Translation from C to D of libsoundio (commit a46b0f2).

Libsoundio is a library for cross-platform real-time audio input and output.

It is licensed under the MIT License. The translation is not affiliated with the original project.

Currently not all backends are translated.

Add this package as a dependency to your project.

dub.sdl:

dependency "libsoundio-d" version="~>1.0.0"

dub.json:

"dependencies": {
	"libsoundio-d": "~>1.0.0"
}

And then use it:

import soundio.api;

void main() {
	SoundIo* soundio = soundio_create();
	soundio_connect(soundio);
	// your app
	soundio_destroy(soundio);
}

The configuration should be automatically selected based on your platform, but you can also choose one explicitly:

• linux
• windows
• dummy

dub.sdl:

subConfiguration "libsoundio-d" "dummy"

dub.json:

"subConfigurations": {
	"libsoundio-d": "dummy"
}

On Linux, you should have ALSA and PulseAudio installed (which you probably have by default, otherwise look up how to install it).

The following version identifiers are used:

• SOUNDIO_HAVE_JACK
• SOUNDIO_HAVE_PULSEAUDIO
• SOUNDIO_HAVE_ALSA
• SOUNDIO_HAVE_COREAUDIO
• SOUNDIO_HAVE_WASAPI

Run the examples

Assuming your current directory is the root of this repository:

dub run libsoundio-d:sine
dub run libsoundio-d:list-devices -- --short
dub run libsoundio-d:microphone -- --latency 0.05
dub run libsoundio-d:record -- output.bin

Run the tests

dub run libsoundio-d:backend-disconnect-recover
dub run libsoundio-d:latency
dub run libsoundio-d:overflow
dub run libsoundio-d:underflow
dub run libsoundio-d:unit-tests

The translation is closely converting C-syntax to D-syntax, no attempt to change the style to idiomatic D has been made. There are a few exceptions where certain constructs had to be changed however.

• ALSA defines certain structs with an unknown size at compile time. There are specific alloca macros that allocate these structs on the stack, and libsoundio uses these. I translated these with malloc and free variants, because alloca has its own share of problems.

• Libsoundio has certain static functions with the same name across backends: both ALSA and Pulse have probe_device and my_flush_events. Since D does not have C's notion of static functions (even private functions emit symbols), this introduces a name clash. Worse, because DMD emits weak symbols, it gives no multiple definition error, but instead silently calls the wrong function: The Pulse backend calls probe_device from the alsa backend instead of its own. This is mitigated by making those functions extern(D) giving them unique names.

• D does not have C bindings of stdatomic.h, so I used equivalent functionality from core.atomic. core.atomic has no direct equivalent of 'flag test and set'. I initially translated as cas (compare and swap), but return value needed to be negated. atomicFetchAdd was added in dmd 2.089 and is not supported in LDC as of 1.22, so I used atomicOp!"+=" instead.

• On Windows InterlockedIncrement and InterlockedDecrement are not in the shipped import library "Kernel32.lib". I replaced it with a corresponding atomicOp!"+=" and atomicOp!"-=".

• On 32-bit Windows, 64-bit atomic operations are not supported. The SoundIoRingBuffer uses ulong for its read and write offset, even on 32-bit. I changed these to a size_t instead.

• Use of fprintf(sderr, ...) on Windows with extern(C) main triggers issue 20532. A custom printf_stderr function was made to work around this.