39.1.1. Audio Device Objects

The audio device objects are returned by open() define the following methods (except control objects which only provide getinfo(), setinfo(), fileno(), and drain()):

audio device.close()

This method explicitly closes the device. It is useful in situations where deleting the object does not immediately close it since there are other references to it. A closed device should not be used again.

audio device.fileno()

Returns the file descriptor associated with the device. This can be used to set up SIGPOLL notification, as described below.

audio device.drain()

This method waits until all pending output is processed and then returns. Calling this method is often not necessary: destroying the object will automatically close the audio device and this will do an implicit drain.

audio device.flush()

This method discards all pending output. It can be used avoid the slow response to a user’s stop request (due to buffering of up to one second of sound).

audio device.getinfo()

This method retrieves status information like input and output volume, etc. and returns it in the form of an audio status object. This object has no methods but it contains a number of attributes describing the current device status. The names and meanings of the attributes are described in <sun/audioio.h> and in the audio(7I) manual page. Member names are slightly different from their C counterparts: a status object is only a single structure. Members of the play substructure have o_ prepended to their name and members of the record structure have i_. So, the C member play.sample_rate is accessed as o_sample_rate, record.gain as i_gain and monitor_gain plainly as monitor_gain.

audio device.ibufcount()

This method returns the number of samples that are buffered on the recording side, i.e. the program will not block on a read() call of so many samples.

audio device.obufcount()

This method returns the number of samples buffered on the playback side. Unfortunately, this number cannot be used to determine a number of samples that can be written without blocking since the kernel output queue length seems to be variable.

audio device.read(size)

This method reads size samples from the audio input and returns them as a Python string. The function blocks until enough data is available.

audio device.setinfo(status)

This method sets the audio device status parameters. The status parameter is an device status object as returned by getinfo() and possibly modified by the program.

audio device.write(samples)

Write is passed a Python string containing audio samples to be played. If there is enough buffer space free it will immediately return, otherwise it will block.

The audio device supports asynchronous notification of various events, through the SIGPOLL signal. Here’s an example of how you might enable this in Python:

def handle_sigpoll(signum, frame):
    print 'I got a SIGPOLL update'

import fcntl, signal, STROPTS

signal.signal(signal.SIGPOLL, handle_sigpoll)
fcntl.ioctl(audio_obj.fileno(), STROPTS.I_SETSIG, STROPTS.S_MSG)