1 /* MP3 decoding plugin for GStreamer using the mpg123 library
2 * Copyright (C) 2012 Carlos Rafael Giani
4 * This library is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU Lesser General Public
6 * License as published by the Free Software Foundation; either
7 * version 2.1 of the License, or (at your option) any later version.
9 * This library is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * Lesser General Public License for more details.
14 * You should have received a copy of the GNU Lesser General Public
15 * License along with this library; if not, write to the Free Software
16 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
20 * SECTION: element-mpg123audiodec
21 * @see_also: lamemp3enc, mad
23 * Audio decoder for MPEG-1 layer 1/2/3 audio data.
26 * <title>Example pipelines</title>
28 * gst-launch filesrc location=music.mp3 ! mpegaudioparse ! mpg123audiodec ! audioconvert ! audioresample ! autoaudiosink
29 * ]| Decode and play the mp3 file
37 #include "gstmpg123audiodec.h"
42 GST_DEBUG_CATEGORY_STATIC (mpg123_debug);
43 #define GST_CAT_DEFAULT mpg123_debug
45 /* Omitted sample formats that mpg123 supports (or at least can support):
46 * - 8bit integer signed
47 * - 8bit integer unsigned
52 * The first four formats are not supported by the GstAudioDecoder base class.
53 * (The internal gst_audio_format_from_caps_structure() call fails.)
55 * The 64bit float issue is tricky. mpg123 actually decodes to "real",
56 * not necessarily to "float".
58 * "real" can be fixed point, 32bit float, 64bit float. There seems to be
59 * no way how to find out which one of them is actually used.
61 * However, in all known installations, "real" equals 32bit float, so that's
64 static GstStaticPadTemplate static_sink_template =
65 GST_STATIC_PAD_TEMPLATE ("sink",
68 GST_STATIC_CAPS ("audio/mpeg, "
69 "mpegversion = (int) { 1 }, "
70 "layer = (int) [ 1, 3 ], "
71 "rate = (int) { 8000, 11025, 12000, 16000, 22050, 24000, 32000, 44100, 48000 }, "
72 "channels = (int) [ 1, 2 ], " "parsed = (boolean) true ")
75 static gboolean gst_mpg123_audio_dec_start (GstAudioDecoder * dec);
76 static gboolean gst_mpg123_audio_dec_stop (GstAudioDecoder * dec);
77 static GstFlowReturn gst_mpg123_audio_dec_push_decoded_bytes (GstMpg123AudioDec
78 * mpg123_decoder, unsigned char const *decoded_bytes,
79 size_t const num_decoded_bytes);
80 static GstFlowReturn gst_mpg123_audio_dec_handle_frame (GstAudioDecoder * dec,
81 GstBuffer * input_buffer);
82 static gboolean gst_mpg123_audio_dec_set_format (GstAudioDecoder * dec,
83 GstCaps * input_caps);
84 static void gst_mpg123_audio_dec_flush (GstAudioDecoder * dec, gboolean hard);
86 G_DEFINE_TYPE (GstMpg123AudioDec, gst_mpg123_audio_dec, GST_TYPE_AUDIO_DECODER);
89 gst_mpg123_audio_dec_class_init (GstMpg123AudioDecClass * klass)
91 GstAudioDecoderClass *base_class;
92 GstElementClass *element_class;
93 GstPadTemplate *src_template, *sink_template;
96 GST_DEBUG_CATEGORY_INIT (mpg123_debug, "mpg123", 0, "mpg123 mp3 decoder");
98 base_class = GST_AUDIO_DECODER_CLASS (klass);
99 element_class = GST_ELEMENT_CLASS (klass);
101 gst_element_class_set_static_metadata (element_class,
102 "mpg123 mp3 decoder",
103 "Codec/Decoder/Audio",
104 "Decodes mp3 streams using the mpg123 library",
105 "Carlos Rafael Giani <dv@pseudoterminal.org>");
107 /* Not using static pad template for srccaps, since the comma-separated list
108 * of formats needs to be created depending on whatever mpg123 supports */
110 const int *format_list;
111 const long *rates_list;
114 GstCaps *src_template_caps;
116 s = g_string_new ("audio/x-raw, ");
118 mpg123_encodings (&format_list, &num);
119 g_string_append (s, "format = { ");
120 for (i = 0; i < num; ++i) {
121 switch (format_list[i]) {
122 case MPG123_ENC_SIGNED_16:
123 g_string_append (s, (i > 0) ? ", " : "");
124 g_string_append (s, GST_AUDIO_NE (S16));
126 case MPG123_ENC_UNSIGNED_16:
127 g_string_append (s, (i > 0) ? ", " : "");
128 g_string_append (s, GST_AUDIO_NE (U16));
130 case MPG123_ENC_SIGNED_24:
131 g_string_append (s, (i > 0) ? ", " : "");
132 g_string_append (s, GST_AUDIO_NE (S24));
134 case MPG123_ENC_UNSIGNED_24:
135 g_string_append (s, (i > 0) ? ", " : "");
136 g_string_append (s, GST_AUDIO_NE (U24));
138 case MPG123_ENC_SIGNED_32:
139 g_string_append (s, (i > 0) ? ", " : "");
140 g_string_append (s, GST_AUDIO_NE (S32));
142 case MPG123_ENC_UNSIGNED_32:
143 g_string_append (s, (i > 0) ? ", " : "");
144 g_string_append (s, GST_AUDIO_NE (U32));
146 case MPG123_ENC_FLOAT_32:
147 g_string_append (s, (i > 0) ? ", " : "");
148 g_string_append (s, GST_AUDIO_NE (F32));
151 GST_DEBUG ("Ignoring mpg123 format %d", format_list[i]);
155 g_string_append (s, " }, ");
157 mpg123_rates (&rates_list, &num);
158 g_string_append (s, "rate = (int) { ");
159 for (i = 0; i < num; ++i) {
160 g_string_append_printf (s, "%s%lu", (i > 0) ? ", " : "", rates_list[i]);
162 g_string_append (s, "}, ");
164 g_string_append (s, "channels = (int) [ 1, 2 ], ");
165 g_string_append (s, "layout = (string) interleaved");
167 src_template_caps = gst_caps_from_string (s->str);
168 src_template = gst_pad_template_new ("src", GST_PAD_SRC, GST_PAD_ALWAYS,
171 g_string_free (s, TRUE);
174 sink_template = gst_static_pad_template_get (&static_sink_template);
176 gst_element_class_add_pad_template (element_class, sink_template);
177 gst_element_class_add_pad_template (element_class, src_template);
179 base_class->start = GST_DEBUG_FUNCPTR (gst_mpg123_audio_dec_start);
180 base_class->stop = GST_DEBUG_FUNCPTR (gst_mpg123_audio_dec_stop);
181 base_class->handle_frame =
182 GST_DEBUG_FUNCPTR (gst_mpg123_audio_dec_handle_frame);
183 base_class->set_format = GST_DEBUG_FUNCPTR (gst_mpg123_audio_dec_set_format);
184 base_class->flush = GST_DEBUG_FUNCPTR (gst_mpg123_audio_dec_flush);
186 error = mpg123_init ();
187 if (G_UNLIKELY (error != MPG123_OK))
188 GST_ERROR ("Could not initialize mpg123 library: %s",
189 mpg123_plain_strerror (error));
191 GST_INFO ("mpg123 library initialized");
196 gst_mpg123_audio_dec_init (GstMpg123AudioDec * mpg123_decoder)
198 mpg123_decoder->handle = NULL;
203 gst_mpg123_audio_dec_start (GstAudioDecoder * dec)
205 GstMpg123AudioDec *mpg123_decoder;
208 mpg123_decoder = GST_MPG123_AUDIO_DEC (dec);
211 mpg123_decoder->handle = mpg123_new (NULL, &error);
212 mpg123_decoder->has_next_audioinfo = FALSE;
213 mpg123_decoder->frame_offset = 0;
215 /* Initially, the mpg123 handle comes with a set of default formats
216 * supported. This clears this set. This is necessary, since only one
217 * format shall be supported (see set_format for more). */
218 mpg123_format_none (mpg123_decoder->handle);
220 /* Built-in mpg123 support for gapless decoding is disabled for now,
221 * since it does not work well with seeking */
222 mpg123_param (mpg123_decoder->handle, MPG123_REMOVE_FLAGS, MPG123_GAPLESS, 0);
223 /* Tells mpg123 to use a small read-ahead buffer for better MPEG sync;
224 * essential for MP3 radio streams */
225 mpg123_param (mpg123_decoder->handle, MPG123_ADD_FLAGS, MPG123_SEEKBUFFER, 0);
226 /* Sets the resync limit to the end of the stream (otherwise mpg123 may give
227 * up on decoding prematurely, especially with mp3 web radios) */
228 mpg123_param (mpg123_decoder->handle, MPG123_RESYNC_LIMIT, -1, 0);
229 /* Don't let mpg123 resample output */
230 mpg123_param (mpg123_decoder->handle, MPG123_REMOVE_FLAGS,
231 MPG123_AUTO_RESAMPLE, 0);
232 /* Don't let mpg123 print messages to stdout/stderr */
233 mpg123_param (mpg123_decoder->handle, MPG123_ADD_FLAGS, MPG123_QUIET, 0);
235 /* Open in feed mode (= encoded data is fed manually into the handle). */
236 error = mpg123_open_feed (mpg123_decoder->handle);
238 if (G_UNLIKELY (error != MPG123_OK)) {
239 GST_ELEMENT_ERROR (dec, LIBRARY, INIT, (NULL),
240 ("%s", mpg123_strerror (mpg123_decoder->handle)));
241 mpg123_close (mpg123_decoder->handle);
242 mpg123_delete (mpg123_decoder->handle);
243 mpg123_decoder->handle = NULL;
247 GST_INFO_OBJECT (dec, "mpg123 decoder started");
254 gst_mpg123_audio_dec_stop (GstAudioDecoder * dec)
256 GstMpg123AudioDec *mpg123_decoder = GST_MPG123_AUDIO_DEC (dec);
258 if (G_LIKELY (mpg123_decoder->handle != NULL)) {
259 mpg123_close (mpg123_decoder->handle);
260 mpg123_delete (mpg123_decoder->handle);
261 mpg123_decoder->handle = NULL;
264 GST_INFO_OBJECT (dec, "mpg123 decoder stopped");
271 gst_mpg123_audio_dec_push_decoded_bytes (GstMpg123AudioDec * mpg123_decoder,
272 unsigned char const *decoded_bytes, size_t const num_decoded_bytes)
274 GstBuffer *output_buffer;
275 GstAudioDecoder *dec;
277 output_buffer = NULL;
278 dec = GST_AUDIO_DECODER (mpg123_decoder);
280 if ((num_decoded_bytes == 0) || (decoded_bytes == NULL)) {
281 /* This occurs in the first few frames, which do not carry data; once
282 * MPG123_AUDIO_DEC_NEW_FORMAT is received, the empty frames stop occurring */
283 GST_DEBUG_OBJECT (mpg123_decoder,
284 "cannot decode yet, need more data -> no output buffer to push");
288 output_buffer = gst_buffer_new_allocate (NULL, num_decoded_bytes, NULL);
290 if (output_buffer == NULL) {
291 /* This is necessary to advance playback in time,
292 * even when nothing was decoded. */
293 return gst_audio_decoder_finish_frame (dec, NULL, 1);
297 if (gst_buffer_map (output_buffer, &info, GST_MAP_WRITE)) {
298 memcpy (info.data, decoded_bytes, num_decoded_bytes);
299 gst_buffer_unmap (output_buffer, &info);
301 GST_ERROR_OBJECT (mpg123_decoder, "gst_buffer_map() returned NULL");
302 gst_buffer_unref (output_buffer);
303 output_buffer = NULL;
306 return gst_audio_decoder_finish_frame (dec, output_buffer, 1);
312 gst_mpg123_audio_dec_handle_frame (GstAudioDecoder * dec,
313 GstBuffer * input_buffer)
315 GstMpg123AudioDec *mpg123_decoder;
317 unsigned char *decoded_bytes;
318 size_t num_decoded_bytes;
319 GstFlowReturn retval;
321 mpg123_decoder = GST_MPG123_AUDIO_DEC (dec);
323 g_assert (mpg123_decoder->handle != NULL);
325 /* The actual decoding */
327 /* feed input data (if there is any) */
328 if (G_LIKELY (input_buffer != NULL)) {
331 if (gst_buffer_map (input_buffer, &info, GST_MAP_READ)) {
332 mpg123_feed (mpg123_decoder->handle, info.data, info.size);
333 gst_buffer_unmap (input_buffer, &info);
335 GST_ERROR_OBJECT (mpg123_decoder, "gst_memory_map() failed");
336 return GST_FLOW_ERROR;
340 /* Try to decode a frame */
341 decoded_bytes = NULL;
342 num_decoded_bytes = 0;
343 decode_error = mpg123_decode_frame (mpg123_decoder->handle,
344 &mpg123_decoder->frame_offset, &decoded_bytes, &num_decoded_bytes);
347 retval = GST_FLOW_OK;
349 switch (decode_error) {
350 case MPG123_NEW_FORMAT:
351 /* As mentioned in gst_mpg123_audio_dec_set_format(), the next audioinfo
352 * is not set immediately; instead, the code waits for mpg123 to take
353 * note of the new format, and then sets the audioinfo. This fixes glitches
354 * with mp3s containing several format headers (for example, first half
355 * using 44.1kHz, second half 32 kHz) */
358 "mpg123 reported a new format -> setting next srccaps");
360 gst_mpg123_audio_dec_push_decoded_bytes (mpg123_decoder, decoded_bytes,
363 /* If there is a next audioinfo, use it, then set has_next_audioinfo to
364 * FALSE, to make sure gst_audio_decoder_set_output_format() isn't called
365 * again until set_format is called by the base class */
366 if (mpg123_decoder->has_next_audioinfo) {
367 if (!gst_audio_decoder_set_output_format (dec,
368 &(mpg123_decoder->next_audioinfo))) {
369 GST_WARNING_OBJECT (dec, "Unable to set output format");
370 retval = GST_FLOW_NOT_NEGOTIATED;
372 mpg123_decoder->has_next_audioinfo = FALSE;
377 case MPG123_NEED_MORE:
379 retval = gst_mpg123_audio_dec_push_decoded_bytes (mpg123_decoder,
380 decoded_bytes, num_decoded_bytes);
384 /* If this happens, then the upstream parser somehow missed the ending
385 * of the bitstream */
386 GST_LOG_OBJECT (dec, "mpg123 is done decoding");
387 gst_mpg123_audio_dec_push_decoded_bytes (mpg123_decoder, decoded_bytes,
389 retval = GST_FLOW_EOS;
394 /* Anything else is considered an error */
396 switch (decode_error) {
398 errcode = mpg123_errcode (mpg123_decoder->handle);
401 errcode = decode_error;
404 case MPG123_BAD_OUTFORMAT:{
405 GstCaps *input_caps =
406 gst_pad_get_current_caps (GST_AUDIO_DECODER_SINK_PAD (dec));
407 GST_ELEMENT_ERROR (dec, STREAM, FORMAT, (NULL),
408 ("Output sample format could not be used when trying to decode frame. "
409 "This is typically caused when the input caps (often the sample "
410 "rate) do not match the actual format of the audio data. "
411 "Input caps: %" GST_PTR_FORMAT, input_caps)
413 gst_caps_unref (input_caps);
417 char const *errmsg = mpg123_plain_strerror (errcode);
418 GST_ERROR_OBJECT (dec, "Reported error: %s", errmsg);
422 retval = GST_FLOW_ERROR;
431 gst_mpg123_audio_dec_set_format (GstAudioDecoder * dec, GstCaps * input_caps)
433 /* Using the parsed information upstream, and the list of allowed caps
434 * downstream, this code tries to find a suitable audio info. It is important
435 * to keep in mind that the rate and number of channels should never deviate
436 * from the one the bitstream has, otherwise mpg123 has to mix channels and/or
437 * resample (and as its docs say, its internal resampler is very crude). The
438 * sample format, however, can be chosen freely, because the MPEG specs do not
439 * mandate any special format. Therefore, rate and number of channels are taken
440 * from upstream (which parsed the MPEG frames, so the input_caps contain
441 * exactly the rate and number of channels the bitstream actually has), while
442 * the sample format is chosen by trying out all caps that are allowed by
443 * downstream. This way, the output is adjusted to what the downstream prefers.
445 * Also, the new output audio info is not set immediately. Instead, it is
446 * considered the "next audioinfo". The code waits for mpg123 to notice the new
447 * format (= when mpg123_decode_frame() returns MPG123_AUDIO_DEC_NEW_FORMAT),
448 * and then sets the next audioinfo. Otherwise, the next audioinfo is set too
449 * soon, which may cause problems with mp3s containing several format headers.
450 * One example would be an mp3 with the first 30 seconds using 44.1 kHz, then
451 * the next 30 seconds using 32 kHz. Rare, but possible.
455 * 1. get rate and channels from input_caps
456 * 2. get allowed caps from src pad
457 * 3. for each structure in allowed caps:
459 * 3.2. if the combination of format with rate and channels is unsupported by
460 * mpg123, go to (3), or exit with error if there are no more structures
462 * 3.3. create next audioinfo out of rate,channels,format, and exit
467 GstMpg123AudioDec *mpg123_decoder;
468 GstCaps *allowed_srccaps;
470 gboolean match_found = FALSE;
472 mpg123_decoder = GST_MPG123_AUDIO_DEC (dec);
474 g_assert (mpg123_decoder->handle != NULL);
476 mpg123_decoder->has_next_audioinfo = FALSE;
478 /* Get rate and channels from input_caps */
480 GstStructure *structure;
481 gboolean err = FALSE;
483 /* Only the first structure is used (multiple
484 * input caps structures don't make sense */
485 structure = gst_caps_get_structure (input_caps, 0);
487 if (!gst_structure_get_int (structure, "rate", &rate)) {
489 GST_ERROR_OBJECT (dec, "Input caps do not have a rate value");
491 if (!gst_structure_get_int (structure, "channels", &channels)) {
493 GST_ERROR_OBJECT (dec, "Input caps do not have a channel value");
500 /* Get the caps that are allowed by downstream */
502 GstCaps *allowed_srccaps_unnorm =
503 gst_pad_get_allowed_caps (GST_AUDIO_DECODER_SRC_PAD (dec));
504 allowed_srccaps = gst_caps_normalize (allowed_srccaps_unnorm);
507 /* Go through all allowed caps, pick the first one that matches */
508 for (structure_nr = 0; structure_nr < gst_caps_get_size (allowed_srccaps);
510 GstStructure *structure;
511 gchar const *format_str;
512 GstAudioFormat format;
515 structure = gst_caps_get_structure (allowed_srccaps, structure_nr);
517 format_str = gst_structure_get_string (structure, "format");
518 if (format_str == NULL) {
519 GST_DEBUG_OBJECT (dec, "Could not get format from src caps");
523 format = gst_audio_format_from_string (format_str);
524 if (format == GST_AUDIO_FORMAT_UNKNOWN) {
525 GST_DEBUG_OBJECT (dec, "Unknown format %s", format_str);
530 case GST_AUDIO_FORMAT_S16:
531 encoding = MPG123_ENC_SIGNED_16;
533 case GST_AUDIO_FORMAT_S24:
534 encoding = MPG123_ENC_SIGNED_24;
536 case GST_AUDIO_FORMAT_S32:
537 encoding = MPG123_ENC_SIGNED_32;
539 case GST_AUDIO_FORMAT_U16:
540 encoding = MPG123_ENC_UNSIGNED_16;
542 case GST_AUDIO_FORMAT_U24:
543 encoding = MPG123_ENC_UNSIGNED_24;
545 case GST_AUDIO_FORMAT_U32:
546 encoding = MPG123_ENC_UNSIGNED_32;
548 case GST_AUDIO_FORMAT_F32:
549 encoding = MPG123_ENC_FLOAT_32;
552 GST_DEBUG_OBJECT (dec,
553 "Format %s in srccaps is not supported", format_str);
560 /* Cleanup old formats & set new one */
561 mpg123_format_none (mpg123_decoder->handle);
562 err = mpg123_format (mpg123_decoder->handle, rate, channels, encoding);
563 if (err != MPG123_OK) {
564 GST_DEBUG_OBJECT (dec,
565 "mpg123 cannot use caps %" GST_PTR_FORMAT
566 " because mpg123_format() failed: %s", structure,
567 mpg123_strerror (mpg123_decoder->handle));
572 gst_audio_info_init (&(mpg123_decoder->next_audioinfo));
573 gst_audio_info_set_format (&(mpg123_decoder->next_audioinfo), format, rate,
575 GST_LOG_OBJECT (dec, "The next audio format is: %s, %u Hz, %u channels",
576 format_str, rate, channels);
577 mpg123_decoder->has_next_audioinfo = TRUE;
584 gst_caps_unref (allowed_srccaps);
591 gst_mpg123_audio_dec_flush (GstAudioDecoder * dec, gboolean hard)
594 GstMpg123AudioDec *mpg123_decoder;
598 GST_LOG_OBJECT (dec, "Flushing decoder");
600 mpg123_decoder = GST_MPG123_AUDIO_DEC (dec);
602 g_assert (mpg123_decoder->handle != NULL);
604 /* Flush by reopening the feed */
605 mpg123_close (mpg123_decoder->handle);
606 error = mpg123_open_feed (mpg123_decoder->handle);
608 if (G_UNLIKELY (error != MPG123_OK)) {
609 GST_ELEMENT_ERROR (dec, LIBRARY, INIT, (NULL),
610 ("Error while reopening mpg123 feed: %s",
611 mpg123_plain_strerror (error)));
612 mpg123_close (mpg123_decoder->handle);
613 mpg123_delete (mpg123_decoder->handle);
614 mpg123_decoder->handle = NULL;
617 mpg123_decoder->has_next_audioinfo = FALSE;
619 /* opening/closing feeds do not affect the format defined by the
620 * mpg123_format() call that was made in gst_mpg123_audio_dec_set_format(),
621 * and since the up/downstream caps are not expected to change here, no
622 * mpg123_format() calls are done */
626 plugin_init (GstPlugin * plugin)
628 return gst_element_register (plugin, "mpg123audiodec",
629 GST_RANK_MARGINAL, gst_mpg123_audio_dec_get_type ());
632 GST_PLUGIN_DEFINE (GST_VERSION_MAJOR,
634 mpg123, "mp3 decoding based on the mpg123 library",
635 plugin_init, VERSION, "LGPL", GST_PACKAGE_NAME, GST_PACKAGE_ORIGIN)