| 1 | # SPDX-License-Identifier: MIT |
| 2 | # This module contains abstractions for the input stream. You don't have to |
| 3 | # looks further, there are no pretty code. |
| 4 | # |
| 5 | # We define two classes here. |
| 6 | # |
| 7 | # Mark(source, line, column) |
| 8 | # It's just a record and its only use is producing nice error messages. |
| 9 | # Parser does not use it for any other purposes. |
| 10 | # |
| 11 | # Reader(source, data) |
| 12 | # Reader determines the encoding of `data` and converts it to unicode. |
| 13 | # Reader provides the following methods and attributes: |
| 14 | # reader.peek(length=1) - return the next `length` characters |
| 15 | # reader.forward(length=1) - move the current position to `length` characters. |
| 16 | # reader.index - the number of the current character. |
| 17 | # reader.line, stream.column - the line and the column of the current character. |
| 18 | |
| 19 | __all__ = ['Reader', 'ReaderError'] |
| 20 | |
| 21 | from .error import YAMLError, Mark |
| 22 | |
| 23 | import codecs, re |
| 24 | |
| 25 | class ReaderError(YAMLError): |
| 26 | |
| 27 | def __init__(self, name, position, character, encoding, reason): |
| 28 | self.name = name |
| 29 | self.character = character |
| 30 | self.position = position |
| 31 | self.encoding = encoding |
| 32 | self.reason = reason |
| 33 | |
| 34 | def __str__(self): |
| 35 | if isinstance(self.character, bytes): |
| 36 | return "'%s' codec can't decode byte #x%02x: %s\n" \ |
| 37 | " in \"%s\", position %d" \ |
| 38 | % (self.encoding, ord(self.character), self.reason, |
| 39 | self.name, self.position) |
| 40 | else: |
| 41 | return "unacceptable character #x%04x: %s\n" \ |
| 42 | " in \"%s\", position %d" \ |
| 43 | % (self.character, self.reason, |
| 44 | self.name, self.position) |
| 45 | |
| 46 | class Reader(object): |
| 47 | # Reader: |
| 48 | # - determines the data encoding and converts it to a unicode string, |
| 49 | # - checks if characters are in allowed range, |
| 50 | # - adds '\0' to the end. |
| 51 | |
| 52 | # Reader accepts |
| 53 | # - a `bytes` object, |
| 54 | # - a `str` object, |
| 55 | # - a file-like object with its `read` method returning `str`, |
| 56 | # - a file-like object with its `read` method returning `unicode`. |
| 57 | |
| 58 | # Yeah, it's ugly and slow. |
| 59 | |
| 60 | def __init__(self, stream): |
| 61 | self.name = None |
| 62 | self.stream = None |
| 63 | self.stream_pointer = 0 |
| 64 | self.eof = True |
| 65 | self.buffer = '' |
| 66 | self.pointer = 0 |
| 67 | self.raw_buffer = None |
| 68 | self.raw_decode = None |
| 69 | self.encoding = None |
| 70 | self.index = 0 |
| 71 | self.line = 0 |
| 72 | self.column = 0 |
| 73 | if isinstance(stream, str): |
| 74 | self.name = "<unicode string>" |
| 75 | self.check_printable(stream) |
| 76 | self.buffer = stream+'\0' |
| 77 | elif isinstance(stream, bytes): |
| 78 | self.name = "<byte string>" |
| 79 | self.raw_buffer = stream |
| 80 | self.determine_encoding() |
| 81 | else: |
| 82 | self.stream = stream |
| 83 | self.name = getattr(stream, 'name', "<file>") |
| 84 | self.eof = False |
| 85 | self.raw_buffer = None |
| 86 | self.determine_encoding() |
| 87 | |
| 88 | def peek(self, index=0): |
| 89 | try: |
| 90 | return self.buffer[self.pointer+index] |
| 91 | except IndexError: |
| 92 | self.update(index+1) |
| 93 | return self.buffer[self.pointer+index] |
| 94 | |
| 95 | def prefix(self, length=1): |
| 96 | if self.pointer+length >= len(self.buffer): |
| 97 | self.update(length) |
| 98 | return self.buffer[self.pointer:self.pointer+length] |
| 99 | |
| 100 | def forward(self, length=1): |
| 101 | if self.pointer+length+1 >= len(self.buffer): |
| 102 | self.update(length+1) |
| 103 | while length: |
| 104 | ch = self.buffer[self.pointer] |
| 105 | self.pointer += 1 |
| 106 | self.index += 1 |
| 107 | if ch in '\n\x85\u2028\u2029' \ |
| 108 | or (ch == '\r' and self.buffer[self.pointer] != '\n'): |
| 109 | self.line += 1 |
| 110 | self.column = 0 |
| 111 | elif ch != '\uFEFF': |
| 112 | self.column += 1 |
| 113 | length -= 1 |
| 114 | |
| 115 | def get_mark(self): |
| 116 | if self.stream is None: |
| 117 | return Mark(self.name, self.index, self.line, self.column, |
| 118 | self.buffer, self.pointer) |
| 119 | else: |
| 120 | return Mark(self.name, self.index, self.line, self.column, |
| 121 | None, None) |
| 122 | |
| 123 | def determine_encoding(self): |
| 124 | while not self.eof and (self.raw_buffer is None or len(self.raw_buffer) < 2): |
| 125 | self.update_raw() |
| 126 | if isinstance(self.raw_buffer, bytes): |
| 127 | if self.raw_buffer.startswith(codecs.BOM_UTF16_LE): |
| 128 | self.raw_decode = codecs.utf_16_le_decode |
| 129 | self.encoding = 'utf-16-le' |
| 130 | elif self.raw_buffer.startswith(codecs.BOM_UTF16_BE): |
| 131 | self.raw_decode = codecs.utf_16_be_decode |
| 132 | self.encoding = 'utf-16-be' |
| 133 | else: |
| 134 | self.raw_decode = codecs.utf_8_decode |
| 135 | self.encoding = 'utf-8' |
| 136 | self.update(1) |
| 137 | |
| 138 | NON_PRINTABLE = re.compile('[^\x09\x0A\x0D\x20-\x7E\x85\xA0-\uD7FF\uE000-\uFFFD]') |
| 139 | def check_printable(self, data): |
| 140 | match = self.NON_PRINTABLE.search(data) |
| 141 | if match: |
| 142 | character = match.group() |
| 143 | position = self.index+(len(self.buffer)-self.pointer)+match.start() |
| 144 | raise ReaderError(self.name, position, ord(character), |
| 145 | 'unicode', "special characters are not allowed") |
| 146 | |
| 147 | def update(self, length): |
| 148 | if self.raw_buffer is None: |
| 149 | return |
| 150 | self.buffer = self.buffer[self.pointer:] |
| 151 | self.pointer = 0 |
| 152 | while len(self.buffer) < length: |
| 153 | if not self.eof: |
| 154 | self.update_raw() |
| 155 | if self.raw_decode is not None: |
| 156 | try: |
| 157 | data, converted = self.raw_decode(self.raw_buffer, |
| 158 | 'strict', self.eof) |
| 159 | except UnicodeDecodeError as exc: |
| 160 | character = self.raw_buffer[exc.start] |
| 161 | if self.stream is not None: |
| 162 | position = self.stream_pointer-len(self.raw_buffer)+exc.start |
| 163 | else: |
| 164 | position = exc.start |
| 165 | raise ReaderError(self.name, position, character, |
| 166 | exc.encoding, exc.reason) |
| 167 | else: |
| 168 | data = self.raw_buffer |
| 169 | converted = len(data) |
| 170 | self.check_printable(data) |
| 171 | self.buffer += data |
| 172 | self.raw_buffer = self.raw_buffer[converted:] |
| 173 | if self.eof: |
| 174 | self.buffer += '\0' |
| 175 | self.raw_buffer = None |
| 176 | break |
| 177 | |
| 178 | def update_raw(self, size=4096): |
| 179 | data = self.stream.read(size) |
| 180 | if self.raw_buffer is None: |
| 181 | self.raw_buffer = data |
| 182 | else: |
| 183 | self.raw_buffer += data |
| 184 | self.stream_pointer += len(data) |
| 185 | if not data: |
| 186 | self.eof = True |
| 187 | |
| 188 | #try: |
| 189 | # import psyco |
| 190 | # psyco.bind(Reader) |
| 191 | #except ImportError: |
| 192 | # pass |
| 193 |