Shortcuts on this page
r m x toggle line displays
j k next/prev highlighted chunk
0 (zero) top of page
1 (one) first highlighted chunk
1# Copyright (c) 2024, pypdf contributors
2# All rights reserved.
3#
4# Redistribution and use in source and binary forms, with or without
5# modification, are permitted provided that the following conditions are
6# met:
7#
8# * Redistributions of source code must retain the above copyright notice,
9# this list of conditions and the following disclaimer.
10# * Redistributions in binary form must reproduce the above copyright notice,
11# this list of conditions and the following disclaimer in the documentation
12# and/or other materials provided with the distribution.
13# * The name of the author may not be used to endorse or promote products
14# derived from this software without specific prior written permission.
15#
16# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
17# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19# ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
20# LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
21# CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
22# SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
23# INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
24# CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
25# ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
26# POSSIBILITY OF SUCH DAMAGE.
28import logging
29from io import BytesIO
30from typing import IO
32from .._utils import (
33 WHITESPACES,
34 WHITESPACES_AS_BYTES,
35 BinaryStreamType,
36 StreamType,
37 logger_warning,
38 read_non_whitespace,
39)
40from ..errors import PdfReadError
42logger = logging.getLogger(__name__)
44# An inline image should be used only for small images (4096 bytes or less),
45# but allow twice this for cases where this has been exceeded.
46BUFFER_SIZE = 8192
49def _check_end_image_marker(stream: StreamType) -> bool:
50 ei_tok = read_non_whitespace(stream)
51 ei_tok += stream.read(2)
52 stream.seek(-3, 1)
53 return ei_tok[:2] == b"EI" and (ei_tok[2:3] == b"" or ei_tok[2:3] in WHITESPACES)
56def extract_inline__ascii_hex_decode(stream: StreamType) -> bytes:
57 """
58 Extract HexEncoded stream from inline image.
59 The stream will be moved onto the EI.
60 """
61 data_out = bytearray()
62 # Read data until delimiter > and EI as backup.
63 while True:
64 data_buffered = read_non_whitespace(stream) + stream.read(BUFFER_SIZE)
65 if not data_buffered:
66 raise PdfReadError("Unexpected end of stream")
67 pos_tok = data_buffered.find(b">")
68 if pos_tok >= 0: # found >
69 data_out += data_buffered[: pos_tok + 1]
70 stream.seek(-len(data_buffered) + pos_tok + 1, 1)
71 break
72 pos_ei = data_buffered.find(b"EI")
73 if pos_ei >= 0: # found EI
74 stream.seek(-len(data_buffered) + pos_ei - 1, 1)
75 c = stream.read(1)
76 while c in WHITESPACES:
77 stream.seek(-2, 1)
78 c = stream.read(1)
79 pos_ei -= 1
80 data_out += data_buffered[:pos_ei]
81 break
82 if len(data_buffered) == 2:
83 data_out += data_buffered
84 raise PdfReadError("Unexpected end of stream")
85 # Neither > nor EI found
86 data_out += data_buffered[:-2]
87 stream.seek(-2, 1)
89 if not _check_end_image_marker(stream):
90 raise PdfReadError("EI stream not found")
91 return bytes(data_out)
94def extract_inline__ascii85_decode(stream: StreamType) -> bytes:
95 """
96 Extract A85 stream from inline image.
97 The stream will be moved onto the EI.
98 """
99 data_out = bytearray()
100 # Read data until delimiter ~>
101 while True:
102 data_buffered = read_non_whitespace(stream) + stream.read(BUFFER_SIZE)
103 if not data_buffered:
104 raise PdfReadError("Unexpected end of stream")
105 pos_tok = data_buffered.find(b"~>")
106 if pos_tok >= 0: # found!
107 data_out += data_buffered[: pos_tok + 2]
108 stream.seek(-len(data_buffered) + pos_tok + 2, 1)
109 break
110 if len(data_buffered) == 2: # end of buffer
111 data_out += data_buffered
112 raise PdfReadError("Unexpected end of stream")
113 data_out += data_buffered[
114 :-2
115 ] # back by one char in case of in the middle of ~>
116 stream.seek(-2, 1)
118 if not _check_end_image_marker(stream):
119 raise PdfReadError("EI stream not found")
120 return bytes(data_out)
123def extract_inline__run_length_decode(stream: StreamType) -> bytes:
124 """
125 Extract RL (RunLengthDecode) stream from inline image.
126 The stream will be moved onto the EI.
127 """
128 data_out = bytearray()
129 # Read data until delimiter 128
130 while True:
131 data_buffered = stream.read(BUFFER_SIZE)
132 if not data_buffered:
133 raise PdfReadError("Unexpected end of stream")
134 pos_tok = data_buffered.find(b"\x80")
135 if pos_tok >= 0: # found
136 # Ideally, we could just use plain run-length decoding here, where 80_16 = 128_10
137 # marks the EOD. But there apparently are cases like in issue #3517, where we have
138 # an inline image with up to 51 EOD markers. In these cases, be resilient here and
139 # use the default `EI` marker detection instead. Please note that this fallback
140 # still omits special `EI` handling within the stream, but for now assume that having
141 # both of these cases occur at the same time is very unlikely (and the image stream
142 # is broken anyway).
143 # For now, do not skip over more than one whitespace character.
144 after_token = data_buffered[pos_tok + 1 : pos_tok + 4]
145 if after_token.startswith(b"EI") or after_token.endswith(b"EI"):
146 data_out += data_buffered[: pos_tok + 1]
147 stream.seek(-len(data_buffered) + pos_tok + 1, 1)
148 else:
149 logger_warning("Early EOD in RunLengthDecode of inline image, using fallback.", __name__)
150 ei_marker = data_buffered.find(b"EI")
151 if ei_marker > 0:
152 data_out += data_buffered[: ei_marker]
153 stream.seek(-len(data_buffered) + ei_marker - 1, 1)
154 break
155 data_out += data_buffered
157 if not _check_end_image_marker(stream):
158 raise PdfReadError("EI stream not found")
159 return bytes(data_out)
162def extract_inline__dct_decode(stream: BinaryStreamType) -> bytes:
163 """
164 Extract DCT (JPEG) stream from inline image.
165 The stream will be moved onto the EI.
166 """
167 def read(length: int) -> bytes:
168 # If 0 bytes are returned, and *size* was not 0, this indicates end of file.
169 # If the object is in non-blocking mode and no bytes are available, `None` is returned.
170 _result = stream.read(length)
171 if _result is None or len(_result) != length:
172 raise PdfReadError("Unexpected end of stream")
173 return _result
175 data_out = bytearray()
176 # Read Blocks of data (ID/Size/data) up to ID=FF/D9
177 # https://www.digicamsoft.com/itu/itu-t81-36.html
178 not_first = False
179 while True:
180 c = read(1)
181 if not_first or (c == b"\xff"):
182 data_out += c
183 if c != b"\xff":
184 continue
185 not_first = True
186 c = read(1)
187 data_out += c
188 if c == b"\xff":
189 stream.seek(-1, 1) # pragma: no cover
190 elif c == b"\x00": # stuffing
191 pass
192 elif c == b"\xd9": # end
193 break
194 elif c in (
195 b"\xc0\xc1\xc2\xc3\xc4\xc5\xc6\xc7\xc9\xca\xcb\xcc\xcd\xce\xcf"
196 b"\xda\xdb\xdc\xdd\xde\xdf"
197 b"\xe0\xe1\xe2\xe3\xe4\xe5\xe6\xe7\xe8\xe9\xea\xeb\xec\xed\xee\xef\xfe"
198 ):
199 c = read(2)
200 data_out += c
201 sz = c[0] * 256 + c[1]
202 data_out += read(sz - 2)
204 if not _check_end_image_marker(stream):
205 raise PdfReadError("EI stream not found")
206 return bytes(data_out)
209def extract_inline_default(stream: StreamType) -> bytes:
210 """Legacy method, used by default"""
211 stream_out = BytesIO()
212 # Read the inline image, while checking for EI (End Image) operator.
213 while True:
214 data_buffered = stream.read(BUFFER_SIZE)
215 if not data_buffered:
216 raise PdfReadError("Unexpected end of stream")
217 pos_ei = data_buffered.find(
218 b"E"
219 ) # We can not look straight for "EI" because it may not have been loaded in the buffer
221 if pos_ei == -1:
222 stream_out.write(data_buffered)
223 else:
224 # Write out everything including E (the one from EI to be removed)
225 stream_out.write(data_buffered[0 : pos_ei + 1])
226 sav_pos_ei = stream_out.tell() - 1
227 # Seek back in the stream to read the E next
228 stream.seek(pos_ei + 1 - len(data_buffered), 1)
229 saved_pos = stream.tell()
230 # Check for End Image
231 tok2 = stream.read(1) # I of "EI"
232 if tok2 != b"I":
233 stream.seek(saved_pos, 0)
234 continue
235 tok3 = stream.read(1) # possible space after "EI"
236 if tok3 not in WHITESPACES:
237 stream.seek(saved_pos, 0)
238 continue
239 while tok3 in WHITESPACES:
240 tok3 = stream.read(1)
241 if data_buffered[pos_ei - 1 : pos_ei] not in WHITESPACES and tok3 not in {
242 b"Q",
243 b"E",
244 }: # for Q or EMC
245 stream.seek(saved_pos, 0)
246 continue
247 if is_followed_by_binary_data(stream):
248 # Inline image contains `EI ` sequence usually marking the end of it, but
249 # is followed by binary data which does not make sense for the actual end.
250 stream.seek(saved_pos, 0)
251 continue
252 # Data contains [\s]EI[\s](Q|EMC): 4 chars are sufficient
253 # remove E(I) wrongly inserted earlier
254 stream.seek(saved_pos - 1, 0)
255 stream_out.truncate(sav_pos_ei)
256 break
258 return stream_out.getvalue()
261def is_followed_by_binary_data(stream: IO[bytes], length: int = 10) -> bool:
262 """
263 Check if the next bytes of the stream look like binary image data or regular page content.
265 This is just some heuristics due to the PDF specification being too imprecise about
266 inline images containing the `EI` marker which would end an image. Starting with PDF 2.0,
267 we finally get a mandatory length field, but with (proper) PDF 2.0 support being very limited
268 everywhere, we should not expect to be able to remove such hacks in the near future - especially
269 considering legacy documents as well.
271 The actual implementation draws some inspiration from
272 https://github.com/itext/itext-java/blob/9.1.0/kernel/src/main/java/com/itextpdf/kernel/pdf/canvas/parser/util/InlineImageParsingUtils.java
273 """
274 position = stream.tell()
275 data = stream.read(length)
276 stream.seek(position)
277 if not data:
278 return False
279 operator_start = None
280 operator_end = None
282 for index, byte in enumerate(data):
283 if byte < 32 and byte not in WHITESPACES_AS_BYTES:
284 # This covers all characters not being displayable directly, although omitting whitespace
285 # to allow for operator detection.
286 return True
287 is_whitespace = byte in WHITESPACES_AS_BYTES
288 if operator_start is None and not is_whitespace:
289 # Interpret all other non-whitespace characters as the start of an operation.
290 operator_start = index
291 if operator_start is not None and is_whitespace:
292 # A whitespace stops an operation.
293 # Assume that having an inline image with tons of whitespace is rather unlikely.
294 operator_end = index
295 break
297 if operator_start is None:
298 # Inline images should not have tons of whitespaces, which would lead to no operator start.
299 return False
300 if operator_end is None:
301 # We probably are inside an operation.
302 operator_end = length
303 operator_length = operator_end - operator_start
304 operator = data[operator_start:operator_end]
305 if operator.startswith(b"/") and operator_length > 1:
306 # Name object.
307 return False
308 if operator.replace(b".", b"").isdigit():
309 # Graphics operator, for example a move. A number (integer or float).
310 return False
311 if operator_length > 3: # noqa: SIM103
312 # Usually, the operators inside a content stream should not have more than three characters,
313 # especially after an inline image.
314 return True
315 return False