/src/kdegraphics-mobipocket/lib/mobipocket.cpp
Line | Count | Source |
1 | | // SPDX-FileCopyrightText: 2008 by Jakub Stachowski <qbast@go2.pl> |
2 | | // SPDX-License-Identifier: GPL-2.0-or-later |
3 | | |
4 | | #include "mobipocket.h" |
5 | | #include "decompressor.h" |
6 | | #include "pdb_p.h" |
7 | | #include "qmobipocket_debug.h" |
8 | | |
9 | | #include <QBuffer> |
10 | | #include <QIODevice> |
11 | | #include <QImageReader> |
12 | | #include <QRegularExpression> |
13 | | #include <QStringConverter> |
14 | | #include <QtEndian> |
15 | | |
16 | | namespace Mobipocket |
17 | | { |
18 | | |
19 | | struct DocumentPrivate |
20 | | { |
21 | | DocumentPrivate(QIODevice *d) |
22 | 0 | : pdb(d) |
23 | 0 | { |
24 | 0 | } |
25 | | PDB pdb; |
26 | | std::unique_ptr<Decompressor> dec; |
27 | | quint16 ntextrecords = 0; |
28 | | quint16 maxRecordSize = 0; |
29 | | bool valid = false; |
30 | | |
31 | | // number of first record holding image. Usually it is directly after end of text, but not always |
32 | | quint16 firstImageRecord = 0; |
33 | | QMap<Document::MetaKey, QString> metadata; |
34 | | QStringDecoder toUtf16; |
35 | | bool drm = false; |
36 | | quint32 extraflags = 0; |
37 | | |
38 | | // index of Thumbnail image in image list. May be specified in EXTH. |
39 | | int thumbnailIndex = -1; |
40 | | // index of Cover image in image list. May be specified in EXTH. |
41 | | int coverIndex = -1; |
42 | | |
43 | | void init(); |
44 | | void findFirstImage(); |
45 | | void parseEXTH(QByteArrayView data); |
46 | | void parseHtmlHead(const QString &data); |
47 | | }; |
48 | | |
49 | | void DocumentPrivate::parseHtmlHead(const QString &data) |
50 | 0 | { |
51 | 0 | static const QRegularExpression title(QLatin1String("<dc:title.*>(.*)</dc:title>"), |
52 | 0 | QRegularExpression::CaseInsensitiveOption | QRegularExpression::InvertedGreedinessOption); |
53 | 0 | static const QRegularExpression author(QLatin1String("<dc:creator.*>(.*)</dc:creator>"), |
54 | 0 | QRegularExpression::CaseInsensitiveOption | QRegularExpression::InvertedGreedinessOption); |
55 | 0 | static const QRegularExpression copyright(QLatin1String("<dc:rights.*>(.*)</dc:rights>"), |
56 | 0 | QRegularExpression::CaseInsensitiveOption | QRegularExpression::InvertedGreedinessOption); |
57 | 0 | static const QRegularExpression subject(QLatin1String("<dc:subject.*>(.*)</dc:subject>"), |
58 | 0 | QRegularExpression::CaseInsensitiveOption | QRegularExpression::InvertedGreedinessOption); |
59 | 0 | static const QRegularExpression description(QLatin1String("<dc:description.*>(.*)</dc:description>"), |
60 | 0 | QRegularExpression::CaseInsensitiveOption | QRegularExpression::InvertedGreedinessOption); |
61 | | |
62 | | // title could have been already taken from MOBI record |
63 | 0 | if (!metadata.contains(Document::Title)) { |
64 | 0 | if (const auto titleMatch = title.match(data); titleMatch.hasMatch()) |
65 | 0 | metadata[Document::Title] = titleMatch.captured(1); |
66 | 0 | } |
67 | 0 | if (const auto authorMatch = author.match(data); authorMatch.hasMatch()) |
68 | 0 | metadata[Document::Author] = authorMatch.captured(1); |
69 | 0 | if (const auto copyrightMatch = copyright.match(data); copyrightMatch.hasMatch()) |
70 | 0 | metadata[Document::Copyright] = copyrightMatch.captured(1); |
71 | 0 | if (const auto subjectMatch = subject.match(data); subjectMatch.hasMatch()) |
72 | 0 | metadata[Document::Subject] = subjectMatch.captured(1); |
73 | 0 | if (const auto descriptionMatch = description.match(data); descriptionMatch.hasMatch()) |
74 | 0 | metadata[Document::Description] = descriptionMatch.captured(1); |
75 | 0 | } |
76 | | |
77 | | namespace |
78 | | { |
79 | | const QVector<QByteArray> getHuffRecords(const PDB &pdb) |
80 | 0 | { |
81 | 0 | const QByteArray header = pdb.getRecord(0); |
82 | 0 | if (header[1] != 'H') { |
83 | 0 | return {}; |
84 | 0 | } |
85 | | |
86 | 0 | quint32 huff_ofs = qFromBigEndian<quint32>(header.constData() + 0x70); |
87 | 0 | quint32 huff_num = qFromBigEndian<quint32>(header.constData() + 0x74); |
88 | | |
89 | | // Check for overflow and out-of-bounds access |
90 | 0 | if (((huff_ofs + huff_num) < huff_num) || ((huff_ofs + huff_num) > pdb.recordCount())) { |
91 | 0 | return {}; |
92 | 0 | } |
93 | | |
94 | 0 | QVector<QByteArray> records(huff_num); |
95 | 0 | for (quint32 i = 0; i < huff_num; i++) { |
96 | 0 | if (auto r = pdb.getRecord(huff_ofs + i); r.isNull()) { |
97 | 0 | return {}; |
98 | 0 | } else { |
99 | 0 | records[i] = r; |
100 | 0 | } |
101 | 0 | } |
102 | 0 | return records; |
103 | 0 | }; |
104 | | } |
105 | | |
106 | | void DocumentPrivate::init() |
107 | 0 | { |
108 | 0 | quint32 encoding = 0; |
109 | |
|
110 | 0 | if (!pdb.isValid()) |
111 | 0 | return; |
112 | 0 | QByteArray mhead = pdb.getRecord(0); |
113 | 0 | if (mhead.isNull() || mhead.size() < 14) |
114 | 0 | return; |
115 | | |
116 | 0 | dec = Decompressor::create(mhead[1], getHuffRecords(pdb)); |
117 | 0 | if ((int)mhead[12] != 0 || (int)mhead[13] != 0) |
118 | 0 | drm = true; |
119 | 0 | if (!dec) |
120 | 0 | return; |
121 | | |
122 | 0 | ntextrecords = qFromBigEndian<quint16>(mhead.constData() + 8); |
123 | 0 | maxRecordSize = qFromBigEndian<quint16>(mhead.constData() + 10); |
124 | 0 | if (mhead.size() > 31) |
125 | 0 | encoding = qFromBigEndian<quint32>(mhead.constData() + 28); |
126 | 0 | if (encoding == 65001) { |
127 | 0 | toUtf16 = QStringDecoder(QStringDecoder::Utf8); |
128 | 0 | } else { |
129 | 0 | toUtf16 = QStringDecoder("windows-1252"); |
130 | 0 | if (!toUtf16.isValid()) { |
131 | 0 | qCWarning(QMOBIPOCKET_LOG) << "Text codec \"windows-1252\" not supported by Qt library, falling back to Latin1"; |
132 | 0 | toUtf16 = QStringDecoder(QStringConverter::Latin1); |
133 | 0 | } |
134 | 0 | } |
135 | 0 | if (mhead.size() >= 92) |
136 | 0 | parseEXTH(mhead); |
137 | |
|
138 | 0 | if (mhead.size() >= 244) { |
139 | 0 | quint32 exthoffs = qFromBigEndian<quint32>(mhead.constData() + 20); |
140 | 0 | if ((exthoffs + 16) > 244) { |
141 | 0 | extraflags = qFromBigEndian<quint32>(mhead.constData() + 240); |
142 | 0 | } |
143 | 0 | } |
144 | | |
145 | | // try getting metadata from HTML if nothing or only title was recovered from MOBI and EXTH records |
146 | 0 | if (metadata.size() < 2 && !drm) |
147 | 0 | parseHtmlHead(toUtf16(dec->decompress(pdb.getRecord(1)))); |
148 | 0 | valid = true; |
149 | 0 | } |
150 | | |
151 | | void DocumentPrivate::findFirstImage() |
152 | 0 | { |
153 | 0 | firstImageRecord = ntextrecords + 1; |
154 | 0 | while (firstImageRecord < pdb.recordCount()) { |
155 | 0 | QByteArray rec = pdb.getRecord(firstImageRecord); |
156 | 0 | if (rec.isNull()) |
157 | 0 | return; |
158 | 0 | QBuffer buf(&rec); |
159 | 0 | buf.open(QIODevice::ReadOnly); |
160 | 0 | QImageReader r(&buf); |
161 | 0 | if (r.canRead()) |
162 | 0 | return; |
163 | 0 | firstImageRecord++; |
164 | 0 | } |
165 | 0 | } |
166 | | |
167 | | void DocumentPrivate::parseEXTH(QByteArrayView data) |
168 | 0 | { |
169 | | // try to get name |
170 | 0 | if (data.size() >= 92) { |
171 | 0 | qint32 nameoffset = qFromBigEndian<quint32>(data.constData() + 84); |
172 | 0 | qint32 namelen = qFromBigEndian<quint32>(data.constData() + 88); |
173 | 0 | if ((nameoffset + namelen) <= data.size()) { |
174 | 0 | metadata[Document::Title] = toUtf16(data.mid(nameoffset, namelen)); |
175 | 0 | } |
176 | 0 | } |
177 | |
|
178 | 0 | quint32 exthoffs = qFromBigEndian<quint32>(data.constData() + 20); |
179 | 0 | if (exthoffs + 28 > quint32(data.size())) { |
180 | 0 | return; |
181 | 0 | } |
182 | | |
183 | 0 | if (data.mid(exthoffs + 16, 4) != "EXTH") |
184 | 0 | return; |
185 | 0 | quint32 records = qFromBigEndian<quint32>(data.constData() + exthoffs + 24); |
186 | 0 | quint32 offset = exthoffs + 28; |
187 | 0 | for (unsigned int i = 0; i < records; i++) { |
188 | 0 | if (offset + 8 > quint32(data.size())) |
189 | 0 | break; |
190 | 0 | quint32 type = qFromBigEndian<quint32>(data.constData() + offset); |
191 | 0 | quint32 len = qFromBigEndian<quint32>(data.constData() + offset + 4); |
192 | 0 | if (offset + len > quint32(data.size())) |
193 | 0 | break; |
194 | 0 | switch (type) { |
195 | 0 | case 100: |
196 | 0 | metadata[Document::Author] = toUtf16(data.mid(offset + 8, len - 8)); |
197 | 0 | break; |
198 | 0 | case 103: |
199 | 0 | metadata[Document::Description] = toUtf16(data.mid(offset + 8, len - 8)); |
200 | 0 | break; |
201 | 0 | case 105: |
202 | 0 | metadata[Document::Subject] = toUtf16(data.mid(offset + 8, len - 8)); |
203 | 0 | break; |
204 | 0 | case 109: |
205 | 0 | metadata[Document::Copyright] = toUtf16(data.mid(offset + 8, len - 8)); |
206 | 0 | break; |
207 | 0 | case 201: |
208 | 0 | coverIndex = qFromBigEndian<quint32>(data.constData() + offset + 8); |
209 | 0 | break; |
210 | 0 | case 202: |
211 | 0 | thumbnailIndex = qFromBigEndian<quint32>(data.constData() + offset + 8); |
212 | 0 | break; |
213 | 0 | default: |
214 | | // ignore |
215 | 0 | break; |
216 | 0 | } |
217 | 0 | offset += len; |
218 | 0 | } |
219 | 0 | } |
220 | | |
221 | | Document::Document(QIODevice *dev) |
222 | 0 | : d(new DocumentPrivate(dev)) |
223 | 0 | { |
224 | 0 | Q_ASSERT(dev->openMode() & QIODevice::ReadOnly); |
225 | 0 | Q_ASSERT(!dev->isSequential()); |
226 | 0 | d->init(); |
227 | 0 | } |
228 | | |
229 | | Document::~Document() |
230 | 0 | { |
231 | 0 | delete d; |
232 | 0 | } |
233 | | |
234 | | namespace |
235 | | { |
236 | | constexpr qsizetype preTrailingDataLength(QByteArrayView data, quint32 flags) |
237 | 0 | { |
238 | 0 | if (flags == 0) { |
239 | 0 | return data.size(); |
240 | 0 | } |
241 | | |
242 | 0 | for (int i = 31; i > 0; i--) { |
243 | 0 | if ((flags & (1u << i)) == 0) { |
244 | 0 | continue; |
245 | 0 | } |
246 | | |
247 | 0 | qsizetype chopN = 0; |
248 | 0 | for (int j = 0; j < 4; j++) { |
249 | 0 | if (j + 1 > data.size()) { |
250 | 0 | return 0; |
251 | 0 | } |
252 | 0 | quint8 l = data.at(data.size() - (j + 1)); |
253 | 0 | chopN |= (l & 0x7f) << (7 * j); |
254 | 0 | if (l & 0x80) { |
255 | 0 | break; |
256 | 0 | } |
257 | 0 | } |
258 | 0 | data.chop(std::min<qsizetype>(chopN, data.size())); |
259 | 0 | } |
260 | 0 | if ((flags & 0x1) && !data.isEmpty()) { |
261 | 0 | quint8 l = data.back() & 0x3; |
262 | 0 | data.chop(std::min<qsizetype>(l + 1, data.size())); |
263 | 0 | } |
264 | 0 | return data.size(); |
265 | 0 | } |
266 | | static_assert(preTrailingDataLength({"0\x00", 2}, 0x0) == 2); |
267 | | static_assert(preTrailingDataLength({"0\x00", 2}, 0x1) == 1); |
268 | | static_assert(preTrailingDataLength({"0\x01", 2}, 0x1) == 0); |
269 | | static_assert(preTrailingDataLength({"0\x02", 2}, 0x1) == 0); |
270 | | static_assert(preTrailingDataLength({"abcd\x03", 5}, 0x1) == 1); |
271 | | static_assert(preTrailingDataLength({"abcd\x81", 5}, 0x2) == 4); |
272 | | static_assert(preTrailingDataLength({"\x02\x01", 2}, 0x2) == 0); |
273 | | static_assert(preTrailingDataLength({"\x80\x02", 2}, 0x2) == 0); |
274 | | static_assert(preTrailingDataLength({"abcd\x85", 5}, 0x2) == 0); |
275 | | static_assert(preTrailingDataLength({"abc\x01\x7f\x82", 6}, 0x2) == 4); |
276 | | static_assert(preTrailingDataLength({"abc\x01\x80\x02", 6}, 0x2) == 4); |
277 | | static_assert(preTrailingDataLength({"abc\x01\x7f\x82", 6}, 0x3) == 2); |
278 | | static_assert(preTrailingDataLength({"abc\x81\x80\x02", 6}, 0x6) == 3); |
279 | | static_assert(preTrailingDataLength({"abc\x00\x81\x81", 6}, 0x7) == 3); |
280 | | } // namespace |
281 | | |
282 | | QString Document::text(int size) const |
283 | 0 | { |
284 | 0 | QByteArray whole; |
285 | 0 | for (int i = 1; i < d->ntextrecords + 1; i++) { |
286 | 0 | auto record = d->pdb.getRecord(i); |
287 | 0 | record.resize(preTrailingDataLength(record, d->extraflags)); |
288 | 0 | QByteArray decompressedRecord = d->dec->decompress(record); |
289 | 0 | whole += decompressedRecord; |
290 | 0 | if (!d->dec->isValid()) { |
291 | 0 | d->valid = false; |
292 | 0 | return QString(); |
293 | 0 | } |
294 | 0 | if (size != -1 && whole.size() > size) |
295 | 0 | break; |
296 | 0 | } |
297 | 0 | return d->toUtf16(whole); |
298 | 0 | } |
299 | | |
300 | | int Document::imageCount() const |
301 | 0 | { |
302 | | // FIXME: don't count FLIS and FCIS records |
303 | 0 | return d->pdb.recordCount() - d->ntextrecords; |
304 | 0 | } |
305 | | |
306 | | bool Document::isValid() const |
307 | 0 | { |
308 | 0 | return d->valid; |
309 | 0 | } |
310 | | |
311 | | QImage Document::getImage(int i) const |
312 | 0 | { |
313 | 0 | if (!d->firstImageRecord) |
314 | 0 | d->findFirstImage(); |
315 | |
|
316 | 0 | if ((i < 0) || (i > std::numeric_limits<quint16>::max()) // |
317 | 0 | || (d->firstImageRecord + i) >= d->pdb.recordCount()) { |
318 | 0 | return {}; |
319 | 0 | } |
320 | | |
321 | 0 | QByteArray rec = d->pdb.getRecord(d->firstImageRecord + i); |
322 | 0 | return (rec.isNull()) ? QImage() : QImage::fromData(rec); |
323 | 0 | } |
324 | | |
325 | | QMap<Document::MetaKey, QString> Document::metadata() const |
326 | 0 | { |
327 | 0 | return d->metadata; |
328 | 0 | } |
329 | | |
330 | | bool Document::hasDRM() const |
331 | 0 | { |
332 | 0 | return d->drm; |
333 | 0 | } |
334 | | |
335 | | QImage Document::thumbnail() const |
336 | 0 | { |
337 | 0 | if (QImage img = getImage(d->thumbnailIndex); !img.isNull()) { |
338 | 0 | return img; |
339 | 0 | } |
340 | | |
341 | | // Fall back to cover image, or return an empty image |
342 | 0 | return getImage(d->coverIndex); |
343 | 0 | } |
344 | | |
345 | | } |