xberg-native-pdf 1.2.7

PDF parsing, extraction and rendering -- the native PDF engine behind xberg.
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

//! Regression coverage for `gather_page_signals`' image classification (GH#1732).
//!
//! Lives beside `open.rs` rather than in `document/tests.rs` (322 KB, close to the
//! project's 500 KiB file-safety limit) for the same reason `pdf/native/images/
//! parallel_tests.rs` gives in the `xberg` crate: it keeps that file from growing
//! further while still sitting next to the code it covers.

use super::PdfDocument;
use crate::extractors::auto::ImageCodecClass;
use crate::extractors::images::decode_call_count;

/// A one-page PDF, `page_side`×`page_side`, with a single uncompressed DeviceRGB
/// image XObject painted at `(inset, inset)` with size `image_side`×`image_side`
/// via `cm`. Uncompressed (no `/Filter`) so the image decodes through the same
/// per-pixel raw-buffer path CCITT/JBIG2/Flate images do (the expensive path
/// GH#1732 is about), without needing a real compressed codestream fixture.
fn build_single_page_pdf_with_raw_image(page_side: u32, image_side: u32, inset: u32) -> Vec<u8> {
    let mut buf = Vec::<u8>::new();
    buf.extend_from_slice(b"%PDF-1.4\n");
    let mut offsets = Vec::new();

    offsets.push(buf.len());
    buf.extend_from_slice(b"1 0 obj\n<</Type /Catalog /Pages 2 0 R>>\nendobj\n");

    offsets.push(buf.len());
    buf.extend_from_slice(b"2 0 obj\n<</Type /Pages /Kids [3 0 R] /Count 1>>\nendobj\n");

    offsets.push(buf.len());
    buf.extend_from_slice(
        format!(
            "3 0 obj\n<</Type /Page /Parent 2 0 R /MediaBox [0 0 {page_side} {page_side}] \
             /Resources <</XObject <</Im0 5 0 R>>>> /Contents 4 0 R>>\nendobj\n"
        )
        .as_bytes(),
    );

    let content = format!("q {image_side} 0 0 {image_side} {inset} {inset} cm /Im0 Do Q");
    offsets.push(buf.len());
    buf.extend_from_slice(format!("4 0 obj\n<</Length {}>>\nstream\n", content.len()).as_bytes());
    buf.extend_from_slice(content.as_bytes());
    buf.extend_from_slice(b"\nendstream\nendobj\n");

    let pixel_count = (image_side as usize) * (image_side as usize) * 3;
    offsets.push(buf.len());
    buf.extend_from_slice(
        format!(
            "5 0 obj\n<</Type /XObject /Subtype /Image /Width {image_side} /Height {image_side} \
             /ColorSpace /DeviceRGB /BitsPerComponent 8 /Length {pixel_count}>>\nstream\n"
        )
        .as_bytes(),
    );
    buf.extend((0..pixel_count).map(|i| ((i * 37 + i / 7) % 251) as u8));
    buf.extend_from_slice(b"\nendstream\nendobj\n");

    let xref_offset = buf.len();
    let total_objs = 6;
    buf.extend_from_slice(b"xref\n");
    buf.extend_from_slice(format!("0 {total_objs}\n").as_bytes());
    buf.extend_from_slice(b"0000000000 65535 f \n");
    for off in &offsets {
        buf.extend_from_slice(format!("{off:010} 00000 n \n").as_bytes());
    }
    buf.extend_from_slice(format!("trailer\n<</Size {total_objs} /Root 1 0 R>>\n").as_bytes());
    buf.extend_from_slice(format!("startxref\n{xref_offset}\n%%EOF\n").as_bytes());
    buf
}

/// `classify_page` (via `gather_page_signals`) must decode zero embedded images: it
/// only needs bbox + filter-chain metadata, both available from the cheap Phase 1
/// `page_image_handles` walk. Before the GH#1732 fix this called `extract_images`,
/// which fully decoded every embedded image just to classify its codec.
///
/// Asserts a *delta* around each call rather than an absolute counter value so the
/// test is immune to other tests incrementing the same process-wide-if-it-were-an-
/// AtomicUsize counter; `decode_call_count` is thread-local instead; see its doc
/// comment. The second half proves the counter itself is wired to real decode work
/// (`extract_images` must move it) rather than reading 0 vacuously either way. ~keep
#[test]
fn classify_page_does_not_decode_embedded_images() {
    let pdf = build_single_page_pdf_with_raw_image(200, 150, 25);
    let doc = PdfDocument::from_bytes(pdf).expect("fixture must open");

    let before_classify = decode_call_count();
    let classification = doc.classify_page(0).expect("classify_page must succeed");
    let after_classify = decode_call_count();
    assert_eq!(
        after_classify,
        before_classify,
        "classify_page must not decode any embedded image, got a decode-count delta of {}",
        after_classify - before_classify
    );

    // The page's image still contributes its area/codec to the signals -- avoiding
    // the decode must not make the image invisible to classification.
    assert_eq!(classification.signals.codec, ImageCodecClass::Other);

    let before_extract = decode_call_count();
    let images = doc.extract_images(0).expect("extract_images must succeed");
    let after_extract = decode_call_count();
    assert_eq!(images.len(), 1, "fixture has exactly one embedded image");
    assert_eq!(
        after_extract - before_extract,
        1,
        "extract_images must decode exactly the one embedded image on the page \
         (proves decode_call_count is wired to real decode work, not vacuously 0)"
    );
}

/// `image_area_ratio` computed from the Phase 1 handle's `bbox` (GH#1732's
/// no-decode path) must equal the value the old full-decode path produced: the
/// image is painted at (25,25) sized 150x150 on a 200x200 page, so the
/// intersection area is exactly 150*150 = 22500 and the ratio is 22500/40000.
#[test]
fn classify_page_image_area_ratio_matches_expected_bbox_intersection() {
    let pdf = build_single_page_pdf_with_raw_image(200, 150, 25);
    let doc = PdfDocument::from_bytes(pdf).expect("fixture must open");

    let classification = doc.classify_page(0).expect("classify_page must succeed");
    let expected_ratio = (150.0 * 150.0) / (200.0 * 200.0);
    assert!(
        (classification.signals.image_area_ratio - expected_ratio).abs() < 1e-5,
        "expected image_area_ratio {expected_ratio}, got {}",
        classification.signals.image_area_ratio
    );
}

/// A one-page PDF with no `/XObject` resource and an empty content stream.
fn build_single_page_pdf_without_images(page_side: u32) -> Vec<u8> {
    let mut buf = Vec::<u8>::new();
    buf.extend_from_slice(b"%PDF-1.4\n");
    let mut offsets = Vec::new();

    offsets.push(buf.len());
    buf.extend_from_slice(b"1 0 obj\n<</Type /Catalog /Pages 2 0 R>>\nendobj\n");

    offsets.push(buf.len());
    buf.extend_from_slice(b"2 0 obj\n<</Type /Pages /Kids [3 0 R] /Count 1>>\nendobj\n");

    offsets.push(buf.len());
    buf.extend_from_slice(
        format!(
            "3 0 obj\n<</Type /Page /Parent 2 0 R /MediaBox [0 0 {page_side} {page_side}] \
             /Resources <<>> /Contents 4 0 R>>\nendobj\n"
        )
        .as_bytes(),
    );

    offsets.push(buf.len());
    buf.extend_from_slice(b"4 0 obj\n<</Length 0>>\nstream\n\nendstream\nendobj\n");

    let xref_offset = buf.len();
    let total_objs = 5;
    buf.extend_from_slice(b"xref\n");
    buf.extend_from_slice(format!("0 {total_objs}\n").as_bytes());
    buf.extend_from_slice(b"0000000000 65535 f \n");
    for off in &offsets {
        buf.extend_from_slice(format!("{off:010} 00000 n \n").as_bytes());
    }
    buf.extend_from_slice(format!("trailer\n<</Size {total_objs} /Root 1 0 R>>\n").as_bytes());
    buf.extend_from_slice(format!("startxref\n{xref_offset}\n%%EOF\n").as_bytes());
    buf
}

/// A page with no embedded images must still classify without decoding anything,
/// and must report `ImageCodecClass::None` and a zero image area ratio.
#[test]
fn classify_page_with_no_images_reports_none_codec() {
    let pdf = build_single_page_pdf_without_images(200);
    let doc = PdfDocument::from_bytes(pdf).expect("fixture must open");

    let before = decode_call_count();
    let classification = doc.classify_page(0).expect("classify_page must succeed");
    let after = decode_call_count();
    assert_eq!(after, before, "a page with no images must still decode nothing");
    assert_eq!(classification.signals.codec, ImageCodecClass::None);
    assert_eq!(classification.signals.image_area_ratio, 0.0);
}