Support for multiple sounds on the same stream

This commit is contained in:
Fabio Manganiello 2018-12-26 16:10:14 +01:00
parent c2544524e1
commit 3554968468
3 changed files with 81 additions and 43 deletions

View file

@ -78,9 +78,13 @@ class RedisBackend(Backend):
format(self.queue, self.redis_args))
while not self.should_stop():
msg = self.get_message()
self.logger.info('Received message on the Redis backend: {}'.format(msg))
self.on_message(msg)
try:
msg = self.get_message()
self.logger.info('Received message on the Redis backend: {}'.
format(msg))
self.on_message(msg)
except Exception as e:
self.logger.exception(e)
# vim:sw=4:ts=4:et:

View file

@ -13,8 +13,9 @@ class Sound(object):
STANDARD_A_FREQUENCY = 440.0
STANDARD_A_MIDI_NOTE = 69
_DEFAULT_BLOCKSIZE = 2048
_DEFAULT_BUFSIZE = 20
_DEFAULT_BLOCKSIZE = 1024
_DEFAULT_SYNTH_BUFSIZE = 2
_DEFAULT_FILE_BUFSIZE = 20
_DEFAULT_SAMPLERATE = 44100
midi_note = None
@ -159,9 +160,11 @@ class Mix(object):
through an audio stream to an audio device
"""
_sounds = []
_sounds = None
def __init__(self, *sounds):
self._sounds = []
for sound in sounds:
self.add(sound)
@ -235,4 +238,21 @@ class Mix(object):
return wave
def duration(self):
"""
:returns: The duration of the mix in seconds as duration of its longest
sample, or None if the mixed sample have no duration set
"""
duration = 0
for sound in self._sounds:
if sound.duration is None:
return None
duration = max(duration, sound.duration)
return duration
# vim:sw=4:ts=4:et:

View file

@ -12,7 +12,7 @@ import time
from enum import Enum
from threading import Thread, Event, RLock
from .core import Sound
from .core import Sound, Mix
from platypush.plugins import Plugin, action
@ -41,8 +41,7 @@ class SoundPlugin(Plugin):
def __init__(self, input_device=None, output_device=None,
input_blocksize=Sound._DEFAULT_BLOCKSIZE,
output_blocksize=Sound._DEFAULT_BLOCKSIZE,
playback_bufsize=Sound._DEFAULT_BUFSIZE, *args, **kwargs):
output_blocksize=Sound._DEFAULT_BLOCKSIZE, *args, **kwargs):
"""
:param input_device: Index or name of the default input device. Use :method:`platypush.plugins.sound.query_devices` to get the available devices. Default: system default
:type input_device: int or str
@ -55,9 +54,6 @@ class SoundPlugin(Plugin):
:param output_blocksize: Blocksize to be applied to the output device. Try to increase this value if you get output underflow errors while playing. Default: 2048
:type output_blocksize: int
:param playback_bufsize: Number of audio blocks that will be cached while playing (default: 20)
:type playback_bufsize: int
"""
super().__init__(*args, **kwargs)
@ -66,11 +62,11 @@ class SoundPlugin(Plugin):
self.output_device = output_device
self.input_blocksize = input_blocksize
self.output_blocksize = output_blocksize
self.playback_bufsize = playback_bufsize
self.playback_state = {}
self.playback_state_lock = RLock()
self.playback_paused_changed = {}
self.stream_mixes = {}
self.recording_state = RecordingState.STOPPED
self.recording_state_lock = RLock()
self.recording_paused_changed = Event()
@ -173,7 +169,7 @@ class SoundPlugin(Plugin):
@action
def play(self, file=None, sound=None, device=None, blocksize=None,
bufsize=Sound._DEFAULT_BUFSIZE, samplerate=None, channels=None,
bufsize=None, samplerate=None, channels=None,
stream_index=None):
"""
Plays a sound file (support formats: wav, raw) or a synthetic sound.
@ -201,7 +197,8 @@ class SoundPlugin(Plugin):
`output_blocksize` or 2048)
:type blocksize: int
:param bufsize: Size of the audio buffer (default: 20)
:param bufsize: Size of the audio buffer (default: 20 frames for audio
files, 2 frames for synth sounds)
:type bufsize: int
:param samplerate: Audio samplerate. Default: audio file samplerate if
@ -227,9 +224,17 @@ class SoundPlugin(Plugin):
if blocksize is None:
blocksize = self.output_blocksize
if bufsize is None:
if file:
bufsize = Sound._DEFAULT_FILE_BUFSIZE
else:
bufsize = Sound._DEFAULT_SYNTH_BUFSIZE
q = queue.Queue(maxsize=bufsize)
f = None
t = 0.
is_new_stream = stream_index is None or \
self.active_streams.get(stream_index) is None
if file:
file = os.path.abspath(os.path.expanduser(file))
@ -239,21 +244,28 @@ class SoundPlugin(Plugin):
if device is None:
device = self._get_default_device('output')
if file:
import soundfile as sf
f = sf.SoundFile(file)
if not samplerate:
samplerate = f.samplerate if f else Sound._DEFAULT_SAMPLERATE
if not channels:
channels = f.channels if f else 1
self.logger.info('Starting playback of {} to sound device [{}]'.
format(file or sound, device))
if is_new_stream:
stream_index = self._allocate_stream_index()
if sound:
mix = self.stream_mixes[stream_index]
mix.add(sound)
if not is_new_stream:
return # Let the existing callback handle the new mix
try:
if file:
import soundfile as sf
f = sf.SoundFile(file)
if not samplerate:
samplerate = f.samplerate if f else Sound._DEFAULT_SAMPLERATE
if not channels:
channels = f.channels if f else 1
self.logger.info('Starting playback of {} to sound device [{}]'.
format(file or sound, device))
if sound:
sound = Sound.build(sound)
# Audio queue pre-fill loop
for _ in range(bufsize):
if f:
@ -261,24 +273,26 @@ class SoundPlugin(Plugin):
if not data:
break
else:
duration = mix.duration()
blocktime = float(blocksize / samplerate)
next_t = min(t+blocktime, sound.duration) \
if sound.duration is not None else t+blocktime
next_t = min(t+blocktime, duration) \
if duration is not None else t+blocktime
data = sound.get_wave(t_start=t, t_end=next_t,
data = mix.get_wave(t_start=t, t_end=next_t,
samplerate=samplerate)
t = next_t
if sound.duration is not None and t >= sound.duration:
if duration is not None and t >= duration:
break
q.put_nowait(data) # Pre-fill the audio queue
if stream_index is None:
stream = self.active_streams[stream_index]
completed_callback_event = self.completed_callback_events[stream_index]
if stream is None:
streamtype = sd.RawOutputStream if file else sd.OutputStream
stream_index = self._allocate_stream_index()
completed_callback_event = self.completed_callback_events[stream_index]
stream = streamtype(samplerate=samplerate, blocksize=blocksize,
device=device, channels=channels,
dtype='float32',
@ -289,9 +303,6 @@ class SoundPlugin(Plugin):
finished_callback=completed_callback_event.set)
self._start_playback(stream_index=stream_index, stream=stream)
else:
stream = self.active_streams[stream_index]
completed_callback_event = self.completed_callback_events[stream_index]
with stream:
# Timeout set until we expect all the buffered blocks to
@ -308,15 +319,16 @@ class SoundPlugin(Plugin):
if not data:
break
else:
duration = mix.duration()
blocktime = float(blocksize / samplerate)
next_t = min(t+blocktime, sound.duration) \
if sound.duration is not None else t+blocktime
next_t = min(t+blocktime, duration) \
if duration is not None else t+blocktime
data = sound.get_wave(t_start=t, t_end=next_t,
data = mix.get_wave(t_start=t, t_end=next_t,
samplerate=samplerate)
t = next_t
if sound.duration is not None and t >= sound.duration:
if duration is not None and t >= duration:
break
if self._get_playback_state(stream_index) == \
@ -586,6 +598,7 @@ class SoundPlugin(Plugin):
raise RuntimeError('No stream index available')
self.active_streams[stream_index] = None
self.stream_mixes[stream_index] = Mix()
self.completed_callback_events[stream_index] = \
completed_callback_event if completed_callback_event else Event()
@ -632,6 +645,7 @@ class SoundPlugin(Plugin):
event.wait()
del self.completed_callback_events[i]
del self.active_streams[i]
del self.stream_mixes[i]
self.logger.info('Playback stopped on streams [{}]'.format(
', '.join([str(stream) for stream in