aboutsummaryrefslogtreecommitdiffstats
path: root/core/metadata.cpp
blob: bca750b733b23b97915bc9ae1d2e16663779e2dc (plain) (blame)
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
// SPDX-License-Identifier: GPL-2.0
#include "metadata.h"
#include "xmp_parser.h"
#include "exif.h"
#include "qthelper.h"
#include <QString>
#include <QFile>
#include <QDateTime>

// Weirdly, android builds fail owing to undefined UINT64_MAX
#ifndef UINT64_MAX
#define UINT64_MAX (~0ULL)
#endif

#if QT_VERSION >= QT_VERSION_CHECK(5, 14, 0)
#define SKIP_EMPTY Qt::SkipEmptyParts
#else
#define SKIP_EMPTY QString::SkipEmptyParts
#endif

// The following functions fetch an arbitrary-length _unsigned_ integer from either
// a file or a memory location in big-endian or little-endian mode. The size of the
// integer is passed via a template argument [e.g. getBE<uint16_t>(...)].
// The functions doing file access return a default value on IO error or end-of-file.
// Warning: This code works properly only for unsigned integers. The template parameter
// is not checked and passing a signed integer will silently fail!
template <typename T>
static inline T getBE(const char *buf_in)
{
	constexpr size_t size = sizeof(T);
	// Interpret raw bytes as unsigned char to avoid sign extension for
	// characters in the 0x80...0xff range.
	auto buf = (unsigned const char *)buf_in;
	T ret = 0;
	for (size_t i = 0; i < size; ++i)
		ret = (ret << 8) | buf[i];
	return ret;
}

template <typename T>
static inline T getBE(QFile &f, T def=0)
{
	constexpr size_t size = sizeof(T);
	char buf[size];
	if (f.read(buf, size) != size)
		return def;
	return getBE<T>(buf);
}

template <typename T>
static inline T getLE(const char *buf_in)
{
	constexpr size_t size = sizeof(T);
	// Interpret raw bytes as unsigned char to avoid sign extension for
	// characters in the 0x80...0xff range.
	auto buf = (unsigned const char *)buf_in;
	T ret = 0;
	for (size_t i = 0; i < size; ++i)
		ret |= static_cast<T>(buf[i]) << (i * 8);
	return ret;
}

template <typename T>
static inline T getLE(QFile &f, T def=0)
{
	constexpr size_t size = sizeof(T);
	char buf[size];
	if (f.read(buf, size) != size)
		return def;
	return getLE<T>(buf);
}

static bool parseExif(QFile &f, struct metadata *metadata)
{
	f.seek(0);
	if (getBE<uint16_t>(f) != 0xffd8)
		return false;
	for (;;) {
		switch (getBE<uint16_t>(f)) {
		case 0xffc0:
		case 0xffc2:
		case 0xffc4:
		case 0xffd0 ... 0xffd7:
		case 0xffdb:
		case 0xffdd:
		case 0xffe0:
		case 0xffe2 ... 0xffef:
		case 0xfffe: {
			uint16_t len = getBE<uint16_t>(f);
			if (len < 2)
				return false;
			f.seek(f.pos() + len - 2); // TODO: switch to QFile::skip()
			break;
		}
		case 0xffe1: {
			uint16_t len = getBE<uint16_t>(f);
			if (len < 2)
				return false;
			len -= 2;
			QByteArray data = f.read(len);
			if (data.size() != len)
				return false;
			easyexif::EXIFInfo exif;
			if (exif.parseFromEXIFSegment(reinterpret_cast<const unsigned char *>(data.constData()), len) != PARSE_EXIF_SUCCESS)
				return false;
			metadata->location = create_location(exif.GeoLocation.Latitude, exif.GeoLocation.Longitude);
			metadata->timestamp = exif.epoch();
			return true;
		}
		case 0xffda:
		case 0xffd9:
			// We expect EXIF data before any scan data
			return false;
		default:
			return false;
		}
	}
}

// Parse an embedded XMP block. Note that this is likely generated by
// external tools and therefore we give priority of XMP data over
// native metadata.
static void parseXMP(const char *data, size_t size, metadata *metadata)
{
	if (timestamp_t timestamp = parse_xmp(data, size))
		metadata->timestamp = timestamp;
}

static bool parseMP4(QFile &f, metadata *metadata)
{
	f.seek(0);

	// MP4s and related formats are hierarchical, being made up of "atoms", which can
	// contain other atoms (an interesting interpretation of the term atom).
	// To parse the file, the remaining to-be-parsed bytes of the upper atoms in
	// the parse-tree are tracked in a stack-like structure. This is not strictly
	// necessary, since the level at which an atom is found is insubstantial.
	// Nevertheless, it is an effective and simple way of sanity-checking the file and the
	// parsing routine.
	std::vector<uint64_t> atom_stack;
	atom_stack.reserve(10);

	// For the outmost level, set the atom-size the the maximum value representable in
	// 64-bits, which effectively means parse to the end of file.
	atom_stack.push_back(UINT64_MAX);

	// The first atom of an MP4 or related video is supposed to be of the "ftyp" kind.
	// If such an atom is found as first atom, this function will return true, indicating
	// that the file is a video.
	bool found_ftyp = false;

	while (!f.atEnd() && !atom_stack.empty()) {
		// Parse atom header. The header can have two forms (each character stands for a byte):
		//	lllltttt
		// or
		// 	0001ttttllllllll
		// where "l" stands for length in big-endian mode and "t" for type of the atom.
		// The length includes the 8- or 16-bytes header.
		uint64_t atom_size = getBE<uint32_t>(f, 2);
		int atom_header_size = 8;
		if (atom_size > 1 && atom_size < 8)
			break;
		char type[4];
		if (f.read(type, 4) != 4)
			break;
		if (atom_size == 1) {
			atom_size = getBE<uint64_t>(f);
			atom_header_size = 16;
			if (atom_size < 16)
			       break;
		}
		if (atom_size == 0)
			atom_size = atom_stack.back();
		if (atom_size > atom_stack.back())
			break;
		atom_stack.back() -= atom_size;
		atom_size -= atom_header_size;

		// The first atom must be "ftyp"
		if (!found_ftyp) {
			found_ftyp = !memcmp(type, "ftyp", 4);
			if (!found_ftyp)
				break;
		}

		if (!memcmp(type, "moov", 4) ||
		    !memcmp(type, "trak", 4) ||
		    !memcmp(type, "mdia", 4) ||
		    !memcmp(type, "udta", 4)) {
			// Recurse into "moov", "trak", "mdia" and "udta" atoms
			atom_stack.push_back(atom_size);
			continue;
		} else if (!memcmp(type, "mvhd", 4) && atom_size >= 100 && atom_size < 4096) {
			std::vector<char> data(atom_size);
			if (f.read(&data[0], atom_size) != static_cast<int>(atom_size))
				break;

			timestamp_t timestamp = getBE<uint32_t>(&data[4]);

			// Timestamp is given as seconds since midnight 1904/1/1. To be convertible to the UNIX epoch
			// it must be larger than 2082844800.
			// Note that we only set timestamp if not already set, because we give priority to XMP data.
			if (!metadata->timestamp && timestamp >= 2082844800)
				metadata->timestamp = timestamp - 2082844800;
		} else if (!memcmp(type, "mdhd", 4) && atom_size >= 24 && atom_size < 4096) {
			// Parse "mdhd" (media header).
			// Sanity check: size between 24 and 4096
			std::vector<char> data(atom_size);
			if (f.read(&data[0], atom_size) != static_cast<int>(atom_size))
				break;
			uint64_t timestamp = 0;
			uint32_t timescale = 0;
			uint64_t duration = 0;
			// First byte is version. We know version 0 and 1
			switch (data[0]) {
			case 0:
				timestamp = getBE<uint32_t>(&data[4]);
				timescale = getBE<uint32_t>(&data[12]);
				duration = getBE<uint32_t>(&data[16]);
				break;
			case 1:
				timestamp = getBE<uint64_t>(&data[4]);
				timescale = getBE<uint32_t>(&data[20]);
				duration = getBE<uint64_t>(&data[24]);
				break;
			default:
				// For unknown versions: ignore -> maybe we find a parseable "mdhd" atom later in this file
				break;
			}
			if (timescale > 0)
				metadata->duration.seconds = lrint((double)duration / timescale);
			// Timestamp is given as seconds since midnight 1904/1/1. To be convertible to the UNIX epoch
			// it must be larger than 2082844800.
			// Note that we only set timestamp if not already set, because we give priority to XMP data.
			if (!metadata->timestamp && timestamp >= 2082844800) {
				metadata->timestamp = timestamp - 2082844800;
				// We got our timestamp and duration. Nevertheless, we continue
				// parsing, as there might still be an XMP atom.
			}
		} else if (!memcmp(type, "XMP_", 4) && atom_size > 32 && atom_size < 100000) {
			// Parse embedded XMP data.
			std::vector<char> d(atom_size);
			if (f.read(&d[0], atom_size) != static_cast<int>(atom_size))
				break;

			parseXMP(&d[0], atom_size, metadata);
		} else if (!memcmp(type, "uuid", 4) && atom_size > 32 && atom_size < 100000) {
			// UUID atoms with uid "BE7ACFCB97A942E89C71999491E3AFAC" contain XMP blocks
			// according the JPEG 2000 standard. exiftools produces mp4-style videos with such
			// an UUID atom.
			std::vector<char> d(atom_size);
			if (f.read(&d[0], atom_size) != static_cast<int>(atom_size))
				break;

			static const char xmp_uid[17] = "\xBE\x7A\xCF\xCB\x97\xA9\x42\xE8\x9C\x71\x99\x94\x91\xE3\xAF\xAC";
			if (!memcmp(&d[0], xmp_uid, 16))
				parseXMP(&d[16], atom_size - 16, metadata);
		} else {
			// Jump over unknown atom
			if (!f.seek(f.pos() + atom_size)) // TODO: switch to QFile::skip()
				break;
		}

		// If end of atom is reached, return to outer atom
		while (!atom_stack.empty() && atom_stack.back() == 0)
			atom_stack.pop_back();
	}

	return found_ftyp;
}

static QStringList weekdays = { "mon", "tue", "wed", "thu", "fri", "sat", "sun" };
static QStringList months = { "jan", "feb", "mar", "apr", "may", "jun", "jul", "aug", "sep", "oct", "nov", "dec" };

static bool parseDate(const QString &s_in, timestamp_t &timestamp)
{
	// As a first attempt we're very crude: replace all '/' and '-' by ':'
	// and try to see if this is of the form "yyyy:mm:dd hh:mm:ss".
	// Since AVIs have no unified way of saving dates, we will have
	// to find out empirically what different software produces.
	// Note that we don't want to parse dates without time. That would
	// be too imprecise and in such a case we'd rather go after the
	// file modification date.
	QString s = s_in;
	s.replace('/', ':');
	s.replace('-', ':');
	QDateTime datetime = QDateTime::fromString(s, "yyyy:M:d h:m:s");
	if (datetime.isValid()) {
		// Not knowing any better, we suppose that time is give in UTC
		datetime.setTimeSpec(Qt::UTC);
		timestamp = dateTimeToTimestamp(datetime);
		return true;
	}

	// I've also seen "Weekday Mon  Day hh:mm:ss yyyy"(!)
	QStringList items = s.split(' ', SKIP_EMPTY);
	if (items.size() < 4)
		return false;

	// Skip weekday if any is given
	for (const QString &day: weekdays) {
		if (items[0].startsWith(day, Qt::CaseInsensitive)) {
			items.removeFirst();
			break;
		}
	}
	if (items.size() < 4)
		return false;
	int month;
	for (month = 0; month < 12; ++month)
		if (items[0].startsWith(months[month], Qt::CaseInsensitive))
			break;
	if (month >= 12)
		return false;
	bool ok;
	int day = items[1].toInt(&ok, 10);
	if (!ok)
		return false;
	QTime time = QTime::fromString(items[2], "h:m:s");
	if (!time.isValid())
		return false;
	int year = items[3].toInt(&ok, 10);
	if (!ok)
		return false;
	QDate date(year, month + 1, day);
	if (!date.isValid())
		return false;

	// Not knowing any better, we suppose that time is give in UTC
	datetime = QDateTime(date, time, Qt::UTC);
	if (datetime.isValid()) {
		timestamp = dateTimeToTimestamp(datetime);
		return true;
	}

	return false;
}

static bool parseAVI(QFile &f, metadata *metadata)
{
	f.seek(0);

	// Like MP4s, AVIs are hierarchical, being made up of  "chunks" and "lists",
	// whereby the latter can contain more "chunks" and "lists".
	// All elements are padded to an even-byte value. I.e. if the length of en element
	// is odd, then a padding byte is introduced.
	// To parse the file, the remaining to-be-parsed bytes of the upper lists in
	// the parse-tree are tracked in a stack-like structure. This is not strictly
	// necessary, since the level at which a chunk is found is insubstantial.
	// Nevertheless, it is an effective and simple way of sanity-checking the file and the
	// parsing routine.
	std::vector<uint64_t> list_stack;
	list_stack.reserve(10);

	// For the outmost level, set the chunk-size the the maximum value representable in
	// 64-bits, which effectively means parse to the end of file.
	list_stack.push_back(UINT64_MAX);

	// The first element of an AVI is supposed to be a "RIFF" list.
	// If such a list is found as first element, this function will return true, indicating
	// that the file is a video.
	bool found_riff = false;

	// Find creation date and duration. If we found both, we may quit.
	bool found_date = false;
	bool found_duration = false;
	while (!f.atEnd() && !list_stack.empty() && (!found_date || !found_duration)) {
		// Parse chunk/list header. If the first four bytes are "RIFF" or "LIST", then this
		// is a list. Otherwise, it is an chunk.
		char type[4];
		if (f.read(type, 4) != 4)
			break;

		// The first element must be RIFF
		if (!found_riff) {
			found_riff = !memcmp(type, "RIFF", 4);
			if (!found_riff)
				break;
		}

		uint32_t len = getLE<uint32_t>(f);
		// Elements are always padded to word (16-bit) boundaries
		uint32_t len_in_file = len + (len & 1);
		if (len_in_file + 8 > list_stack.back())
			break;
		list_stack.back() -= len_in_file + 8;

		// Check if this is a list
		if (!memcmp(type, "RIFF", 4) || !memcmp(type, "LIST", 4)) {
			// This is a list
			// The format is as follows:
			//	4 bytes "RIFF" or "LIST"
			//	4 bytes length (not including this and the previous entry)
			//	4 bytes type
			//	n bytes data
			// length includes the 4 bytes type
			if (len < 4)
				break;
			char list_type[4];
			if (f.read(list_type, 4) != 4)
				break;

			if (!memcmp(list_type, "AVI ", 4) || !memcmp(list_type, "hdrl", 4) ||
			    !memcmp(list_type, "strl", 4) || !memcmp(list_type, "INFO", 4)) {
				// Recurse into "AVI ", "hdrl", "strl" and "INFO" lists
				list_stack.push_back(len_in_file - 4);
				continue;
			} else {
				// Skip other lists
				if (!f.seek(f.pos() + len_in_file - 4)) // TODO: switch to QFile::skip()
					break;
			}
		} else if (!memcmp(type, "strh", 4) && !found_duration) {
			// The stream header contains the duration information. We will just assume that
			// the stream header is the correct one.
			// Before reading, sanity-check the length.
			if (len < 48 || len > 4096)
				break;
			std::vector<char> data(len_in_file);
			if (f.read(data.data(), len_in_file) != len_in_file)
				break;
			double scale = getLE<uint32_t>(&data[20]);
			double rate = getLE<uint32_t>(&data[24]);
			double start = getLE<uint32_t>(&data[28]);
			double length = getLE<uint32_t>(&data[32]);
			double duration = (start + length) * scale / rate;
			metadata->duration.seconds = lrint(duration);
			found_duration = true;
		} else if (!memcmp(type, "IDIT", 4) || !memcmp(type, "ICRD", 4)) {
			// "IDIT" of "ICRD" chunks may contain the creation date/time of the file
			// First, sanity-check the length.
			if (len > 4096)
				break;
			std::vector<char> data(len_in_file);
			if (f.read(data.data(), len_in_file) != len_in_file)
				break;
			QString idit = QString::fromUtf8(data.data(), len);
			// In my test file, the string contained a '\0' terminator. Remove it.
			idit.remove(QChar(0));
			found_date = parseDate(idit, metadata->timestamp);
		} else {
			if (!f.seek(f.pos() + len_in_file)) // TODO: switch to QFile::skip()
				break;
		}

		// If end of current list is reached, return to outer list
		while (!list_stack.empty() && list_stack.back() == 0)
			list_stack.pop_back();
	}
	return found_riff;
}

static bool parseASF(QFile &f, metadata *metadata)
{
	f.seek(0);

	// Parse the header of the header object:
	//	id				(16 bytes)
	//	size				 (8 bytes)
	//	number of header objects	 (4 bytes)
	//	reserved			 (2 bytes)
	//	------------------------------------------
	//	total				(30 bytes)
	char header[30];
	if (f.read(header, 30) != 30)
		return false;

	// Check if this is indeed an ASF header.
	if (memcmp(&header[0], "\x30\x26\xb2\x75\x8e\x66\xcf\x11\xa6\xd9\x00\xaa\x00\x62\xce\x6c", 16) != 0)
		return false;

	uint64_t header_len = getLE<uint64_t>(&header[16]);
	uint32_t num = getLE<uint32_t>(&header[24]);

	// Sanity check
	if (header_len <= 30 || num > 10000)
		return false;
	header_len -= 30;

	// Read through all the header objects
	for (uint32_t i = 0; i < num && header_len > 24; ++i) {
		// Each objects starts with the same header:
		//	id			(16 bytes)
		//	size			 (8 bytes)
		char data[24];
		if (f.read(data, 24) != 24)
			return false;

		uint64_t object_len = getLE<uint64_t>(&data[16]);
		// Sanity check
		if (object_len < 24 || object_len > header_len)
			return false;

		header_len -= object_len;
		object_len -= 24;
		if (!memcmp(data, "\xa1\xdc\xab\x8c\x47\xa9\xcf\x11\x8e\xe4\x0\xc0\xc\x20\x53\x65", 16)) {
			// This is a file properties object. The interesting data are:
			//	quadword (64 bit) at byte 24: creation date in 100-nanoseconds since Jan. 1, 1601.
			//	quadword (64 bit) at byte 40: duration in 100-nanoseconds.
			//	quadword (64 bit) at byte 56: offset in msec (to be subtracted from duration)
			// But first a sanity check:
			if (object_len < 80 || object_len > 4096)
				break;

			std::vector<char> v(object_len);
			if (f.read(v.data(), object_len) != (int)object_len)
				break;

			uint64_t creation_date = getLE<uint64_t>(&v[24]);
			// OK - first convert to seconds
			creation_date /= 10000000;
			// Check if this is during the UNIX epoch and convert into epoch
			if (creation_date <= 11644473600)
				metadata->timestamp = 0;		// Can't determine creation date, sorry!
			else
				metadata->timestamp = creation_date - 11644473600;

			uint64_t duration = getLE<uint64_t>(&v[40]);
			uint64_t offset = getLE<uint64_t>(&v[56]);
			metadata->duration.seconds = lrint(duration / 10000000.0 - offset / 1000.0);

			// We found everything that we wanted -> return success
			return true;
		} else {
			// Skip over unknown object
			if (!f.seek(f.pos() + object_len)) // TODO: switch to QFile::skip()
				break;
		}
	}

	// We didn't find a file properties object. According to the ASF specification, this is
	// *not* a valid ASF-file. Return failure accordingly.
	return false;
}

extern "C" mediatype_t get_metadata(const char *filename_in, metadata *data)
{
	data->timestamp = 0;
	data->duration.seconds = 0;
	data->location.lat.udeg = 0;
	data->location.lon.udeg = 0;

	QString filename = localFilePath(QString(filename_in));
	QFile f(filename);
	if (!f.open(QIODevice::ReadOnly))
		return MEDIATYPE_IO_ERROR;

	mediatype_t res = MEDIATYPE_UNKNOWN;
	if (parseExif(f, data))
		res = MEDIATYPE_PICTURE;
	else if(parseMP4(f, data))
		res = MEDIATYPE_VIDEO;
	else if(parseAVI(f, data))
		res = MEDIATYPE_VIDEO;
	else if(parseASF(f, data))
		res = MEDIATYPE_VIDEO;

	// If we couldn't get a creation date from the file (for example AVI files don't
	// have a standard way of storing this datum), use the file creation date of the file.
	if (data->timestamp == 0)
#if QT_VERSION >= QT_VERSION_CHECK(5, 10, 0)
		data->timestamp = dateTimeToTimestamp(QFileInfo(filename).birthTime());
#else
		data->timestamp = dateTimeToTimestamp(QFileInfo(filename).created());
#endif
	return res;
}

extern "C" timestamp_t picture_get_timestamp(const char *filename)
{
	struct metadata data;
	get_metadata(filename, &data);
	return data.timestamp;
}