pdf2 1.8.0

A PDF 2.0 (ISO 32000-2) reader, writer and text extractor.
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
//! Parser for PDF objects, built on [`crate::lexer`].
//!
//! [`parse_object`] turns a byte slice into a direct [`Object`];
//! [`parse_indirect_object`] reads the `n g obj … endobj` wrapper and the
//! `stream` … `endstream` payload. The cross-reference machinery that finds
//! those offsets is layered on top in Q3.

use alloc::vec::Vec;

use alloc::collections::VecDeque;

use crate::error::{Error, Result};
use crate::lexer::{Lexer, Token};
use crate::object::{Dictionary, Object, ObjectId, Stream};

/// Maximum container nesting. Hostile input must not recurse without bound,
/// which M10 pins with an error rather than a stack overflow.
pub const MAX_DEPTH: usize = 128;

/// Resolves a `/Length` that is an indirect object.
///
/// The parser cannot follow references on its own — that needs the object
/// table — so [`crate::Document`] supplies this when it parses a stream whose
/// `/Length` is `n g R`. Without a resolver the payload is delimited by
/// scanning for `endstream` instead.
pub trait LengthResolver: core::fmt::Debug {
    /// The byte length of the object `id` points at.
    fn resolve_length(&self, id: ObjectId) -> Result<usize>;
}

/// Parse exactly one object from `input`.
///
/// Only whitespace and comments may follow the object; anything else is an
/// error, so a caller that expects one value cannot silently drop the rest.
pub fn parse_object(input: &[u8]) -> Result<Object> {
    let mut parser = Parser::new(input);
    let object = parser.parse_object()?;
    if matches!(parser.peek_token()?, Token::Eof) {
        Ok(object)
    } else {
        Err(Error::Malformed {
            what: "trailing data after object",
            offset: Some(parser.position()),
        })
    }
}

/// An object together with the identity it was stored under.
#[derive(Debug, Clone, PartialEq)]
pub struct IndirectObject {
    /// The object number and generation from its `n g obj` header.
    pub id: ObjectId,
    /// The value, which may be a [`crate::object::Stream`].
    pub object: Object,
}

/// Parse exactly one indirect object from `input`.
///
/// The whole slice must be consumed: `n g obj … endobj`, optionally with a
/// `stream` payload whose `/Length` must agree with the bytes found.
pub fn parse_indirect_object(input: &[u8]) -> Result<IndirectObject> {
    let mut parser = Parser::new(input);
    let indirect = parser.parse_indirect_object()?;
    if matches!(parser.peek_token()?, Token::Eof) {
        Ok(indirect)
    } else {
        Err(Error::Malformed {
            what: "trailing data after endobj",
            offset: Some(parser.position()),
        })
    }
}

/// A token cursor that assembles [`Object`]s.
///
/// Two tokens of lookahead are buffered, which is what `n g R` reference
/// detection and the dictionary/array terminators need.
#[derive(Debug, Clone)]
pub struct Parser<'a> {
    lexer: Lexer<'a>,
    buffered: VecDeque<Token>,
    depth: usize,
    resolver: Option<&'a dyn LengthResolver>,
}

impl<'a> Parser<'a> {
    /// A parser over `input`.
    pub fn new(input: &'a [u8]) -> Self {
        Self {
            lexer: Lexer::new(input),
            buffered: VecDeque::new(),
            depth: 0,
            resolver: None,
        }
    }

    /// A parser that can resolve an indirect `/Length`.
    pub fn with_resolver(input: &'a [u8], resolver: &'a dyn LengthResolver) -> Self {
        Self {
            lexer: Lexer::new(input),
            buffered: VecDeque::new(),
            depth: 0,
            resolver: Some(resolver),
        }
    }

    /// Offset of the next token in the input.
    pub fn position(&self) -> usize {
        self.lexer.position()
    }

    /// The whole byte slice this parser reads.
    pub fn content(&self) -> &'a [u8] {
        self.lexer.input()
    }

    /// Offset just past the last token consumed.
    pub fn raw_position(&self) -> usize {
        self.lexer.raw_position()
    }

    /// Jump the cursor to `offset`, dropping any lookahead.
    ///
    /// Content-stream interpretation uses this to skip inline image data,
    /// which cannot be tokenized.
    pub fn seek(&mut self, offset: usize) {
        self.buffered.clear();
        self.lexer.seek(offset);
    }

    /// Skip trivia and return the next token without consuming it.
    pub fn peek_token(&mut self) -> Result<&Token> {
        self.fill(1)?;
        match self.buffered.front() {
            Some(token) => Ok(token),
            None => Err(Error::malformed("parser buffer")),
        }
    }

    /// Consume and return the next token.
    pub fn next_token(&mut self) -> Result<Token> {
        self.fill(1)?;
        self.buffered
            .pop_front()
            .ok_or_else(|| Error::malformed("parser buffer"))
    }

    /// Consume the next token, which must be `expected`.
    pub fn expect(&mut self, expected: &Token) -> Result<()> {
        let token = self.next_token()?;
        if &token == expected {
            Ok(())
        } else {
            Err(Error::Malformed {
                what: "expected token",
                offset: Some(self.lexer.raw_position()),
            })
        }
    }

    /// Parse `n g obj … endobj`, including a `stream` payload.
    ///
    /// Streams are read by byte offset, not as tokens: their payload is opaque
    /// and may contain anything, so `/Length` decides where the data ends and
    /// `endstream` is checked afterwards to catch a wrong length.
    pub fn parse_indirect_object(&mut self) -> Result<IndirectObject> {
        let offset = self.position();
        let number = match self.next_token()? {
            Token::Integer(number) => number,
            _ => {
                return Err(Error::Malformed {
                    what: "object number",
                    offset: Some(offset),
                });
            }
        };
        let generation = match self.next_token()? {
            Token::Integer(generation) => generation,
            _ => {
                return Err(Error::Malformed {
                    what: "object generation",
                    offset: Some(offset),
                });
            }
        };
        if number < 1 {
            return Err(Error::Malformed {
                what: "object number must be positive",
                offset: Some(offset),
            });
        }
        let number = u32::try_from(number).map_err(|_| Error::Malformed {
            what: "object number",
            offset: Some(offset),
        })?;
        let generation = u16::try_from(generation).map_err(|_| Error::Malformed {
            what: "object generation",
            offset: Some(offset),
        })?;
        self.expect_keyword(b"obj")?;
        let object = self.parse_value_or_stream()?;
        self.expect_keyword(b"endobj")?;
        Ok(IndirectObject {
            id: ObjectId::new(number, generation),
            object,
        })
    }

    fn expect_keyword(&mut self, keyword: &[u8]) -> Result<()> {
        match self.next_token()? {
            Token::Keyword(found) if found == keyword => Ok(()),
            _ => Err(Error::Malformed {
                what: "expected keyword",
                offset: Some(self.lexer.raw_position()),
            }),
        }
    }

    /// A direct object, followed by stream data when the value is a dictionary
    /// and the `stream` keyword comes next.
    fn parse_value_or_stream(&mut self) -> Result<Object> {
        let object = self.parse_object()?;
        let Object::Dictionary(dict) = object else {
            return Ok(object);
        };
        if !matches!(self.peek_token()?, Token::Keyword(keyword) if keyword == b"stream") {
            return Ok(Object::Dictionary(dict));
        }
        self.next_token()?; // `stream`
        self.parse_stream(dict)
    }

    fn parse_stream(&mut self, dict: Dictionary) -> Result<Object> {
        let input = self.lexer.input();
        let start = stream_data_start(self.lexer.raw_position(), input)?;
        let (data_end, scanned) = match stream_length(&dict)? {
            StreamLength::Direct(length) => (checked_end(start, length)?, None),
            StreamLength::Indirect(id) => match self.resolver {
                Some(resolver) => (checked_end(start, resolver.resolve_length(id)?)?, None),
                // Standalone parsing has no object table, so the payload is
                // delimited by its `endstream` marker instead.
                None => {
                    let (data_end, after) = scan_endstream(input, start)?;
                    (data_end, Some(after))
                }
            },
        };
        if data_end > input.len() {
            return Err(Error::unexpected_eof("stream data"));
        }
        let after = match scanned {
            Some(after) => after,
            None => endstream_after(input, data_end)?,
        };
        let data = input[start..data_end].to_vec();
        self.lexer.seek(after);
        Ok(Object::Stream(Stream::new(dict, data)))
    }

    /// Parse one direct object.
    pub fn parse_object(&mut self) -> Result<Object> {
        if self.depth >= MAX_DEPTH {
            return Err(Error::malformed("nesting depth"));
        }
        self.depth += 1;
        let object = self.parse_object_inner();
        self.depth -= 1;
        object
    }

    fn fill(&mut self, count: usize) -> Result<()> {
        while self.buffered.len() < count {
            let token = self.lexer.next_token()?;
            let eof = token == Token::Eof;
            self.buffered.push_back(token);
            if eof {
                // Keep returning EOF instead of reading past the end forever.
                while self.buffered.len() < count {
                    self.buffered.push_back(Token::Eof);
                }
            }
        }
        Ok(())
    }

    fn peek_at(&mut self, index: usize) -> Result<&Token> {
        self.fill(index + 1)?;
        match self.buffered.get(index) {
            Some(token) => Ok(token),
            None => Err(Error::malformed("parser buffer")),
        }
    }

    fn parse_object_inner(&mut self) -> Result<Object> {
        match self.next_token()? {
            Token::Integer(value) => self.parse_integer_or_reference(value),
            Token::Real(value) => Ok(Object::Real(value)),
            Token::String(string) => Ok(Object::String(string)),
            Token::Name(name) => Ok(Object::Name(name)),
            Token::ArrayStart => self.parse_array(),
            Token::DictionaryStart => self.parse_dictionary(),
            Token::Keyword(keyword) => match keyword.as_slice() {
                b"true" => Ok(Object::Boolean(true)),
                b"false" => Ok(Object::Boolean(false)),
                b"null" => Ok(Object::Null),
                _ => Err(Error::Malformed {
                    what: "keyword",
                    offset: Some(self.lexer.raw_position()),
                }),
            },
            Token::ArrayEnd | Token::DictionaryEnd => Err(Error::Malformed {
                what: "unbalanced delimiter",
                offset: Some(self.lexer.raw_position()),
            }),
            Token::Eof => Err(Error::unexpected_eof("object")),
        }
    }

    /// `12` alone is an integer, `12 0 R` is a reference.
    fn parse_integer_or_reference(&mut self, value: i64) -> Result<Object> {
        let is_reference = matches!(self.peek_at(0)?, Token::Integer(_))
            && matches!(self.peek_at(1)?, Token::Keyword(keyword) if keyword == b"R");
        if !is_reference {
            return Ok(Object::Integer(value));
        }
        let generation = match self.next_token()? {
            Token::Integer(generation) => generation,
            _ => return Err(Error::malformed("reference generation")),
        };
        self.expect(&Token::Keyword(b"R".to_vec()))?;
        let number = u32::try_from(value).map_err(|_| Error::Malformed {
            what: "reference object number",
            offset: None,
        })?;
        let generation = u16::try_from(generation).map_err(|_| Error::Malformed {
            what: "reference generation number",
            offset: None,
        })?;
        Ok(Object::Reference(ObjectId::new(number, generation)))
    }

    fn parse_array(&mut self) -> Result<Object> {
        let mut items = Vec::new();
        loop {
            match self.peek_at(0)? {
                Token::ArrayEnd => {
                    self.next_token()?;
                    return Ok(Object::Array(items));
                }
                Token::Eof => return Err(Error::unexpected_eof("array")),
                _ => items.push(self.parse_object()?),
            }
        }
    }

    fn parse_dictionary(&mut self) -> Result<Object> {
        let mut dict = Dictionary::new();
        loop {
            match self.peek_at(0)? {
                Token::DictionaryEnd => {
                    self.next_token()?;
                    return Ok(Object::Dictionary(dict));
                }
                Token::Name(_) => {
                    let name = match self.next_token()? {
                        Token::Name(name) => name,
                        _ => return Err(Error::malformed("dictionary key")),
                    };
                    let value = self.parse_object()?;
                    dict.insert(name, value);
                }
                Token::Eof => return Err(Error::unexpected_eof("dictionary")),
                _ => {
                    return Err(Error::Malformed {
                        what: "dictionary key",
                        offset: Some(self.lexer.position()),
                    });
                }
            }
        }
    }
}

/// What a stream dictionary says about the payload length.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
enum StreamLength {
    /// A direct integer, usable immediately.
    Direct(usize),
    /// An indirect object that only [`LengthResolver`] can turn into a number.
    Indirect(ObjectId),
}

fn stream_length(dict: &Dictionary) -> Result<StreamLength> {
    match dict.get(b"Length") {
        Some(Object::Integer(length)) => usize::try_from(*length)
            .map(StreamLength::Direct)
            .map_err(|_| Error::Malformed {
                what: "stream length must not be negative",
                offset: None,
            }),
        Some(Object::Reference(id)) => Ok(StreamLength::Indirect(*id)),
        Some(_) => Err(Error::Malformed {
            what: "stream length must be an integer",
            offset: None,
        }),
        None => Err(Error::Malformed {
            what: "stream without /Length",
            offset: None,
        }),
    }
}

/// The offset just past the end-of-line that follows the `stream` keyword.
fn stream_data_start(after_keyword: usize, input: &[u8]) -> Result<usize> {
    let mut start = after_keyword;
    match input.get(start) {
        Some(b'\r') => {
            start += 1;
            if input.get(start) == Some(&b'\n') {
                start += 1;
            }
        }
        Some(b'\n') => start += 1,
        _ => {
            return Err(Error::Malformed {
                what: "stream keyword without an end-of-line",
                offset: Some(start),
            });
        }
    }
    Ok(start)
}

fn checked_end(start: usize, length: usize) -> Result<usize> {
    start.checked_add(length).ok_or(Error::Malformed {
        what: "stream length overflow",
        offset: Some(start),
    })
}

const ENDSTREAM: &[u8] = b"endstream";

/// Verify that `end` is followed (after an optional end-of-line) by
/// `endstream`, and return the offset just past that keyword.
fn endstream_after(input: &[u8], end: usize) -> Result<usize> {
    let mut after = end;
    match input.get(after) {
        Some(b'\r') => {
            after += 1;
            if input.get(after) == Some(&b'\n') {
                after += 1;
            }
        }
        Some(b'\n') => after += 1,
        _ => {}
    }
    if input
        .get(after..)
        .is_some_and(|rest| rest.starts_with(ENDSTREAM))
    {
        Ok(after + ENDSTREAM.len())
    } else {
        // A wrong length shows up here: the bytes after the payload must reach
        // `endstream`, otherwise the dictionary and the data disagree.
        Err(Error::Malformed {
            what: "stream length does not end at endstream",
            offset: Some(end),
        })
    }
}

/// Delimit the payload by its `endstream` marker, for streams whose `/Length`
/// cannot be resolved. Returns `(data_end, offset after endstream)`.
fn scan_endstream(input: &[u8], start: usize) -> Result<(usize, usize)> {
    let found = input
        .get(start..)
        .and_then(|rest| {
            rest.windows(ENDSTREAM.len())
                .position(|window| window == ENDSTREAM)
        })
        .map(|position| start + position)
        .ok_or(Error::unexpected_eof("endstream"))?;
    let mut data_end = found;
    if data_end > start && input[data_end - 1] == b'\n' {
        data_end -= 1;
    }
    if data_end > start && input[data_end - 1] == b'\r' {
        data_end -= 1;
    }
    Ok((data_end, found + ENDSTREAM.len()))
}

#[cfg(test)]
mod tests {
    use super::{
        IndirectObject, LengthResolver, MAX_DEPTH, Parser, parse_indirect_object, parse_object,
    };
    use crate::error::Error;
    use crate::object::{Dictionary, Object, ObjectId};

    fn parsed(input: &str) -> Object {
        parse_object(input.as_bytes()).unwrap_or_else(|error| panic!("{input}: {error}"))
    }

    #[test]
    fn scalars_and_keywords_parse() {
        assert_eq!(parsed("null"), Object::Null);
        assert_eq!(parsed("true"), Object::Boolean(true));
        assert_eq!(parsed("false"), Object::Boolean(false));
        assert_eq!(parsed("-17"), Object::Integer(-17));
        assert_eq!(parsed("2.375"), Object::Real(2.375));
        assert_eq!(parsed("/Page"), Object::name("Page"));
        assert_eq!(parsed("(text)"), Object::string("text"));
        assert_eq!(parsed("  \n 7 \r\n"), Object::Integer(7));
    }

    #[test]
    fn references_need_both_numbers_and_r() {
        assert_eq!(parsed("12 0 R"), Object::reference(ObjectId::new(12, 0)));
        assert_eq!(
            parsed("7 65535 R"),
            Object::reference(ObjectId::new(7, 65535))
        );
        assert_eq!(
            parsed("[1 0 R 2]"),
            Object::array([Object::reference(ObjectId::new(1, 0)), Object::Integer(2)])
        );
        assert_eq!(
            parsed("[1 2 3]"),
            Object::array([Object::Integer(1), Object::Integer(2), Object::Integer(3)]),
            "runs of integers are not mistaken for references"
        );
        // Two integers without `R` stay two integers; the second one is
        // trailing data at top level, so it must be rejected.
        assert!(parse_object(b"12 0").is_err());
    }

    #[test]
    fn arrays_and_dictionaries_nest() {
        assert_eq!(parsed("[]"), Object::array(Vec::new()));
        assert_eq!(
            parsed("[[1] [2 3]]"),
            Object::array([
                Object::array([Object::Integer(1)]),
                Object::array([Object::Integer(2), Object::Integer(3)])
            ])
        );
        assert_eq!(parsed("<<>>"), Object::Dictionary(Dictionary::new()));
        let page = parsed("<< /Type /Page /Count 3 /Kids [1 0 R] >>");
        let dict = page.as_dict().expect("dictionary");
        assert_eq!(dict.get(b"Type"), Some(&Object::name("Page")));
        assert_eq!(dict.get(b"Count"), Some(&Object::Integer(3)));
        assert_eq!(
            dict.get(b"Kids"),
            Some(&Object::array([Object::reference(ObjectId::new(1, 0))]))
        );
    }

    #[test]
    fn dictionary_keys_keep_their_order() {
        let page = parsed("<< /Zed 1 /Alpha 2 /Mid 3 >>");
        let dict = page.as_dict().expect("dictionary");
        let names: Vec<&[u8]> = dict.iter().map(|(name, _)| name).collect();
        assert_eq!(names, vec![&b"Zed"[..], &b"Alpha"[..], &b"Mid"[..]]);
    }

    #[test]
    fn duplicate_dictionary_keys_keep_the_first_position() {
        let dict = parsed("<< /A 1 /B 2 /A 3 >>");
        let dict = dict.as_dict().expect("dictionary");
        assert_eq!(dict.len(), 2);
        assert_eq!(dict.get(b"A"), Some(&Object::Integer(3)));
    }

    #[test]
    fn malformed_input_returns_an_error() {
        let cases = [
            "",
            "(",
            "[1 2",
            "<< /A >>",
            "<< 1 2 >>",
            "]",
            ">>",
            "<< /A 1 >",
            "/Bad#zz",
            "1.2.3",
            "obj",
            "12 R",
        ];
        for case in cases {
            let error = parse_object(case.as_bytes()).expect_err(case);
            assert!(
                !error.to_string().is_empty(),
                "{case} produced an error without a message"
            );
        }
    }

    #[test]
    fn trailing_tokens_are_rejected() {
        assert!(matches!(parse_object(b"1 2"), Err(Error::Malformed { .. })));
        assert!(matches!(
            parse_object(b"<< >> << >>"),
            Err(Error::Malformed { .. })
        ));
    }

    #[test]
    fn deep_nesting_is_an_error_not_a_stack_overflow() {
        let depth = MAX_DEPTH + 10;
        let mut input = "[".repeat(depth);
        input.push_str(&"]".repeat(depth));
        assert!(matches!(
            parse_object(input.as_bytes()),
            Err(Error::Malformed { .. })
        ));
        // Just below the limit still parses.
        let depth = MAX_DEPTH - 1;
        let mut input = "[".repeat(depth);
        input.push_str(&"]".repeat(depth));
        assert!(parse_object(input.as_bytes()).is_ok());
    }

    #[test]
    fn truncation_never_panics() {
        let document = b"<< /Type /Page /Kids [1 0 R 2 0 R] /Count 2 >>";
        for end in 0..document.len() {
            // Either an error or a complete object; never a panic.
            let _ = parse_object(&document[..end]);
        }
    }

    fn indirect(input: &str) -> IndirectObject {
        parse_indirect_object(input.as_bytes()).unwrap_or_else(|error| panic!("{input}: {error}"))
    }

    #[test]
    fn indirect_objects_carry_their_identity() {
        let object = indirect("1 0 obj\n<< /Type /Catalog >>\nendobj\n");
        assert_eq!(object.id, ObjectId::new(1, 0));
        assert_eq!(
            object.object,
            Object::Dictionary({
                let mut dict = Dictionary::new();
                dict.insert("Type", Object::name("Catalog"));
                dict
            })
        );
        assert_eq!(
            indirect("2147483647 65535 obj 42 endobj").id,
            ObjectId::new(2147483647, 65535)
        );
        assert_eq!(indirect("7 0 obj 42 endobj").object, Object::Integer(42));
    }

    #[test]
    fn streams_take_their_bytes_from_length() {
        let object = indirect("3 0 obj\n<< /Length 5 >>\nstream\nhello\nendstream\nendobj\n");
        let stream = object.object.as_stream().expect("stream");
        assert_eq!(stream.data(), b"hello");
        assert_eq!(stream.dict().get(b"Length"), Some(&Object::Integer(5)));
    }

    #[test]
    fn stream_data_is_opaque_and_may_contain_endstream() {
        let input = "4 0 obj\n<< /Length 9 >>\nstream\nendstream\nendstream\nendobj\n";
        let object = indirect(input);
        let stream = object.object.as_stream().expect("stream");
        assert_eq!(stream.data(), b"endstream");
    }

    #[test]
    fn stream_payload_keeps_binary_bytes() {
        let mut input = Vec::new();
        input.extend_from_slice(b"5 0 obj\n<< /Length 4 >>\nstream\r\n");
        input.extend_from_slice(&[0x00, 0xff, b')', b'>']);
        input.extend_from_slice(b"\r\nendstream\r\nendobj\r\n");
        let object = parse_indirect_object(&input).expect("stream");
        let stream = object.object.as_stream().expect("stream");
        assert_eq!(stream.data(), &[0x00, 0xff, b')', b'>']);
    }

    #[test]
    fn a_wrong_length_is_an_error() {
        // Too short: three of the five bytes, so `endstream` is not next.
        let short = b"3 0 obj\n<< /Length 3 >>\nstream\nhello\nendstream\nendobj\n";
        let error = parse_indirect_object(short).expect_err("short length");
        assert!(error.to_string().contains("endstream"), "{error}");

        // Too long: the length runs past the available bytes.
        let long = b"3 0 obj\n<< /Length 500 >>\nstream\nhello\nendstream\nendobj\n";
        assert!(matches!(
            parse_indirect_object(long),
            Err(Error::UnexpectedEof { .. })
        ));
    }

    #[test]
    fn stream_length_must_be_a_number_or_a_reference() {
        let missing = b"3 0 obj\n<< /Type /XObject >>\nstream\nhello\nendstream\nendobj\n";
        assert!(parse_indirect_object(missing).is_err());

        let name = b"3 0 obj\n<< /Length /Five >>\nstream\nhello\nendstream\nendobj\n";
        assert!(parse_indirect_object(name).is_err());

        let negative = b"3 0 obj\n<< /Length -1 >>\nstream\nhello\nendstream\nendobj\n";
        assert!(parse_indirect_object(negative).is_err());
    }

    #[test]
    fn an_indirect_length_is_scanned_without_a_document() {
        let input = b"3 0 obj\n<< /Length 9 0 R >>\nstream\nhello\nendstream\nendobj\n";
        let object = parse_indirect_object(input).expect("scanned stream");
        let stream = object.object.as_stream().expect("stream");
        assert_eq!(stream.data(), b"hello");
    }

    #[derive(Debug)]
    struct FixedLength(usize);

    impl LengthResolver for FixedLength {
        fn resolve_length(&self, _id: ObjectId) -> crate::Result<usize> {
            Ok(self.0)
        }
    }

    #[test]
    fn an_indirect_length_can_be_resolved_through_the_object_table() {
        let input = b"3 0 obj\n<< /Length 9 0 R >>\nstream\nhello\nendstream\nendobj\n";
        let resolver = FixedLength(5);
        let mut parser = Parser::with_resolver(input, &resolver);
        let object = parser.parse_indirect_object().expect("resolved stream");
        assert_eq!(object.object.as_stream().expect("stream").data(), b"hello");

        // A length that disagrees with the payload is still an error.
        let wrong = FixedLength(3);
        let mut parser = Parser::with_resolver(input, &wrong);
        assert!(parser.parse_indirect_object().is_err());
    }

    #[test]
    fn a_stream_keyword_needs_an_end_of_line() {
        let input = b"3 0 obj\n<< /Length 5 >>\nstreamhello\nendstream\nendobj\n";
        assert!(matches!(
            parse_indirect_object(input),
            Err(Error::Malformed { .. })
        ));
    }

    #[test]
    fn indirect_object_wrappers_are_enforced() {
        for input in [
            &b"obj 42 endobj"[..],
            b"1 0 obj 42",
            b"1 0 obj 42 endobj 43",
            b"1 obj 42 endobj",
            b"1 0 42 endobj",
            b"0 0 obj 42 endobj",
            b"-1 0 obj 42 endobj",
            b"4294967296 0 obj 42 endobj",
            b"1 65536 obj 42 endobj",
        ] {
            assert!(
                parse_indirect_object(input).is_err(),
                "{} must not parse",
                String::from_utf8_lossy(input)
            );
        }
    }

    #[test]
    fn truncated_indirect_objects_never_panic() {
        let document =
            b"6 0 obj\n<< /Type /XObject /Length 5 >>\nstream\nhello\nendstream\nendobj\n";
        for end in 0..document.len() {
            let _ = parse_indirect_object(&document[..end]);
        }
        assert!(parse_indirect_object(document).is_ok());
    }
}