2 * Copyright (C) 2011 Mark Nauwelaerts <mark.nauwelaerts@collabora.co.uk>.
3 * Copyright (C) 2011 Nokia Corporation. All rights reserved.
4 * Contact: Stefan Kost <stefan.kost@nokia.com>
6 * This library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Library General Public
8 * License as published by the Free Software Foundation; either
9 * version 2 of the License, or (at your option) any later version.
11 * This library is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Library General Public License for more details.
16 * You should have received a copy of the GNU Library General Public
17 * License along with this library; if not, write to the
18 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
19 * Boston, MA 02111-1307, USA.
23 * SECTION:gstaudioencoder
24 * @short_description: Base class for audio encoders
25 * @see_also: #GstBaseTransform
28 * This base class is for audio encoders turning raw audio samples into
31 * GstAudioEncoder and subclass should cooperate as follows.
34 * <itemizedlist><title>Configuration</title>
36 * Initially, GstAudioEncoder calls @start when the encoder element
37 * is activated, which allows subclass to perform any global setup.
40 * GstAudioEncoder calls @set_format to inform subclass of the format
41 * of input audio data that it is about to receive. Subclass should
42 * setup for encoding and configure various base class parameters
43 * appropriately, notably those directing desired input data handling.
44 * While unlikely, it might be called more than once, if changing input
45 * parameters require reconfiguration.
48 * GstAudioEncoder calls @stop at end of all processing.
52 * As of configuration stage, and throughout processing, GstAudioEncoder
53 * maintains various parameters that provide required context,
54 * e.g. describing the format of input audio data.
55 * Conversely, subclass can and should configure these context parameters
56 * to inform base class of its expectation w.r.t. buffer handling.
59 * <title>Data processing</title>
61 * Base class gathers input sample data (as directed by the context's
62 * frame_samples and frame_max) and provides this to subclass' @handle_frame.
65 * If codec processing results in encoded data, subclass should call
66 * @gst_audio_encoder_finish_frame to have encoded data pushed
67 * downstream. Alternatively, it might also call to indicate dropped
68 * (non-encoded) samples.
71 * Just prior to actually pushing a buffer downstream,
72 * it is passed to @pre_push.
75 * During the parsing process GstAudioEncoderClass will handle both
76 * srcpad and sinkpad events. Sink events will be passed to subclass
77 * if @event callback has been provided.
82 * <itemizedlist><title>Shutdown phase</title>
84 * GstAudioEncoder class calls @stop to inform the subclass that data
85 * parsing will be stopped.
91 * Subclass is responsible for providing pad template caps for
92 * source and sink pads. The pads need to be named "sink" and "src". It also
93 * needs to set the fixed caps on srcpad, when the format is ensured. This
94 * is typically when base class calls subclass' @set_format function, though
95 * it might be delayed until calling @gst_audio_encoder_finish_frame.
97 * In summary, above process should have subclass concentrating on
98 * codec data processing while leaving other matters to base class,
99 * such as most notably timestamp handling. While it may exert more control
100 * in this area (see e.g. @pre_push), it is very much not recommended.
102 * In particular, base class will either favor tracking upstream timestamps
103 * (at the possible expense of jitter) or aim to arrange for a perfect stream of
104 * output timestamps, depending on #GstAudioEncoder:perfect-timestamp.
105 * However, in the latter case, the input may not be so perfect or ideal, which
106 * is handled as follows. An input timestamp is compared with the expected
107 * timestamp as dictated by input sample stream and if the deviation is less
108 * than #GstAudioEncoder:tolerance, the deviation is discarded.
109 * Otherwise, it is considered a discontuinity and subsequent output timestamp
110 * is resynced to the new position after performing configured discontinuity
111 * processing. In the non-perfect-timestamp case, an upstream variation
112 * exceeding tolerance only leads to marking DISCONT on subsequent outgoing
113 * (while timestamps are adjusted to upstream regardless of variation).
114 * While DISCONT is also marked in the perfect-timestamp case, this one
115 * optionally (see #GstAudioEncoder:hard-resync)
116 * performs some additional steps, such as clipping of (early) input samples
117 * or draining all currently remaining input data, depending on the direction
118 * of the discontuinity.
120 * If perfect timestamps are arranged, it is also possible to request baseclass
121 * (usually set by subclass) to provide additional buffer metadata (in OFFSET
122 * and OFFSET_END) fields according to granule defined semantics currently
123 * needed by oggmux. Specifically, OFFSET is set to granulepos (= sample count
124 * including buffer) and OFFSET_END to corresponding timestamp (as determined
125 * by same sample count and sample rate).
127 * Things that subclass need to take care of:
129 * <listitem><para>Provide pad templates</para></listitem>
131 * Set source pad caps when appropriate
134 * Inform base class of buffer processing needs using context's
135 * frame_samples and frame_bytes.
138 * Set user-configurable properties to sane defaults for format and
139 * implementing codec at hand, e.g. those controlling timestamp behaviour
140 * and discontinuity processing.
143 * Accept data in @handle_frame and provide encoded results to
144 * @gst_audio_encoder_finish_frame.
154 #define GST_USE_UNSTABLE_API
155 #include "gstaudioencoder.h"
156 #include <gst/base/gstadapter.h>
157 #include <gst/audio/audio.h>
158 #include <gst/pbutils/descriptions.h>
164 GST_DEBUG_CATEGORY_STATIC (gst_audio_encoder_debug);
165 #define GST_CAT_DEFAULT gst_audio_encoder_debug
167 #define GST_AUDIO_ENCODER_GET_PRIVATE(obj) \
168 (G_TYPE_INSTANCE_GET_PRIVATE ((obj), GST_TYPE_AUDIO_ENCODER, \
169 GstAudioEncoderPrivate))
180 #define DEFAULT_PERFECT_TS FALSE
181 #define DEFAULT_GRANULE FALSE
182 #define DEFAULT_HARD_RESYNC FALSE
183 #define DEFAULT_TOLERANCE 40000000
185 typedef struct _GstAudioEncoderContext
191 gint frame_samples_min, frame_samples_max;
194 /* MT-protected (with LOCK) */
195 GstClockTime min_latency;
196 GstClockTime max_latency;
197 } GstAudioEncoderContext;
199 struct _GstAudioEncoderPrivate
201 /* activation status */
204 /* input base/first ts as basis for output ts;
205 * kept nearly constant for perfect_ts,
206 * otherwise resyncs to upstream ts */
207 GstClockTime base_ts;
208 /* corresponding base granulepos */
210 /* input samples processed and sent downstream so far (w.r.t. base_ts) */
213 /* currently collected sample data */
215 /* offset in adapter up to which already supplied to encoder */
217 /* mark outgoing discont */
219 /* to guess duration of drained data */
220 GstClockTime last_duration;
222 /* subclass provided data in processing round */
224 /* subclass gave all it could already */
226 /* subclass currently being forcibly drained */
229 /* output bps estimatation */
230 /* global in samples seen */
232 /* global bytes sent out */
235 /* context storage */
236 GstAudioEncoderContext ctx;
241 gboolean hard_resync;
246 /* pending serialized sink events, will be sent from finish_frame() */
247 GList *pending_events;
251 static GstElementClass *parent_class = NULL;
253 static void gst_audio_encoder_class_init (GstAudioEncoderClass * klass);
254 static void gst_audio_encoder_init (GstAudioEncoder * parse,
255 GstAudioEncoderClass * klass);
258 gst_audio_encoder_get_type (void)
260 static GType audio_encoder_type = 0;
262 if (!audio_encoder_type) {
263 static const GTypeInfo audio_encoder_info = {
264 sizeof (GstAudioEncoderClass),
265 (GBaseInitFunc) NULL,
266 (GBaseFinalizeFunc) NULL,
267 (GClassInitFunc) gst_audio_encoder_class_init,
270 sizeof (GstAudioEncoder),
272 (GInstanceInitFunc) gst_audio_encoder_init,
274 const GInterfaceInfo preset_interface_info = {
275 NULL, /* interface_init */
276 NULL, /* interface_finalize */
277 NULL /* interface_data */
280 audio_encoder_type = g_type_register_static (GST_TYPE_ELEMENT,
281 "GstAudioEncoder", &audio_encoder_info, G_TYPE_FLAG_ABSTRACT);
283 g_type_add_interface_static (audio_encoder_type, GST_TYPE_PRESET,
284 &preset_interface_info);
286 return audio_encoder_type;
289 static void gst_audio_encoder_finalize (GObject * object);
290 static void gst_audio_encoder_reset (GstAudioEncoder * enc, gboolean full);
292 static void gst_audio_encoder_set_property (GObject * object,
293 guint prop_id, const GValue * value, GParamSpec * pspec);
294 static void gst_audio_encoder_get_property (GObject * object,
295 guint prop_id, GValue * value, GParamSpec * pspec);
297 static gboolean gst_audio_encoder_sink_activate_push (GstPad * pad,
300 static gboolean gst_audio_encoder_sink_event (GstPad * pad, GstEvent * event);
301 static gboolean gst_audio_encoder_sink_setcaps (GstPad * pad, GstCaps * caps);
302 static GstFlowReturn gst_audio_encoder_chain (GstPad * pad, GstBuffer * buffer);
303 static gboolean gst_audio_encoder_src_query (GstPad * pad, GstQuery * query);
304 static gboolean gst_audio_encoder_sink_query (GstPad * pad, GstQuery * query);
305 static const GstQueryType *gst_audio_encoder_get_query_types (GstPad * pad);
306 static GstCaps *gst_audio_encoder_sink_getcaps (GstPad * pad);
310 gst_audio_encoder_class_init (GstAudioEncoderClass * klass)
312 GObjectClass *gobject_class;
314 gobject_class = G_OBJECT_CLASS (klass);
315 parent_class = g_type_class_peek_parent (klass);
317 GST_DEBUG_CATEGORY_INIT (gst_audio_encoder_debug, "audioencoder", 0,
318 "audio encoder base class");
320 g_type_class_add_private (klass, sizeof (GstAudioEncoderPrivate));
322 gobject_class->set_property = gst_audio_encoder_set_property;
323 gobject_class->get_property = gst_audio_encoder_get_property;
325 gobject_class->finalize = GST_DEBUG_FUNCPTR (gst_audio_encoder_finalize);
328 g_object_class_install_property (gobject_class, PROP_PERFECT_TS,
329 g_param_spec_boolean ("perfect-timestamp", "Perfect Timestamps",
330 "Favour perfect timestamps over tracking upstream timestamps",
331 DEFAULT_PERFECT_TS, G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
332 g_object_class_install_property (gobject_class, PROP_GRANULE,
333 g_param_spec_boolean ("mark-granule", "Granule Marking",
334 "Apply granule semantics to buffer metadata (implies perfect-timestamp)",
335 DEFAULT_GRANULE, G_PARAM_READABLE | G_PARAM_STATIC_STRINGS));
336 g_object_class_install_property (gobject_class, PROP_HARD_RESYNC,
337 g_param_spec_boolean ("hard-resync", "Hard Resync",
338 "Perform clipping and sample flushing upon discontinuity",
339 DEFAULT_HARD_RESYNC, G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
340 g_object_class_install_property (gobject_class, PROP_TOLERANCE,
341 g_param_spec_int64 ("tolerance", "Tolerance",
342 "Consider discontinuity if timestamp jitter/imperfection exceeds tolerance (ns)",
343 0, G_MAXINT64, DEFAULT_TOLERANCE,
344 G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
348 gst_audio_encoder_init (GstAudioEncoder * enc, GstAudioEncoderClass * bclass)
350 GstPadTemplate *pad_template;
352 GST_DEBUG_OBJECT (enc, "gst_audio_encoder_init");
354 enc->priv = GST_AUDIO_ENCODER_GET_PRIVATE (enc);
356 /* only push mode supported */
358 gst_element_class_get_pad_template (GST_ELEMENT_CLASS (bclass), "sink");
359 g_return_if_fail (pad_template != NULL);
360 enc->sinkpad = gst_pad_new_from_template (pad_template, "sink");
361 gst_pad_set_event_function (enc->sinkpad,
362 GST_DEBUG_FUNCPTR (gst_audio_encoder_sink_event));
363 gst_pad_set_setcaps_function (enc->sinkpad,
364 GST_DEBUG_FUNCPTR (gst_audio_encoder_sink_setcaps));
365 gst_pad_set_getcaps_function (enc->sinkpad,
366 GST_DEBUG_FUNCPTR (gst_audio_encoder_sink_getcaps));
367 gst_pad_set_query_function (enc->sinkpad,
368 GST_DEBUG_FUNCPTR (gst_audio_encoder_sink_query));
369 gst_pad_set_chain_function (enc->sinkpad,
370 GST_DEBUG_FUNCPTR (gst_audio_encoder_chain));
371 gst_pad_set_activatepush_function (enc->sinkpad,
372 GST_DEBUG_FUNCPTR (gst_audio_encoder_sink_activate_push));
373 gst_element_add_pad (GST_ELEMENT (enc), enc->sinkpad);
375 GST_DEBUG_OBJECT (enc, "sinkpad created");
377 /* and we don't mind upstream traveling stuff that much ... */
379 gst_element_class_get_pad_template (GST_ELEMENT_CLASS (bclass), "src");
380 g_return_if_fail (pad_template != NULL);
381 enc->srcpad = gst_pad_new_from_template (pad_template, "src");
382 gst_pad_set_query_function (enc->srcpad,
383 GST_DEBUG_FUNCPTR (gst_audio_encoder_src_query));
384 gst_pad_set_query_type_function (enc->srcpad,
385 GST_DEBUG_FUNCPTR (gst_audio_encoder_get_query_types));
386 gst_pad_use_fixed_caps (enc->srcpad);
387 gst_element_add_pad (GST_ELEMENT (enc), enc->srcpad);
388 GST_DEBUG_OBJECT (enc, "src created");
390 enc->priv->adapter = gst_adapter_new ();
392 g_static_rec_mutex_init (&enc->stream_lock);
394 /* property default */
395 enc->priv->granule = DEFAULT_GRANULE;
396 enc->priv->perfect_ts = DEFAULT_PERFECT_TS;
397 enc->priv->hard_resync = DEFAULT_HARD_RESYNC;
398 enc->priv->tolerance = DEFAULT_TOLERANCE;
401 gst_audio_encoder_reset (enc, TRUE);
402 GST_DEBUG_OBJECT (enc, "init ok");
406 gst_audio_encoder_reset (GstAudioEncoder * enc, gboolean full)
408 GST_AUDIO_ENCODER_STREAM_LOCK (enc);
410 GST_LOG_OBJECT (enc, "reset full %d", full);
413 enc->priv->active = FALSE;
414 enc->priv->samples_in = 0;
415 enc->priv->bytes_out = 0;
416 gst_audio_info_clear (&enc->priv->ctx.info);
417 memset (&enc->priv->ctx, 0, sizeof (enc->priv->ctx));
420 gst_tag_list_free (enc->priv->tags);
421 enc->priv->tags = NULL;
423 g_list_foreach (enc->priv->pending_events, (GFunc) gst_event_unref, NULL);
424 g_list_free (enc->priv->pending_events);
425 enc->priv->pending_events = NULL;
428 gst_segment_init (&enc->segment, GST_FORMAT_TIME);
430 gst_adapter_clear (enc->priv->adapter);
431 enc->priv->got_data = FALSE;
432 enc->priv->drained = TRUE;
433 enc->priv->offset = 0;
434 enc->priv->base_ts = GST_CLOCK_TIME_NONE;
435 enc->priv->base_gp = -1;
436 enc->priv->samples = 0;
437 enc->priv->discont = FALSE;
439 GST_AUDIO_ENCODER_STREAM_UNLOCK (enc);
443 gst_audio_encoder_finalize (GObject * object)
445 GstAudioEncoder *enc = GST_AUDIO_ENCODER (object);
447 g_object_unref (enc->priv->adapter);
449 g_static_rec_mutex_free (&enc->stream_lock);
451 G_OBJECT_CLASS (parent_class)->finalize (object);
455 * gst_audio_encoder_finish_frame:
456 * @enc: a #GstAudioEncoder
457 * @buffer: encoded data
458 * @samples: number of samples (per channel) represented by encoded data
460 * Collects encoded data and/or pushes encoded data downstream.
461 * Source pad caps must be set when this is called. Depending on the nature
462 * of the (framing of) the format, subclass can decide whether to push
463 * encoded data directly or to collect various "frames" in a single buffer.
464 * Note that the latter behaviour is recommended whenever the format is allowed,
465 * as it incurs no additional latency and avoids otherwise generating a
466 * a multitude of (small) output buffers. If not explicitly pushed,
467 * any available encoded data is pushed at the end of each processing cycle,
468 * i.e. which encodes as much data as available input data allows.
470 * If @samples < 0, then best estimate is all samples provided to encoder
471 * (subclass) so far. @buf may be NULL, in which case next number of @samples
472 * are considered discarded, e.g. as a result of discontinuous transmission,
473 * and a discontinuity is marked (note that @buf == NULL => push == TRUE).
475 * Returns: a #GstFlowReturn that should be escalated to caller (of caller)
480 gst_audio_encoder_finish_frame (GstAudioEncoder * enc, GstBuffer * buf,
483 GstAudioEncoderClass *klass;
484 GstAudioEncoderPrivate *priv;
485 GstAudioEncoderContext *ctx;
486 GstFlowReturn ret = GST_FLOW_OK;
488 klass = GST_AUDIO_ENCODER_GET_CLASS (enc);
490 ctx = &enc->priv->ctx;
492 /* subclass should know what it is producing by now */
493 g_return_val_if_fail (GST_PAD_CAPS (enc->srcpad) != NULL, GST_FLOW_ERROR);
494 /* subclass should not hand us no data */
495 g_return_val_if_fail (buf == NULL || GST_BUFFER_SIZE (buf) > 0,
498 GST_AUDIO_ENCODER_STREAM_LOCK (enc);
500 if (G_UNLIKELY (enc->priv->tags)) {
503 /* add codec info to pending tags */
504 tags = enc->priv->tags;
505 /* no more pending */
506 enc->priv->tags = NULL;
507 gst_pb_utils_add_codec_description_to_tag_list (tags, GST_TAG_CODEC,
508 GST_PAD_CAPS (enc->srcpad));
509 gst_pb_utils_add_codec_description_to_tag_list (tags, GST_TAG_AUDIO_CODEC,
510 GST_PAD_CAPS (enc->srcpad));
511 GST_DEBUG_OBJECT (enc, "sending tags %" GST_PTR_FORMAT, tags);
512 gst_element_found_tags_for_pad (GST_ELEMENT (enc), enc->srcpad, tags);
515 GST_LOG_OBJECT (enc, "accepting %d bytes encoded data as %d samples",
516 buf ? GST_BUFFER_SIZE (buf) : -1, samples);
518 /* mark subclass still alive and providing */
519 priv->got_data = TRUE;
521 if (priv->pending_events) {
522 GList *pending_events, *l;
524 pending_events = priv->pending_events;
525 priv->pending_events = NULL;
527 GST_DEBUG_OBJECT (enc, "Pushing pending events");
528 for (l = priv->pending_events; l; l = l->next)
529 gst_pad_push_event (enc->srcpad, l->data);
530 g_list_free (pending_events);
533 /* remove corresponding samples from input */
535 samples = (enc->priv->offset / ctx->info.bpf);
537 if (G_LIKELY (samples)) {
538 /* track upstream ts if so configured */
539 if (!enc->priv->perfect_ts) {
540 guint64 ts, distance;
542 ts = gst_adapter_prev_timestamp (priv->adapter, &distance);
543 g_assert (distance % ctx->info.bpf == 0);
544 distance /= ctx->info.bpf;
545 GST_LOG_OBJECT (enc, "%" G_GUINT64_FORMAT " samples past prev_ts %"
546 GST_TIME_FORMAT, distance, GST_TIME_ARGS (ts));
547 GST_LOG_OBJECT (enc, "%" G_GUINT64_FORMAT " samples past base_ts %"
548 GST_TIME_FORMAT, priv->samples, GST_TIME_ARGS (priv->base_ts));
549 /* when draining adapter might be empty and no ts to offer */
550 if (GST_CLOCK_TIME_IS_VALID (ts) && ts != priv->base_ts) {
551 GstClockTimeDiff diff;
552 GstClockTime old_ts, next_ts;
554 /* passed into another buffer;
555 * mild check for discontinuity and only mark if so */
557 gst_util_uint64_scale (distance, GST_SECOND, ctx->info.rate);
558 old_ts = priv->base_ts +
559 gst_util_uint64_scale (priv->samples, GST_SECOND, ctx->info.rate);
560 diff = GST_CLOCK_DIFF (next_ts, old_ts);
561 GST_LOG_OBJECT (enc, "ts diff %d ms", (gint) (diff / GST_MSECOND));
562 /* only mark discontinuity if beyond tolerance */
563 if (G_UNLIKELY (diff < -enc->priv->tolerance ||
564 diff > enc->priv->tolerance)) {
565 GST_DEBUG_OBJECT (enc, "marked discont");
566 priv->discont = TRUE;
568 if (diff > GST_SECOND / ctx->info.rate / 2 ||
569 diff < -GST_SECOND / ctx->info.rate / 2) {
570 GST_LOG_OBJECT (enc, "new upstream ts %" GST_TIME_FORMAT
571 " at distance %" G_GUINT64_FORMAT, GST_TIME_ARGS (ts), distance);
572 /* re-sync to upstream ts */
574 priv->samples = distance;
576 GST_LOG_OBJECT (enc, "new upstream ts only introduces jitter");
580 /* advance sample view */
581 if (G_UNLIKELY (samples * ctx->info.bpf > priv->offset)) {
582 if (G_LIKELY (!priv->force)) {
583 /* no way we can let this pass */
584 g_assert_not_reached ();
589 if (samples * ctx->info.bpf >= gst_adapter_available (priv->adapter))
590 gst_adapter_clear (priv->adapter);
592 gst_adapter_flush (priv->adapter, samples * ctx->info.bpf);
595 gst_adapter_flush (priv->adapter, samples * ctx->info.bpf);
596 priv->offset -= samples * ctx->info.bpf;
597 /* avoid subsequent stray prev_ts */
598 if (G_UNLIKELY (gst_adapter_available (priv->adapter) == 0))
599 gst_adapter_clear (priv->adapter);
601 /* sample count advanced below after buffer handling */
605 if (G_LIKELY (buf)) {
606 GST_LOG_OBJECT (enc, "taking %d bytes for output", GST_BUFFER_SIZE (buf));
607 buf = gst_buffer_make_metadata_writable (buf);
610 gst_buffer_set_caps (buf, GST_PAD_CAPS (enc->srcpad));
611 if (G_LIKELY (GST_CLOCK_TIME_IS_VALID (priv->base_ts))) {
612 /* FIXME ? lookahead could lead to weird ts and duration ?
613 * (particularly if not in perfect mode) */
614 /* mind sample rounding and produce perfect output */
615 GST_BUFFER_TIMESTAMP (buf) = priv->base_ts +
616 gst_util_uint64_scale (priv->samples - ctx->lookahead, GST_SECOND,
618 GST_DEBUG_OBJECT (enc, "out samples %d", samples);
619 if (G_LIKELY (samples > 0)) {
620 priv->samples += samples;
621 GST_BUFFER_DURATION (buf) = priv->base_ts +
622 gst_util_uint64_scale (priv->samples - ctx->lookahead, GST_SECOND,
623 ctx->info.rate) - GST_BUFFER_TIMESTAMP (buf);
624 priv->last_duration = GST_BUFFER_DURATION (buf);
626 /* duration forecast in case of handling remainder;
627 * the last one is probably like the previous one ... */
628 GST_BUFFER_DURATION (buf) = priv->last_duration;
630 if (priv->base_gp >= 0) {
632 /* FIXME: in longer run, muxer should take care of this ... */
633 /* offset_end = granulepos for ogg muxer */
634 GST_BUFFER_OFFSET_END (buf) = priv->base_gp + priv->samples -
635 enc->priv->ctx.lookahead;
636 /* offset = timestamp corresponding to granulepos for ogg muxer */
637 GST_BUFFER_OFFSET (buf) =
638 GST_FRAMES_TO_CLOCK_TIME (GST_BUFFER_OFFSET_END (buf),
641 GST_BUFFER_OFFSET (buf) = priv->bytes_out;
642 GST_BUFFER_OFFSET_END (buf) = priv->bytes_out + GST_BUFFER_SIZE (buf);
646 priv->bytes_out += GST_BUFFER_SIZE (buf);
648 if (G_UNLIKELY (priv->discont)) {
649 GST_LOG_OBJECT (enc, "marking discont");
650 GST_BUFFER_FLAG_SET (buf, GST_BUFFER_FLAG_DISCONT);
651 priv->discont = FALSE;
654 if (klass->pre_push) {
655 /* last chance for subclass to do some dirty stuff */
656 ret = klass->pre_push (enc, &buf);
657 if (ret != GST_FLOW_OK || !buf) {
658 GST_DEBUG_OBJECT (enc, "subclass returned %s, buf %p",
659 gst_flow_get_name (ret), buf);
661 gst_buffer_unref (buf);
666 GST_LOG_OBJECT (enc, "pushing buffer of size %d with ts %" GST_TIME_FORMAT
667 ", duration %" GST_TIME_FORMAT, GST_BUFFER_SIZE (buf),
668 GST_TIME_ARGS (GST_BUFFER_TIMESTAMP (buf)),
669 GST_TIME_ARGS (GST_BUFFER_DURATION (buf)));
671 ret = gst_pad_push (enc->srcpad, buf);
672 GST_LOG_OBJECT (enc, "buffer pushed: %s", gst_flow_get_name (ret));
674 /* merely advance samples, most work for that already done above */
675 priv->samples += samples;
679 GST_AUDIO_ENCODER_STREAM_UNLOCK (enc);
686 GST_ELEMENT_ERROR (enc, STREAM, ENCODE,
687 ("received more encoded samples %d than provided %d",
688 samples, priv->offset / ctx->info.bpf), (NULL));
690 gst_buffer_unref (buf);
691 ret = GST_FLOW_ERROR;
696 /* adapter tracking idea:
697 * - start of adapter corresponds with what has already been encoded
698 * (i.e. really returned by encoder subclass)
699 * - start + offset is what needs to be fed to subclass next */
701 gst_audio_encoder_push_buffers (GstAudioEncoder * enc, gboolean force)
703 GstAudioEncoderClass *klass;
704 GstAudioEncoderPrivate *priv;
705 GstAudioEncoderContext *ctx;
708 GstFlowReturn ret = GST_FLOW_OK;
710 klass = GST_AUDIO_ENCODER_GET_CLASS (enc);
712 g_return_val_if_fail (klass->handle_frame != NULL, GST_FLOW_ERROR);
715 ctx = &enc->priv->ctx;
717 while (ret == GST_FLOW_OK) {
720 av = gst_adapter_available (priv->adapter);
722 g_assert (priv->offset <= av);
726 ctx->frame_samples_min >
727 0 ? ctx->frame_samples_min * ctx->info.bpf : av;
728 GST_LOG_OBJECT (enc, "available: %d, needed: %d, force: %d", av, need,
731 if ((need > av) || !av) {
732 if (G_UNLIKELY (force)) {
742 if (ctx->frame_samples_max > 0)
743 need = MIN (av, ctx->frame_samples_max * ctx->info.bpf);
745 if (ctx->frame_samples_min == ctx->frame_samples_max) {
746 /* if we have some extra metadata,
747 * provide for integer multiple of frames to allow for better granularity
749 if (ctx->frame_samples_min > 0 && need) {
750 if (ctx->frame_max > 1)
751 need = need * MIN ((av / need), ctx->frame_max);
752 else if (ctx->frame_max == 0)
753 need = need * (av / need);
758 buf = gst_buffer_new ();
759 GST_BUFFER_DATA (buf) = (guint8 *)
760 gst_adapter_peek (priv->adapter, priv->offset + need) + priv->offset;
761 GST_BUFFER_SIZE (buf) = need;
764 GST_LOG_OBJECT (enc, "providing subclass with %d bytes at offset %d",
767 /* mark this already as consumed,
768 * which it should be when subclass gives us data in exchange for samples */
769 priv->offset += need;
770 priv->samples_in += need / ctx->info.bpf;
772 priv->got_data = FALSE;
773 ret = klass->handle_frame (enc, buf);
776 gst_buffer_unref (buf);
778 /* no data to feed, no leftover provided, then bail out */
779 if (G_UNLIKELY (!buf && !priv->got_data)) {
780 priv->drained = TRUE;
781 GST_LOG_OBJECT (enc, "no more data drained from subclass");
790 gst_audio_encoder_drain (GstAudioEncoder * enc)
792 if (enc->priv->drained)
795 return gst_audio_encoder_push_buffers (enc, TRUE);
799 gst_audio_encoder_set_base_gp (GstAudioEncoder * enc)
803 if (!enc->priv->granule)
806 /* use running time for granule */
807 /* incoming data is clipped, so a valid input should yield a valid output */
808 ts = gst_segment_to_running_time (&enc->segment, GST_FORMAT_TIME,
810 if (GST_CLOCK_TIME_IS_VALID (ts)) {
812 GST_CLOCK_TIME_TO_FRAMES (enc->priv->base_ts, enc->priv->ctx.info.rate);
813 GST_DEBUG_OBJECT (enc, "new base gp %" G_GINT64_FORMAT, enc->priv->base_gp);
815 /* should reasonably have a valid base,
816 * otherwise start at 0 if we did not already start there earlier */
817 if (enc->priv->base_gp < 0) {
818 enc->priv->base_gp = 0;
819 GST_DEBUG_OBJECT (enc, "new base gp %" G_GINT64_FORMAT,
826 gst_audio_encoder_chain (GstPad * pad, GstBuffer * buffer)
828 GstAudioEncoder *enc;
829 GstAudioEncoderPrivate *priv;
830 GstAudioEncoderContext *ctx;
831 GstFlowReturn ret = GST_FLOW_OK;
834 enc = GST_AUDIO_ENCODER (GST_OBJECT_PARENT (pad));
837 ctx = &enc->priv->ctx;
839 GST_AUDIO_ENCODER_STREAM_LOCK (enc);
841 /* should know what is coming by now */
846 "received buffer of size %d with ts %" GST_TIME_FORMAT
847 ", duration %" GST_TIME_FORMAT, GST_BUFFER_SIZE (buffer),
848 GST_TIME_ARGS (GST_BUFFER_TIMESTAMP (buffer)),
849 GST_TIME_ARGS (GST_BUFFER_DURATION (buffer)));
851 /* input shoud be whole number of sample frames */
852 if (GST_BUFFER_SIZE (buffer) % ctx->info.bpf)
855 #ifndef GST_DISABLE_GST_DEBUG
857 GstClockTime duration;
858 GstClockTimeDiff diff;
860 /* verify buffer duration */
861 duration = gst_util_uint64_scale (GST_BUFFER_SIZE (buffer), GST_SECOND,
862 ctx->info.rate * ctx->info.bpf);
863 diff = GST_CLOCK_DIFF (duration, GST_BUFFER_DURATION (buffer));
864 if (GST_BUFFER_DURATION (buffer) != GST_CLOCK_TIME_NONE &&
865 (diff > GST_SECOND / ctx->info.rate / 2 ||
866 diff < -GST_SECOND / ctx->info.rate / 2)) {
867 GST_DEBUG_OBJECT (enc, "incoming buffer had incorrect duration %"
868 GST_TIME_FORMAT ", expected duration %" GST_TIME_FORMAT,
869 GST_TIME_ARGS (GST_BUFFER_DURATION (buffer)),
870 GST_TIME_ARGS (duration));
875 discont = GST_BUFFER_FLAG_IS_SET (buffer, GST_BUFFER_FLAG_DISCONT);
876 if (G_UNLIKELY (discont)) {
877 GST_LOG_OBJECT (buffer, "marked discont");
878 enc->priv->discont = discont;
881 /* clip to segment */
882 /* NOTE: slightly painful linking -laudio only for this one ... */
883 buffer = gst_audio_buffer_clip (buffer, &enc->segment, ctx->info.rate,
885 if (G_UNLIKELY (!buffer)) {
886 GST_DEBUG_OBJECT (buffer, "no data after clipping to segment");
891 "buffer after segment clipping has size %d with ts %" GST_TIME_FORMAT
892 ", duration %" GST_TIME_FORMAT, GST_BUFFER_SIZE (buffer),
893 GST_TIME_ARGS (GST_BUFFER_TIMESTAMP (buffer)),
894 GST_TIME_ARGS (GST_BUFFER_DURATION (buffer)));
896 if (!GST_CLOCK_TIME_IS_VALID (priv->base_ts)) {
897 priv->base_ts = GST_BUFFER_TIMESTAMP (buffer);
898 GST_DEBUG_OBJECT (enc, "new base ts %" GST_TIME_FORMAT,
899 GST_TIME_ARGS (priv->base_ts));
900 gst_audio_encoder_set_base_gp (enc);
903 /* check for continuity;
904 * checked elsewhere in non-perfect case */
905 if (enc->priv->perfect_ts) {
906 GstClockTimeDiff diff = 0;
907 GstClockTime next_ts = 0;
909 if (GST_BUFFER_TIMESTAMP_IS_VALID (buffer) &&
910 GST_CLOCK_TIME_IS_VALID (priv->base_ts)) {
913 samples = priv->samples +
914 gst_adapter_available (priv->adapter) / ctx->info.bpf;
915 next_ts = priv->base_ts +
916 gst_util_uint64_scale (samples, GST_SECOND, ctx->info.rate);
917 GST_LOG_OBJECT (enc, "buffer is %" G_GUINT64_FORMAT
918 " samples past base_ts %" GST_TIME_FORMAT
919 ", expected ts %" GST_TIME_FORMAT, samples,
920 GST_TIME_ARGS (priv->base_ts), GST_TIME_ARGS (next_ts));
921 diff = GST_CLOCK_DIFF (next_ts, GST_BUFFER_TIMESTAMP (buffer));
922 GST_LOG_OBJECT (enc, "ts diff %d ms", (gint) (diff / GST_MSECOND));
923 /* if within tolerance,
924 * discard buffer ts and carry on producing perfect stream,
925 * otherwise clip or resync to ts */
926 if (G_UNLIKELY (diff < -enc->priv->tolerance ||
927 diff > enc->priv->tolerance)) {
928 GST_DEBUG_OBJECT (enc, "marked discont");
933 /* do some fancy tweaking in hard resync case */
934 if (discont && enc->priv->hard_resync) {
938 GST_WARNING_OBJECT (enc, "Buffer is older than expected ts %"
939 GST_TIME_FORMAT ". Clipping buffer", GST_TIME_ARGS (next_ts));
942 GST_CLOCK_TIME_TO_FRAMES (-diff, ctx->info.rate) * ctx->info.bpf;
943 if (diff_bytes >= GST_BUFFER_SIZE (buffer)) {
944 gst_buffer_unref (buffer);
947 buffer = gst_buffer_make_metadata_writable (buffer);
948 GST_BUFFER_DATA (buffer) += diff_bytes;
949 GST_BUFFER_SIZE (buffer) -= diff_bytes;
951 GST_BUFFER_TIMESTAMP (buffer) += diff;
952 /* care even less about duration after this */
954 /* drain stuff prior to resync */
955 gst_audio_encoder_drain (enc);
959 priv->base_ts += diff;
960 gst_audio_encoder_set_base_gp (enc);
961 priv->discont |= discont;
964 gst_adapter_push (enc->priv->adapter, buffer);
965 /* new stuff, so we can push subclass again */
966 enc->priv->drained = FALSE;
968 ret = gst_audio_encoder_push_buffers (enc, FALSE);
971 GST_LOG_OBJECT (enc, "chain leaving");
973 GST_AUDIO_ENCODER_STREAM_UNLOCK (enc);
980 GST_ELEMENT_ERROR (enc, CORE, NEGOTIATION, (NULL),
981 ("encoder not initialized"));
982 gst_buffer_unref (buffer);
983 ret = GST_FLOW_NOT_NEGOTIATED;
988 GST_ELEMENT_ERROR (enc, STREAM, ENCODE, (NULL),
989 ("buffer size %d not a multiple of %d", GST_BUFFER_SIZE (buffer),
991 gst_buffer_unref (buffer);
992 ret = GST_FLOW_ERROR;
998 audio_info_is_equal (GstAudioInfo * from, GstAudioInfo * to)
1002 if (from->finfo == NULL || to->finfo == NULL)
1004 if (GST_AUDIO_INFO_FORMAT (from) != GST_AUDIO_INFO_FORMAT (to))
1006 if (GST_AUDIO_INFO_RATE (from) != GST_AUDIO_INFO_RATE (to))
1008 if (GST_AUDIO_INFO_CHANNELS (from) != GST_AUDIO_INFO_CHANNELS (to))
1010 if (GST_AUDIO_INFO_CHANNELS (from) > 64)
1012 return memcmp (from->position, to->position,
1013 GST_AUDIO_INFO_CHANNELS (from) * sizeof (to->position[0]));
1017 gst_audio_encoder_sink_setcaps (GstPad * pad, GstCaps * caps)
1019 GstAudioEncoder *enc;
1020 GstAudioEncoderClass *klass;
1021 GstAudioEncoderContext *ctx;
1022 GstAudioInfo *state, *old_state;
1023 gboolean res = TRUE, changed = FALSE;
1026 enc = GST_AUDIO_ENCODER (GST_PAD_PARENT (pad));
1027 klass = GST_AUDIO_ENCODER_GET_CLASS (enc);
1029 /* subclass must do something here ... */
1030 g_return_val_if_fail (klass->set_format != NULL, FALSE);
1032 ctx = &enc->priv->ctx;
1035 GST_AUDIO_ENCODER_STREAM_LOCK (enc);
1037 GST_DEBUG_OBJECT (enc, "caps: %" GST_PTR_FORMAT, caps);
1039 if (!gst_caps_is_fixed (caps))
1042 /* adjust ts tracking to new sample rate */
1043 old_rate = GST_AUDIO_INFO_RATE (state);
1044 if (GST_CLOCK_TIME_IS_VALID (enc->priv->base_ts) && old_rate) {
1045 enc->priv->base_ts +=
1046 GST_FRAMES_TO_CLOCK_TIME (enc->priv->samples, old_rate);
1047 enc->priv->samples = 0;
1050 old_state = gst_audio_info_copy (state);
1051 if (!gst_audio_info_from_caps (state, caps))
1054 changed = !audio_info_is_equal (state, old_state);
1055 gst_audio_info_free (old_state);
1058 GstClockTime old_min_latency;
1059 GstClockTime old_max_latency;
1061 /* drain any pending old data stuff */
1062 gst_audio_encoder_drain (enc);
1064 /* context defaults */
1065 enc->priv->ctx.frame_samples_min = 0;
1066 enc->priv->ctx.frame_samples_max = 0;
1067 enc->priv->ctx.frame_max = 0;
1068 enc->priv->ctx.lookahead = 0;
1070 /* element might report latency */
1071 GST_OBJECT_LOCK (enc);
1072 old_min_latency = ctx->min_latency;
1073 old_max_latency = ctx->max_latency;
1074 GST_OBJECT_UNLOCK (enc);
1076 if (klass->set_format)
1077 res = klass->set_format (enc, state);
1079 /* notify if new latency */
1080 GST_OBJECT_LOCK (enc);
1081 if ((ctx->min_latency > 0 && ctx->min_latency != old_min_latency) ||
1082 (ctx->max_latency > 0 && ctx->max_latency != old_max_latency)) {
1083 GST_OBJECT_UNLOCK (enc);
1084 /* post latency message on the bus */
1085 gst_element_post_message (GST_ELEMENT (enc),
1086 gst_message_new_latency (GST_OBJECT (enc)));
1087 GST_OBJECT_LOCK (enc);
1089 GST_OBJECT_UNLOCK (enc);
1091 GST_DEBUG_OBJECT (enc, "new audio format identical to configured format");
1096 GST_AUDIO_ENCODER_STREAM_UNLOCK (enc);
1103 GST_WARNING_OBJECT (enc, "rejected caps %" GST_PTR_FORMAT, caps);
1110 * gst_audio_encoder_proxy_getcaps:
1111 * @enc: a #GstAudioEncoder
1112 * @caps: initial caps
1114 * Returns caps that express @caps (or sink template caps if @caps == NULL)
1115 * restricted to channel/rate combinations supported by downstream elements
1118 * Returns: a #GstCaps owned by caller
1123 gst_audio_encoder_proxy_getcaps (GstAudioEncoder * enc, GstCaps * caps)
1125 const GstCaps *templ_caps;
1126 GstCaps *allowed = NULL;
1127 GstCaps *fcaps, *filter_caps;
1130 /* we want to be able to communicate to upstream elements like audioconvert
1131 * and audioresample any rate/channel restrictions downstream (e.g. muxer
1132 * only accepting certain sample rates) */
1133 templ_caps = caps ? caps : gst_pad_get_pad_template_caps (enc->sinkpad);
1134 allowed = gst_pad_get_allowed_caps (enc->srcpad);
1135 if (!allowed || gst_caps_is_empty (allowed) || gst_caps_is_any (allowed)) {
1136 fcaps = gst_caps_copy (templ_caps);
1140 GST_LOG_OBJECT (enc, "template caps %" GST_PTR_FORMAT, templ_caps);
1141 GST_LOG_OBJECT (enc, "allowed caps %" GST_PTR_FORMAT, allowed);
1143 filter_caps = gst_caps_new_empty ();
1145 for (i = 0; i < gst_caps_get_size (templ_caps); i++) {
1148 q_name = gst_structure_get_name_id (gst_caps_get_structure (templ_caps, i));
1150 /* pick rate + channel fields from allowed caps */
1151 for (j = 0; j < gst_caps_get_size (allowed); j++) {
1152 const GstStructure *allowed_s = gst_caps_get_structure (allowed, j);
1156 s = gst_structure_id_empty_new (q_name);
1157 if ((val = gst_structure_get_value (allowed_s, "rate")))
1158 gst_structure_set_value (s, "rate", val);
1159 if ((val = gst_structure_get_value (allowed_s, "channels")))
1160 gst_structure_set_value (s, "channels", val);
1161 /* following might also make sense for some encoded formats,
1163 if ((val = gst_structure_get_value (allowed_s, "width")))
1164 gst_structure_set_value (s, "width", val);
1165 if ((val = gst_structure_get_value (allowed_s, "depth")))
1166 gst_structure_set_value (s, "depth", val);
1167 if ((val = gst_structure_get_value (allowed_s, "endianness")))
1168 gst_structure_set_value (s, "endianness", val);
1169 if ((val = gst_structure_get_value (allowed_s, "signed")))
1170 gst_structure_set_value (s, "signed", val);
1171 if ((val = gst_structure_get_value (allowed_s, "channel-positions")))
1172 gst_structure_set_value (s, "channel-positions", val);
1174 gst_caps_merge_structure (filter_caps, s);
1178 fcaps = gst_caps_intersect (filter_caps, templ_caps);
1179 gst_caps_unref (filter_caps);
1182 gst_caps_replace (&allowed, NULL);
1184 GST_LOG_OBJECT (enc, "proxy caps %" GST_PTR_FORMAT, fcaps);
1190 gst_audio_encoder_sink_getcaps (GstPad * pad)
1192 GstAudioEncoder *enc;
1193 GstAudioEncoderClass *klass;
1196 enc = GST_AUDIO_ENCODER (gst_pad_get_parent (pad));
1197 klass = GST_AUDIO_ENCODER_GET_CLASS (enc);
1198 g_assert (pad == enc->sinkpad);
1201 caps = klass->getcaps (enc);
1203 caps = gst_audio_encoder_proxy_getcaps (enc, NULL);
1204 gst_object_unref (enc);
1206 GST_LOG_OBJECT (enc, "returning caps %" GST_PTR_FORMAT, caps);
1212 gst_audio_encoder_sink_eventfunc (GstAudioEncoder * enc, GstEvent * event)
1214 GstAudioEncoderClass *klass;
1215 gboolean handled = FALSE;
1217 klass = GST_AUDIO_ENCODER_GET_CLASS (enc);
1219 switch (GST_EVENT_TYPE (event)) {
1220 case GST_EVENT_NEWSEGMENT:
1223 gdouble rate, arate;
1224 gint64 start, stop, time;
1227 gst_event_parse_new_segment_full (event, &update, &rate, &arate, &format,
1228 &start, &stop, &time);
1230 if (format == GST_FORMAT_TIME) {
1231 GST_DEBUG_OBJECT (enc, "received TIME NEW_SEGMENT %" GST_TIME_FORMAT
1232 " -- %" GST_TIME_FORMAT ", time %" GST_TIME_FORMAT
1233 ", rate %g, applied_rate %g",
1234 GST_TIME_ARGS (start), GST_TIME_ARGS (stop), GST_TIME_ARGS (time),
1237 GST_DEBUG_OBJECT (enc, "received NEW_SEGMENT %" G_GINT64_FORMAT
1238 " -- %" G_GINT64_FORMAT ", time %" G_GINT64_FORMAT
1239 ", rate %g, applied_rate %g", start, stop, time, rate, arate);
1240 GST_DEBUG_OBJECT (enc, "unsupported format; ignoring");
1244 GST_AUDIO_ENCODER_STREAM_LOCK (enc);
1245 /* finish current segment */
1246 gst_audio_encoder_drain (enc);
1247 /* reset partially for new segment */
1248 gst_audio_encoder_reset (enc, FALSE);
1249 /* and follow along with segment */
1250 gst_segment_set_newsegment_full (&enc->segment, update, rate, arate,
1251 format, start, stop, time);
1252 GST_AUDIO_ENCODER_STREAM_UNLOCK (enc);
1256 case GST_EVENT_FLUSH_START:
1259 case GST_EVENT_FLUSH_STOP:
1260 GST_AUDIO_ENCODER_STREAM_LOCK (enc);
1261 /* discard any pending stuff */
1262 /* TODO route through drain ?? */
1263 if (!enc->priv->drained && klass->flush)
1265 /* and get (re)set for the sequel */
1266 gst_audio_encoder_reset (enc, FALSE);
1268 g_list_foreach (enc->priv->pending_events, (GFunc) gst_event_unref, NULL);
1269 g_list_free (enc->priv->pending_events);
1270 enc->priv->pending_events = NULL;
1271 GST_AUDIO_ENCODER_STREAM_UNLOCK (enc);
1276 GST_AUDIO_ENCODER_STREAM_LOCK (enc);
1277 gst_audio_encoder_drain (enc);
1278 GST_AUDIO_ENCODER_STREAM_UNLOCK (enc);
1285 gst_event_parse_tag (event, &tags);
1286 tags = gst_tag_list_copy (tags);
1287 gst_event_unref (event);
1288 gst_tag_list_remove_tag (tags, GST_TAG_CODEC);
1289 gst_tag_list_remove_tag (tags, GST_TAG_AUDIO_CODEC);
1290 event = gst_event_new_tag (tags);
1292 GST_OBJECT_LOCK (enc);
1293 enc->priv->pending_events =
1294 g_list_append (enc->priv->pending_events, event);
1295 GST_OBJECT_UNLOCK (enc);
1308 gst_audio_encoder_sink_event (GstPad * pad, GstEvent * event)
1310 GstAudioEncoder *enc;
1311 GstAudioEncoderClass *klass;
1312 gboolean handled = FALSE;
1313 gboolean ret = TRUE;
1315 enc = GST_AUDIO_ENCODER (gst_pad_get_parent (pad));
1316 klass = GST_AUDIO_ENCODER_GET_CLASS (enc);
1318 GST_DEBUG_OBJECT (enc, "received event %d, %s", GST_EVENT_TYPE (event),
1319 GST_EVENT_TYPE_NAME (event));
1322 handled = klass->event (enc, event);
1325 handled = gst_audio_encoder_sink_eventfunc (enc, event);
1328 /* Forward non-serialized events and EOS/FLUSH_STOP immediately.
1329 * For EOS this is required because no buffer or serialized event
1330 * will come after EOS and nothing could trigger another
1331 * _finish_frame() call.
1333 * For FLUSH_STOP this is required because it is expected
1334 * to be forwarded immediately and no buffers are queued anyway.
1336 if (!GST_EVENT_IS_SERIALIZED (event)
1337 || GST_EVENT_TYPE (event) == GST_EVENT_EOS
1338 || GST_EVENT_TYPE (event) == GST_EVENT_FLUSH_STOP) {
1339 ret = gst_pad_event_default (pad, event);
1341 GST_AUDIO_ENCODER_STREAM_LOCK (enc);
1342 enc->priv->pending_events =
1343 g_list_append (enc->priv->pending_events, event);
1344 GST_AUDIO_ENCODER_STREAM_UNLOCK (enc);
1349 GST_DEBUG_OBJECT (enc, "event handled");
1351 gst_object_unref (enc);
1356 gst_audio_encoder_sink_query (GstPad * pad, GstQuery * query)
1358 gboolean res = TRUE;
1359 GstAudioEncoder *enc;
1361 enc = GST_AUDIO_ENCODER (gst_pad_get_parent (pad));
1363 switch (GST_QUERY_TYPE (query)) {
1364 case GST_QUERY_FORMATS:
1366 gst_query_set_formats (query, 3,
1367 GST_FORMAT_TIME, GST_FORMAT_BYTES, GST_FORMAT_DEFAULT);
1371 case GST_QUERY_CONVERT:
1373 GstFormat src_fmt, dest_fmt;
1374 gint64 src_val, dest_val;
1376 gst_query_parse_convert (query, &src_fmt, &src_val, &dest_fmt, &dest_val);
1377 if (!(res = gst_audio_info_convert (&enc->priv->ctx.info,
1378 src_fmt, src_val, dest_fmt, &dest_val)))
1380 gst_query_set_convert (query, src_fmt, src_val, dest_fmt, dest_val);
1384 res = gst_pad_query_default (pad, query);
1389 gst_object_unref (enc);
1393 static const GstQueryType *
1394 gst_audio_encoder_get_query_types (GstPad * pad)
1396 static const GstQueryType gst_audio_encoder_src_query_types[] = {
1404 return gst_audio_encoder_src_query_types;
1408 * gst_audio_encoded_audio_convert:
1409 * @fmt: audio format of the encoded audio
1410 * @bytes: number of encoded bytes
1411 * @samples: number of encoded samples
1412 * @src_format: source format
1413 * @src_value: source value
1414 * @dest_format: destination format
1415 * @dest_value: destination format
1417 * Helper function to convert @src_value in @src_format to @dest_value in
1418 * @dest_format for encoded audio data. Conversion is possible between
1419 * BYTE and TIME format by using estimated bitrate based on
1420 * @samples and @bytes (and @fmt).
1424 /* FIXME: make gst_audio_encoded_audio_convert() public? */
1426 gst_audio_encoded_audio_convert (GstAudioInfo * fmt,
1427 gint64 bytes, gint64 samples, GstFormat src_format,
1428 gint64 src_value, GstFormat * dest_format, gint64 * dest_value)
1430 gboolean res = FALSE;
1432 g_return_val_if_fail (dest_format != NULL, FALSE);
1433 g_return_val_if_fail (dest_value != NULL, FALSE);
1435 if (G_UNLIKELY (src_format == *dest_format || src_value == 0 ||
1438 *dest_value = src_value;
1442 if (samples == 0 || bytes == 0 || fmt->rate == 0) {
1443 GST_DEBUG ("not enough metadata yet to convert");
1449 switch (src_format) {
1450 case GST_FORMAT_BYTES:
1451 switch (*dest_format) {
1452 case GST_FORMAT_TIME:
1453 *dest_value = gst_util_uint64_scale (src_value,
1454 GST_SECOND * samples, bytes);
1461 case GST_FORMAT_TIME:
1462 switch (*dest_format) {
1463 case GST_FORMAT_BYTES:
1464 *dest_value = gst_util_uint64_scale (src_value, bytes,
1465 samples * GST_SECOND);
1480 /* FIXME ? are any of these queries (other than latency) an encoder's business
1481 * also, the conversion stuff might seem to make sense, but seems to not mind
1482 * segment stuff etc at all
1483 * Supposedly that's backward compatibility ... */
1485 gst_audio_encoder_src_query (GstPad * pad, GstQuery * query)
1487 GstAudioEncoder *enc;
1489 gboolean res = FALSE;
1491 enc = GST_AUDIO_ENCODER (GST_PAD_PARENT (pad));
1492 peerpad = gst_pad_get_peer (GST_PAD (enc->sinkpad));
1494 GST_LOG_OBJECT (enc, "handling query: %" GST_PTR_FORMAT, query);
1496 switch (GST_QUERY_TYPE (query)) {
1497 case GST_QUERY_POSITION:
1499 GstFormat fmt, req_fmt;
1502 if ((res = gst_pad_peer_query (enc->sinkpad, query))) {
1503 GST_LOG_OBJECT (enc, "returning peer response");
1508 GST_LOG_OBJECT (enc, "no peer");
1512 gst_query_parse_position (query, &req_fmt, NULL);
1513 fmt = GST_FORMAT_TIME;
1514 if (!(res = gst_pad_query_position (peerpad, &fmt, &pos)))
1517 if ((res = gst_pad_query_convert (peerpad, fmt, pos, &req_fmt, &val))) {
1518 gst_query_set_position (query, req_fmt, val);
1522 case GST_QUERY_DURATION:
1524 GstFormat fmt, req_fmt;
1527 if ((res = gst_pad_peer_query (enc->sinkpad, query))) {
1528 GST_LOG_OBJECT (enc, "returning peer response");
1533 GST_LOG_OBJECT (enc, "no peer");
1537 gst_query_parse_duration (query, &req_fmt, NULL);
1538 fmt = GST_FORMAT_TIME;
1539 if (!(res = gst_pad_query_duration (peerpad, &fmt, &dur)))
1542 if ((res = gst_pad_query_convert (peerpad, fmt, dur, &req_fmt, &val))) {
1543 gst_query_set_duration (query, req_fmt, val);
1547 case GST_QUERY_FORMATS:
1549 gst_query_set_formats (query, 2, GST_FORMAT_TIME, GST_FORMAT_BYTES);
1553 case GST_QUERY_CONVERT:
1555 GstFormat src_fmt, dest_fmt;
1556 gint64 src_val, dest_val;
1558 gst_query_parse_convert (query, &src_fmt, &src_val, &dest_fmt, &dest_val);
1559 if (!(res = gst_audio_encoded_audio_convert (&enc->priv->ctx.info,
1560 enc->priv->bytes_out, enc->priv->samples_in, src_fmt, src_val,
1561 &dest_fmt, &dest_val)))
1563 gst_query_set_convert (query, src_fmt, src_val, dest_fmt, dest_val);
1566 case GST_QUERY_LATENCY:
1568 if ((res = gst_pad_peer_query (enc->sinkpad, query))) {
1570 GstClockTime min_latency, max_latency;
1572 gst_query_parse_latency (query, &live, &min_latency, &max_latency);
1573 GST_DEBUG_OBJECT (enc, "Peer latency: live %d, min %"
1574 GST_TIME_FORMAT " max %" GST_TIME_FORMAT, live,
1575 GST_TIME_ARGS (min_latency), GST_TIME_ARGS (max_latency));
1577 GST_OBJECT_LOCK (enc);
1578 /* add our latency */
1579 if (min_latency != -1)
1580 min_latency += enc->priv->ctx.min_latency;
1581 if (max_latency != -1)
1582 max_latency += enc->priv->ctx.max_latency;
1583 GST_OBJECT_UNLOCK (enc);
1585 gst_query_set_latency (query, live, min_latency, max_latency);
1590 res = gst_pad_query_default (pad, query);
1594 gst_object_unref (peerpad);
1599 gst_audio_encoder_set_property (GObject * object, guint prop_id,
1600 const GValue * value, GParamSpec * pspec)
1602 GstAudioEncoder *enc;
1604 enc = GST_AUDIO_ENCODER (object);
1607 case PROP_PERFECT_TS:
1608 if (enc->priv->granule && !g_value_get_boolean (value))
1609 GST_WARNING_OBJECT (enc, "perfect-timestamp can not be set FALSE "
1610 "while granule handling is enabled");
1612 enc->priv->perfect_ts = g_value_get_boolean (value);
1614 case PROP_HARD_RESYNC:
1615 enc->priv->hard_resync = g_value_get_boolean (value);
1617 case PROP_TOLERANCE:
1618 enc->priv->tolerance = g_value_get_int64 (value);
1621 G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
1627 gst_audio_encoder_get_property (GObject * object, guint prop_id,
1628 GValue * value, GParamSpec * pspec)
1630 GstAudioEncoder *enc;
1632 enc = GST_AUDIO_ENCODER (object);
1635 case PROP_PERFECT_TS:
1636 g_value_set_boolean (value, enc->priv->perfect_ts);
1639 g_value_set_boolean (value, enc->priv->granule);
1641 case PROP_HARD_RESYNC:
1642 g_value_set_boolean (value, enc->priv->hard_resync);
1644 case PROP_TOLERANCE:
1645 g_value_set_int64 (value, enc->priv->tolerance);
1648 G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
1654 gst_audio_encoder_activate (GstAudioEncoder * enc, gboolean active)
1656 GstAudioEncoderClass *klass;
1657 gboolean result = FALSE;
1659 klass = GST_AUDIO_ENCODER_GET_CLASS (enc);
1661 g_return_val_if_fail (!enc->priv->granule || enc->priv->perfect_ts, FALSE);
1663 GST_DEBUG_OBJECT (enc, "activate %d", active);
1667 if (enc->priv->tags)
1668 gst_tag_list_free (enc->priv->tags);
1669 enc->priv->tags = gst_tag_list_new ();
1671 if (!enc->priv->active && klass->start)
1672 result = klass->start (enc);
1674 /* We must make sure streaming has finished before resetting things
1675 * and calling the ::stop vfunc */
1676 GST_PAD_STREAM_LOCK (enc->sinkpad);
1677 GST_PAD_STREAM_UNLOCK (enc->sinkpad);
1679 if (enc->priv->active && klass->stop)
1680 result = klass->stop (enc);
1683 gst_audio_encoder_reset (enc, TRUE);
1685 GST_DEBUG_OBJECT (enc, "activate return: %d", result);
1691 gst_audio_encoder_sink_activate_push (GstPad * pad, gboolean active)
1693 gboolean result = TRUE;
1694 GstAudioEncoder *enc;
1696 enc = GST_AUDIO_ENCODER (gst_pad_get_parent (pad));
1698 GST_DEBUG_OBJECT (enc, "sink activate push %d", active);
1700 result = gst_audio_encoder_activate (enc, active);
1703 enc->priv->active = active;
1705 GST_DEBUG_OBJECT (enc, "sink activate push return: %d", result);
1707 gst_object_unref (enc);
1712 * gst_audio_encoder_get_audio_info:
1713 * @enc: a #GstAudioEncoder
1715 * Returns: a #GstAudioInfo describing the input audio format
1720 gst_audio_encoder_get_audio_info (GstAudioEncoder * enc)
1722 g_return_val_if_fail (GST_IS_AUDIO_ENCODER (enc), NULL);
1724 return &enc->priv->ctx.info;
1728 * gst_audio_encoder_set_frame_samples_min:
1729 * @enc: a #GstAudioEncoder
1730 * @num: number of samples per frame
1732 * Sets number of samples (per channel) subclass needs to be handed,
1733 * at least or will be handed all available if 0.
1735 * If an exact number of samples is required, gst_audio_encoder_set_frame_samples_max()
1736 * must be called with the same number.
1741 gst_audio_encoder_set_frame_samples_min (GstAudioEncoder * enc, gint num)
1743 g_return_if_fail (GST_IS_AUDIO_ENCODER (enc));
1745 enc->priv->ctx.frame_samples_min = num;
1749 * gst_audio_encoder_get_frame_samples_min:
1750 * @enc: a #GstAudioEncoder
1752 * Returns: currently minimum requested samples per frame
1757 gst_audio_encoder_get_frame_samples_min (GstAudioEncoder * enc)
1759 g_return_val_if_fail (GST_IS_AUDIO_ENCODER (enc), 0);
1761 return enc->priv->ctx.frame_samples_min;
1765 * gst_audio_encoder_set_frame_samples_max:
1766 * @enc: a #GstAudioEncoder
1767 * @num: number of samples per frame
1769 * Sets number of samples (per channel) subclass needs to be handed,
1770 * at most or will be handed all available if 0.
1772 * If an exact number of samples is required, gst_audio_encoder_set_frame_samples_min()
1773 * must be called with the same number.
1778 gst_audio_encoder_set_frame_samples_max (GstAudioEncoder * enc, gint num)
1780 g_return_if_fail (GST_IS_AUDIO_ENCODER (enc));
1782 enc->priv->ctx.frame_samples_max = num;
1786 * gst_audio_encoder_get_frame_samples_min:
1787 * @enc: a #GstAudioEncoder
1789 * Returns: currently maximum requested samples per frame
1794 gst_audio_encoder_get_frame_samples_max (GstAudioEncoder * enc)
1796 g_return_val_if_fail (GST_IS_AUDIO_ENCODER (enc), 0);
1798 return enc->priv->ctx.frame_samples_max;
1802 * gst_audio_encoder_set_frame_max:
1803 * @enc: a #GstAudioEncoder
1804 * @num: number of frames
1806 * Sets max number of frames accepted at once (assumed minimally 1).
1807 * Requires @frame_samples_min and @frame_samples_max to be the equal.
1812 gst_audio_encoder_set_frame_max (GstAudioEncoder * enc, gint num)
1814 g_return_if_fail (GST_IS_AUDIO_ENCODER (enc));
1816 enc->priv->ctx.frame_max = num;
1820 * gst_audio_encoder_get_frame_max:
1821 * @enc: a #GstAudioEncoder
1823 * Returns: currently configured maximum handled frames
1828 gst_audio_encoder_get_frame_max (GstAudioEncoder * enc)
1830 g_return_val_if_fail (GST_IS_AUDIO_ENCODER (enc), 0);
1832 return enc->priv->ctx.frame_max;
1836 * gst_audio_encoder_set_lookahead:
1837 * @enc: a #GstAudioEncoder
1840 * Sets encoder lookahead (in units of input rate samples)
1845 gst_audio_encoder_set_lookahead (GstAudioEncoder * enc, gint num)
1847 g_return_if_fail (GST_IS_AUDIO_ENCODER (enc));
1849 enc->priv->ctx.lookahead = num;
1853 * gst_audio_encoder_get_lookahead:
1854 * @enc: a #GstAudioEncoder
1856 * Returns: currently configured encoder lookahead
1859 gst_audio_encoder_get_lookahead (GstAudioEncoder * enc)
1861 g_return_val_if_fail (GST_IS_AUDIO_ENCODER (enc), 0);
1863 return enc->priv->ctx.lookahead;
1867 * gst_audio_encoder_set_latency:
1868 * @enc: a #GstAudioEncoder
1869 * @min: minimum latency
1870 * @max: maximum latency
1872 * Sets encoder latency.
1877 gst_audio_encoder_set_latency (GstAudioEncoder * enc,
1878 GstClockTime min, GstClockTime max)
1880 g_return_if_fail (GST_IS_AUDIO_ENCODER (enc));
1882 GST_OBJECT_LOCK (enc);
1883 enc->priv->ctx.min_latency = min;
1884 enc->priv->ctx.max_latency = max;
1885 GST_OBJECT_UNLOCK (enc);
1889 * gst_audio_encoder_get_latency:
1890 * @enc: a #GstAudioEncoder
1891 * @min: (out) (allow-none): a pointer to storage to hold minimum latency
1892 * @max: (out) (allow-none): a pointer to storage to hold maximum latency
1894 * Sets the variables pointed to by @min and @max to the currently configured
1900 gst_audio_encoder_get_latency (GstAudioEncoder * enc,
1901 GstClockTime * min, GstClockTime * max)
1903 g_return_if_fail (GST_IS_AUDIO_ENCODER (enc));
1905 GST_OBJECT_LOCK (enc);
1907 *min = enc->priv->ctx.min_latency;
1909 *max = enc->priv->ctx.max_latency;
1910 GST_OBJECT_UNLOCK (enc);
1914 * gst_audio_encoder_set_mark_granule:
1915 * @enc: a #GstAudioEncoder
1916 * @enabled: new state
1918 * Enable or disable encoder granule handling.
1925 gst_audio_encoder_set_mark_granule (GstAudioEncoder * enc, gboolean enabled)
1927 g_return_if_fail (GST_IS_AUDIO_ENCODER (enc));
1929 GST_LOG_OBJECT (enc, "enabled: %d", enabled);
1931 GST_OBJECT_LOCK (enc);
1932 enc->priv->granule = enabled;
1933 GST_OBJECT_UNLOCK (enc);
1937 * gst_audio_encoder_get_mark_granule:
1938 * @enc: a #GstAudioEncoder
1940 * Queries if the encoder will handle granule marking.
1942 * Returns: TRUE if granule marking is enabled.
1949 gst_audio_encoder_get_mark_granule (GstAudioEncoder * enc)
1953 g_return_val_if_fail (GST_IS_AUDIO_ENCODER (enc), FALSE);
1955 GST_OBJECT_LOCK (enc);
1956 result = enc->priv->granule;
1957 GST_OBJECT_UNLOCK (enc);
1963 * gst_audio_encoder_set_perfect_timestamp:
1964 * @enc: a #GstAudioEncoder
1965 * @enabled: new state
1967 * Enable or disable encoder perfect output timestamp preference.
1974 gst_audio_encoder_set_perfect_timestamp (GstAudioEncoder * enc,
1977 g_return_if_fail (GST_IS_AUDIO_ENCODER (enc));
1979 GST_LOG_OBJECT (enc, "enabled: %d", enabled);
1981 GST_OBJECT_LOCK (enc);
1982 enc->priv->perfect_ts = enabled;
1983 GST_OBJECT_UNLOCK (enc);
1987 * gst_audio_encoder_get_perfect_timestamp:
1988 * @enc: a #GstAudioEncoder
1990 * Queries encoder perfect timestamp behaviour.
1992 * Returns: TRUE if pefect timestamp setting enabled.
1999 gst_audio_encoder_get_perfect_timestamp (GstAudioEncoder * enc)
2003 g_return_val_if_fail (GST_IS_AUDIO_ENCODER (enc), FALSE);
2005 GST_OBJECT_LOCK (enc);
2006 result = enc->priv->perfect_ts;
2007 GST_OBJECT_UNLOCK (enc);
2013 * gst_audio_encoder_set_hard_sync:
2014 * @enc: a #GstAudioEncoder
2015 * @enabled: new state
2017 * Sets encoder hard resync handling.
2024 gst_audio_encoder_set_hard_resync (GstAudioEncoder * enc, gboolean enabled)
2026 g_return_if_fail (GST_IS_AUDIO_ENCODER (enc));
2028 GST_LOG_OBJECT (enc, "enabled: %d", enabled);
2030 GST_OBJECT_LOCK (enc);
2031 enc->priv->hard_resync = enabled;
2032 GST_OBJECT_UNLOCK (enc);
2036 * gst_audio_encoder_get_hard_sync:
2037 * @enc: a #GstAudioEncoder
2039 * Queries encoder's hard resync setting.
2041 * Returns: TRUE if hard resync is enabled.
2048 gst_audio_encoder_get_hard_resync (GstAudioEncoder * enc)
2052 g_return_val_if_fail (GST_IS_AUDIO_ENCODER (enc), FALSE);
2054 GST_OBJECT_LOCK (enc);
2055 result = enc->priv->hard_resync;
2056 GST_OBJECT_UNLOCK (enc);
2062 * gst_audio_encoder_set_tolerance:
2063 * @enc: a #GstAudioEncoder
2064 * @tolerance: new tolerance
2066 * Configures encoder audio jitter tolerance threshold.
2073 gst_audio_encoder_set_tolerance (GstAudioEncoder * enc, gint64 tolerance)
2075 g_return_if_fail (GST_IS_AUDIO_ENCODER (enc));
2077 GST_OBJECT_LOCK (enc);
2078 enc->priv->tolerance = tolerance;
2079 GST_OBJECT_UNLOCK (enc);
2083 * gst_audio_encoder_get_tolerance:
2084 * @enc: a #GstAudioEncoder
2086 * Queries current audio jitter tolerance threshold.
2088 * Returns: encoder audio jitter tolerance threshold.
2095 gst_audio_encoder_get_tolerance (GstAudioEncoder * enc)
2099 g_return_val_if_fail (GST_IS_AUDIO_ENCODER (enc), 0);
2101 GST_OBJECT_LOCK (enc);
2102 result = enc->priv->tolerance;
2103 GST_OBJECT_UNLOCK (enc);
2109 * gst_audio_encoder_merge_tags:
2110 * @enc: a #GstAudioEncoder
2111 * @tags: a #GstTagList to merge
2112 * @mode: the #GstTagMergeMode to use
2114 * Adds tags to so-called pending tags, which will be processed
2115 * before pushing out data downstream.
2117 * Note that this is provided for convenience, and the subclass is
2118 * not required to use this and can still do tag handling on its own,
2119 * although it should be aware that baseclass already takes care
2120 * of the usual CODEC/AUDIO_CODEC tags.
2127 gst_audio_encoder_merge_tags (GstAudioEncoder * enc,
2128 const GstTagList * tags, GstTagMergeMode mode)
2132 g_return_if_fail (GST_IS_AUDIO_ENCODER (enc));
2133 g_return_if_fail (tags == NULL || GST_IS_TAG_LIST (tags));
2135 GST_OBJECT_LOCK (enc);
2137 GST_DEBUG_OBJECT (enc, "merging tags %" GST_PTR_FORMAT, tags);
2138 otags = enc->priv->tags;
2139 enc->priv->tags = gst_tag_list_merge (enc->priv->tags, tags, mode);
2141 gst_tag_list_free (otags);
2142 GST_OBJECT_UNLOCK (enc);