pdf2 1.1.0

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
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
//! Stream filters (§7.4): Flate, LZW, ASCIIHex, ASCII85, RunLength and the
//! predictor functions that Flate/LZW payloads may be wrapped in.
//!
//! Decoding is pure: [`decode_chain`] takes bytes and returns bytes, so a
//! filter can be tested without a document. Resolving indirect `/Filter`,
//! `/DecodeParms` or `/Length` entries is [`crate::Document`]'s job.

use alloc::vec;
use alloc::vec::Vec;

use crate::error::{Error, Result};
use crate::lexer::{hex_value, is_whitespace};
use crate::object::{Dictionary, Object, Stream};

/// A filter named by `/Filter`.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum Filter {
    /// `/FlateDecode` — zlib-wrapped DEFLATE.
    FlateDecode,
    /// `/LZWDecode` — LZW with variable code width.
    LzwDecode,
    /// `/ASCIIHexDecode` — hex digits, `>` terminates.
    AsciiHexDecode,
    /// `/ASCII85Decode` — base-85, `~>` terminates.
    Ascii85Decode,
    /// `/RunLengthDecode` — literal and repeat runs.
    RunLengthDecode,
}

impl Filter {
    /// Map a `/Filter` name to a filter.
    pub fn from_name(name: &[u8]) -> Result<Self> {
        match name {
            b"FlateDecode" | b"Fl" => Ok(Self::FlateDecode),
            b"LZWDecode" | b"LZW" => Ok(Self::LzwDecode),
            b"ASCIIHexDecode" | b"AHx" => Ok(Self::AsciiHexDecode),
            b"ASCII85Decode" | b"A85" => Ok(Self::Ascii85Decode),
            b"RunLengthDecode" | b"RL" => Ok(Self::RunLengthDecode),
            other => Err(Error::Malformed {
                what: match other {
                    b"DCTDecode" | b"JPXDecode" | b"CCITTFaxDecode" | b"JBIG2Decode" => {
                        "image filters are out of scope"
                    }
                    b"Crypt" => "the /Crypt filter is Q9's job",
                    _ => "unknown /Filter name",
                },
                offset: None,
            }),
        }
    }

    /// The canonical `/Filter` name.
    pub fn name(self) -> &'static str {
        match self {
            Self::FlateDecode => "FlateDecode",
            Self::LzwDecode => "LZWDecode",
            Self::AsciiHexDecode => "ASCIIHexDecode",
            Self::Ascii85Decode => "ASCII85Decode",
            Self::RunLengthDecode => "RunLengthDecode",
        }
    }
}

/// One entry of the filter chain: the filter plus its `/DecodeParms`.
#[derive(Debug, Clone, PartialEq)]
pub struct FilterSpec {
    /// The filter to apply.
    pub filter: Filter,
    /// The dictionary from `/DecodeParms` for this filter, if any.
    pub parms: Option<Dictionary>,
}

impl FilterSpec {
    /// Read `/Filter` and `/DecodeParms` from a stream dictionary.
    ///
    /// Only direct values are followed; an indirect `/Filter` or
    /// `/DecodeParms` needs [`crate::Document::decode_stream`] to resolve it
    /// first.
    pub fn from_dict(dict: &Dictionary) -> Result<Vec<Self>> {
        let filters = match dict.get(b"Filter") {
            None | Some(Object::Null) => return Ok(Vec::new()),
            Some(Object::Name(name)) => vec![Filter::from_name(name)?],
            Some(Object::Array(items)) => items
                .iter()
                .map(|item| match item {
                    Object::Name(name) => Filter::from_name(name),
                    _ => Err(Error::Malformed {
                        what: "/Filter array entries must be names",
                        offset: None,
                    }),
                })
                .collect::<Result<Vec<_>>>()?,
            Some(Object::Reference(_)) => {
                return Err(Error::Malformed {
                    what: "indirect /Filter needs the document",
                    offset: None,
                })
            }
            Some(_) => {
                return Err(Error::Malformed {
                    what: "/Filter must be a name or an array of names",
                    offset: None,
                })
            }
        };
        let parms = match dict.get(b"DecodeParms") {
            None | Some(Object::Null) => Vec::new(),
            Some(Object::Dictionary(parms)) => vec![Some(parms.clone())],
            Some(Object::Array(items)) => items
                .iter()
                .map(|item| match item {
                    Object::Null => Ok(None),
                    Object::Dictionary(parms) => Ok(Some(parms.clone())),
                    _ => Err(Error::Malformed {
                        what: "/DecodeParms array entries must be dictionaries or null",
                        offset: None,
                    }),
                })
                .collect::<Result<Vec<_>>>()?,
            Some(Object::Reference(_)) => {
                return Err(Error::Malformed {
                    what: "indirect /DecodeParms needs the document",
                    offset: None,
                })
            }
            Some(_) => {
                return Err(Error::Malformed {
                    what: "/DecodeParms must be a dictionary or an array",
                    offset: None,
                })
            }
        };
        Ok(filters
            .into_iter()
            .enumerate()
            .map(|(index, filter)| FilterSpec {
                filter,
                parms: parms.get(index).cloned().flatten(),
            })
            .collect())
    }
}

/// Apply a filter chain in order: the first `/Filter` is the outermost.
pub fn decode_chain(data: &[u8], chain: &[FilterSpec]) -> Result<Vec<u8>> {
    let mut current = data.to_vec();
    for spec in chain {
        current = decode_filter(spec.filter, &current, spec.parms.as_ref())?;
    }
    Ok(current)
}

/// Apply one filter, then its predictor when the filter has one.
pub fn decode_filter(filter: Filter, data: &[u8], parms: Option<&Dictionary>) -> Result<Vec<u8>> {
    let decoded = match filter {
        Filter::FlateDecode => flate_decode(data)?,
        Filter::LzwDecode => lzw_decode(data, early_change(parms)?)?,
        Filter::AsciiHexDecode => ascii_hex_decode(data)?,
        Filter::Ascii85Decode => ascii85_decode(data)?,
        Filter::RunLengthDecode => run_length_decode(data)?,
    };
    match filter {
        Filter::FlateDecode | Filter::LzwDecode => {
            let predictor = Predictor::from_dict(parms)?;
            apply_predictor(&decoded, predictor)
        }
        _ => Ok(decoded),
    }
}

impl Stream {
    /// Decode this stream's payload with its own `/Filter` and `/DecodeParms`.
    ///
    /// Fails when either entry is indirect — use
    /// [`crate::Document::decode_stream`] for that case.
    pub fn decode(&self) -> Result<Vec<u8>> {
        let chain = FilterSpec::from_dict(self.dict())?;
        decode_chain(self.data(), &chain)
    }
}

/// `/DecodeParms` predictor settings, with the defaults from §7.4.4.4.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub struct Predictor {
    /// 1 = none, 2 = TIFF, 10–15 = PNG filters.
    pub predictor: u8,
    /// Colour components per pixel, default 1.
    pub colors: u8,
    /// Bits per component, default 8.
    pub bits_per_component: u8,
    /// Bytes per row divided by the pixel size, default 1.
    pub columns: u16,
}

impl Default for Predictor {
    fn default() -> Self {
        Self {
            predictor: 1,
            colors: 1,
            bits_per_component: 8,
            columns: 1,
        }
    }
}

impl Predictor {
    /// Read predictor settings from a `/DecodeParms` dictionary.
    pub fn from_dict(parms: Option<&Dictionary>) -> Result<Self> {
        let Some(parms) = parms else {
            return Ok(Self::default());
        };
        let mut predictor = Self::default();
        if let Some(Object::Integer(value)) = parms.get(b"Predictor") {
            predictor.predictor = u8::try_from(*value).map_err(|_| Error::Malformed {
                what: "/Predictor out of range",
                offset: None,
            })?;
        }
        if let Some(Object::Integer(value)) = parms.get(b"Colors") {
            predictor.colors = u8::try_from(*value).map_err(|_| Error::Malformed {
                what: "/Colors out of range",
                offset: None,
            })?;
        }
        if let Some(Object::Integer(value)) = parms.get(b"BitsPerComponent") {
            predictor.bits_per_component = u8::try_from(*value).map_err(|_| Error::Malformed {
                what: "/BitsPerComponent out of range",
                offset: None,
            })?;
        }
        if let Some(Object::Integer(value)) = parms.get(b"Columns") {
            predictor.columns = u16::try_from(*value).map_err(|_| Error::Malformed {
                what: "/Columns out of range",
                offset: None,
            })?;
        }
        Ok(predictor)
    }
}

fn flate_decode(data: &[u8]) -> Result<Vec<u8>> {
    // PDF Flate streams are zlib-wrapped, but some writers emit raw DEFLATE.
    if let Ok(out) = miniz_oxide::inflate::decompress_to_vec_zlib(data) {
        return Ok(out);
    }
    miniz_oxide::inflate::decompress_to_vec(data).map_err(|_| Error::Malformed {
        what: "FlateDecode payload",
        offset: None,
    })
}

fn early_change(parms: Option<&Dictionary>) -> Result<bool> {
    match parms.and_then(|parms| parms.get(b"EarlyChange")) {
        None => Ok(true),
        Some(Object::Integer(0)) => Ok(false),
        Some(Object::Integer(1)) => Ok(true),
        Some(_) => Err(Error::Malformed {
            what: "/EarlyChange must be 0 or 1",
            offset: None,
        }),
    }
}

fn ascii_hex_decode(data: &[u8]) -> Result<Vec<u8>> {
    let mut out = Vec::new();
    let mut high: Option<u8> = None;
    for (offset, &byte) in data.iter().enumerate() {
        if byte == b'>' {
            break;
        }
        if is_whitespace(byte) {
            continue;
        }
        let value = hex_value(byte).ok_or(Error::UnexpectedByte {
            byte,
            offset,
            what: "ASCIIHexDecode payload",
        })?;
        match high.take() {
            Some(high) => out.push((high << 4) | value),
            None => high = Some(value),
        }
    }
    if let Some(high) = high {
        // An odd final digit is the high nibble of a byte padded with 0.
        out.push(high << 4);
    }
    Ok(out)
}

fn ascii85_decode(data: &[u8]) -> Result<Vec<u8>> {
    let mut out = Vec::new();
    let mut group = [0_u8; 5];
    let mut count = 0;
    let start = if data.starts_with(b"<~") { 2 } else { 0 };
    let mut offset = start;
    while offset < data.len() {
        let byte = data[offset];
        offset += 1;
        if byte == b'~' {
            break;
        }
        if is_whitespace(byte) {
            continue;
        }
        if byte == b'z' {
            if count != 0 {
                return Err(Error::Malformed {
                    what: "ASCII85Decode `z` inside a group",
                    offset: None,
                });
            }
            out.extend_from_slice(&[0, 0, 0, 0]);
            continue;
        }
        if !(b'!'..=b'u').contains(&byte) {
            return Err(Error::UnexpectedByte {
                byte,
                offset: offset - 1,
                what: "ASCII85Decode payload",
            });
        }
        group[count] = byte - b'!';
        count += 1;
        if count == 5 {
            push_ascii85_group(&mut out, &group, 4)?;
            count = 0;
        }
    }
    if count > 0 {
        if count == 1 {
            return Err(Error::Malformed {
                what: "ASCII85Decode final group of one character",
                offset: None,
            });
        }
        for slot in group.iter_mut().skip(count) {
            *slot = 84; // `u`, the largest digit, pads the group
        }
        push_ascii85_group(&mut out, &group, count - 1)?;
    }
    Ok(out)
}

fn push_ascii85_group(out: &mut Vec<u8>, group: &[u8; 5], emit: usize) -> Result<()> {
    let mut value = 0_u64;
    for &digit in group {
        value = value * 85 + u64::from(digit);
    }
    let value = u32::try_from(value).map_err(|_| Error::Malformed {
        what: "ASCII85Decode group overflows 32 bits",
        offset: None,
    })?;
    out.extend_from_slice(&value.to_be_bytes()[..emit]);
    Ok(())
}

fn run_length_decode(data: &[u8]) -> Result<Vec<u8>> {
    let mut out = Vec::new();
    let mut offset = 0;
    while offset < data.len() {
        let length = data[offset];
        offset += 1;
        match length {
            // 128 ends the data.
            128 => break,
            // A length up to 127 is followed by that many plus one literal bytes.
            0..=127 => {
                let count = usize::from(length) + 1;
                let chunk = data
                    .get(offset..offset + count)
                    .ok_or(Error::unexpected_eof("RunLengthDecode literal run"))?;
                out.extend_from_slice(chunk);
                offset += count;
            }
            // 129–255 repeats the next byte 257 minus the length times.
            _ => {
                let byte = *data
                    .get(offset)
                    .ok_or(Error::unexpected_eof("RunLengthDecode repeat run"))?;
                offset += 1;
                out.resize(out.len() + (257 - usize::from(length)), byte);
            }
        }
    }
    Ok(out)
}

const LZW_CLEAR: u16 = 256;
const LZW_EOD: u16 = 257;
const LZW_MAX_CODE: usize = 4096;

fn lzw_decode(data: &[u8], early_change: bool) -> Result<Vec<u8>> {
    let mut out = Vec::new();
    let mut table: Vec<Vec<u8>> = (0..=255_u16).map(|byte| vec![byte as u8]).collect();
    table.push(Vec::new()); // 256: clear table
    table.push(Vec::new()); // 257: end of data
    let mut code_width = 9_u8;
    let mut previous: Option<Vec<u8>> = None;
    let mut bit = 0_usize;
    while let Some(code) = read_lzw_code(data, &mut bit, code_width) {
        if code == LZW_EOD {
            break;
        }
        if code == LZW_CLEAR {
            table.truncate(258);
            code_width = 9;
            previous = None;
            continue;
        }
        let sequence = match table.get(usize::from(code)) {
            Some(entry) if !entry.is_empty() => entry.clone(),
            // The one-code-ahead case: the code names the sequence about to be
            // added, which is the previous sequence plus its own first byte.
            _ if usize::from(code) == table.len() => match &previous {
                Some(previous) => {
                    let mut sequence = previous.clone();
                    sequence.push(previous[0]);
                    sequence
                }
                None => {
                    return Err(Error::Malformed {
                        what: "LZWDecode code without a base sequence",
                        offset: None,
                    })
                }
            },
            _ => {
                return Err(Error::Malformed {
                    what: "LZWDecode code out of range",
                    offset: None,
                })
            }
        };
        out.extend_from_slice(&sequence);
        if let Some(previous) = previous {
            if table.len() < LZW_MAX_CODE {
                let mut entry = previous;
                entry.push(sequence[0]);
                table.push(entry);
                if table.len() + usize::from(early_change) == 512 {
                    code_width = 10;
                } else if table.len() + usize::from(early_change) == 1024 {
                    code_width = 11;
                } else if table.len() + usize::from(early_change) == 2048 {
                    code_width = 12;
                }
            }
        }
        previous = Some(sequence);
    }
    Ok(out)
}

fn read_lzw_code(data: &[u8], bit: &mut usize, width: u8) -> Option<u16> {
    if *bit + usize::from(width) > data.len() * 8 {
        return None;
    }
    let mut value = 0_u16;
    for _ in 0..width {
        let byte = data[*bit / 8];
        let bit_in_byte = 7 - (*bit % 8);
        value = (value << 1) | u16::from((byte >> bit_in_byte) & 1);
        *bit += 1;
    }
    Some(value)
}

fn apply_predictor(data: &[u8], predictor: Predictor) -> Result<Vec<u8>> {
    match predictor.predictor {
        1 => Ok(data.to_vec()),
        2 => tiff_predictor(data, predictor),
        10..=15 => png_predictor(data, predictor),
        _ => Err(Error::Malformed {
            what: "unsupported /Predictor",
            offset: None,
        }),
    }
}

fn png_predictor(data: &[u8], predictor: Predictor) -> Result<Vec<u8>> {
    let colors = usize::from(predictor.colors.max(1));
    let bits = usize::from(predictor.bits_per_component.max(1));
    let columns = usize::from(predictor.columns.max(1));
    let bpp = (colors * bits).div_ceil(8);
    let row_bytes = (columns * colors * bits).div_ceil(8);
    let mut out = Vec::with_capacity(data.len());
    let mut previous = vec![0_u8; row_bytes];
    let mut offset = 0;
    while offset < data.len() {
        let filter = data[offset];
        offset += 1;
        let row = data
            .get(offset..offset + row_bytes)
            .ok_or(Error::unexpected_eof("PNG predictor row"))?;
        offset += row_bytes;
        let mut current = vec![0_u8; row_bytes];
        for index in 0..row_bytes {
            let left = if index >= bpp {
                current[index - bpp]
            } else {
                0
            };
            let up = previous[index];
            let up_left = if index >= bpp {
                previous[index - bpp]
            } else {
                0
            };
            current[index] = match filter {
                0 => row[index],
                1 => row[index].wrapping_add(left),
                2 => row[index].wrapping_add(up),
                3 => row[index].wrapping_add(((u16::from(left) + u16::from(up)) / 2) as u8),
                4 => row[index].wrapping_add(paeth(left, up, up_left)),
                _ => {
                    return Err(Error::Malformed {
                        what: "PNG predictor row filter must be 0–4",
                        offset: None,
                    })
                }
            };
        }
        out.extend_from_slice(&current);
        previous = current;
    }
    Ok(out)
}

fn paeth(left: u8, up: u8, up_left: u8) -> u8 {
    let left = i32::from(left);
    let up = i32::from(up);
    let up_left = i32::from(up_left);
    let prediction = left + up - up_left;
    let distance_left = (prediction - left).abs();
    let distance_up = (prediction - up).abs();
    let distance_up_left = (prediction - up_left).abs();
    if distance_left <= distance_up && distance_left <= distance_up_left {
        left as u8
    } else if distance_up <= distance_up_left {
        up as u8
    } else {
        up_left as u8
    }
}

fn tiff_predictor(data: &[u8], predictor: Predictor) -> Result<Vec<u8>> {
    if predictor.bits_per_component != 8 {
        return Err(Error::Malformed {
            what: "TIFF predictor needs 8 bits per component",
            offset: None,
        });
    }
    let colors = usize::from(predictor.colors.max(1));
    let columns = usize::from(predictor.columns.max(1));
    let row_bytes = colors * columns;
    let mut out = data.to_vec();
    for row in out.chunks_mut(row_bytes) {
        for index in colors..row.len() {
            row[index] = row[index].wrapping_add(row[index - colors]);
        }
    }
    Ok(out)
}

#[cfg(test)]
mod tests {
    use super::{
        apply_predictor, ascii85_decode, ascii_hex_decode, decode_chain, lzw_decode,
        run_length_decode, Filter, FilterSpec, Predictor,
    };
    use crate::error::Error;
    use crate::object::{Dictionary, Object};

    fn parms(entries: &[(&str, Object)]) -> Dictionary {
        let mut dict = Dictionary::new();
        for (name, value) in entries {
            dict.insert(name.as_bytes(), value.clone());
        }
        dict
    }

    #[test]
    fn filter_names_map_to_filters() {
        assert_eq!(
            Filter::from_name(b"FlateDecode").expect("flate"),
            Filter::FlateDecode
        );
        assert_eq!(
            Filter::from_name(b"Fl").expect("flate"),
            Filter::FlateDecode
        );
        assert_eq!(
            Filter::from_name(b"LZWDecode").expect("lzw"),
            Filter::LzwDecode
        );
        assert_eq!(
            Filter::from_name(b"ASCIIHexDecode").expect("ahx"),
            Filter::AsciiHexDecode
        );
        assert_eq!(
            Filter::from_name(b"ASCII85Decode").expect("a85"),
            Filter::Ascii85Decode
        );
        assert_eq!(
            Filter::from_name(b"RunLengthDecode").expect("rl"),
            Filter::RunLengthDecode
        );
        assert_eq!(Filter::FlateDecode.name(), "FlateDecode");
        let image = Filter::from_name(b"DCTDecode").expect_err("image filter");
        assert!(image.to_string().contains("out of scope"), "{image}");
        assert!(Filter::from_name(b"Nope").is_err());
    }

    #[test]
    fn ascii_hex_decodes_with_and_without_the_terminator() {
        assert_eq!(ascii_hex_decode(b"48656C6C6F>").expect("hex"), b"Hello");
        assert_eq!(ascii_hex_decode(b"48 65 6c 6C 6f").expect("hex"), b"Hello");
        assert_eq!(
            ascii_hex_decode(b"414").expect("odd digit"),
            vec![0x41, 0x40]
        );
        assert_eq!(ascii_hex_decode(b">").expect("empty"), Vec::<u8>::new());
        assert!(ascii_hex_decode(b"48G").is_err());
    }

    #[test]
    fn ascii85_decodes_groups_shortcuts_and_partials() {
        // "Hello" as ASCII85 is "87cURDZ".
        assert_eq!(ascii85_decode(b"87cURDZ~>").expect("ascii85"), b"Hello");
        assert_eq!(ascii85_decode(b"87cURDZ").expect("no terminator"), b"Hello");
        assert_eq!(ascii85_decode(b"@:E^").expect("three bytes"), b"abc");
        assert_eq!(ascii85_decode(b"z~>").expect("z shortcut"), [0, 0, 0, 0]);
        assert_eq!(ascii85_decode(b"<~87cURDZ~>").expect("prefix"), b"Hello");
        assert_eq!(ascii85_decode(b"").expect("empty"), Vec::<u8>::new());
        assert!(ascii85_decode(b"87{UR").is_err(), "character out of range");
        assert_eq!(ascii85_decode(b"87").expect("partial").len(), 1);
        assert!(ascii85_decode(b"8").is_err(), "single trailing character");
        assert!(ascii85_decode(b"zZ~>").is_err(), "z inside a group");
    }

    #[test]
    fn run_length_decodes_literal_and_repeat_runs() {
        // 2 literal bytes "AB", then 3 copies of 'x' (257-254=3), then EOD.
        assert_eq!(
            run_length_decode(&[1, b'A', b'B', 254, b'x', 128]).expect("runlength"),
            b"ABxxx"
        );
        assert_eq!(run_length_decode(&[128]).expect("empty"), Vec::<u8>::new());
        assert!(
            run_length_decode(&[5, b'A']).is_err(),
            "truncated literal run"
        );
        assert!(run_length_decode(&[254]).is_err(), "truncated repeat run");
    }

    #[test]
    fn lzw_decodes_a_hand_computed_vector() {
        // 9-bit MSB-first codes 65 ('A'), 66 ('B'), 67 ('C'), 257 (EOD).
        let data = [0x20, 0x90, 0x88, 0x70, 0x10];
        assert_eq!(lzw_decode(&data, true).expect("lzw"), b"ABC");
        assert_eq!(lzw_decode(&data, false).expect("lzw"), b"ABC");
    }

    #[test]
    fn lzw_handles_clear_codes_and_growing_code_widths() {
        // Produced by an independent encoder: clear code, 9-bit codes, width
        // growth at the EarlyChange boundaries, end-of-data. Decoding 360 bytes
        // of text exercises the 9 -> 10 bit step.
        let text: Vec<u8> = b"The quick brown fox jumps over the lazy dog. ".repeat(8);
        let encoded = bytes(concat!(
            "80150d065101c4ea69319ac4062391bcee6e10198de78101a8ea6d381cc406f3",
            "b194e4203a40c406c309e8f220321bcce2e1040a09068442a190e8844a29168c",
            "46a391e9048a4926944aa592e82c1e130b86c3e23138ac5e331b8ec7e43049fc",
            "9e532b96c8a614799d2a6d4d9cd42795391c96ad42accbe8d32a4cd69938a7ce",
            "ea53eb3506b144addb2694b9bd3a75519ed52eb57a1d6ad748bdd7ee17fb1dd2",
            "8184b4d166388af5befd62b9e0b1f68808"
        ));
        assert_eq!(lzw_decode(&encoded, true).expect("lzw"), text);
    }

    fn bytes(hex: &str) -> Vec<u8> {
        hex.as_bytes()
            .chunks(2)
            .map(|pair| {
                u8::from_str_radix(core::str::from_utf8(pair).expect("hex"), 16).expect("hex")
            })
            .collect()
    }

    #[test]
    fn png_predictors_undo_every_row_filter() {
        // Two rows of three bytes with explicit PNG filter bytes.
        let none = [0, 1, 2, 3, 0, 4, 5, 6];
        let sub = [1, 1, 1, 1, 1, 4, 1, 1];
        let up = [2, 1, 2, 3, 2, 3, 3, 3];
        let average = [3, 1, 2, 2, 3, 4, 2, 2];
        let paeth = [4, 1, 1, 1, 4, 3, 1, 1];
        let settings = Predictor {
            predictor: 15,
            colors: 1,
            bits_per_component: 8,
            columns: 3,
        };
        for encoded in [none, sub, up, average, paeth] {
            assert_eq!(
                apply_predictor(&encoded, settings).expect("predictor"),
                vec![1, 2, 3, 4, 5, 6],
                "encoded {encoded:?}"
            );
        }
        assert!(
            apply_predictor(&[9, 1, 2, 3], settings).is_err(),
            "bad filter byte"
        );
    }

    #[test]
    fn tiff_predictor_is_horizontal_differencing() {
        let encoded = [10, 10, 10, 1, 0, 0];
        let settings = Predictor {
            predictor: 2,
            colors: 1,
            bits_per_component: 8,
            columns: 3,
        };
        assert_eq!(
            apply_predictor(&encoded, settings).expect("tiff"),
            vec![10, 20, 30, 1, 1, 1]
        );
        let rgb = Predictor {
            predictor: 2,
            colors: 3,
            bits_per_component: 8,
            columns: 2,
        };
        assert_eq!(
            apply_predictor(&[1, 2, 3, 10, 10, 10], rgb).expect("tiff rgb"),
            vec![1, 2, 3, 11, 12, 13]
        );
        let unsupported = Predictor {
            predictor: 2,
            bits_per_component: 4,
            ..Predictor::default()
        };
        assert!(apply_predictor(&[0, 0], unsupported).is_err());
        assert!(apply_predictor(
            &[0],
            Predictor {
                predictor: 7,
                ..Predictor::default()
            }
        )
        .is_err());
    }

    #[test]
    fn decoders_report_bad_payloads() {
        assert!(matches!(
            super::decode_filter(Filter::FlateDecode, b"not deflate at all", None),
            Err(Error::Malformed { .. })
        ));
        assert!(matches!(
            super::decode_filter(Filter::Ascii85Decode, b"87{UR", None),
            Err(Error::UnexpectedByte { .. })
        ));
    }

    #[test]
    fn filter_specs_read_names_arrays_and_parms() {
        let dict = parms(&[("Filter", Object::name("FlateDecode"))]);
        let specs = FilterSpec::from_dict(&dict).expect("single filter");
        assert_eq!(specs.len(), 1);
        assert_eq!(specs[0].filter, Filter::FlateDecode);
        assert!(specs[0].parms.is_none());

        let dict = parms(&[
            (
                "Filter",
                Object::array([Object::name("ASCII85Decode"), Object::name("FlateDecode")]),
            ),
            (
                "DecodeParms",
                Object::array([
                    Object::Null,
                    Object::Dictionary(parms(&[("Predictor", Object::Integer(12))])),
                ]),
            ),
        ]);
        let specs = FilterSpec::from_dict(&dict).expect("filter array");
        assert_eq!(specs.len(), 2);
        assert_eq!(specs[0].filter, Filter::Ascii85Decode);
        assert!(specs[0].parms.is_none());
        assert_eq!(specs[1].filter, Filter::FlateDecode);
        assert!(specs[1].parms.is_some());

        let indirect = parms(&[("Filter", Object::reference(crate::ObjectId::new(9, 0)))]);
        assert!(FilterSpec::from_dict(&indirect).is_err());
        let bad_entry = parms(&[("Filter", Object::array([Object::Integer(1)]))]);
        assert!(FilterSpec::from_dict(&bad_entry).is_err());
    }

    #[test]
    fn chains_apply_in_order() {
        // "Hello" -> ASCII85 -> and back again through a one-filter chain.
        let encoded = b"87cURDZ~>";
        let chain = [FilterSpec {
            filter: Filter::Ascii85Decode,
            parms: None,
        }];
        assert_eq!(decode_chain(encoded, &chain).expect("chain"), b"Hello");
        assert_eq!(decode_chain(b"Hello", &[]).expect("empty chain"), b"Hello");
    }

    #[test]
    fn predictor_defaults_are_inert() {
        assert_eq!(
            Predictor::from_dict(None).expect("defaults"),
            Predictor::default()
        );
        assert_eq!(
            apply_predictor(b"abc", Predictor::default()).expect("none"),
            b"abc"
        );
    }
}