harumi 0.2.2

Overlay searchable CJK text on PDFs, extract text, merge/split pages — pure Rust, zero C dependencies
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

use lopdf::{Dictionary, Object, ObjectId, Stream};

use crate::error::{Error, Result};

pub(crate) struct PreparedImage {
    pub width: u32,
    pub height: u32,
    pub data: ImageData,
}

pub(crate) enum ImageData {
    /// Original JPEG bytes — embedded as-is with DCTDecode.
    Jpeg(Vec<u8>),
    /// Decoded raw RGB bytes (3 bytes/pixel, top-to-bottom, left-to-right). Fully opaque.
    Rgb { bytes: Vec<u8> },
    /// Decoded raw RGB bytes plus a separate alpha channel (1 byte/pixel).
    /// Embedded as an Image XObject with a DeviceGray SMask sub-object.
    RgbWithAlpha { rgb: Vec<u8>, alpha: Vec<u8> },
}

/// Prepare an image for PDF embedding.
///
/// JPEG files are embedded without decoding (DCTDecode filter).
/// Other formats are decoded via the `image` crate:
/// - Fully opaque PNG → `ImageData::Rgb` (no SMask needed)
/// - PNG with any transparent pixel → `ImageData::RgbWithAlpha` (PDF SMask)
pub(crate) fn prepare(bytes: &[u8]) -> Result<PreparedImage> {
    if bytes.starts_with(b"\xff\xd8\xff") {
        let (w, h) = parse_jpeg_dims(bytes)?;
        return Ok(PreparedImage { width: w, height: h, data: ImageData::Jpeg(bytes.to_vec()) });
    }

    // Check dimensions before decoding to prevent OOM on crafted inputs.
    // `into_dimensions()` reads only the image header — no pixel allocation.
    let (w, h) = image::ImageReader::new(std::io::Cursor::new(bytes))
        .with_guessed_format()
        .map_err(|e| Error::ImageDecode(e.to_string()))?
        .into_dimensions()
        .map_err(|e| Error::ImageDecode(e.to_string()))?;
    let pixel_count = w as u64 * h as u64;
    if pixel_count > 200_000_000 {
        return Err(Error::InvalidInput(format!(
            "image too large: {w}x{h} = {} pixels (limit 200 MP)", pixel_count
        )));
    }

    let img = image::load_from_memory(bytes)
        .map_err(|e| Error::ImageDecode(e.to_string()))?;
    let rgba = img.to_rgba8();

    let has_alpha = rgba.pixels().any(|p| p[3] < 255);
    if has_alpha {
        let mut rgb   = Vec::with_capacity((pixel_count * 3) as usize);
        let mut alpha = Vec::with_capacity(pixel_count as usize);
        for p in rgba.pixels() {
            rgb.extend_from_slice(&[p[0], p[1], p[2]]);
            alpha.push(p[3]);
        }
        Ok(PreparedImage { width: w, height: h, data: ImageData::RgbWithAlpha { rgb, alpha } })
    } else {
        let mut rgb = Vec::with_capacity((pixel_count * 3) as usize);
        for p in rgba.pixels() {
            rgb.extend_from_slice(&[p[0], p[1], p[2]]);
        }
        Ok(PreparedImage { width: w, height: h, data: ImageData::Rgb { bytes: rgb } })
    }
}

/// Add an Image XObject to the lopdf document and return its object ID.
pub(crate) fn embed_xobject(doc: &mut lopdf::Document, img: PreparedImage) -> Result<ObjectId> {
    match img.data {
        ImageData::Jpeg(bytes) => {
            let mut dict = Dictionary::new();
            dict.set("Type", Object::Name(b"XObject".to_vec()));
            dict.set("Subtype", Object::Name(b"Image".to_vec()));
            dict.set("Width", Object::Integer(img.width as i64));
            dict.set("Height", Object::Integer(img.height as i64));
            dict.set("ColorSpace", Object::Name(b"DeviceRGB".to_vec()));
            dict.set("BitsPerComponent", Object::Integer(8));
            dict.set("Filter", Object::Name(b"DCTDecode".to_vec()));
            Ok(doc.add_object(Object::Stream(Stream::new(dict, bytes))))
        }
        ImageData::Rgb { bytes } => {
            let mut dict = Dictionary::new();
            dict.set("Type", Object::Name(b"XObject".to_vec()));
            dict.set("Subtype", Object::Name(b"Image".to_vec()));
            dict.set("Width", Object::Integer(img.width as i64));
            dict.set("Height", Object::Integer(img.height as i64));
            dict.set("ColorSpace", Object::Name(b"DeviceRGB".to_vec()));
            dict.set("BitsPerComponent", Object::Integer(8));
            let mut stream = Stream::new(dict, bytes);
            let _ = stream.compress();
            Ok(doc.add_object(Object::Stream(stream)))
        }
        ImageData::RgbWithAlpha { rgb, alpha } => {
            // SMask sub-object: grayscale image carrying the alpha channel.
            // Not registered in /Resources — referenced only by the main image dict.
            let mut smask_dict = Dictionary::new();
            smask_dict.set("Type",             Object::Name(b"XObject".to_vec()));
            smask_dict.set("Subtype",          Object::Name(b"Image".to_vec()));
            smask_dict.set("Width",            Object::Integer(img.width as i64));
            smask_dict.set("Height",           Object::Integer(img.height as i64));
            smask_dict.set("ColorSpace",       Object::Name(b"DeviceGray".to_vec()));
            smask_dict.set("BitsPerComponent", Object::Integer(8));
            let mut smask_stream = Stream::new(smask_dict, alpha);
            let _ = smask_stream.compress();
            let smask_id = doc.add_object(Object::Stream(smask_stream));

            let mut dict = Dictionary::new();
            dict.set("Type",             Object::Name(b"XObject".to_vec()));
            dict.set("Subtype",          Object::Name(b"Image".to_vec()));
            dict.set("Width",            Object::Integer(img.width as i64));
            dict.set("Height",           Object::Integer(img.height as i64));
            dict.set("ColorSpace",       Object::Name(b"DeviceRGB".to_vec()));
            dict.set("BitsPerComponent", Object::Integer(8));
            dict.set("SMask",            Object::Reference(smask_id));
            let mut stream = Stream::new(dict, rgb);
            let _ = stream.compress();
            Ok(doc.add_object(Object::Stream(stream)))
        }
    }
}

/// Returns the content stream fragment that renders the image at `rect`.
///
/// `rect` = `[x, y, width, height]` in PDF points (origin bottom-left).
/// The PDF cm operator maps the unit square to `rect`.
pub(crate) fn image_stream(xobj_name: &str, rect: &[f32; 4], gs_name: &str) -> Vec<u8> {
    format!(
        "q\n/{gs} gs\n{w:.4} 0 0 {h:.4} {x:.4} {y:.4} cm\n/{name} Do\nQ\n",
        gs = gs_name,
        w = rect[2], h = rect[3],
        x = rect[0], y = rect[1],
        name = xobj_name,
    )
    .into_bytes()
}

/// Parse JPEG dimensions by scanning SOF markers (FF C0–CF, except DHT/DAC/etc.).
fn parse_jpeg_dims(data: &[u8]) -> Result<(u32, u32)> {
    let mut i = 2; // skip SOI marker (FF D8)
    while i < data.len() {
        if data[i] != 0xFF {
            return Err(Error::ImageDecode("malformed JPEG: expected marker".into()));
        }
        // Skip 0xFF fill bytes (JPEG spec §B.1.1.2: any number of 0xFF may precede a marker).
        while i + 1 < data.len() && data[i + 1] == 0xFF {
            i += 1;
        }
        if i + 1 >= data.len() {
            break;
        }
        let marker = data[i + 1];
        // SOF markers carry image dimensions; skip others.
        if matches!(
            marker,
            0xC0 | 0xC1 | 0xC2 | 0xC3 | 0xC5 | 0xC6 | 0xC7 | 0xC9 | 0xCA | 0xCB | 0xCD | 0xCE
                | 0xCF
        ) && i + 8 < data.len()
        {
            let h = u16::from_be_bytes([data[i + 5], data[i + 6]]) as u32;
            let w = u16::from_be_bytes([data[i + 7], data[i + 8]]) as u32;
            if w > 0 && h > 0 {
                return Ok((w, h));
            }
        }
        // Stand-alone markers (RST0-RST7, SOI, EOI, TEM) carry no length field.
        if matches!(marker, 0xD0..=0xD9 | 0x01) {
            i += 2;
            continue;
        }
        if i + 3 >= data.len() {
            break;
        }
        let seg_len = u16::from_be_bytes([data[i + 2], data[i + 3]]) as usize;
        if seg_len < 2 {
            break;
        }
        i += 2 + seg_len;
    }
    Err(Error::ImageDecode("JPEG: could not find SOF marker with valid dimensions".into()))
}

#[cfg(test)]
mod tests {
    use super::*;

    fn make_jpeg_with_sof(w: u16, h: u16, fill_bytes: usize) -> Vec<u8> {
        // SOI + optional 0xFF fill padding + SOF0 + EOI
        let mut data = vec![0xFF, 0xD8]; // SOI
        // APP0-like junk segment so the parser skips it
        data.extend_from_slice(&[0xFF, 0xE0, 0x00, 0x10]); // marker + length=16
        data.extend_from_slice(&[0u8; 14]); // padding up to length
        // 0xFF fill bytes before SOF0 (the Mi-3 case)
        for _ in 0..fill_bytes {
            data.push(0xFF);
        }
        // SOF0: FF C0 len(2) precision(1) height(2) width(2) components(1)
        data.extend_from_slice(&[0xFF, 0xC0, 0x00, 0x11, 0x08]);
        data.extend_from_slice(&h.to_be_bytes());
        data.extend_from_slice(&w.to_be_bytes());
        data.push(0x03); // 3 components
        data.extend_from_slice(&[0u8; 12]); // component data
        data.extend_from_slice(&[0xFF, 0xD9]); // EOI
        data
    }

    #[test]
    fn jpeg_dims_no_fill_bytes() {
        let data = make_jpeg_with_sof(640, 480, 0);
        assert_eq!(parse_jpeg_dims(&data).unwrap(), (640, 480));
    }

    #[test]
    fn jpeg_dims_with_fill_bytes() {
        // JPEG spec §B.1.1.2: 0xFF padding before the marker byte is legal
        let data = make_jpeg_with_sof(320, 240, 3);
        assert_eq!(parse_jpeg_dims(&data).unwrap(), (320, 240));
    }
}