gstreamer/gst/audioresample/gstaudioresample.c

844 lines
26 KiB
C
Raw Normal View History

/* GStreamer
* Copyright (C) 1999 Erik Walthinsen <omega@cse.ogi.edu>
* Copyright (C) 2003,2004 David A. Schleef <ds@schleef.org>
*
* 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.
*/
/* Element-Checklist-Version: 5 */
/**
* SECTION:element-audioresample
*
* <refsect2>
* Audioresample resamples raw audio buffers to different sample rates using
* a configurable windowing function to enhance quality.
* <title>Example launch line</title>
* <para>
* <programlisting>
* gst-launch -v filesrc location=sine.ogg ! oggdemux ! vorbisdec ! audioconvert ! audioresample ! audio/x-raw-int, rate=8000 ! alsasink
* </programlisting>
* Decode an Ogg/Vorbis downsample to 8Khz and play sound through alsa.
* To create the Ogg/Vorbis file refer to the documentation of vorbisenc.
* </para>
* </refsect2>
*
* Last reviewed on 2006-03-02 (0.10.4)
*/
#ifdef HAVE_CONFIG_H
#include "config.h"
#endif
#include <string.h>
#include <math.h>
/*#define DEBUG_ENABLED */
#include "gstaudioresample.h"
#include <gst/audio/audio.h>
#include <gst/base/gstbasetransform.h>
Correct all relevant warnings found by the sparse semantic code analyzer. This include marking several symbols static... Original commit message from CVS: * ext/alsa/gstalsamixertrack.c: (gst_alsa_mixer_track_get_type): * ext/alsa/gstalsasink.c: (set_hwparams): * ext/alsa/gstalsasrc.c: (set_hwparams): * ext/gio/gstgio.c: (gst_gio_uri_handler_get_uri): * ext/ogg/gstoggmux.h: * ext/ogg/gstogmparse.c: * gst-libs/gst/audio/audio.c: * gst-libs/gst/fft/kiss_fft_f64.c: (kiss_fft_f64_alloc): * gst-libs/gst/pbutils/missing-plugins.c: (gst_missing_uri_sink_message_new), (gst_missing_element_message_new), (gst_missing_decoder_message_new), (gst_missing_encoder_message_new): * gst-libs/gst/rtp/gstbasertppayload.c: * gst-libs/gst/rtp/gstrtcpbuffer.c: (gst_rtcp_packet_bye_get_reason): * gst/audioconvert/gstaudioconvert.c: * gst/audioresample/gstaudioresample.c: * gst/ffmpegcolorspace/imgconvert.c: * gst/playback/test.c: (gen_video_element), (gen_audio_element): * gst/typefind/gsttypefindfunctions.c: * gst/videoscale/vs_4tap.c: * gst/videoscale/vs_4tap.h: * sys/v4l/gstv4lelement.c: * sys/v4l/gstv4lsrc.c: (gst_v4lsrc_get_any_caps): * sys/v4l/v4l_calls.c: * sys/v4l/v4lsrc_calls.c: (gst_v4lsrc_capture_init), (gst_v4lsrc_try_capture): * sys/ximage/ximagesink.c: (gst_ximagesink_check_xshm_calls), (gst_ximagesink_ximage_new): * sys/xvimage/xvimagesink.c: (gst_xvimagesink_check_xshm_calls), (gst_xvimagesink_xvimage_new): * tests/check/elements/audioconvert.c: * tests/check/elements/audioresample.c: (fail_unless_perfect_stream): * tests/check/elements/audiotestsrc.c: (setup_audiotestsrc): * tests/check/elements/decodebin.c: * tests/check/elements/gdpdepay.c: (setup_gdpdepay), (setup_gdpdepay_streamheader): * tests/check/elements/gdppay.c: (setup_gdppay), (GST_START_TEST), (setup_gdppay_streamheader): * tests/check/elements/gnomevfssink.c: (setup_gnomevfssink): * tests/check/elements/multifdsink.c: (setup_multifdsink): * tests/check/elements/textoverlay.c: * tests/check/elements/videorate.c: (setup_videorate): * tests/check/elements/videotestsrc.c: (setup_videotestsrc): * tests/check/elements/volume.c: (setup_volume): * tests/check/elements/vorbisdec.c: (setup_vorbisdec): * tests/check/elements/vorbistag.c: * tests/check/generic/clock-selection.c: * tests/check/generic/states.c: (setup), (teardown): * tests/check/libs/cddabasesrc.c: * tests/check/libs/video.c: * tests/check/pipelines/gio.c: * tests/check/pipelines/oggmux.c: * tests/check/pipelines/simple-launch-lines.c: (simple_launch_lines_suite): * tests/check/pipelines/streamheader.c: * tests/check/pipelines/theoraenc.c: * tests/check/pipelines/vorbisdec.c: * tests/check/pipelines/vorbisenc.c: * tests/examples/seek/scrubby.c: * tests/examples/seek/seek.c: (query_positions_elems), (query_positions_pads): * tests/icles/stress-xoverlay.c: (myclock): Correct all relevant warnings found by the sparse semantic code analyzer. This include marking several symbols static, using NULL instead of 0 for pointers and using "foo (void)" instead of "foo ()" for declarations. * win32/common/libgstrtp.def: Add gst_rtp_buffer_set_extension_data to the symbol definition file.
2008-03-03 06:04:02 +00:00
GST_DEBUG_CATEGORY_STATIC (audioresample_debug);
#define GST_CAT_DEFAULT audioresample_debug
/* elementfactory information */
make GstElementDetails const Original commit message from CVS: * ext/alsa/gstalsamixerelement.c: * ext/alsa/gstalsasrc.c: * ext/cdparanoia/gstcdparanoiasrc.c: * ext/gnomevfs/gstgnomevfssink.c: * ext/gnomevfs/gstgnomevfssrc.c: * ext/ogg/gstoggdemux.c: * ext/ogg/gstoggmux.c: * ext/ogg/gstoggparse.c: * ext/ogg/gstogmparse.c: * ext/pango/gstclockoverlay.c: * ext/pango/gsttextoverlay.c: * ext/pango/gsttextrender.c: * ext/pango/gsttimeoverlay.c: * ext/theora/theoradec.c: * ext/theora/theoraenc.c: * ext/vorbis/vorbisdec.c: * ext/vorbis/vorbisenc.c: * gst-libs/gst/audio/gstaudiofilter.c: * gst-libs/gst/audio/gstaudiofiltertemplate.c: * gst/audioconvert/gstaudioconvert.c: * gst/audiorate/gstaudiorate.c: * gst/audioresample/gstaudioresample.c: * gst/audiotestsrc/gstaudiotestsrc.c: * gst/ffmpegcolorspace/gstffmpegcolorspace.c: * gst/playback/gstdecodebin.c: * gst/playback/gstplaybin.c: * gst/playback/gststreamselector.c: * gst/subparse/gstsubparse.c: * gst/tcp/gstmultifdsink.c: * gst/tcp/gsttcpclientsink.c: * gst/tcp/gsttcpclientsrc.c: * gst/tcp/gsttcpserversink.c: * gst/tcp/gsttcpserversrc.c: * gst/typefind/gsttypefindfunctions.c: (plugin_init): * gst/videorate/gstvideorate.c: * gst/videoscale/gstvideoscale.c: * gst/videotestsrc/gstvideotestsrc.c: * gst/volume/gstvolume.c: * sys/v4l/gstv4ljpegsrc.c: * sys/v4l/gstv4lmjpegsink.c: * sys/v4l/gstv4lmjpegsrc.c: * sys/v4l/gstv4lsrc.c: * sys/ximage/ximagesink.c: * sys/xvimage/xvimagesink.c: * tests/check/libs/cddabasesrc.c: make GstElementDetails const
2006-04-28 19:46:36 +00:00
static const GstElementDetails gst_audioresample_details =
GST_ELEMENT_DETAILS ("Audio scaler",
"Filter/Converter/Audio",
"Resample audio",
"David Schleef <ds@schleef.org>");
#define DEFAULT_FILTERLEN 16
enum
{
PROP_0,
PROP_FILTERLEN
};
#define SUPPORTED_CAPS \
GST_STATIC_CAPS ( \
"audio/x-raw-int, " \
"rate = (int) [ 1, MAX ], " \
"channels = (int) [ 1, MAX ], " \
"endianness = (int) BYTE_ORDER, " \
"width = (int) 16, " \
"depth = (int) 16, " \
"signed = (boolean) true;" \
"audio/x-raw-int, " \
"rate = (int) [ 1, MAX ], " \
"channels = (int) [ 1, MAX ], " \
"endianness = (int) BYTE_ORDER, " \
"width = (int) 32, " \
"depth = (int) 32, " \
"signed = (boolean) true;" \
"audio/x-raw-float, " \
"rate = (int) [ 1, MAX ], " \
"channels = (int) [ 1, MAX ], " \
"endianness = (int) BYTE_ORDER, " \
"width = (int) 32; " \
"audio/x-raw-float, " \
"rate = (int) [ 1, MAX ], " \
"channels = (int) [ 1, MAX ], " \
"endianness = (int) BYTE_ORDER, " \
"width = (int) 64" \
)
static GstStaticPadTemplate gst_audioresample_sink_template =
GST_STATIC_PAD_TEMPLATE ("sink",
GST_PAD_SINK, GST_PAD_ALWAYS, SUPPORTED_CAPS);
static GstStaticPadTemplate gst_audioresample_src_template =
GST_STATIC_PAD_TEMPLATE ("src",
GST_PAD_SRC, GST_PAD_ALWAYS, SUPPORTED_CAPS);
static void gst_audioresample_set_property (GObject * object,
guint prop_id, const GValue * value, GParamSpec * pspec);
static void gst_audioresample_get_property (GObject * object,
guint prop_id, GValue * value, GParamSpec * pspec);
/* vmethods */
static gboolean audioresample_get_unit_size (GstBaseTransform * base,
GstCaps * caps, guint * size);
static GstCaps *audioresample_transform_caps (GstBaseTransform * base,
GstPadDirection direction, GstCaps * caps);
static gboolean audioresample_transform_size (GstBaseTransform * trans,
GstPadDirection direction, GstCaps * incaps, guint insize,
GstCaps * outcaps, guint * outsize);
static gboolean audioresample_set_caps (GstBaseTransform * base,
GstCaps * incaps, GstCaps * outcaps);
static GstFlowReturn audioresample_pushthrough (GstAudioresample *
audioresample);
static GstFlowReturn audioresample_transform (GstBaseTransform * base,
GstBuffer * inbuf, GstBuffer * outbuf);
static gboolean audioresample_event (GstBaseTransform * base, GstEvent * event);
static gboolean audioresample_start (GstBaseTransform * base);
static gboolean audioresample_stop (GstBaseTransform * base);
static gboolean audioresample_query (GstPad * pad, GstQuery * query);
static const GstQueryType *audioresample_query_type (GstPad * pad);
#define DEBUG_INIT(bla) \
GST_DEBUG_CATEGORY_INIT (audioresample_debug, "audioresample", 0, "audio resampling element");
GST_BOILERPLATE_FULL (GstAudioresample, gst_audioresample, GstBaseTransform,
GST_TYPE_BASE_TRANSFORM, DEBUG_INIT);
static void
gst_audioresample_base_init (gpointer g_class)
{
GstElementClass *gstelement_class = GST_ELEMENT_CLASS (g_class);
gst_element_class_add_pad_template (gstelement_class,
gst_static_pad_template_get (&gst_audioresample_src_template));
gst_element_class_add_pad_template (gstelement_class,
gst_static_pad_template_get (&gst_audioresample_sink_template));
gst_element_class_set_details (gstelement_class, &gst_audioresample_details);
}
static void
gst_audioresample_class_init (GstAudioresampleClass * klass)
{
GObjectClass *gobject_class;
gobject_class = (GObjectClass *) klass;
gobject_class->set_property = gst_audioresample_set_property;
gobject_class->get_property = gst_audioresample_get_property;
g_object_class_install_property (gobject_class, PROP_FILTERLEN,
Use G_PARAM_STATIC_STRINGS everywhere for GParamSpecs that use static strings (i.e. all). This gives us less memory u... Original commit message from CVS: * configure.ac: * ext/alsa/gstalsamixerelement.c: (gst_alsa_mixer_element_class_init): * ext/alsa/gstalsasink.c: (gst_alsasink_class_init): * ext/alsa/gstalsasrc.c: (gst_alsasrc_class_init): * ext/cdparanoia/gstcdparanoiasrc.c: (gst_cd_paranoia_src_class_init): * ext/gio/gstgiosink.c: (gst_gio_sink_class_init): * ext/gio/gstgiosrc.c: (gst_gio_src_class_init): * ext/gio/gstgiostreamsink.c: (gst_gio_stream_sink_class_init): * ext/gio/gstgiostreamsrc.c: (gst_gio_stream_src_class_init): * ext/gnomevfs/gstgnomevfssink.c: (gst_gnome_vfs_sink_class_init): * ext/gnomevfs/gstgnomevfssrc.c: (gst_gnome_vfs_src_class_init): * ext/ogg/gstoggmux.c: (gst_ogg_mux_class_init): * ext/pango/gsttextoverlay.c: (gst_text_overlay_class_init): * ext/pango/gsttextrender.c: (gst_text_render_class_init): * ext/theora/theoradec.c: (gst_theora_dec_class_init): * ext/theora/theoraenc.c: (gst_theora_enc_class_init): * ext/theora/theoraparse.c: (gst_theora_parse_class_init): * ext/vorbis/vorbisenc.c: (gst_vorbis_enc_class_init): * gst-libs/gst/audio/gstaudiofiltertemplate.c: (gst_audio_filter_template_class_init): * gst-libs/gst/audio/gstbaseaudiosink.c: (gst_base_audio_sink_class_init): * gst-libs/gst/audio/gstbaseaudiosrc.c: (gst_base_audio_src_class_init): * gst-libs/gst/cdda/gstcddabasesrc.c: (gst_cdda_base_src_class_init): * gst-libs/gst/interfaces/mixertrack.c: (gst_mixer_track_class_init): * gst-libs/gst/rtp/gstbasertpdepayload.c: (gst_base_rtp_depayload_class_init): * gst-libs/gst/rtp/gstbasertppayload.c: (gst_basertppayload_class_init): * gst/audioconvert/gstaudioconvert.c: (gst_audio_convert_class_init): * gst/audiorate/gstaudiorate.c: (gst_audio_rate_class_init): * gst/audioresample/gstaudioresample.c: (gst_audioresample_class_init): * gst/audiotestsrc/gstaudiotestsrc.c: (gst_audio_test_src_class_init): * gst/gdp/gstgdppay.c: (gst_gdp_pay_class_init): * gst/playback/gstdecodebin2.c: (gst_decode_bin_class_init): * gst/playback/gstplaybasebin.c: (gst_play_base_bin_class_init), (preroll_unlinked): * gst/playback/gstplaybin.c: (gst_play_bin_class_init): * gst/playback/gstplaybin2.c: (gst_play_bin_class_init): * gst/playback/gstplaysink.c: (gst_play_sink_class_init): * gst/playback/gstqueue2.c: (gst_queue_class_init): * gst/playback/gststreaminfo.c: (gst_stream_info_class_init): * gst/playback/gststreamselector.c: (gst_selector_pad_class_init), (gst_stream_selector_class_init): * gst/playback/gsturidecodebin.c: (gst_uri_decode_bin_class_init): * gst/subparse/gstsubparse.c: (gst_sub_parse_class_init): * gst/tcp/gstmultifdsink.c: (gst_multi_fd_sink_class_init): * gst/tcp/gsttcpclientsink.c: (gst_tcp_client_sink_class_init): * gst/tcp/gsttcpclientsrc.c: (gst_tcp_client_src_class_init): * gst/tcp/gsttcpserversink.c: (gst_tcp_server_sink_class_init): * gst/tcp/gsttcpserversrc.c: (gst_tcp_server_src_class_init): * gst/videorate/gstvideorate.c: (gst_video_rate_class_init): * gst/videoscale/gstvideoscale.c: (gst_video_scale_class_init): * gst/videotestsrc/gstvideotestsrc.c: (gst_video_test_src_class_init): * gst/volume/gstvolume.c: (gst_volume_class_init): * sys/v4l/gstv4lelement.c: (gst_v4lelement_class_init): * sys/v4l/gstv4lmjpegsink.c: (gst_v4lmjpegsink_class_init): * sys/v4l/gstv4lmjpegsrc.c: (gst_v4lmjpegsrc_class_init): * sys/v4l/gstv4lsrc.c: (gst_v4lsrc_class_init): * sys/ximage/ximagesink.c: (gst_ximagesink_class_init): * sys/xvimage/xvimagesink.c: (gst_xvimagesink_class_init): Use G_PARAM_STATIC_STRINGS everywhere for GParamSpecs that use static strings (i.e. all). This gives us less memory usage, fewer allocations and thus less memory defragmentation. Depend on core CVS for this. Fixes bug #523806.
2008-03-22 15:00:47 +00:00
g_param_spec_int ("filter-length", "filter length",
"Length of the resample filter", 0, G_MAXINT, DEFAULT_FILTERLEN,
G_PARAM_READWRITE | G_PARAM_CONSTRUCT | G_PARAM_STATIC_STRINGS));
GST_BASE_TRANSFORM_CLASS (klass)->start =
GST_DEBUG_FUNCPTR (audioresample_start);
GST_BASE_TRANSFORM_CLASS (klass)->stop =
GST_DEBUG_FUNCPTR (audioresample_stop);
GST_BASE_TRANSFORM_CLASS (klass)->transform_size =
GST_DEBUG_FUNCPTR (audioresample_transform_size);
GST_BASE_TRANSFORM_CLASS (klass)->get_unit_size =
GST_DEBUG_FUNCPTR (audioresample_get_unit_size);
GST_BASE_TRANSFORM_CLASS (klass)->transform_caps =
GST_DEBUG_FUNCPTR (audioresample_transform_caps);
GST_BASE_TRANSFORM_CLASS (klass)->set_caps =
GST_DEBUG_FUNCPTR (audioresample_set_caps);
GST_BASE_TRANSFORM_CLASS (klass)->transform =
GST_DEBUG_FUNCPTR (audioresample_transform);
GST_BASE_TRANSFORM_CLASS (klass)->event =
GST_DEBUG_FUNCPTR (audioresample_event);
GST_BASE_TRANSFORM_CLASS (klass)->passthrough_on_same_caps = TRUE;
}
static void
gst_audioresample_init (GstAudioresample * audioresample,
GstAudioresampleClass * klass)
{
GstBaseTransform *trans;
trans = GST_BASE_TRANSFORM (audioresample);
/* buffer alloc passthrough is too impossible. FIXME, it
* is trivial in the passthrough case. */
gst_pad_set_bufferalloc_function (trans->sinkpad, NULL);
audioresample->filter_length = DEFAULT_FILTERLEN;
audioresample->need_discont = FALSE;
gst_pad_set_query_function (trans->srcpad, audioresample_query);
gst_pad_set_query_type_function (trans->srcpad, audioresample_query_type);
}
/* vmethods */
static gboolean
audioresample_start (GstBaseTransform * base)
{
GstAudioresample *audioresample = GST_AUDIORESAMPLE (base);
audioresample->resample = resample_new ();
audioresample->ts_offset = -1;
audioresample->offset = -1;
audioresample->next_ts = -1;
resample_set_filter_length (audioresample->resample,
audioresample->filter_length);
return TRUE;
}
static gboolean
audioresample_stop (GstBaseTransform * base)
{
GstAudioresample *audioresample = GST_AUDIORESAMPLE (base);
if (audioresample->resample) {
resample_free (audioresample->resample);
audioresample->resample = NULL;
}
gst_caps_replace (&audioresample->sinkcaps, NULL);
gst_caps_replace (&audioresample->srccaps, NULL);
return TRUE;
}
static gboolean
audioresample_get_unit_size (GstBaseTransform * base, GstCaps * caps,
guint * size)
{
gint width, channels;
GstStructure *structure;
gboolean ret;
g_assert (size);
/* this works for both float and int */
structure = gst_caps_get_structure (caps, 0);
ret = gst_structure_get_int (structure, "width", &width);
ret &= gst_structure_get_int (structure, "channels", &channels);
g_return_val_if_fail (ret, FALSE);
*size = width * channels / 8;
return TRUE;
}
static GstCaps *
audioresample_transform_caps (GstBaseTransform * base,
GstPadDirection direction, GstCaps * caps)
{
GstCaps *res;
GstStructure *structure;
/* transform caps gives one single caps so we can just replace
* the rate property with our range. */
res = gst_caps_copy (caps);
structure = gst_caps_get_structure (res, 0);
gst_structure_set (structure, "rate", GST_TYPE_INT_RANGE, 1, G_MAXINT, NULL);
return res;
}
static gboolean
resample_set_state_from_caps (ResampleState * state, GstCaps * incaps,
GstCaps * outcaps, gint * channels, gint * inrate, gint * outrate)
{
GstStructure *structure;
gboolean ret;
gint myinrate, myoutrate;
int mychannels;
gint width, depth;
ResampleFormat format;
GST_DEBUG ("incaps %" GST_PTR_FORMAT ", outcaps %"
GST_PTR_FORMAT, incaps, outcaps);
structure = gst_caps_get_structure (incaps, 0);
/* get width */
ret = gst_structure_get_int (structure, "width", &width);
if (!ret)
goto no_width;
/* figure out the format */
if (g_str_equal (gst_structure_get_name (structure), "audio/x-raw-float")) {
if (width == 32)
format = RESAMPLE_FORMAT_F32;
else if (width == 64)
format = RESAMPLE_FORMAT_F64;
else
goto wrong_depth;
} else {
/* for int, depth and width must be the same */
ret = gst_structure_get_int (structure, "depth", &depth);
if (!ret || width != depth)
goto not_equal;
if (width == 16)
format = RESAMPLE_FORMAT_S16;
else if (width == 32)
format = RESAMPLE_FORMAT_S32;
else
goto wrong_depth;
}
ret = gst_structure_get_int (structure, "rate", &myinrate);
ret &= gst_structure_get_int (structure, "channels", &mychannels);
if (!ret)
goto no_in_rate_channels;
structure = gst_caps_get_structure (outcaps, 0);
ret = gst_structure_get_int (structure, "rate", &myoutrate);
if (!ret)
goto no_out_rate;
if (channels)
*channels = mychannels;
if (inrate)
*inrate = myinrate;
if (outrate)
*outrate = myoutrate;
resample_set_format (state, format);
resample_set_n_channels (state, mychannels);
resample_set_input_rate (state, myinrate);
resample_set_output_rate (state, myoutrate);
return TRUE;
/* ERRORS */
no_width:
{
GST_DEBUG ("failed to get width from caps");
return FALSE;
}
not_equal:
{
GST_DEBUG ("width %d and depth %d must be the same", width, depth);
return FALSE;
}
wrong_depth:
{
GST_DEBUG ("unknown depth %d found", depth);
return FALSE;
}
no_in_rate_channels:
{
GST_DEBUG ("could not get input rate and channels");
return FALSE;
}
no_out_rate:
{
GST_DEBUG ("could not get output rate");
return FALSE;
}
}
static gboolean
audioresample_transform_size (GstBaseTransform * base,
GstPadDirection direction, GstCaps * caps, guint size, GstCaps * othercaps,
guint * othersize)
{
GstAudioresample *audioresample = GST_AUDIORESAMPLE (base);
ResampleState *state;
GstCaps *srccaps, *sinkcaps;
gboolean use_internal = FALSE; /* whether we use the internal state */
gboolean ret = TRUE;
GST_LOG_OBJECT (base, "asked to transform size %d in direction %s",
size, direction == GST_PAD_SINK ? "SINK" : "SRC");
if (direction == GST_PAD_SINK) {
sinkcaps = caps;
srccaps = othercaps;
} else {
sinkcaps = othercaps;
srccaps = caps;
}
/* if the caps are the ones that _set_caps got called with; we can use
* our own state; otherwise we'll have to create a state */
if (gst_caps_is_equal (sinkcaps, audioresample->sinkcaps) &&
gst_caps_is_equal (srccaps, audioresample->srccaps)) {
use_internal = TRUE;
state = audioresample->resample;
} else {
GST_DEBUG_OBJECT (audioresample,
"caps are not the set caps, creating state");
state = resample_new ();
resample_set_filter_length (state, audioresample->filter_length);
resample_set_state_from_caps (state, sinkcaps, srccaps, NULL, NULL, NULL);
}
if (direction == GST_PAD_SINK) {
/* asked to convert size of an incoming buffer */
*othersize = resample_get_output_size_for_input (state, size);
} else {
/* asked to convert size of an outgoing buffer */
*othersize = resample_get_input_size_for_output (state, size);
}
g_assert (*othersize % state->sample_size == 0);
/* we make room for one extra sample, given that the resampling filter
* can output an extra one for non-integral i_rate/o_rate */
GST_LOG_OBJECT (base, "transformed size %d to %d", size, *othersize);
if (!use_internal) {
resample_free (state);
}
return ret;
}
static gboolean
audioresample_set_caps (GstBaseTransform * base, GstCaps * incaps,
GstCaps * outcaps)
{
gboolean ret;
gint inrate, outrate;
int channels;
GstAudioresample *audioresample = GST_AUDIORESAMPLE (base);
GST_DEBUG_OBJECT (base, "incaps %" GST_PTR_FORMAT ", outcaps %"
GST_PTR_FORMAT, incaps, outcaps);
ret = resample_set_state_from_caps (audioresample->resample, incaps, outcaps,
&channels, &inrate, &outrate);
g_return_val_if_fail (ret, FALSE);
audioresample->channels = channels;
GST_DEBUG_OBJECT (audioresample, "set channels to %d", channels);
audioresample->i_rate = inrate;
GST_DEBUG_OBJECT (audioresample, "set i_rate to %d", inrate);
audioresample->o_rate = outrate;
GST_DEBUG_OBJECT (audioresample, "set o_rate to %d", outrate);
/* save caps so we can short-circuit in the size_transform if the caps
* are the same */
gst_caps_replace (&audioresample->sinkcaps, incaps);
gst_caps_replace (&audioresample->srccaps, outcaps);
return TRUE;
}
static gboolean
audioresample_event (GstBaseTransform * base, GstEvent * event)
{
GstAudioresample *audioresample;
audioresample = GST_AUDIORESAMPLE (base);
switch (GST_EVENT_TYPE (event)) {
case GST_EVENT_FLUSH_START:
break;
case GST_EVENT_FLUSH_STOP:
resample_input_flush (audioresample->resample);
audioresample->ts_offset = -1;
audioresample->next_ts = -1;
audioresample->offset = -1;
break;
case GST_EVENT_NEWSEGMENT:
resample_input_pushthrough (audioresample->resample);
audioresample_pushthrough (audioresample);
audioresample->ts_offset = -1;
audioresample->next_ts = -1;
audioresample->offset = -1;
break;
case GST_EVENT_EOS:
resample_input_eos (audioresample->resample);
audioresample_pushthrough (audioresample);
break;
default:
break;
}
parent_class->event (base, event);
return TRUE;
}
static GstFlowReturn
audioresample_do_output (GstAudioresample * audioresample, GstBuffer * outbuf)
{
int outsize;
int outsamples;
ResampleState *r;
r = audioresample->resample;
outsize = resample_get_output_size (r);
GST_LOG_OBJECT (audioresample, "audioresample can give me %d bytes", outsize);
/* protect against mem corruption */
if (outsize > GST_BUFFER_SIZE (outbuf)) {
GST_WARNING_OBJECT (audioresample,
"overriding audioresample's outsize %d with outbuffer's size %d",
outsize, GST_BUFFER_SIZE (outbuf));
outsize = GST_BUFFER_SIZE (outbuf);
}
/* catch possibly wrong size differences */
if (GST_BUFFER_SIZE (outbuf) - outsize > r->sample_size) {
GST_WARNING_OBJECT (audioresample,
"audioresample's outsize %d too far from outbuffer's size %d",
outsize, GST_BUFFER_SIZE (outbuf));
}
outsize = resample_get_output_data (r, GST_BUFFER_DATA (outbuf), outsize);
outsamples = outsize / r->sample_size;
GST_LOG_OBJECT (audioresample, "resample gave me %d bytes or %d samples",
outsize, outsamples);
GST_BUFFER_OFFSET (outbuf) = audioresample->offset;
GST_BUFFER_TIMESTAMP (outbuf) = audioresample->next_ts;
if (audioresample->ts_offset != -1) {
audioresample->offset += outsamples;
audioresample->ts_offset += outsamples;
audioresample->next_ts =
gst_util_uint64_scale_int (audioresample->ts_offset, GST_SECOND,
audioresample->o_rate);
GST_BUFFER_OFFSET_END (outbuf) = audioresample->offset;
/* we calculate DURATION as the difference between "next" timestamp
* and current timestamp so we ensure a contiguous stream, instead of
* having rounding errors. */
GST_BUFFER_DURATION (outbuf) = audioresample->next_ts -
GST_BUFFER_TIMESTAMP (outbuf);
} else {
/* no valid offset know, we can still sortof calculate the duration though */
GST_BUFFER_DURATION (outbuf) =
gst_util_uint64_scale_int (outsamples, GST_SECOND,
audioresample->o_rate);
}
/* check for possible mem corruption */
if (outsize > GST_BUFFER_SIZE (outbuf)) {
/* this is an error that when it happens, would need fixing in the
* resample library; we told it we wanted only GST_BUFFER_SIZE (outbuf),
* and it gave us more ! */
GST_WARNING_OBJECT (audioresample,
"audioresample, you memory corrupting bastard. "
"you gave me outsize %d while my buffer was size %d",
outsize, GST_BUFFER_SIZE (outbuf));
return GST_FLOW_ERROR;
}
/* catch possibly wrong size differences */
if (GST_BUFFER_SIZE (outbuf) - outsize > r->sample_size) {
GST_WARNING_OBJECT (audioresample,
"audioresample's written outsize %d too far from outbuffer's size %d",
outsize, GST_BUFFER_SIZE (outbuf));
}
GST_BUFFER_SIZE (outbuf) = outsize;
if (G_UNLIKELY (audioresample->need_discont)) {
GST_DEBUG_OBJECT (audioresample,
"marking this buffer with the DISCONT flag");
GST_BUFFER_FLAG_SET (outbuf, GST_BUFFER_FLAG_DISCONT);
audioresample->need_discont = FALSE;
}
GST_LOG_OBJECT (audioresample, "transformed to buffer of %d bytes, ts %"
GST_TIME_FORMAT ", duration %" GST_TIME_FORMAT ", offset %"
G_GINT64_FORMAT ", offset_end %" G_GINT64_FORMAT,
outsize, GST_TIME_ARGS (GST_BUFFER_TIMESTAMP (outbuf)),
GST_TIME_ARGS (GST_BUFFER_DURATION (outbuf)),
GST_BUFFER_OFFSET (outbuf), GST_BUFFER_OFFSET_END (outbuf));
return GST_FLOW_OK;
}
static gboolean
audioresample_check_discont (GstAudioresample * audioresample,
GstClockTime timestamp)
{
if (timestamp != GST_CLOCK_TIME_NONE &&
audioresample->prev_ts != GST_CLOCK_TIME_NONE &&
audioresample->prev_duration != GST_CLOCK_TIME_NONE &&
timestamp != audioresample->prev_ts + audioresample->prev_duration) {
/* Potentially a discontinuous buffer. However, it turns out that many
* elements generate imperfect streams due to rounding errors, so we permit
* a small error (up to one sample) without triggering a filter
* flush/restart (if triggered incorrectly, this will be audible) */
GstClockTimeDiff diff = timestamp -
(audioresample->prev_ts + audioresample->prev_duration);
if (ABS (diff) > GST_SECOND / audioresample->i_rate) {
GST_WARNING_OBJECT (audioresample,
"encountered timestamp discontinuity of %" G_GINT64_FORMAT, diff);
return TRUE;
}
}
return FALSE;
}
static GstFlowReturn
audioresample_transform (GstBaseTransform * base, GstBuffer * inbuf,
GstBuffer * outbuf)
{
GstAudioresample *audioresample;
ResampleState *r;
guchar *data, *datacopy;
gulong size;
GstClockTime timestamp;
audioresample = GST_AUDIORESAMPLE (base);
r = audioresample->resample;
data = GST_BUFFER_DATA (inbuf);
size = GST_BUFFER_SIZE (inbuf);
timestamp = GST_BUFFER_TIMESTAMP (inbuf);
GST_LOG_OBJECT (audioresample, "transforming buffer of %ld bytes, ts %"
GST_TIME_FORMAT ", duration %" GST_TIME_FORMAT ", offset %"
G_GINT64_FORMAT ", offset_end %" G_GINT64_FORMAT,
size, GST_TIME_ARGS (timestamp),
GST_TIME_ARGS (GST_BUFFER_DURATION (inbuf)),
GST_BUFFER_OFFSET (inbuf), GST_BUFFER_OFFSET_END (inbuf));
/* check for timestamp discontinuities and flush/reset if needed */
if (G_UNLIKELY (audioresample_check_discont (audioresample, timestamp))) {
/* Flush internal samples */
audioresample_pushthrough (audioresample);
/* Inform downstream element about discontinuity */
audioresample->need_discont = TRUE;
/* We want to recalculate the offset */
audioresample->ts_offset = -1;
}
if (audioresample->ts_offset == -1) {
/* if we don't know the initial offset yet, calculate it based on the
* input timestamp. */
if (GST_CLOCK_TIME_IS_VALID (timestamp)) {
GstClockTime stime;
/* offset used to calculate the timestamps. We use the sample offset for
* this to make it more accurate. We want the first buffer to have the
* same timestamp as the incoming timestamp. */
audioresample->next_ts = timestamp;
audioresample->ts_offset =
gst_util_uint64_scale_int (timestamp, r->o_rate, GST_SECOND);
/* offset used to set as the buffer offset, this offset is always
* relative to the stream time, note that timestamp is not... */
stime = (timestamp - base->segment.start) + base->segment.time;
audioresample->offset =
gst_util_uint64_scale_int (stime, r->o_rate, GST_SECOND);
}
}
audioresample->prev_ts = timestamp;
audioresample->prev_duration = GST_BUFFER_DURATION (inbuf);
/* need to memdup, resample takes ownership. */
datacopy = g_memdup (data, size);
resample_add_input_data (r, datacopy, size, g_free, datacopy);
return audioresample_do_output (audioresample, outbuf);
}
/* push remaining data in the buffers out */
static GstFlowReturn
audioresample_pushthrough (GstAudioresample * audioresample)
{
int outsize;
ResampleState *r;
GstBuffer *outbuf;
GstFlowReturn res = GST_FLOW_OK;
GstBaseTransform *trans;
r = audioresample->resample;
outsize = resample_get_output_size (r);
if (outsize == 0) {
GST_DEBUG_OBJECT (audioresample, "no internal buffers needing flush");
goto done;
}
trans = GST_BASE_TRANSFORM (audioresample);
res = gst_pad_alloc_buffer (trans->srcpad, GST_BUFFER_OFFSET_NONE, outsize,
GST_PAD_CAPS (trans->srcpad), &outbuf);
if (G_UNLIKELY (res != GST_FLOW_OK)) {
GST_WARNING_OBJECT (audioresample, "failed allocating buffer of %d bytes",
outsize);
goto done;
}
res = audioresample_do_output (audioresample, outbuf);
if (G_UNLIKELY (res != GST_FLOW_OK))
goto done;
res = gst_pad_push (trans->srcpad, outbuf);
done:
return res;
}
static gboolean
audioresample_query (GstPad * pad, GstQuery * query)
{
GstAudioresample *audioresample =
GST_AUDIORESAMPLE (gst_pad_get_parent (pad));
GstBaseTransform *trans = GST_BASE_TRANSFORM (audioresample);
gboolean res = TRUE;
switch (GST_QUERY_TYPE (query)) {
case GST_QUERY_LATENCY:
{
GstClockTime min, max;
gboolean live;
guint64 latency;
GstPad *peer;
gint rate = audioresample->i_rate;
gint resampler_latency = audioresample->filter_length / 2;
if (gst_base_transform_is_passthrough (trans))
resampler_latency = 0;
if ((peer = gst_pad_get_peer (trans->sinkpad))) {
if ((res = gst_pad_query (peer, query))) {
gst_query_parse_latency (query, &live, &min, &max);
GST_DEBUG ("Peer latency: min %"
GST_TIME_FORMAT " max %" GST_TIME_FORMAT,
GST_TIME_ARGS (min), GST_TIME_ARGS (max));
/* add our own latency */
if (rate != 0 && resampler_latency != 0)
latency =
gst_util_uint64_scale (resampler_latency, GST_SECOND, rate);
else
latency = 0;
GST_DEBUG ("Our latency: %" GST_TIME_FORMAT, GST_TIME_ARGS (latency));
min += latency;
if (max != GST_CLOCK_TIME_NONE)
max += latency;
GST_DEBUG ("Calculated total latency : min %"
GST_TIME_FORMAT " max %" GST_TIME_FORMAT,
GST_TIME_ARGS (min), GST_TIME_ARGS (max));
gst_query_set_latency (query, live, min, max);
}
gst_object_unref (peer);
}
break;
}
default:
res = gst_pad_query_default (pad, query);
break;
}
gst_object_unref (audioresample);
return res;
}
static const GstQueryType *
audioresample_query_type (GstPad * pad)
{
static const GstQueryType types[] = {
GST_QUERY_LATENCY,
0
};
return types;
}
static void
gst_audioresample_set_property (GObject * object, guint prop_id,
const GValue * value, GParamSpec * pspec)
{
GstAudioresample *audioresample;
audioresample = GST_AUDIORESAMPLE (object);
switch (prop_id) {
case PROP_FILTERLEN:
audioresample->filter_length = g_value_get_int (value);
GST_DEBUG_OBJECT (GST_ELEMENT (audioresample), "new filter length %d",
audioresample->filter_length);
if (audioresample->resample) {
resample_set_filter_length (audioresample->resample,
audioresample->filter_length);
gst_element_post_message (GST_ELEMENT (audioresample),
gst_message_new_latency (GST_OBJECT (audioresample)));
}
break;
default:
G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
break;
}
}
static void
gst_audioresample_get_property (GObject * object, guint prop_id,
GValue * value, GParamSpec * pspec)
{
GstAudioresample *audioresample;
audioresample = GST_AUDIORESAMPLE (object);
switch (prop_id) {
case PROP_FILTERLEN:
g_value_set_int (value, audioresample->filter_length);
break;
default:
G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
break;
}
}
static gboolean
plugin_init (GstPlugin * plugin)
{
resample_init ();
if (!gst_element_register (plugin, "audioresample", GST_RANK_PRIMARY,
GST_TYPE_AUDIORESAMPLE)) {
return FALSE;
}
return TRUE;
}
GST_PLUGIN_DEFINE (GST_VERSION_MAJOR,
GST_VERSION_MINOR,
"audioresample",
"Resamples audio", plugin_init, VERSION, "LGPL", GST_PACKAGE_NAME,
GST_PACKAGE_ORIGIN);