gstreamer/ext/mad/gstmad.c

/* GStreamer
 * Copyright (C) <1999> Erik Walthinsen <omega@cse.ogi.edu>
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Library General Public
 * License as published by the Free Software Foundation; either
 * version 2 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Library General Public License for more details.
 *
 * You should have received a copy of the GNU Library General Public
 * License along with this library; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 02111-1307, USA.
 */

/**
 * SECTION:element-mad
 * @see_also: lame
 *
 * MP3 audio decoder.
 *
 * <refsect2>
 * <title>Example pipelines</title>
 * |[
 * gst-launch filesrc location=music.mp3 ! mpegaudioparse ! mad ! audioconvert ! audioresample ! autoaudiosink
 * ]| Decode and play the mp3 file
 * </refsect2>
 */

#ifdef HAVE_CONFIG_H
#include "config.h"
#endif

#include <stdlib.h>
#include <string.h>
#include "gstmad.h"
#include <gst/audio/audio.h>

enum
{
  ARG_0,
  ARG_HALF,
  ARG_IGNORE_CRC
};

GST_DEBUG_CATEGORY_STATIC (mad_debug);
#define GST_CAT_DEFAULT mad_debug

static GstStaticPadTemplate mad_src_template_factory =
GST_STATIC_PAD_TEMPLATE ("src",
    GST_PAD_SRC,
    GST_PAD_ALWAYS,
    GST_STATIC_CAPS ("audio/x-raw, "
        "format = (string) " GST_AUDIO_NE (S32) ", "
        "layout = (string) interleaved, "
        "rate = (int) { 8000, 11025, 12000, 16000, 22050, 24000, 32000, 44100, 48000 }, "
        "channels = (int) [ 1, 2 ]")
    );

/* FIXME: make three caps, for mpegversion 1, 2 and 2.5 */
static GstStaticPadTemplate mad_sink_template_factory =
GST_STATIC_PAD_TEMPLATE ("sink",
    GST_PAD_SINK,
    GST_PAD_ALWAYS,
    GST_STATIC_CAPS ("audio/mpeg, "
        "mpegversion = (int) 1, "
        "layer = (int) [ 1, 3 ], "
        "rate = (int) { 8000, 11025, 12000, 16000, 22050, 24000, 32000, 44100, 48000 }, "
        "channels = (int) [ 1, 2 ]")
    );


static gboolean gst_mad_start (GstAudioDecoder * dec);
static gboolean gst_mad_stop (GstAudioDecoder * dec);
static gboolean gst_mad_parse (GstAudioDecoder * dec, GstAdapter * adapter,
    gint * offset, gint * length);
static GstFlowReturn gst_mad_handle_frame (GstAudioDecoder * dec,
    GstBuffer * buffer);
static void gst_mad_flush (GstAudioDecoder * dec, gboolean hard);

static void gst_mad_set_property (GObject * object, guint prop_id,
    const GValue * value, GParamSpec * pspec);
static void gst_mad_get_property (GObject * object, guint prop_id,
    GValue * value, GParamSpec * pspec);

G_DEFINE_TYPE (GstMad, gst_mad, GST_TYPE_AUDIO_DECODER);

static void
gst_mad_class_init (GstMadClass * klass)
{
  GObjectClass *gobject_class = (GObjectClass *) klass;
  GstElementClass *element_class = (GstElementClass *) klass;
  GstAudioDecoderClass *base_class = (GstAudioDecoderClass *) klass;

  base_class->start = GST_DEBUG_FUNCPTR (gst_mad_start);
  base_class->stop = GST_DEBUG_FUNCPTR (gst_mad_stop);
  base_class->parse = GST_DEBUG_FUNCPTR (gst_mad_parse);
  base_class->handle_frame = GST_DEBUG_FUNCPTR (gst_mad_handle_frame);
  base_class->flush = GST_DEBUG_FUNCPTR (gst_mad_flush);

  gobject_class->set_property = gst_mad_set_property;
  gobject_class->get_property = gst_mad_get_property;

  /* init properties */
  /* currently, string representations are used, we might want to change that */
  /* FIXME: descriptions need to be more technical,
   * default values and ranges need to be selected right */
  g_object_class_install_property (gobject_class, ARG_HALF,
      g_param_spec_boolean ("half", "Half", "Generate PCM at 1/2 sample rate",
          FALSE, G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
  g_object_class_install_property (gobject_class, ARG_IGNORE_CRC,
      g_param_spec_boolean ("ignore-crc", "Ignore CRC", "Ignore CRC errors",
          TRUE, G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));

  gst_element_class_add_pad_template (element_class,
      gst_static_pad_template_get (&mad_sink_template_factory));
  gst_element_class_add_pad_template (element_class,
      gst_static_pad_template_get (&mad_src_template_factory));

  gst_element_class_set_details_simple (element_class, "mad mp3 decoder",
      "Codec/Decoder/Audio",
      "Uses mad code to decode mp3 streams", "Wim Taymans <wim@fluendo.com>");
}

static void
gst_mad_init (GstMad * mad)
{
  GstAudioDecoder *dec;

  dec = GST_AUDIO_DECODER (mad);
  gst_audio_decoder_set_tolerance (dec, 20 * GST_MSECOND);

  mad->half = FALSE;
  mad->ignore_crc = TRUE;
}

static gboolean
gst_mad_start (GstAudioDecoder * dec)
{
  GstMad *mad = GST_MAD (dec);
  guint options = 0;

  GST_DEBUG_OBJECT (dec, "start");
  mad_stream_init (&mad->stream);
  mad_frame_init (&mad->frame);
  mad_synth_init (&mad->synth);
  mad->rate = 0;
  mad->channels = 0;
  mad->caps_set = FALSE;
  mad->frame.header.samplerate = 0;
  if (mad->ignore_crc)
    options |= MAD_OPTION_IGNORECRC;
  if (mad->half)
    options |= MAD_OPTION_HALFSAMPLERATE;
  mad_stream_options (&mad->stream, options);
  mad->header.mode = -1;
  mad->header.emphasis = -1;

  /* call upon legacy upstream byte support (e.g. seeking) */
  gst_audio_decoder_set_byte_time (dec, TRUE);

  return TRUE;
}

static gboolean
gst_mad_stop (GstAudioDecoder * dec)
{
  GstMad *mad = GST_MAD (dec);

  GST_DEBUG_OBJECT (dec, "stop");
  mad_synth_finish (&mad->synth);
  mad_frame_finish (&mad->frame);
  mad_stream_finish (&mad->stream);

  return TRUE;
}

static inline gint32
scale (mad_fixed_t sample)
{
#if MAD_F_FRACBITS < 28
  /* round */
  sample += (1L << (28 - MAD_F_FRACBITS - 1));
#endif

  /* clip */
  if (sample >= MAD_F_ONE)
    sample = MAD_F_ONE - 1;
  else if (sample < -MAD_F_ONE)
    sample = -MAD_F_ONE;

#if MAD_F_FRACBITS < 28
  /* quantize */
  sample >>= (28 - MAD_F_FRACBITS);
#endif

  /* convert from 29 bits to 32 bits */
  return (gint32) (sample << 3);
}

/* internal function to check if the header has changed and thus the
 * caps need to be reset.  Only call during normal mode, not resyncing */
static void
gst_mad_check_caps_reset (GstMad * mad)
{
  guint nchannels;
  guint rate;

  nchannels = MAD_NCHANNELS (&mad->frame.header);

#if MAD_VERSION_MINOR <= 12
  rate = mad->header.sfreq;
#else
  rate = mad->frame.header.samplerate;
#endif

  /* rate and channels are not supposed to change in a continuous stream,
   * so check this first before doing anything */

  /* only set caps if they weren't already set for this continuous stream */
  if (!gst_pad_has_current_caps (GST_AUDIO_DECODER_SRC_PAD (mad))
      || mad->channels != nchannels || mad->rate != rate) {
    GstCaps *caps;

    if (mad->caps_set) {
      GST_DEBUG_OBJECT (mad, "Header changed from %d Hz/%d ch to %d Hz/%d ch, "
          "failed sync after seek ?", mad->rate, mad->channels, rate,
          nchannels);
      /* we're conservative on stream changes. However, our *initial* caps
       * might have been wrong as well - mad ain't perfect in syncing. So,
       * we count caps changes and change if we pass a limit treshold (3). */
      if (nchannels != mad->pending_channels || rate != mad->pending_rate) {
        mad->times_pending = 0;
        mad->pending_channels = nchannels;
        mad->pending_rate = rate;
      }
      if (++mad->times_pending < 3)
        return;
    }

    if (mad->stream.options & MAD_OPTION_HALFSAMPLERATE)
      rate >>= 1;

    /* we set the caps even when the pad is not connected so they
     * can be gotten for streaminfo */
    caps = gst_caps_new_simple ("audio/x-raw",
        "format", G_TYPE_STRING, GST_AUDIO_NE (S32),
        "layout", G_TYPE_STRING, "interleaved",
        "signed", G_TYPE_BOOLEAN, TRUE,
        "width", G_TYPE_INT, 32,
        "depth", G_TYPE_INT, 32,
        "rate", G_TYPE_INT, rate, "channels", G_TYPE_INT, nchannels, NULL);

    if (nchannels > 1)
      gst_caps_set_simple (caps, "channel-mask", GST_TYPE_BITMASK,
          GST_AUDIO_CHANNEL_POSITION_MASK (FRONT_LEFT) |
          GST_AUDIO_CHANNEL_POSITION_MASK (FRONT_RIGHT), NULL);

    gst_audio_decoder_set_outcaps (GST_AUDIO_DECODER (mad), caps);
    gst_caps_unref (caps);

    mad->caps_set = TRUE;
    mad->channels = nchannels;
    mad->rate = rate;
  }
}

/* FIXME: does this work properly at all? filesrc ! mad ! pulsesink fails */
static GstFlowReturn
gst_mad_parse (GstAudioDecoder * dec, GstAdapter * adapter,
    gint * _offset, gint * len)
{
  GstMad *mad;
  GstFlowReturn ret = GST_FLOW_EOS;
  gint av, size, offset, prev_offset, consumed = 0;
  const guint8 *data;

  mad = GST_MAD (dec);

  /* we basically let mad library do parsing,
   * and translate that back to baseclass.
   * if a frame is found (and also decoded), subsequent handle_frame
   * only needs to synthesize it */

  prev_offset = -1;
  offset = 0;
  av = gst_adapter_available (adapter);
  data = gst_adapter_map (adapter, av);

  while (offset < av) {
    size = MIN (MAD_BUFFER_MDLEN * 3, av - offset);

    /* check for mad asking too much */
    if (offset == prev_offset) {
      if (G_UNLIKELY (offset + size < av)) {
        /* mad should not do this, so really fatal */
        GST_ELEMENT_ERROR (mad, STREAM, DECODE, (NULL),
            ("mad claims to need more data than %u bytes", size));
        ret = GST_FLOW_ERROR;
        goto exit;
      } else {
        break;
      }
    }

    /* only feed that much to mad at a time */
    mad_stream_buffer (&mad->stream, data + offset, size);
    prev_offset = offset;

    while (offset - prev_offset < size) {
      consumed = 0;

      GST_LOG_OBJECT (mad, "decoding the header now");
      if (mad_header_decode (&mad->frame.header, &mad->stream) == -1) {
        if (mad->stream.error == MAD_ERROR_BUFLEN) {
          GST_LOG_OBJECT (mad,
              "not enough data in tempbuffer (%d), breaking to get more", size);
          break;
        } else {
          GST_WARNING_OBJECT (mad, "mad_header_decode had an error: %s",
              mad_stream_errorstr (&mad->stream));
        }
      }

      GST_LOG_OBJECT (mad, "parsing and decoding one frame now");
      if (mad_frame_decode (&mad->frame, &mad->stream) == -1) {
        GST_LOG_OBJECT (mad, "got error %d", mad->stream.error);

        /* not enough data, need to wait for next buffer? */
        if (mad->stream.error == MAD_ERROR_BUFLEN) {
          if (mad->stream.next_frame == data) {
            GST_LOG_OBJECT (mad,
                "not enough data in tempbuffer (%d), breaking to get more",
                size);
            break;
          } else {
            GST_LOG_OBJECT (mad, "sync error, flushing unneeded data");
            goto flush;
          }
        } else if (mad->stream.error == MAD_ERROR_BADDATAPTR) {
          /* Flush data */
          goto flush;
        } else {
          GST_WARNING_OBJECT (mad, "mad_frame_decode had an error: %s",
              mad_stream_errorstr (&mad->stream));
          if (!MAD_RECOVERABLE (mad->stream.error)) {
            /* well, all may be well enough bytes later on ... */
            GST_AUDIO_DECODER_ERROR (mad, 1, STREAM, DECODE, (NULL),
                ("mad error: %s", mad_stream_errorstr (&mad->stream)), ret);
            /* so make sure we really move along ... */
            if (!offset)
              offset++;
            goto exit;
          } else {
            const guint8 *before_sync, *after_sync;

            mad_frame_mute (&mad->frame);
            mad_synth_mute (&mad->synth);
            before_sync = mad->stream.ptr.byte;
            if (mad_stream_sync (&mad->stream) != 0)
              GST_WARNING_OBJECT (mad, "mad_stream_sync failed");
            after_sync = mad->stream.ptr.byte;
            /* a succesful resync should make us drop bytes as consumed, so
             * calculate from the byte pointers before and after resync */
            consumed = after_sync - before_sync;
            GST_DEBUG_OBJECT (mad, "resynchronization consumes %d bytes",
                consumed);
            GST_DEBUG_OBJECT (mad, "synced to data: 0x%0x 0x%0x",
                *mad->stream.ptr.byte, *(mad->stream.ptr.byte + 1));

            mad_stream_sync (&mad->stream);
            /* recoverable errors pass */
            goto flush;
          }
        }
      } else {
        /* decoding ok; found frame */
        ret = GST_FLOW_OK;
      }
    flush:
      if (consumed == 0) {
        consumed = mad->stream.next_frame - (data + offset);
        g_assert (consumed >= 0);
      }

      if (ret == GST_FLOW_OK)
        goto exit;

      offset += consumed;
    }
  }

exit:

  gst_adapter_unmap (adapter);

  *_offset = offset;
  *len = consumed;

  return ret;
}

static GstFlowReturn
gst_mad_handle_frame (GstAudioDecoder * dec, GstBuffer * buffer)
{
  GstMad *mad;
  GstFlowReturn ret = GST_FLOW_EOS;
  GstBuffer *outbuffer;
  guint nsamples;
  gint32 *outdata;
  mad_fixed_t const *left_ch, *right_ch;

  mad = GST_MAD (dec);

  /* no fancy draining */
  if (G_UNLIKELY (!buffer))
    return GST_FLOW_OK;

  /* _parse prepared a frame */
  nsamples = MAD_NSBSAMPLES (&mad->frame.header) *
      (mad->stream.options & MAD_OPTION_HALFSAMPLERATE ? 16 : 32);
  GST_LOG_OBJECT (mad, "mad frame with %d samples", nsamples);

  /* arrange for initial caps before pushing data,
   * and update later on if needed */
  gst_mad_check_caps_reset (mad);

  mad_synth_frame (&mad->synth, &mad->frame);
  left_ch = mad->synth.pcm.samples[0];
  right_ch = mad->synth.pcm.samples[1];

  outbuffer = gst_buffer_new_and_alloc (nsamples * mad->channels * 4);

  outdata = gst_buffer_map (outbuffer, NULL, NULL, GST_MAP_WRITE);

  /* output sample(s) in 16-bit signed native-endian PCM */
  if (mad->channels == 1) {
    gint count = nsamples;

    while (count--) {
      *outdata++ = scale (*left_ch++) & 0xffffffff;
    }
  } else {
    gint count = nsamples;

    while (count--) {
      *outdata++ = scale (*left_ch++) & 0xffffffff;
      *outdata++ = scale (*right_ch++) & 0xffffffff;
    }
  }

  gst_buffer_unmap (outbuffer, outdata, -1);

  ret = gst_audio_decoder_finish_frame (dec, outbuffer, 1);

  return ret;
}

static void
gst_mad_flush (GstAudioDecoder * dec, gboolean hard)
{
  GstMad *mad;

  mad = GST_MAD (dec);
  if (hard) {
    mad_frame_mute (&mad->frame);
    mad_synth_mute (&mad->synth);
  }
}

static void
gst_mad_set_property (GObject * object, guint prop_id,
    const GValue * value, GParamSpec * pspec)
{
  GstMad *mad;

  mad = GST_MAD (object);

  switch (prop_id) {
    case ARG_HALF:
      mad->half = g_value_get_boolean (value);
      break;
    case ARG_IGNORE_CRC:
      mad->ignore_crc = g_value_get_boolean (value);
      break;
    default:
      G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
      break;
  }
}

static void
gst_mad_get_property (GObject * object, guint prop_id,
    GValue * value, GParamSpec * pspec)
{
  GstMad *mad;

  mad = GST_MAD (object);

  switch (prop_id) {
    case ARG_HALF:
      g_value_set_boolean (value, mad->half);
      break;
    case ARG_IGNORE_CRC:
      g_value_set_boolean (value, mad->ignore_crc);
      break;
    default:
      G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
      break;
  }
}

/* plugin initialisation */

static gboolean
plugin_init (GstPlugin * plugin)
{
  GST_DEBUG_CATEGORY_INIT (mad_debug, "mad", 0, "mad mp3 decoding");

  /* FIXME 0.11: rename to something better like madmp3dec or madmpegaudiodec
   * or so? */
  return gst_element_register (plugin, "mad", GST_RANK_SECONDARY,
      gst_mad_get_type ());
}

GST_PLUGIN_DEFINE (GST_VERSION_MAJOR,
    GST_VERSION_MINOR,
    "mad",
    "mp3 decoding based on the mad library",
    plugin_init, VERSION, "GPL", GST_PACKAGE_NAME, GST_PACKAGE_ORIGIN);