | + ); |
| 256 | + } |
| 257 | + _position = end; |
| 258 | + } |
| 259 | + } |
| 260 | +} |
0 commit comments