libsoundio

C99 library providing cross-platform audio input and output. The API is suitable for real-time software such as digital audio workstations as well as consumer software such as music players.

This library is an abstraction; however in the delicate balance between performance and power, and API convenience, the scale is tipped closer to the former. Features that only exist in some sound backends are exposed.

The goal of this library is to be the only resource needed to implement top quality audio playback and capture on desktop and laptop systems. This includes detailed documentation explaining how audio works on each supported backend, how they are abstracted to provide the libsoundio API, and what assumptions you can and cannot make in order to guarantee consistent, reliable behavior on every platform.

This project is a work-in-progress.

Features and Limitations

• Supported backends:
  • JACK
  • PulseAudio
  • ALSA
  • (planned) CoreAudio
  • Dummy (silence)
  • (planned) WASAPI
  • (planned) ASIO
• C library. Depends only on the respective backend API libraries and libc. Does not depend on libstdc++, and does not have exceptions, run-time type information, or setjmp.
• Errors are communicated via return codes, not logging to stdio. This is one of my many complaints against PortAudio.
• Supports channel layouts (also known as channel maps), important for surround sound applications.
• Ability to monitor devices and get an event when available devices change.
• Ability to get an event when the backend is disconnected, for example when the JACK server or PulseAudio server shuts down.
• Detects which input device is default and which output device is default.
• Ability to connect to multiple backends at once. For example you could have an ALSA device open and a JACK device open at the same time.
• Meticulously checks all return codes and memory allocations and uses meaningful error codes.

Synopsis

Complete program to emit a sine wave over the default device using the best backend:

#include <soundio/soundio.h>

#include <stdio.h>
#include <stdarg.h>
#include <stdlib.h>
#include <string.h>
#include <math.h>

__attribute__ ((cold))
__attribute__ ((noreturn))
__attribute__ ((format (printf, 1, 2)))
static void panic(const char *format, ...) {
    va_list ap;
    va_start(ap, format);
    vfprintf(stderr, format, ap);
    fprintf(stderr, "\n");
    va_end(ap);
    abort();
}

static const float PI = 3.1415926535f;
static float seconds_offset = 0.0f;
static void write_callback(struct SoundIoOutStream *outstream, int requested_frame_count) {
    float float_sample_rate = outstream->sample_rate;
    float seconds_per_frame = 1.0f / float_sample_rate;
    int err;

    for (;;) {
        int frame_count = requested_frame_count;

        struct SoundIoChannelArea *areas;
        if ((err = soundio_outstream_begin_write(outstream, &areas, &frame_count)))
            panic("%s", soundio_strerror(err));

        if (!frame_count)
            break;

        const struct SoundIoChannelLayout *layout = &outstream->layout;

        float pitch = 440.0f;
        float radians_per_second = pitch * 2.0f * PI;
        for (int frame = 0; frame < frame_count; frame += 1) {
            float sample = sinf((seconds_offset + frame * seconds_per_frame) * radians_per_second);
            for (int channel = 0; channel < layout->channel_count; channel += 1) {
                float *ptr = (float*)(areas[channel].ptr + areas[channel].step * frame);
                *ptr = sample;
            }
        }
        seconds_offset += seconds_per_frame * frame_count;

        if ((err = soundio_outstream_end_write(outstream, frame_count)))
            panic("%s", soundio_strerror(err));

        requested_frame_count -= frame_count;
        if (requested_frame_count <= 0)
            break;
    }
}

int main(int argc, char **argv) {
    struct SoundIo *soundio = soundio_create();
    if (!soundio)
        panic("out of memory");

    if ((err = soundio_connect(soundio)))
        panic("error connecting: %s", soundio_strerror(err));

    int default_out_device_index = soundio_get_default_output_device_index(soundio);
    if (default_out_device_index < 0)
        panic("no output device found");

    struct SoundIoDevice *device = soundio_get_output_device(soundio, default_out_device_index);
    if (!device)
        panic("out of memory");

    fprintf(stderr, "Output device: %s: %s\n", device->name, device->description);

    struct SoundIoOutStream *outstream = soundio_outstream_create(device);
    outstream->format = SoundIoFormatFloat32NE;
    outstream->write_callback = write_callback;

    if ((err = soundio_outstream_open(outstream)))
        panic("unable to open device: %s", soundio_strerror(err));

    if (outstream->layout_error)
        fprintf(stderr, "unable to set channel layout: %s\n", soundio_strerror(outstream->layout_error));

    if ((err = soundio_outstream_start(outstream)))
        panic("unable to start device: %s", soundio_strerror(err));

    for (;;)
        soundio_wait_events(soundio);

    soundio_outstream_destroy(outstream);
    soundio_device_unref(device);
    soundio_destroy(soundio);
    return 0;
}

Backend Priority

When you use soundio_connect, libsoundio tries these backends in order. If unable to connect to that backend, due to the backend not being installed, or the server not running, or the platform is wrong, the next backend is tried.

0. JACK
1. PulseAudio
2. ALSA (Linux)
3. CoreAudio (OSX)
4. WASAPI (Windows)
5. ASIO (Windows)
6. Dummy

If you don't like this order, you can use soundio_connect_backend to explicitly choose a backend to connect to. You can use soundio_backend_count and soundio_get_backend to get the list of available backends.

For complete API documentation, see src/soundio.h.

Contributing

libsoundio is programmed in a tiny subset of C++11:

• No STL.
• No new or delete.
• No class. All fields in structs are public.
• No exceptions or run-time type information.
• No references.
• No linking against libstdc++.

Do not be fooled - this is a C library, not a C++ library. We just take advantage of a select few C++11 compiler features such as templates, and then link against libc.

Building

Install the dependencies:

• cmake
• ALSA library (optional)
• libjack2 (optional)
• libpulseaudio (optional)

mkdir build
cd build
cmake ..
make
sudo make install

Building for Windows

You can build libsoundio with mxe. Follow the requirements section to install the packages necessary on your system. Then somewhere on your file system:

git clone https://github.com/mxe/mxe
cd mxe
make gcc

Then in the libsoundio source directory (replace "/path/to/mxe" with the appropriate path):

mkdir build-win
cd build-win
cmake .. -DCMAKE_TOOLCHAIN_FILE=/path/to/mxe/usr/i686-w64-mingw32.static/share/cmake/mxe-conf.cmake
make

Running the Tests

make test

For more detailed output:

make
./unit_tests

To see test coverage, install lcov, run make coverage and then view coverage/index.html in a browser.

Roadmap

0. implement CoreAudio (OSX) backend, get examples working
1. Add some builtin channel layouts from https://developer.apple.com/library/mac/documentation/MusicAudio/Reference/CoreAudioDataTypesRef/#//apple_ref/doc/constant_group/Audio_Channel_Layout_Tags
2. Make sure sending bogus device id results in "SoundIoErrorNoSuchDevice" on each backend
3. Make sure PulseAudio can handle refresh devices crashing before block_until_have_devices
4. CoreAudio exposes a list of min/max pairs of supported sample rates. libsoundio should do the same.
5. implement WASAPI (Windows) backend, get examples working
6. implement ASIO (Windows) backend, get examples working
7. Integrate into libgroove and test with Groove Basin
8. clear buffer maybe could take an argument to say how many frames to not clear
9. Verify that JACK xrun callback context is the same as process callback. If not, might need to hav xrun callback set a flag and have process callback call the underflow callback.
10. Create a test for pausing and resuming input and output streams.
11. Create a test for the latency / synchronization API.
    • Input is an audio file and some events indexed at particular frame - when listening the events should line up exactly with a beat or visual indicator, even when the latency is large.
    • Play the audio file, have the user press an input right at the beat. Find out what the frame index it thinks the user pressed it at and make sure that is correct.
12. Create a test for input stream overflow handling.
13. Allow calling functions from outside the callbacks as long as they first call lock and then unlock when done.
14. Should pause/resume be callable from outside the callbacks?
15. clean up API and improve documentation
    • make sure every function which can return an error documents which errors it can return
16. use a documentation generator and host the docs somewhere
17. -fvisibility=hidden and then explicitly export stuff, or explicitly make the unexported stuff private
18. add len arguments to APIs that have char *
    • replace strdup with soundio_str_dupe
19. Support PulseAudio proplist properties for main context and streams
20. Expose JACK options in jack_client_open
21. custom allocator support
22. mlock memory which is accessed in the real time path
23. make rtprio warning a callback and have existing behavior be the default callback
24. write detailed docs on buffer underflows explaining when they occur, what state changes are related to them, and how to recover from them.
25. Consider testing on FreeBSD
26. PulseAudio idea: when prebuf gets set to 0 need to pass PA_STREAM_START_CORKED. In PulseAudio, to get buffer duration and period duration, fill the buffer with silence before starting, start the stream corked, and have the callback be a callback that just provides silence. Once soundio_outstream_start is called, switch to the real callback, then call pa_stream_flush, then uncork the stream.
27. In ALSA do we need to wake up the poll when destroying the in or out stream?
28. Detect PulseAudio server going offline and emit on_backend_disconnect.

Planned Uses for libsoundio

• Genesis
• libgroove (Groove Basin)