FS#11964. Rework replaygain handling to save metadata buffer and binsize. Remove...
[kugel-rb.git] / apps / metadata / id3tags.c
blobc1d9cb202034be3c0f4cd7bd88def2d55181cf31
1 /***************************************************************************
2 * __________ __ ___.
3 * Open \______ \ ____ ____ | | _\_ |__ _______ ___
4 * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ /
5 * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < <
6 * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \
7 * \/ \/ \/ \/ \/
8 * $Id$
10 * Copyright (C) 2002 by Daniel Stenberg
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version 2
15 * of the License, or (at your option) any later version.
17 * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
18 * KIND, either express or implied.
20 ****************************************************************************/
22 * Parts of this code has been stolen from the Ample project and was written
23 * by David H�deman. It has since been extended and enhanced pretty much by
24 * all sorts of friendly Rockbox people.
28 /* tagResolver and associated code copyright 2003 Thomas Paul Diffenbach
31 #include <stdio.h>
32 #include <stdlib.h>
33 #include <string.h>
34 #include <errno.h>
35 #include <stdbool.h>
36 #include <stddef.h>
37 #include <ctype.h>
38 #include "string-extra.h"
39 #include "config.h"
40 #include "file.h"
41 #include "logf.h"
42 #include "system.h"
43 #include "replaygain.h"
44 #include "rbunicode.h"
46 #include "metadata.h"
47 #include "mp3data.h"
48 #if CONFIG_CODEC == SWCODEC
49 #include "metadata_common.h"
50 #endif
51 #include "metadata_parsers.h"
53 static unsigned long unsync(unsigned long b0,
54 unsigned long b1,
55 unsigned long b2,
56 unsigned long b3)
58 return (((long)(b0 & 0x7F) << (3*7)) |
59 ((long)(b1 & 0x7F) << (2*7)) |
60 ((long)(b2 & 0x7F) << (1*7)) |
61 ((long)(b3 & 0x7F) << (0*7)));
64 static const char* const genres[] = {
65 "Blues", "Classic Rock", "Country", "Dance", "Disco", "Funk", "Grunge",
66 "Hip-Hop", "Jazz", "Metal", "New Age", "Oldies", "Other", "Pop", "R&B",
67 "Rap", "Reggae", "Rock", "Techno", "Industrial", "Alternative", "Ska",
68 "Death Metal", "Pranks", "Soundtrack", "Euro-Techno", "Ambient", "Trip-Hop",
69 "Vocal", "Jazz+Funk", "Fusion", "Trance", "Classical", "Instrumental",
70 "Acid", "House", "Game", "Sound Clip", "Gospel", "Noise", "AlternRock",
71 "Bass", "Soul", "Punk", "Space", "Meditative", "Instrumental Pop",
72 "Instrumental Rock", "Ethnic", "Gothic", "Darkwave", "Techno-Industrial",
73 "Electronic", "Pop-Folk", "Eurodance", "Dream", "Southern Rock", "Comedy",
74 "Cult", "Gangsta", "Top 40", "Christian Rap", "Pop/Funk", "Jungle",
75 "Native American", "Cabaret", "New Wave", "Psychadelic", "Rave",
76 "Showtunes", "Trailer", "Lo-Fi", "Tribal", "Acid Punk", "Acid Jazz",
77 "Polka", "Retro", "Musical", "Rock & Roll", "Hard Rock",
79 /* winamp extensions */
80 "Folk", "Folk-Rock", "National Folk", "Swing", "Fast Fusion", "Bebob",
81 "Latin", "Revival", "Celtic", "Bluegrass", "Avantgarde", "Gothic Rock",
82 "Progressive Rock", "Psychedelic Rock", "Symphonic Rock", "Slow Rock",
83 "Big Band", "Chorus", "Easy Listening", "Acoustic", "Humour", "Speech",
84 "Chanson", "Opera", "Chamber Music", "Sonata", "Symphony", "Booty Bass",
85 "Primus", "Porn Groove", "Satire", "Slow Jam", "Club", "Tango", "Samba",
86 "Folklore", "Ballad", "Power Ballad", "Rhythmic Soul", "Freestyle",
87 "Duet", "Punk Rock", "Drum Solo", "A capella", "Euro-House", "Dance Hall",
88 "Goa", "Drum & Bass", "Club-House", "Hardcore", "Terror", "Indie",
89 "BritPop", "Negerpunk", "Polsk Punk", "Beat", "Christian Gangsta Rap",
90 "Heavy Metal", "Black Metal", "Crossover", "Contemporary Christian",
91 "Christian Rock", "Merengue", "Salsa", "Thrash Metal", "Anime", "Jpop",
92 "Synthpop"
95 char* id3_get_num_genre(unsigned int genre_num)
97 if (genre_num < ARRAYLEN(genres))
98 return (char*)genres[genre_num];
99 return NULL;
103 HOW TO ADD ADDITIONAL ID3 VERSION 2 TAGS
104 Code and comments by Thomas Paul Diffenbach
106 To add another ID3v2 Tag, do the following:
107 1. add a char* named for the tag to struct mp3entry in id3.h,
108 (I (tpd) prefer to use char* rather than ints, even for what seems like
109 numerical values, for cases where a number won't do, e.g.,
110 YEAR: "circa 1765", "1790/1977" (composed/performed), "28 Feb 1969"
111 TRACK: "1/12", "1 of 12", GENRE: "Freeform genre name"
112 Text is more flexible, and as the main use of id3 data is to
113 display it, converting it to an int just means reconverting to
114 display it, at a runtime cost.)
116 2. If any special processing beyond copying the tag value from the Id3
117 block to the struct mp3entry is rrequired (such as converting to an
118 int), write a function to perform this special processing.
120 This function's prototype must match that of
121 typedef tagPostProcessFunc, that is it must be:
122 int func( struct mp3entry*, char* tag, int bufferpos )
123 the first argument is a pointer to the current mp3entry structure the
124 second argument is a pointer to the null terminated string value of the
125 tag found the third argument is the offset of the next free byte in the
126 mp3entry's buffer your function should return the corrected offset; if
127 you don't lengthen or shorten the tag string, you can return the third
128 argument unchanged.
130 Unless you have a good reason no to, make the function static.
131 TO JUST COPY THE TAG NO SPECIAL PROCESSING FUNCTION IS NEEDED.
133 3. add one or more entries to the tagList array, using the format:
134 char* ID3 Tag symbolic name -- see the ID3 specification for these,
135 sizeof() that name minus 1,
136 offsetof( struct mp3entry, variable_name_in_struct_mp3entry ),
137 pointer to your special processing function or NULL
138 if you need no special processing
139 flag indicating if this tag is binary or textual
140 Many ID3 symbolic names come in more than one form. You can add both
141 forms, each referencing the same variable in struct mp3entry.
142 If both forms are present, the last found will be used.
143 Note that the offset can be zero, in which case no entry will be set
144 in the mp3entry struct; the frame is still read into the buffer and
145 the special processing function is called (several times, if there
146 are several frames with the same name).
148 4. Alternately, use the TAG_LIST_ENTRY macro with
149 ID3 tag symbolic name,
150 variable in struct mp3entry,
151 special processing function address
153 5. Add code to wps-display.c function get_tag to assign a printf-like
154 format specifier for the tag */
156 /* Structure for ID3 Tag extraction information */
157 struct tag_resolver {
158 const char* tag;
159 int tag_length;
160 size_t offset;
161 int (*ppFunc)(struct mp3entry*, char* tag, int bufferpos);
162 bool binary;
165 static bool global_ff_found;
167 static int unsynchronize(char* tag, int len, bool *ff_found)
169 int i;
170 unsigned char c;
171 unsigned char *rp, *wp;
173 wp = rp = (unsigned char *)tag;
175 rp = (unsigned char *)tag;
176 for(i = 0;i < len;i++) {
177 /* Read the next byte and write it back, but don't increment the
178 write pointer */
179 c = *rp++;
180 *wp = c;
181 if(*ff_found) {
182 /* Increment the write pointer if it isn't an unsynch pattern */
183 if(c != 0)
184 wp++;
185 *ff_found = false;
186 } else {
187 if(c == 0xff)
188 *ff_found = true;
189 wp++;
192 return (long)wp - (long)tag;
195 static int unsynchronize_frame(char* tag, int len)
197 bool ff_found = false;
199 return unsynchronize(tag, len, &ff_found);
202 static int read_unsynched(int fd, void *buf, int len)
204 int i;
205 int rc;
206 int remaining = len;
207 char *wp;
208 char *rp;
210 wp = buf;
212 while(remaining) {
213 rp = wp;
214 rc = read(fd, rp, remaining);
215 if(rc <= 0)
216 return rc;
218 i = unsynchronize(wp, remaining, &global_ff_found);
219 remaining -= i;
220 wp += i;
223 return len;
226 static int skip_unsynched(int fd, int len)
228 int rc;
229 int remaining = len;
230 int rlen;
231 char buf[32];
233 while(remaining) {
234 rlen = MIN(sizeof(buf), (unsigned int)remaining);
235 rc = read(fd, buf, rlen);
236 if(rc <= 0)
237 return rc;
239 remaining -= unsynchronize(buf, rlen, &global_ff_found);
242 return len;
245 /* parse numeric value from string */
246 static int parsetracknum( struct mp3entry* entry, char* tag, int bufferpos )
248 entry->tracknum = atoi( tag );
249 return bufferpos;
252 /* parse numeric value from string */
253 static int parsediscnum( struct mp3entry* entry, char* tag, int bufferpos )
255 entry->discnum = atoi( tag );
256 return bufferpos;
259 /* parse numeric value from string */
260 static int parseyearnum( struct mp3entry* entry, char* tag, int bufferpos )
262 entry->year = atoi( tag );
263 return bufferpos;
266 /* parse numeric genre from string, version 2.2 and 2.3 */
267 static int parsegenre( struct mp3entry* entry, char* tag, int bufferpos )
269 /* Use bufferpos to hold current position in entry->id3v2buf. */
270 bufferpos = tag - entry->id3v2buf;
272 if(entry->id3version >= ID3_VER_2_4) {
273 /* In version 2.4 and up, there are no parentheses, and the genre frame
274 is a list of strings, either numbers or text. */
276 /* Is it a number? */
277 if(isdigit(tag[0])) {
278 entry->genre_string = id3_get_num_genre(atoi( tag ));
279 return bufferpos;
280 } else {
281 entry->genre_string = tag;
282 return bufferpos + strlen(tag) + 1;
284 } else {
285 if( tag[0] == '(' && tag[1] != '(' ) {
286 entry->genre_string = id3_get_num_genre(atoi( tag + 1 ));
287 return bufferpos;
289 else {
290 entry->genre_string = tag;
291 return bufferpos + strlen(tag) + 1;
296 #ifdef HAVE_ALBUMART
297 /* parse embed albumart */
298 static int parsealbumart( struct mp3entry* entry, char* tag, int bufferpos )
300 entry->embed_albumart = false;
302 /* we currently don't support unsynchronizing albumart */
303 if (entry->albumart.type == AA_TYPE_UNSYNC)
304 return bufferpos;
306 entry->albumart.type = AA_TYPE_UNKNOWN;
308 char *start = tag;
309 /* skip text encoding */
310 tag += 1;
312 if (memcmp(tag, "image/", 6) == 0)
314 /* ID3 v2.3+ */
315 tag += 6;
316 if (strcmp(tag, "jpeg") == 0)
318 entry->albumart.type = AA_TYPE_JPG;
319 tag += 5;
321 else if (strcmp(tag, "png") == 0)
323 entry->albumart.type = AA_TYPE_PNG;
324 tag += 4;
327 else
329 /* ID3 v2.2 */
330 if (memcmp(tag, "JPG", 3) == 0)
331 entry->albumart.type = AA_TYPE_JPG;
332 else if (memcmp(tag, "PNG", 3) == 0)
333 entry->albumart.type = AA_TYPE_PNG;
334 tag += 3;
337 if (entry->albumart.type != AA_TYPE_UNKNOWN)
339 /* skip picture type */
340 tag += 1;
341 /* skip description */
342 tag = strchr(tag, '\0') + 1;
343 /* fixup offset&size for image data */
344 entry->albumart.pos += tag - start;
345 entry->albumart.size -= tag - start;
346 entry->embed_albumart = true;
348 /* return bufferpos as we didn't store anything in id3v2buf */
349 return bufferpos;
351 #endif
353 /* parse user defined text, looking for album artist and replaygain
354 * information.
356 static int parseuser( struct mp3entry* entry, char* tag, int bufferpos )
358 char* value = NULL;
359 int desc_len = strlen(tag);
360 int length = 0;
362 if ((tag - entry->id3v2buf + desc_len + 2) < bufferpos) {
363 /* At least part of the value was read, so we can safely try to
364 * parse it */
365 value = tag + desc_len + 1;
367 if (!strcasecmp(tag, "ALBUM ARTIST")) {
368 length = strlen(value) + 1;
369 strlcpy(tag, value, length);
370 entry->albumartist = tag;
371 #if CONFIG_CODEC == SWCODEC
372 } else {
373 /* Call parse_replaygain(). */
374 parse_replaygain(tag, value, entry);
375 #endif
379 return tag - entry->id3v2buf + length;
382 #if CONFIG_CODEC == SWCODEC
383 /* parse RVA2 binary data and convert to replaygain information. */
384 static int parserva2( struct mp3entry* entry, char* tag, int bufferpos)
386 int desc_len = strlen(tag);
387 int start_pos = tag - entry->id3v2buf;
388 int end_pos = start_pos + desc_len + 5;
389 unsigned char* value = tag + desc_len + 1;
391 /* Only parse RVA2 replaygain tags if tag version == 2.4 and channel
392 * type is master volume.
394 if (entry->id3version == ID3_VER_2_4 && end_pos < bufferpos
395 && *value++ == 1) {
396 long gain = 0;
397 long peak = 0;
398 long peakbits;
399 long peakbytes;
400 bool album = false;
402 /* The RVA2 specification is unclear on some things (id string and
403 * peak volume), but this matches how Quod Libet use them.
406 gain = (int16_t) ((value[0] << 8) | value[1]);
407 value += 2;
408 peakbits = *value++;
409 peakbytes = (peakbits + 7) / 8;
411 /* Only use the topmost 24 bits for peak volume */
412 if (peakbytes > 3) {
413 peakbytes = 3;
416 /* Make sure the peak bits were read */
417 if (end_pos + peakbytes < bufferpos) {
418 long shift = ((8 - (peakbits & 7)) & 7) + (3 - peakbytes) * 8;
420 for ( ; peakbytes; peakbytes--) {
421 peak <<= 8;
422 peak += *value++;
425 peak <<= shift;
427 if (peakbits > 24) {
428 peak += *value >> (8 - shift);
432 if (strcasecmp(tag, "album") == 0) {
433 album = true;
434 } else if (strcasecmp(tag, "track") != 0) {
435 /* Only accept non-track values if we don't have any previous
436 * value.
438 if (entry->track_gain != 0) {
439 return start_pos;
443 parse_replaygain_int(album, gain, peak * 2, entry);
446 return start_pos;
448 #endif
450 static int parsembtid( struct mp3entry* entry, char* tag, int bufferpos )
452 char* value = NULL;
453 int desc_len = strlen(tag);
454 /*DEBUGF("MBID len: %d\n", desc_len);*/
455 /* Musicbrainz track IDs are always 36 chars long */
456 const size_t mbtid_len = 36;
458 if ((tag - entry->id3v2buf + desc_len + 2) < bufferpos)
460 value = tag + desc_len + 1;
462 if (strcasecmp(tag, "http://musicbrainz.org") == 0)
464 if (mbtid_len == strlen(value))
466 entry->mb_track_id = value;
467 return bufferpos + mbtid_len + 1;
472 return bufferpos;
475 static const struct tag_resolver taglist[] = {
476 { "TPE1", 4, offsetof(struct mp3entry, artist), NULL, false },
477 { "TP1", 3, offsetof(struct mp3entry, artist), NULL, false },
478 { "TIT2", 4, offsetof(struct mp3entry, title), NULL, false },
479 { "TT2", 3, offsetof(struct mp3entry, title), NULL, false },
480 { "TALB", 4, offsetof(struct mp3entry, album), NULL, false },
481 { "TAL", 3, offsetof(struct mp3entry, album), NULL, false },
482 { "TRK", 3, offsetof(struct mp3entry, track_string), &parsetracknum, false },
483 { "TPOS", 4, offsetof(struct mp3entry, disc_string), &parsediscnum, false },
484 { "TPA", 3, offsetof(struct mp3entry, disc_string), &parsediscnum, false },
485 { "TRCK", 4, offsetof(struct mp3entry, track_string), &parsetracknum, false },
486 { "TDRC", 4, offsetof(struct mp3entry, year_string), &parseyearnum, false },
487 { "TYER", 4, offsetof(struct mp3entry, year_string), &parseyearnum, false },
488 { "TYE", 3, offsetof(struct mp3entry, year_string), &parseyearnum, false },
489 { "TCOM", 4, offsetof(struct mp3entry, composer), NULL, false },
490 { "TCM", 3, offsetof(struct mp3entry, composer), NULL, false },
491 { "TPE2", 4, offsetof(struct mp3entry, albumartist), NULL, false },
492 { "TP2", 3, offsetof(struct mp3entry, albumartist), NULL, false },
493 { "TIT1", 4, offsetof(struct mp3entry, grouping), NULL, false },
494 { "TT1", 3, offsetof(struct mp3entry, grouping), NULL, false },
495 { "COMM", 4, offsetof(struct mp3entry, comment), NULL, false },
496 { "COM", 3, offsetof(struct mp3entry, comment), NULL, false },
497 { "TCON", 4, offsetof(struct mp3entry, genre_string), &parsegenre, false },
498 { "TCO", 3, offsetof(struct mp3entry, genre_string), &parsegenre, false },
499 #ifdef HAVE_ALBUMART
500 { "APIC", 4, 0, &parsealbumart, true },
501 { "PIC", 3, 0, &parsealbumart, true },
502 #endif
503 { "TXXX", 4, 0, &parseuser, false },
504 #if CONFIG_CODEC == SWCODEC
505 { "RVA2", 4, 0, &parserva2, true },
506 #endif
507 { "UFID", 4, 0, &parsembtid, false },
510 #define TAGLIST_SIZE ((int)ARRAYLEN(taglist))
512 /* Get the length of an ID3 string in the given encoding. Returns the length
513 * in bytes, including end nil, or -1 if the encoding is unknown.
515 static int unicode_len(char encoding, const void* string)
517 int len = 0;
519 if (encoding == 0x01 || encoding == 0x02) {
520 char first;
521 const char *s = string;
522 /* string might be unaligned, so using short* can crash on ARM and SH1 */
523 do {
524 first = *s++;
525 } while ((first | *s++) != 0);
527 len = s - (const char*) string;
528 } else {
529 len = strlen((char*) string) + 1;
532 return len;
535 /* Checks to see if the passed in string is a 16-bit wide Unicode v2
536 string. If it is, we convert it to a UTF-8 string. If it's not unicode,
537 we convert from the default codepage */
538 static int unicode_munge(char* string, char* utf8buf, int *len) {
539 long tmp;
540 bool le = false;
541 int i = 0;
542 unsigned char *str = (unsigned char *)string;
543 int templen = 0;
544 unsigned char* utf8 = (unsigned char *)utf8buf;
546 switch (str[0]) {
547 case 0x00: /* Type 0x00 is ordinary ISO 8859-1 */
548 str++;
549 (*len)--;
550 utf8 = iso_decode(str, utf8, -1, *len);
551 *utf8 = 0;
552 *len = (unsigned long)utf8 - (unsigned long)utf8buf;
553 break;
555 case 0x01: /* Unicode with or without BOM */
556 case 0x02:
557 (*len)--;
558 str++;
560 /* Handle frames with more than one string
561 (needed for TXXX frames).*/
562 do {
563 tmp = bytes2int(0, 0, str[0], str[1]);
565 /* Now check if there is a BOM
566 (zero-width non-breaking space, 0xfeff)
567 and if it is in little or big endian format */
568 if(tmp == 0xfffe) { /* Little endian? */
569 le = true;
570 str += 2;
571 (*len)-=2;
572 } else if(tmp == 0xfeff) { /* Big endian? */
573 str += 2;
574 (*len)-=2;
575 } else
576 /* If there is no BOM (which is a specification violation),
577 let's try to guess it. If one of the bytes is 0x00, it is
578 probably the most significant one. */
579 if(str[1] == 0)
580 le = true;
582 do {
583 if(le)
584 utf8 = utf16LEdecode(str, utf8, 1);
585 else
586 utf8 = utf16BEdecode(str, utf8, 1);
588 str+=2;
589 i += 2;
590 } while((str[0] || str[1]) && (i < *len));
592 *utf8++ = 0; /* Terminate the string */
593 templen += (strlen(&utf8buf[templen]) + 1);
594 str += 2;
595 i+=2;
596 } while(i < *len);
597 *len = templen - 1;
598 break;
600 case 0x03: /* UTF-8 encoded string */
601 for(i=0; i < *len; i++)
602 utf8[i] = str[i+1];
603 (*len)--;
604 break;
606 default: /* Plain old string */
607 utf8 = iso_decode(str, utf8, -1, *len);
608 *utf8 = 0;
609 *len = (unsigned long)utf8 - (unsigned long)utf8buf;
610 break;
612 return 0;
616 * Sets the title of an MP3 entry based on its ID3v1 tag.
618 * Arguments: file - the MP3 file to scen for a ID3v1 tag
619 * entry - the entry to set the title in
621 * Returns: true if a title was found and created, else false
623 bool setid3v1title(int fd, struct mp3entry *entry)
625 unsigned char buffer[128];
626 static const char offsets[] = {3, 33, 63, 97, 93, 125, 127};
627 int i, j;
628 unsigned char* utf8;
630 if (-1 == lseek(fd, -128, SEEK_END))
631 return false;
633 if (read(fd, buffer, sizeof buffer) != sizeof buffer)
634 return false;
636 if (strncmp((char *)buffer, "TAG", 3))
637 return false;
639 entry->id3v1len = 128;
640 entry->id3version = ID3_VER_1_0;
642 for (i=0; i < (int)sizeof offsets; i++) {
643 unsigned char* ptr = (unsigned char *)buffer + offsets[i];
645 switch(i) {
646 case 0:
647 case 1:
648 case 2:
649 /* kill trailing space in strings */
650 for (j=29; j && (ptr[j]==0 || ptr[j]==' '); j--)
651 ptr[j] = 0;
652 /* convert string to utf8 */
653 utf8 = (unsigned char *)entry->id3v1buf[i];
654 utf8 = iso_decode(ptr, utf8, -1, 30);
655 /* make sure string is terminated */
656 *utf8 = 0;
657 break;
659 case 3:
660 /* kill trailing space in strings */
661 for (j=27; j && (ptr[j]==0 || ptr[j]==' '); j--)
662 ptr[j] = 0;
663 /* convert string to utf8 */
664 utf8 = (unsigned char *)entry->id3v1buf[3];
665 utf8 = iso_decode(ptr, utf8, -1, 28);
666 /* make sure string is terminated */
667 *utf8 = 0;
668 break;
670 case 4:
671 ptr[4] = 0;
672 entry->year = atoi((char *)ptr);
673 break;
675 case 5:
676 /* id3v1.1 uses last two bytes of comment field for track
677 number: first must be 0 and second is track num */
678 if (!ptr[0] && ptr[1]) {
679 entry->tracknum = ptr[1];
680 entry->id3version = ID3_VER_1_1;
682 break;
684 case 6:
685 /* genre */
686 entry->genre_string = id3_get_num_genre(ptr[0]);
687 break;
691 entry->title = entry->id3v1buf[0];
692 entry->artist = entry->id3v1buf[1];
693 entry->album = entry->id3v1buf[2];
694 entry->comment = entry->id3v1buf[3];
696 return true;
701 * Sets the title of an MP3 entry based on its ID3v2 tag.
703 * Arguments: file - the MP3 file to scan for a ID3v2 tag
704 * entry - the entry to set the title in
706 * Returns: true if a title was found and created, else false
708 void setid3v2title(int fd, struct mp3entry *entry)
710 int minframesize;
711 int size;
712 long bufferpos = 0, totframelen, framelen;
713 char header[10];
714 char tmp[4];
715 unsigned char version;
716 char *buffer = entry->id3v2buf;
717 int bytesread = 0;
718 int buffersize = sizeof(entry->id3v2buf);
719 unsigned char global_flags;
720 int flags;
721 bool global_unsynch = false;
722 bool unsynch = false;
723 int i, j;
724 int rc;
725 #if CONFIG_CODEC == SWCODEC
726 bool itunes_gapless = false;
727 #endif
729 global_ff_found = false;
731 /* Bail out if the tag is shorter than 10 bytes */
732 if(entry->id3v2len < 10)
733 return;
735 /* Read the ID3 tag version from the header */
736 lseek(fd, 0, SEEK_SET);
737 if(10 != read(fd, header, 10))
738 return;
740 /* Get the total ID3 tag size */
741 size = entry->id3v2len - 10;
743 version = header[3];
744 switch ( version ) {
745 case 2:
746 version = ID3_VER_2_2;
747 minframesize = 8;
748 break;
750 case 3:
751 version = ID3_VER_2_3;
752 minframesize = 12;
753 break;
755 case 4:
756 version = ID3_VER_2_4;
757 minframesize = 12;
758 break;
760 default:
761 /* unsupported id3 version */
762 return;
764 entry->id3version = version;
765 entry->tracknum = entry->year = entry->discnum = 0;
766 entry->title = entry->artist = entry->album = NULL; /* FIXME incomplete */
768 global_flags = header[5];
770 /* Skip the extended header if it is present */
771 if(global_flags & 0x40) {
772 if(version == ID3_VER_2_3) {
773 if(10 != read(fd, header, 10))
774 return;
775 /* The 2.3 extended header size doesn't include the header size
776 field itself. Also, it is not unsynched. */
777 framelen =
778 bytes2int(header[0], header[1], header[2], header[3]) + 4;
780 /* Skip the rest of the header */
781 lseek(fd, framelen - 10, SEEK_CUR);
784 if(version >= ID3_VER_2_4) {
785 if(4 != read(fd, header, 4))
786 return;
788 /* The 2.4 extended header size does include the entire header,
789 so here we can just skip it. This header is unsynched. */
790 framelen = unsync(header[0], header[1],
791 header[2], header[3]);
793 lseek(fd, framelen - 4, SEEK_CUR);
797 /* Is unsynchronization applied? */
798 if(global_flags & 0x80) {
799 global_unsynch = true;
803 * We must have at least minframesize bytes left for the
804 * remaining frames to be interesting
806 while (size >= minframesize && bufferpos < buffersize - 1) {
807 flags = 0;
809 /* Read frame header and check length */
810 if(version >= ID3_VER_2_3) {
811 if(global_unsynch && version <= ID3_VER_2_3)
812 rc = read_unsynched(fd, header, 10);
813 else
814 rc = read(fd, header, 10);
815 if(rc != 10)
816 return;
817 /* Adjust for the 10 bytes we read */
818 size -= 10;
820 flags = bytes2int(0, 0, header[8], header[9]);
822 if (version >= ID3_VER_2_4) {
823 framelen = unsync(header[4], header[5],
824 header[6], header[7]);
825 } else {
826 /* version .3 files don't use synchsafe ints for
827 * size */
828 framelen = bytes2int(header[4], header[5],
829 header[6], header[7]);
831 } else {
832 if(6 != read(fd, header, 6))
833 return;
834 /* Adjust for the 6 bytes we read */
835 size -= 6;
837 framelen = bytes2int(0, header[3], header[4], header[5]);
840 logf("framelen = %ld, flags = 0x%04x", framelen, flags);
841 if(framelen == 0){
842 if (header[0] == 0 && header[1] == 0 && header[2] == 0)
843 return;
844 else
845 continue;
848 unsynch = false;
850 if(flags)
852 if (version >= ID3_VER_2_4) {
853 if(flags & 0x0040) { /* Grouping identity */
854 lseek(fd, 1, SEEK_CUR); /* Skip 1 byte */
855 framelen--;
857 } else {
858 if(flags & 0x0020) { /* Grouping identity */
859 lseek(fd, 1, SEEK_CUR); /* Skip 1 byte */
860 framelen--;
864 if(flags & 0x000c) /* Compression or encryption */
866 /* Skip it */
867 size -= framelen;
868 lseek(fd, framelen, SEEK_CUR);
869 continue;
872 if(flags & 0x0002) /* Unsynchronization */
873 unsynch = true;
875 if (version >= ID3_VER_2_4) {
876 if(flags & 0x0001) { /* Data length indicator */
877 if(4 != read(fd, tmp, 4))
878 return;
880 /* We don't need the data length */
881 framelen -= 4;
886 if (framelen == 0)
887 continue;
889 if (framelen < 0)
890 return;
892 /* Keep track of the remaining frame size */
893 totframelen = framelen;
895 /* If the frame is larger than the remaining buffer space we try
896 to read as much as would fit in the buffer */
897 if(framelen >= buffersize - bufferpos)
898 framelen = buffersize - bufferpos - 1;
900 /* Limit the maximum length of an id3 data item to ID3V2_MAX_ITEM_SIZE
901 bytes. This reduces the chance that the available buffer is filled
902 by single metadata items like large comments. */
903 if (ID3V2_MAX_ITEM_SIZE < framelen)
904 framelen = ID3V2_MAX_ITEM_SIZE;
906 logf("id3v2 frame: %.4s", header);
908 /* Check for certain frame headers
910 'size' is the amount of frame bytes remaining. We decrement it by
911 the amount of bytes we read. If we fail to read as many bytes as
912 we expect, we assume that we can't read from this file, and bail
913 out.
915 For each frame. we will iterate over the list of supported tags,
916 and read the tag into entry's buffer. All tags will be kept as
917 strings, for cases where a number won't do, e.g., YEAR: "circa
918 1765", "1790/1977" (composed/performed), "28 Feb 1969" TRACK:
919 "1/12", "1 of 12", GENRE: "Freeform genre name" Text is more
920 flexible, and as the main use of id3 data is to display it,
921 converting it to an int just means reconverting to display it, at a
922 runtime cost.
924 For tags that the current code does convert to ints, a post
925 processing function will be called via a pointer to function. */
927 for (i=0; i<TAGLIST_SIZE; i++) {
928 const struct tag_resolver* tr = &taglist[i];
929 char** ptag = tr->offset ? (char**) (((char*)entry) + tr->offset)
930 : NULL;
931 char* tag;
933 /* Only ID3_VER_2_2 uses frames with three-character names. */
934 if (((version == ID3_VER_2_2) && (tr->tag_length != 3))
935 || ((version > ID3_VER_2_2) && (tr->tag_length != 4))) {
936 continue;
939 if( !memcmp( header, tr->tag, tr->tag_length ) ) {
941 /* found a tag matching one in tagList, and not yet filled */
942 tag = buffer + bufferpos;
944 if(global_unsynch && version <= ID3_VER_2_3)
945 bytesread = read_unsynched(fd, tag, framelen);
946 else
947 bytesread = read(fd, tag, framelen);
949 if( bytesread != framelen )
950 return;
952 size -= bytesread;
954 if(unsynch || (global_unsynch && version >= ID3_VER_2_4))
955 bytesread = unsynchronize_frame(tag, bytesread);
957 /* the COMM frame has a 3 char field to hold an ISO-639-1
958 * language string and an optional short description;
959 * remove them so unicode_munge can work correctly
962 if((tr->tag_length == 4 && !memcmp( header, "COMM", 4)) ||
963 (tr->tag_length == 3 && !memcmp( header, "COM", 3))) {
964 int offset;
965 if(!strncmp(tag+4, "iTun", 4)) {
966 #if CONFIG_CODEC == SWCODEC
967 /* check for iTunes gapless information */
968 if(!strncmp(tag+4, "iTunSMPB", 8))
969 itunes_gapless = true;
970 else
971 #endif
972 /* ignore other with iTunes tags */
973 break;
976 offset = 3 + unicode_len(*tag, tag + 4);
977 if(bytesread > offset) {
978 bytesread -= offset;
979 memmove(tag + 1, tag + 1 + offset, bytesread - 1);
983 /* Attempt to parse Unicode string only if the tag contents
984 aren't binary */
985 if(!tr->binary) {
986 /* UTF-8 could potentially be 3 times larger */
987 /* so we need to create a new buffer */
988 char utf8buf[(3 * bytesread) + 1];
990 unicode_munge( tag, utf8buf, &bytesread );
992 if(bytesread >= buffersize - bufferpos)
993 bytesread = buffersize - bufferpos - 1;
995 for (j = 0; j < bytesread; j++)
996 tag[j] = utf8buf[j];
998 /* remove trailing spaces */
999 while ( bytesread > 0 && isspace(tag[bytesread-1]))
1000 bytesread--;
1003 tag[bytesread] = 0;
1004 bufferpos += bytesread + 1;
1006 #if CONFIG_CODEC == SWCODEC
1007 /* parse the tag if it contains iTunes gapless info */
1008 if (itunes_gapless)
1010 itunes_gapless = false;
1011 entry->lead_trim = get_itunes_int32(tag, 1);
1012 entry->tail_trim = get_itunes_int32(tag, 2);
1014 #endif
1016 /* Note that parser functions sometimes set *ptag to NULL, so
1017 * the "!*ptag" check here doesn't always have the desired
1018 * effect. Should the parser functions (parsegenre in
1019 * particular) be updated to handle the case of being called
1020 * multiple times, or should the "*ptag" check be removed?
1022 if (ptag && !*ptag)
1023 *ptag = tag;
1025 #ifdef HAVE_ALBUMART
1026 /* albumart */
1027 if ((!entry->embed_albumart) &&
1028 ((tr->tag_length == 4 && !memcmp( header, "APIC", 4)) ||
1029 (tr->tag_length == 3 && !memcmp( header, "PIC" , 3))))
1031 if (unsynch || (global_unsynch && version <= ID3_VER_2_3))
1032 entry->albumart.type = AA_TYPE_UNSYNC;
1033 else
1035 entry->albumart.pos = lseek(fd, 0, SEEK_CUR) - framelen;
1036 entry->albumart.size = totframelen;
1037 entry->albumart.type = AA_TYPE_UNKNOWN;
1040 #endif
1041 if( tr->ppFunc )
1042 bufferpos = tr->ppFunc(entry, tag, bufferpos);
1044 /* Seek to the next frame */
1045 if(framelen < totframelen)
1046 lseek(fd, totframelen - framelen, SEEK_CUR);
1047 break;
1051 if( i == TAGLIST_SIZE ) {
1052 /* no tag in tagList was found, or it was a repeat.
1053 skip it using the total size */
1055 if(global_unsynch && version <= ID3_VER_2_3) {
1056 size -= skip_unsynched(fd, totframelen);
1057 } else {
1058 size -= totframelen;
1059 if( lseek(fd, totframelen, SEEK_CUR) == -1 )
1060 return;
1067 * Calculates the size of the ID3v2 tag.
1069 * Arguments: file - the file to search for a tag.
1071 * Returns: the size of the tag or 0 if none was found
1073 int getid3v2len(int fd)
1075 char buf[6];
1076 int offset;
1078 /* Make sure file has a ID3 tag */
1079 if((-1 == lseek(fd, 0, SEEK_SET)) ||
1080 (read(fd, buf, 6) != 6) ||
1081 (strncmp(buf, "ID3", strlen("ID3")) != 0))
1082 offset = 0;
1084 /* Now check what the ID3v2 size field says */
1085 else
1086 if(read(fd, buf, 4) != 4)
1087 offset = 0;
1088 else
1089 offset = unsync(buf[0], buf[1], buf[2], buf[3]) + 10;
1091 logf("ID3V2 Length: 0x%x", offset);
1092 return offset;
1095 #ifdef DEBUG_STANDALONE
1097 char *secs2str(int ms)
1099 static char buffer[32];
1100 int secs = ms/1000;
1101 ms %= 1000;
1102 snprintf(buffer, sizeof(buffer), "%d:%02d.%d", secs/60, secs%60, ms/100);
1103 return buffer;
1106 int main(int argc, char **argv)
1108 int i;
1109 for(i=1; i<argc; i++) {
1110 struct mp3entry mp3;
1111 mp3.album = "Bogus";
1112 if(mp3info(&mp3, argv[i], false)) {
1113 printf("Failed to get %s\n", argv[i]);
1114 return 0;
1117 printf("****** File: %s\n"
1118 " Title: %s\n"
1119 " Artist: %s\n"
1120 " Album: %s\n"
1121 " Genre: %s (%d) \n"
1122 " Composer: %s\n"
1123 " Year: %s (%d)\n"
1124 " Track: %s (%d)\n"
1125 " Length: %s / %d s\n"
1126 " Bitrate: %d\n"
1127 " Frequency: %d\n",
1128 argv[i],
1129 mp3.title?mp3.title:"<blank>",
1130 mp3.artist?mp3.artist:"<blank>",
1131 mp3.album?mp3.album:"<blank>",
1132 mp3.genre_string?mp3.genre_string:"<blank>",
1133 mp3.genre,
1134 mp3.composer?mp3.composer:"<blank>",
1135 mp3.year_string?mp3.year_string:"<blank>",
1136 mp3.year,
1137 mp3.track_string?mp3.track_string:"<blank>",
1138 mp3.tracknum,
1139 secs2str(mp3.length),
1140 mp3.length/1000,
1141 mp3.bitrate,
1142 mp3.frequency);
1145 return 0;
1148 #endif