pdf-syntax 0.5.0

A low-level crate for reading PDF files.
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
//! Reading the pages of a PDF document.

use crate::content::{TypedIter, UntypedIter};
use crate::object::Array;
use crate::object::Dict;
use crate::object::Name;
use crate::object::Rect;
use crate::object::Stream;
use crate::object::dict::keys::*;
use crate::object::{Object, ObjectLike};
use crate::reader::ReaderContext;
use crate::sync::OnceLock;
use crate::util::FloatExt;
use crate::xref::XRef;
use alloc::boxed::Box;
use alloc::vec;
use alloc::vec::Vec;
use core::ops::Deref;
use log::warn;

/// Attributes that can be inherited.
#[derive(Debug, Clone)]
struct PagesContext {
    media_box: Option<Rect>,
    crop_box: Option<Rect>,
    rotate: Option<u32>,
}

impl PagesContext {
    fn new() -> Self {
        Self {
            media_box: None,
            crop_box: None,
            rotate: None,
        }
    }
}

/// A structure holding the pages of a PDF document.
pub struct Pages<'a> {
    pages: Vec<Page<'a>>,
    xref: &'a XRef,
}

impl<'a> Pages<'a> {
    /// Create a new `Pages` object.
    pub(crate) fn new(
        pages_dict: &Dict<'a>,
        ctx: &ReaderContext<'a>,
        xref: &'a XRef,
    ) -> Option<Self> {
        let mut pages = vec![];
        let pages_ctx = PagesContext::new();
        resolve_pages(
            pages_dict,
            &mut pages,
            pages_ctx,
            Resources::new(Dict::empty(), None, ctx),
        )?;

        Some(Self { pages, xref })
    }

    /// Create a new `Pages` object by bruteforce-searching.
    ///
    /// Of course this could result in the order of pages being messed up, but
    /// this is still better than nothing.
    pub(crate) fn new_brute_force(ctx: &ReaderContext<'a>, xref: &'a XRef) -> Option<Self> {
        let mut pages = vec![];

        for object in xref.objects() {
            if let Some(dict) = object.into_dict()
                && let Some(page) = Page::new(
                    &dict,
                    &PagesContext::new(),
                    Resources::new(Dict::empty(), None, ctx),
                    true,
                )
            {
                pages.push(page);
            }
        }

        if pages.is_empty() {
            return None;
        }

        Some(Self { pages, xref })
    }

    /// Return the xref table (of the document the pages belong to).   
    pub fn xref(&self) -> &'a XRef {
        self.xref
    }
}

impl<'a> Deref for Pages<'a> {
    type Target = [Page<'a>];

    fn deref(&self) -> &Self::Target {
        &self.pages
    }
}

/// Maximum depth for recursive page tree traversal.
/// Prevents stack overflow on malformed PDFs with deeply nested or circular page trees.
const MAX_PAGE_TREE_DEPTH: usize = 256;

/// Maximum number of pages to collect from the page tree.
/// Prevents exponential blowup from "page tree bombs" where shared Kids nodes
/// cause the same subtrees to be visited multiple times (e.g., 8 KB PDF → 2^27 pages).
const MAX_PAGE_COUNT: usize = 100_000;

fn resolve_pages<'a>(
    pages_dict: &Dict<'a>,
    entries: &mut Vec<Page<'a>>,
    ctx: PagesContext,
    resources: Resources<'a>,
) -> Option<()> {
    resolve_pages_depth(pages_dict, entries, ctx, resources, 0)
}

fn resolve_pages_depth<'a>(
    pages_dict: &Dict<'a>,
    entries: &mut Vec<Page<'a>>,
    mut ctx: PagesContext,
    resources: Resources<'a>,
    depth: usize,
) -> Option<()> {
    if depth > MAX_PAGE_TREE_DEPTH {
        log::warn!("Page tree depth exceeds {MAX_PAGE_TREE_DEPTH}, stopping traversal");
        return None;
    }

    if let Some(media_box) = pages_dict.get::<Rect>(MEDIA_BOX) {
        ctx.media_box = Some(media_box);
    }

    if let Some(crop_box) = pages_dict.get::<Rect>(CROP_BOX) {
        ctx.crop_box = Some(crop_box);
    }

    // /Rotate may be negative (e.g. -90), which is valid per the PDF spec.
    // Normalise to [0, 360) via Euclidean remainder so u32 storage is safe.
    if let Some(rotate) = pages_dict.get::<i32>(ROTATE) {
        ctx.rotate = Some(rotate.rem_euclid(360) as u32);
    }

    let resources = Resources::from_parent(
        pages_dict.get::<Dict<'_>>(RESOURCES).unwrap_or_default(),
        resources.clone(),
    );

    let kids = pages_dict.get::<Array<'a>>(KIDS)?;

    for dict in kids.iter::<Dict<'_>>() {
        if entries.len() >= MAX_PAGE_COUNT {
            log::warn!("Page count exceeds {MAX_PAGE_COUNT}, stopping page tree traversal");
            return Some(());
        }

        match dict.get::<Name>(TYPE).as_deref() {
            Some(PAGES) => {
                resolve_pages_depth(&dict, entries, ctx.clone(), resources.clone(), depth + 1);
            }
            // Let's be lenient and assume it's a `Page` in case it's `None` or something else
            // (see corpus test case 0083781).
            _ => {
                if let Some(page) = Page::new(&dict, &ctx, resources.clone(), false) {
                    entries.push(page);
                }
            }
        }
    }

    Some(())
}

/// The rotation of the page.
#[derive(Debug, Copy, Clone)]
pub enum Rotation {
    /// No rotation.
    None,
    /// A rotation of 90 degrees.
    Horizontal,
    /// A rotation of 180 degrees.
    Flipped,
    /// A rotation of 270 degrees.
    FlippedHorizontal,
}

/// A PDF page.
pub struct Page<'a> {
    inner: Dict<'a>,
    media_box: Rect,
    crop_box: Rect,
    rotation: Rotation,
    page_streams: OnceLock<Option<Vec<u8>>>,
    resources: Resources<'a>,
    ctx: ReaderContext<'a>,
}

impl<'a> Page<'a> {
    fn new(
        dict: &Dict<'a>,
        ctx: &PagesContext,
        resources: Resources<'a>,
        brute_force: bool,
    ) -> Option<Self> {
        // In general, pages without content are allowed, but in case we are brute-forcing
        // we ignore them.
        if brute_force && !dict.contains_key(CONTENTS) {
            return None;
        }

        let media_box = dict
            .get::<Rect>(MEDIA_BOX)
            .or(ctx.media_box)
            .unwrap_or(US_LETTER);

        let crop_box = dict
            .get::<Rect>(CROP_BOX)
            .or(ctx.crop_box)
            .unwrap_or(media_box);

        let rotation = match dict
            .get::<i32>(ROTATE)
            .map(|r| r.rem_euclid(360) as u32)
            .or(ctx.rotate)
            .unwrap_or(0)
        {
            0 => Rotation::None,
            90 => Rotation::Horizontal,
            180 => Rotation::Flipped,
            270 => Rotation::FlippedHorizontal,
            _ => Rotation::None,
        };

        let ctx = resources.ctx.clone();
        let resources = Resources::from_parent(
            dict.get::<Dict<'_>>(RESOURCES).unwrap_or_default(),
            resources,
        );

        Some(Self {
            inner: dict.clone(),
            media_box,
            crop_box,
            rotation,
            page_streams: OnceLock::new(),
            resources,
            ctx,
        })
    }

    fn operations_impl(&self) -> Option<UntypedIter<'_>> {
        let stream = self.page_stream()?;
        let iter = UntypedIter::new(stream);

        Some(iter)
    }

    /// Return the decoded content stream of the page.
    pub fn page_stream(&self) -> Option<&[u8]> {
        let convert_single = |s: Stream<'_>| {
            let data = s.decoded().ok()?;
            Some(data.to_vec())
        };

        self.page_streams
            .get_or_init(|| {
                if let Some(stream) = self.inner.get::<Stream<'_>>(CONTENTS) {
                    convert_single(stream)
                } else if let Some(array) = self.inner.get::<Array<'_>>(CONTENTS) {
                    let streams = array.iter::<Stream<'_>>().flat_map(convert_single);

                    let mut collected = vec![];

                    for stream in streams {
                        collected.extend(stream);
                        // Streams must have at least one whitespace in-between.
                        collected.push(b' ');
                    }

                    Some(collected)
                } else {
                    warn!("contents entry of page was neither stream nor array of streams");

                    None
                }
            })
            .as_ref()
            .map(|d| d.as_slice())
    }

    /// Get the resources of the page.
    pub fn resources(&self) -> &Resources<'a> {
        &self.resources
    }

    /// Get the media box of the page.
    pub fn media_box(&self) -> Rect {
        self.media_box
    }

    /// Get the rotation of the page.
    pub fn rotation(&self) -> Rotation {
        self.rotation
    }

    /// Get the crop box of the page.
    pub fn crop_box(&self) -> Rect {
        self.crop_box
    }

    /// Return the intersection of crop box and media box.
    pub fn intersected_crop_box(&self) -> Rect {
        self.crop_box().intersect(self.media_box())
    }

    /// Return the base dimensions of the page used for the canvas size.
    ///
    /// When the CropBox origin is within the MediaBox (i.e. CropBox.x0 >=
    /// MediaBox.x0 and CropBox.y0 >= MediaBox.y0), the canvas is sized to
    /// intersect(CropBox, MediaBox).  This matches MuPDF's behaviour for
    /// spec-violating PDFs where CropBox.y1 > MediaBox.y1 (e.g. gen-271:
    /// CropBox=[0,0,595,793.7] vs MediaBox=[0,0,612,792] — using raw
    /// CropBox gives 1654px height, a 4px vertical content offset, and
    /// SSIM 0.49; intersecting gives 1650px matching MuPDF exactly).
    ///
    /// When CropBox extends below the MediaBox origin (gen-802 style:
    /// CropBox=[0,0,684,864] vs MediaBox=[36,36,648,828]), MuPDF still uses
    /// the full CropBox dimensions, so we do too. (#544, #558, gen-271)
    pub fn base_dimensions(&self) -> (f32, f32) {
        let crop_box = self.crop_box();
        let media_box = self.media_box();

        // Clip to MediaBox only when the CropBox origin lies within the
        // MediaBox (both axes).  When the CropBox extends below the MediaBox
        // origin MuPDF uses the raw CropBox, so preserve that behaviour.
        let effective = if crop_box.x0 >= media_box.x0 && crop_box.y0 >= media_box.y0 {
            crop_box.intersect(media_box)
        } else {
            crop_box
        };

        if (effective.width() as f32).is_nearly_zero()
            || (effective.height() as f32).is_nearly_zero()
        {
            (US_LETTER.width() as f32, US_LETTER.height() as f32)
        } else {
            (
                effective.width().max(1.0) as f32,
                effective.height().max(1.0) as f32,
            )
        }
    }

    /// Return the with and height of the page that should be assumed when rendering the page.
    ///
    /// Depending on the document, it is either based on the media box or the crop box
    /// of the page. In addition to that, it also takes the rotation of the page into account.
    pub fn render_dimensions(&self) -> (f32, f32) {
        let (mut base_width, mut base_height) = self.base_dimensions();

        if matches!(
            self.rotation(),
            Rotation::Horizontal | Rotation::FlippedHorizontal
        ) {
            core::mem::swap(&mut base_width, &mut base_height);
        }

        (base_width, base_height)
    }

    /// Return an untyped iterator over the operators of the page's content stream.
    pub fn operations(&self) -> UntypedIter<'_> {
        self.operations_impl().unwrap_or(UntypedIter::empty())
    }

    /// Get the raw dictionary of the page.
    pub fn raw(&self) -> &Dict<'a> {
        &self.inner
    }

    /// Get the xref table (of the document the page belongs to).
    pub fn xref(&self) -> &'a XRef {
        self.ctx.xref()
    }

    /// Return a typed iterator over the operators of the page's content stream.
    pub fn typed_operations(&self) -> TypedIter<'_> {
        TypedIter::from_untyped(self.operations())
    }

    /// Return the annotation dictionaries for this page, if any.
    pub fn annots(&self) -> Vec<Dict<'a>> {
        self.inner
            .get::<Array<'_>>(crate::object::dict::keys::ANNOTS)
            .map(|arr| arr.iter::<Dict<'_>>().collect())
            .unwrap_or_default()
    }
}

/// A structure keeping track of the resources of a page.
#[derive(Clone, Debug)]
pub struct Resources<'a> {
    parent: Option<Box<Self>>,
    ctx: ReaderContext<'a>,
    /// The raw dictionary of external graphics states.
    pub ext_g_states: Dict<'a>,
    /// The raw dictionary of fonts.
    pub fonts: Dict<'a>,
    /// The raw dictionary of properties.
    pub properties: Dict<'a>,
    /// The raw dictionary of color spaces.
    pub color_spaces: Dict<'a>,
    /// The raw dictionary of x objects.
    pub x_objects: Dict<'a>,
    /// The raw dictionary of patterns.
    pub patterns: Dict<'a>,
    /// The raw dictionary of shadings.
    pub shadings: Dict<'a>,
}

impl<'a> Resources<'a> {
    /// Create a new `Resources` object from a dictionary with a parent.
    pub fn from_parent(resources: Dict<'a>, parent: Self) -> Self {
        let ctx = parent.ctx.clone();

        Self::new(resources, Some(parent), &ctx)
    }

    /// Create a new `Resources` object.
    pub(crate) fn new(resources: Dict<'a>, parent: Option<Self>, ctx: &ReaderContext<'a>) -> Self {
        let ext_g_states = resources.get::<Dict<'_>>(EXT_G_STATE).unwrap_or_default();
        let fonts = resources.get::<Dict<'_>>(FONT).unwrap_or_default();
        let color_spaces = resources.get::<Dict<'_>>(COLORSPACE).unwrap_or_default();
        let x_objects = resources.get::<Dict<'_>>(XOBJECT).unwrap_or_default();
        let patterns = resources.get::<Dict<'_>>(PATTERN).unwrap_or_default();
        let shadings = resources.get::<Dict<'_>>(SHADING).unwrap_or_default();
        let properties = resources.get::<Dict<'_>>(PROPERTIES).unwrap_or_default();

        let parent = parent.map(Box::new);

        Self {
            parent,
            ext_g_states,
            fonts,
            color_spaces,
            properties,
            x_objects,
            patterns,
            shadings,
            ctx: ctx.clone(),
        }
    }

    fn get_resource<T: ObjectLike<'a>>(&self, name: Name, dict: &Dict<'a>) -> Option<T> {
        dict.get::<T>(name.deref())
    }

    /// Get the parent in the resource, chain, if available.
    pub fn parent(&self) -> Option<&Self> {
        self.parent.as_deref()
    }

    /// Get an external graphics state by name.
    pub fn get_ext_g_state(&self, name: Name) -> Option<Dict<'a>> {
        self.get_resource::<Dict<'_>>(name.clone(), &self.ext_g_states)
            .or_else(|| self.parent.as_ref().and_then(|p| p.get_ext_g_state(name)))
    }

    /// Get a color space by name.
    pub fn get_color_space(&self, name: Name) -> Option<Object<'a>> {
        self.get_resource::<Object<'_>>(name.clone(), &self.color_spaces)
            .or_else(|| self.parent.as_ref().and_then(|p| p.get_color_space(name)))
    }

    /// Get a font by name.
    pub fn get_font(&self, name: Name) -> Option<Dict<'a>> {
        self.get_resource::<Dict<'_>>(name.clone(), &self.fonts)
            .or_else(|| self.parent.as_ref().and_then(|p| p.get_font(name)))
    }

    /// Get a pattern by name.
    pub fn get_pattern(&self, name: Name) -> Option<Object<'a>> {
        self.get_resource::<Object<'_>>(name.clone(), &self.patterns)
            .or_else(|| self.parent.as_ref().and_then(|p| p.get_pattern(name)))
    }

    /// Get an x object by name.
    pub fn get_x_object(&self, name: Name) -> Option<Stream<'a>> {
        self.get_resource::<Stream<'_>>(name.clone(), &self.x_objects)
            .or_else(|| self.parent.as_ref().and_then(|p| p.get_x_object(name)))
    }

    /// Get a shading by name.
    pub fn get_shading(&self, name: Name) -> Option<Object<'a>> {
        self.get_resource::<Object<'_>>(name.clone(), &self.shadings)
            .or_else(|| self.parent.as_ref().and_then(|p| p.get_shading(name)))
    }
}

// <https://github.com/apache/pdfbox/blob/a53a70db16ea3133994120bcf1e216b9e760c05b/pdfbox/src/main/java/org/apache/pdfbox/pdmodel/common/PDRectangle.java#L38>
const POINTS_PER_INCH: f64 = 72.0;
const POINTS_PER_MM: f64 = 1.0 / (10.0 * 2.54) * POINTS_PER_INCH;

/// The dimension of an A4 page (kept for completeness).
pub const A4: Rect = Rect {
    x0: 0.0,
    y0: 0.0,
    x1: 210.0 * POINTS_PER_MM,
    y1: 297.0 * POINTS_PER_MM,
};

/// US Letter (8.5×11 in) — used as fallback when no MediaBox is present.
///
/// Old PDF 1.0/1.1 documents (especially US-government publications from the
/// 1990s) omit MediaBox entirely and assume a US-Letter canvas, which matches
/// the PostScript default and MuPDF's behaviour.  Using A4 here causes a
/// ~5 % scale mismatch that tanks SSIM against MuPDF's oracle renders.
/// See render_mupdf_oracle failures gen-059, gen-069, …  (#544)
const US_LETTER: Rect = Rect {
    x0: 0.0,
    y0: 0.0,
    x1: 8.5 * POINTS_PER_INCH,
    y1: 11.0 * POINTS_PER_INCH,
};

pub(crate) mod cached {
    use crate::page::Pages;
    use crate::reader::ReaderContext;
    use crate::xref::XRef;
    use core::ops::Deref;

    // Keep in sync with the implementation in `sync`. We duplicate it here
    // to make it more visible since we have unsafe code here.
    #[cfg(feature = "std")]
    pub(crate) use std::sync::Arc;

    #[cfg(not(feature = "std"))]
    pub(crate) use alloc::rc::Rc as Arc;

    pub(crate) struct CachedPages {
        pages: Pages<'static>,
        // NOTE: `pages` references the data in `xref`, so it's important that `xref`
        // appears after `pages` in the struct definition to ensure correct drop order.
        _xref: Arc<XRef>,
    }

    impl CachedPages {
        pub(crate) fn new(xref: Arc<XRef>) -> Option<Self> {
            // SAFETY:
            // - The XRef's location is stable in memory:
            //   - We wrapped it in a `Arc` (or `Rc` in `no_std`), which implements `StableDeref`.
            //   - The struct owns the `Arc`, ensuring that the inner value is not dropped during the whole
            //     duration.
            // - The internal 'static lifetime is not leaked because its rewritten
            //   to the self-lifetime in `pages()`.
            let xref_reference: &'static XRef = unsafe { core::mem::transmute(xref.deref()) };

            let ctx = ReaderContext::new(xref_reference, false);
            let pages = xref_reference
                .get_with(xref.trailer_data().pages_ref, &ctx)
                .and_then(|p| Pages::new(&p, &ctx, xref_reference))
                .or_else(|| Pages::new_brute_force(&ctx, xref_reference))?;

            Some(Self { pages, _xref: xref })
        }

        pub(crate) fn get(&self) -> &Pages<'_> {
            &self.pages
        }
    }
}