mach/src/sysaudio/coreaudio.zig
Stephen Gutekanst bca1543391 src/sysaudio: move mach-sysaudio@ce8ab30dd300b822224d14997c58c06520b642c9 package to here
Helps hexops/mach#1165

Signed-off-by: Stephen Gutekanst <stephen@hexops.com>
2024-03-05 00:22:22 -07:00

769 lines
27 KiB
Zig

const std = @import("std");
const builtin = @import("builtin");
const main = @import("main.zig");
const backends = @import("backends.zig");
const util = @import("util.zig");
const c = @cImport({
@cInclude("CoreAudio/CoreAudio.h");
@cInclude("AudioUnit/AudioUnit.h");
});
const avaudio = @import("objc").avf_audio.avaudio;
const native_endian = builtin.cpu.arch.endian();
var is_darling = false;
const default_sample_rate = 44_100; // Hz
pub const Context = struct {
allocator: std.mem.Allocator,
devices_info: util.DevicesInfo,
pub fn init(allocator: std.mem.Allocator, options: main.Context.Options) !backends.Context {
_ = options;
if (std.fs.accessAbsolute("/usr/lib/darling", .{})) {
is_darling = true;
} else |_| {}
const ctx = try allocator.create(Context);
errdefer allocator.destroy(ctx);
ctx.* = .{
.allocator = allocator,
.devices_info = util.DevicesInfo.init(),
};
return .{ .coreaudio = ctx };
}
pub fn deinit(ctx: *Context) void {
for (ctx.devices_info.list.items) |d|
freeDevice(ctx.allocator, d);
ctx.devices_info.list.deinit(ctx.allocator);
ctx.allocator.destroy(ctx);
}
pub fn refresh(ctx: *Context) !void {
for (ctx.devices_info.list.items) |d|
freeDevice(ctx.allocator, d);
ctx.devices_info.clear();
var prop_address = c.AudioObjectPropertyAddress{
.mSelector = c.kAudioHardwarePropertyDevices,
.mScope = c.kAudioObjectPropertyScopeGlobal,
.mElement = c.kAudioObjectPropertyElementMain,
};
var io_size: u32 = 0;
if (c.AudioObjectGetPropertyDataSize(
c.kAudioObjectSystemObject,
&prop_address,
0,
null,
&io_size,
) != c.noErr) {
return error.OpeningDevice;
}
const devices_count = io_size / @sizeOf(c.AudioObjectID);
if (devices_count == 0) return;
const devs = try ctx.allocator.alloc(c.AudioObjectID, devices_count);
defer ctx.allocator.free(devs);
if (c.AudioObjectGetPropertyData(
c.kAudioObjectSystemObject,
&prop_address,
0,
null,
&io_size,
@as(*anyopaque, @ptrCast(devs)),
) != c.noErr) {
return error.OpeningDevice;
}
var default_input_id: c.AudioObjectID = undefined;
var default_output_id: c.AudioObjectID = undefined;
io_size = @sizeOf(c.AudioObjectID);
if (c.AudioHardwareGetProperty(
c.kAudioHardwarePropertyDefaultInputDevice,
&io_size,
&default_input_id,
) != c.noErr) {
return error.OpeningDevice;
}
io_size = @sizeOf(c.AudioObjectID);
if (c.AudioHardwareGetProperty(
c.kAudioHardwarePropertyDefaultOutputDevice,
&io_size,
&default_output_id,
) != c.noErr) {
return error.OpeningDevice;
}
for (devs) |id| {
const buf_list = try ctx.allocator.create(c.AudioBufferList);
defer ctx.allocator.destroy(buf_list);
for (std.meta.tags(main.Device.Mode)) |mode| {
io_size = 0;
prop_address.mSelector = c.kAudioDevicePropertyStreamConfiguration;
prop_address.mScope = switch (mode) {
.playback => c.kAudioObjectPropertyScopeOutput,
.capture => c.kAudioObjectPropertyScopeInput,
};
if (c.AudioObjectGetPropertyDataSize(
id,
&prop_address,
0,
null,
&io_size,
) != c.noErr) {
continue;
}
if (c.AudioObjectGetPropertyData(
id,
&prop_address,
0,
null,
&io_size,
buf_list,
) != c.noErr) {
return error.OpeningDevice;
}
if (buf_list.mBuffers[0].mNumberChannels == 0) break;
const audio_buffer_list_property_address = c.AudioObjectPropertyAddress{
.mSelector = c.kAudioDevicePropertyStreamConfiguration,
.mScope = switch (mode) {
.playback => c.kAudioDevicePropertyScopeOutput,
.capture => c.kAudioDevicePropertyScopeInput,
},
.mElement = c.kAudioObjectPropertyElementMain,
};
var output_audio_buffer_list: c.AudioBufferList = undefined;
var audio_buffer_list_size: c_uint = undefined;
if (c.AudioObjectGetPropertyDataSize(
id,
&audio_buffer_list_property_address,
0,
null,
&audio_buffer_list_size,
) != c.noErr) {
return error.OpeningDevice;
}
if (c.AudioObjectGetPropertyData(
id,
&prop_address,
0,
null,
&audio_buffer_list_size,
&output_audio_buffer_list,
) != c.noErr) {
return error.OpeningDevice;
}
var output_channel_count: usize = 0;
for (0..output_audio_buffer_list.mNumberBuffers) |mBufferIndex| {
output_channel_count += output_audio_buffer_list.mBuffers[mBufferIndex].mNumberChannels;
}
const channels = try ctx.allocator.alloc(main.ChannelPosition, output_channel_count);
prop_address.mSelector = c.kAudioDevicePropertyNominalSampleRate;
io_size = @sizeOf(f64);
var sample_rate: f64 = undefined;
if (c.AudioObjectGetPropertyData(
id,
&prop_address,
0,
null,
&io_size,
&sample_rate,
) != c.noErr) {
return error.OpeningDevice;
}
io_size = @sizeOf([*]const u8);
if (c.AudioDeviceGetPropertyInfo(
id,
0,
0,
c.kAudioDevicePropertyDeviceName,
&io_size,
null,
) != c.noErr) {
return error.OpeningDevice;
}
const name = try ctx.allocator.allocSentinel(u8, io_size, 0);
errdefer ctx.allocator.free(name);
if (c.AudioDeviceGetProperty(
id,
0,
0,
c.kAudioDevicePropertyDeviceName,
&io_size,
name.ptr,
) != c.noErr) {
return error.OpeningDevice;
}
const id_str = try std.fmt.allocPrintZ(ctx.allocator, "{d}", .{id});
errdefer ctx.allocator.free(id_str);
const dev = main.Device{
.id = id_str,
.name = name,
.mode = mode,
.channels = channels,
.formats = &.{ .i16, .i32, .f32 },
.sample_rate = .{
.min = @as(u24, @intFromFloat(@floor(sample_rate))),
.max = @as(u24, @intFromFloat(@floor(sample_rate))),
},
};
try ctx.devices_info.list.append(ctx.allocator, dev);
if (id == default_output_id and mode == .playback) {
ctx.devices_info.default_output = ctx.devices_info.list.items.len - 1;
}
if (id == default_input_id and mode == .capture) {
ctx.devices_info.default_input = ctx.devices_info.list.items.len - 1;
}
}
}
}
pub fn devices(ctx: Context) []const main.Device {
return ctx.devices_info.list.items;
}
pub fn defaultDevice(ctx: Context, mode: main.Device.Mode) ?main.Device {
return ctx.devices_info.default(mode);
}
pub fn createPlayer(ctx: *Context, device: main.Device, writeFn: main.WriteFn, options: main.StreamOptions) !backends.Player {
const player = try ctx.allocator.create(Player);
errdefer ctx.allocator.destroy(player);
// obtain an AudioOutputUnit using an AUHAL component description
var component_desc = c.AudioComponentDescription{
.componentType = c.kAudioUnitType_Output,
.componentSubType = c.kAudioUnitSubType_HALOutput,
.componentManufacturer = c.kAudioUnitManufacturer_Apple,
.componentFlags = 0,
.componentFlagsMask = 0,
};
const component = c.AudioComponentFindNext(null, &component_desc);
if (component == null) return error.OpeningDevice;
// instantiate the audio unit
var audio_unit: c.AudioComponentInstance = undefined;
if (c.AudioComponentInstanceNew(component, &audio_unit) != c.noErr) return error.OpeningDevice;
// Initialize the AUHAL before making any changes or using it. Note that an AUHAL may need
// to be initialized twice, e.g. before and after making changes to it, as an AUHAL needs to
// be initialized *before* anything is done to it.
if (c.AudioUnitInitialize(audio_unit) != c.noErr) return error.OpeningDevice;
errdefer _ = c.AudioUnitUninitialize(audio_unit);
const device_id = std.fmt.parseInt(c.AudioDeviceID, device.id, 10) catch return error.OpeningDevice;
if (c.AudioUnitSetProperty(
audio_unit,
c.kAudioOutputUnitProperty_CurrentDevice,
c.kAudioUnitScope_Input,
0,
&device_id,
@sizeOf(c.AudioDeviceID),
) != c.noErr) {
return error.OpeningDevice;
}
const stream_desc = try createStreamDesc(options.format, options.sample_rate orelse default_sample_rate, device.channels.len);
if (c.AudioUnitSetProperty(
audio_unit,
c.kAudioUnitProperty_StreamFormat,
c.kAudioUnitScope_Input,
0,
&stream_desc,
@sizeOf(c.AudioStreamBasicDescription),
) != c.noErr) {
return error.OpeningDevice;
}
const render_callback = c.AURenderCallbackStruct{
.inputProc = Player.renderCallback,
.inputProcRefCon = player,
};
if (c.AudioUnitSetProperty(
audio_unit,
c.kAudioUnitProperty_SetRenderCallback,
c.kAudioUnitScope_Input,
0,
&render_callback,
@sizeOf(c.AURenderCallbackStruct),
) != c.noErr) {
return error.OpeningDevice;
}
player.* = .{
.allocator = ctx.allocator,
.audio_unit = audio_unit.?,
.is_paused = false,
.vol = 1.0,
.writeFn = writeFn,
.user_data = options.user_data,
.channels = device.channels,
.format = options.format,
.sample_rate = options.sample_rate orelse default_sample_rate,
};
return .{ .coreaudio = player };
}
pub fn createRecorder(ctx: *Context, device: main.Device, readFn: main.ReadFn, options: main.StreamOptions) !backends.Recorder {
// Request permission to record via requestRecordPermission. If permission was previously
// granted, it will immediately return. Otherwise the function will block, the OS will display
// an "<App> wants to access the Microphone. [Allow] [Deny]" menu, then the callback will be
// invoked and requestRecordPermission will return.
const audio_session = avaudio.AVAudioSession.sharedInstance();
const PermissionContext = void;
const perm_ctx: PermissionContext = {};
audio_session.requestRecordPermission(perm_ctx, (struct {
pub fn callback(perm_ctx_2: PermissionContext, permission_granted: bool) void {
_ = permission_granted;
_ = perm_ctx_2;
// Note: in the event permission was NOT granted by the user, we could capture that here
// and surface it as an error.
//
// However, in this situation the OS will simply replace all audio samples with zero-value
// (silence) ones - so there's no harm for us in doing nothing in that case: the user would
// find out we're recording only silence, and they would need to correct it in their System
// Preferences by granting the app permission to record.
}
}).callback);
const recorder = try ctx.allocator.create(Recorder);
errdefer ctx.allocator.destroy(recorder);
const device_id = std.fmt.parseInt(c.AudioDeviceID, device.id, 10) catch return error.OpeningDevice;
var io_size: u32 = 0;
var prop_address = c.AudioObjectPropertyAddress{
.mSelector = c.kAudioDevicePropertyStreamConfiguration,
.mScope = c.kAudioObjectPropertyScopeInput,
.mElement = c.kAudioObjectPropertyElementMain,
};
if (c.AudioObjectGetPropertyDataSize(
device_id,
&prop_address,
0,
null,
&io_size,
) != c.noErr) {
return error.OpeningDevice;
}
std.debug.assert(io_size == @sizeOf(c.AudioBufferList));
const buf_list = try ctx.allocator.create(c.AudioBufferList);
errdefer ctx.allocator.destroy(buf_list);
if (c.AudioObjectGetPropertyData(
device_id,
&prop_address,
0,
null,
&io_size,
@as(*anyopaque, @ptrCast(buf_list)),
) != c.noErr) {
return error.OpeningDevice;
}
// obtain an AudioOutputUnit using an AUHAL component description
var component_desc = c.AudioComponentDescription{
.componentType = c.kAudioUnitType_Output,
.componentSubType = c.kAudioUnitSubType_HALOutput,
.componentManufacturer = c.kAudioUnitManufacturer_Apple,
.componentFlags = 0,
.componentFlagsMask = 0,
};
const component = c.AudioComponentFindNext(null, &component_desc);
if (component == null) return error.OpeningDevice;
// instantiate the audio unit
var audio_unit: c.AudioComponentInstance = undefined;
if (c.AudioComponentInstanceNew(component, &audio_unit) != c.noErr) return error.OpeningDevice;
// Initialize the AUHAL before making any changes or using it. Note that an AUHAL may need
// to be initialized twice, e.g. before and after making changes to it, as an AUHAL needs to
// be initialized *before* anything is done to it.
if (c.AudioUnitInitialize(audio_unit) != c.noErr) return error.OpeningDevice;
errdefer _ = c.AudioUnitUninitialize(audio_unit);
// To obtain the device input, we must enable IO on the input scope of the Audio Unit. Input
// must be explicitly enabled with kAudioOutputUnitProperty_EnableIO on Element 1 of the
// AUHAL, and this *must be done before* setting the AUHAL's current device. We must also
// disable IO on the output scope of the AUHAL, since it can be used for both.
// Enable AUHAL input
const enable_io: u32 = 1;
const au_element_input: u32 = 1;
if (c.AudioUnitSetProperty(
audio_unit,
c.kAudioOutputUnitProperty_EnableIO,
c.kAudioUnitScope_Input,
au_element_input,
&enable_io,
@sizeOf(u32),
) != c.noErr) {
return error.OpeningDevice;
}
// Disable AUHAL output
const disable_io: u32 = 0;
const au_element_output: u32 = 0;
if (c.AudioUnitSetProperty(
audio_unit,
c.kAudioOutputUnitProperty_EnableIO,
c.kAudioUnitScope_Output,
au_element_output,
&disable_io,
@sizeOf(u32),
) != c.noErr) {
return error.OpeningDevice;
}
// Set the audio device to be Audio Unit's current device. A device can only be associated
// with an AUHAL after enabling IO.
if (c.AudioUnitSetProperty(
audio_unit,
c.kAudioOutputUnitProperty_CurrentDevice,
c.kAudioUnitScope_Global,
au_element_output,
&device_id,
@sizeOf(c.AudioDeviceID),
) != c.noErr) {
return error.OpeningDevice;
}
// Register the capture callback for the AUHAL; this will be called when the AUHAL has
// received new data from the input device.
const capture_callback = c.AURenderCallbackStruct{
.inputProc = Recorder.captureCallback,
.inputProcRefCon = recorder,
};
if (c.AudioUnitSetProperty(
audio_unit,
c.kAudioOutputUnitProperty_SetInputCallback,
c.kAudioUnitScope_Global,
au_element_output,
&capture_callback,
@sizeOf(c.AURenderCallbackStruct),
) != c.noErr) {
return error.OpeningDevice;
}
// Set the desired output format.
const sample_rate = blk: {
if (options.sample_rate) |rate| {
if (rate < device.sample_rate.min or rate > device.sample_rate.max) return error.OpeningDevice;
break :blk rate;
}
break :blk device.sample_rate.max;
};
const stream_desc = try createStreamDesc(options.format, sample_rate, device.channels.len);
if (c.AudioUnitSetProperty(
audio_unit,
c.kAudioUnitProperty_StreamFormat,
c.kAudioUnitScope_Output,
au_element_input,
&stream_desc,
@sizeOf(c.AudioStreamBasicDescription),
) != c.noErr) {
return error.OpeningDevice;
}
// Now that we are done with modifying the AUHAL, initialize it once more to ensure that it
// is ready to use.
if (c.AudioUnitInitialize(audio_unit) != c.noErr) return error.OpeningDevice;
errdefer _ = c.AudioUnitUninitialize(audio_unit);
recorder.* = .{
.allocator = ctx.allocator,
.audio_unit = audio_unit.?,
.is_paused = false,
.vol = 1.0,
.buf_list = buf_list,
.readFn = readFn,
.user_data = options.user_data,
.channels = device.channels,
.format = options.format,
.sample_rate = sample_rate,
};
return .{ .coreaudio = recorder };
}
};
pub const Player = struct {
allocator: std.mem.Allocator,
audio_unit: c.AudioUnit,
is_paused: bool,
vol: f32,
writeFn: main.WriteFn,
user_data: ?*anyopaque,
channels: []main.ChannelPosition,
format: main.Format,
sample_rate: u24,
pub fn renderCallback(
player_opaque: ?*anyopaque,
action_flags: [*c]c.AudioUnitRenderActionFlags,
time_stamp: [*c]const c.AudioTimeStamp,
bus_number: u32,
frames_left: u32,
buf: [*c]c.AudioBufferList,
) callconv(.C) c.OSStatus {
_ = action_flags;
_ = time_stamp;
_ = bus_number;
_ = frames_left;
const player = @as(*Player, @ptrCast(@alignCast(player_opaque.?)));
const frames = buf.*.mBuffers[0].mDataByteSize;
player.writeFn(player.user_data, @as([*]u8, @ptrCast(buf.*.mBuffers[0].mData.?))[0..frames]);
return c.noErr;
}
pub fn deinit(player: *Player) void {
_ = c.AudioOutputUnitStop(player.audio_unit);
_ = c.AudioUnitUninitialize(player.audio_unit);
_ = c.AudioComponentInstanceDispose(player.audio_unit);
player.allocator.destroy(player);
}
pub fn start(player: *Player) !void {
try player.play();
}
pub fn play(player: *Player) !void {
if (c.AudioOutputUnitStart(player.audio_unit) != c.noErr) {
return error.CannotPlay;
}
player.is_paused = false;
}
pub fn pause(player: *Player) !void {
if (c.AudioOutputUnitStop(player.audio_unit) != c.noErr) {
return error.CannotPause;
}
player.is_paused = true;
}
pub fn paused(player: *Player) bool {
return player.is_paused;
}
pub fn setVolume(player: *Player, vol: f32) !void {
if (c.AudioUnitSetParameter(
player.audio_unit,
c.kHALOutputParam_Volume,
c.kAudioUnitScope_Global,
0,
vol,
0,
) != c.noErr) {
if (is_darling) return;
return error.CannotSetVolume;
}
}
pub fn volume(player: *Player) !f32 {
var vol: f32 = 0;
if (c.AudioUnitGetParameter(
player.audio_unit,
c.kHALOutputParam_Volume,
c.kAudioUnitScope_Global,
0,
&vol,
) != c.noErr) {
if (is_darling) return 1;
return error.CannotGetVolume;
}
return vol;
}
};
pub const Recorder = struct {
allocator: std.mem.Allocator,
audio_unit: c.AudioUnit,
is_paused: bool,
vol: f32,
buf_list: *c.AudioBufferList,
m_data: ?[]u8 = null,
readFn: main.ReadFn,
user_data: ?*anyopaque,
channels: []main.ChannelPosition,
format: main.Format,
sample_rate: u24,
pub fn captureCallback(
recorder_opaque: ?*anyopaque,
action_flags: [*c]c.AudioUnitRenderActionFlags,
time_stamp: [*c]const c.AudioTimeStamp,
bus_number: u32,
num_frames: u32,
buffer_list: [*c]c.AudioBufferList,
) callconv(.C) c.OSStatus {
_ = buffer_list;
const recorder = @as(*Recorder, @ptrCast(@alignCast(recorder_opaque.?)));
// We want interleaved multi-channel audio, when multiple channels are available-so we'll
// only use a single buffer. If we wanted non-interleaved audio we would use multiple
// buffers.
var m_buffer = &recorder.buf_list.*.mBuffers[0];
// Ensure our buffer matches the size needed for the render operation. Note that the buffer
// may grow (in the case of multi-channel audio during the first render callback) or shrink
// in e.g. the event of the device being unplugged and the default input device switching.
const new_len = num_frames * recorder.format.frameSize(@intCast(recorder.channels.len));
if (recorder.m_data == null or recorder.m_data.?.len != new_len) {
if (recorder.m_data) |old| recorder.allocator.free(old);
recorder.m_data = recorder.allocator.alloc(u8, new_len) catch return c.noErr;
}
recorder.buf_list.*.mNumberBuffers = 1;
m_buffer.mData = recorder.m_data.?.ptr;
m_buffer.mDataByteSize = @intCast(recorder.m_data.?.len);
m_buffer.mNumberChannels = @intCast(recorder.channels.len);
const err_no = c.AudioUnitRender(
recorder.audio_unit,
action_flags,
time_stamp,
bus_number,
num_frames,
recorder.buf_list,
);
if (err_no != c.noErr) {
// TODO: err_no here is rather helpful, we should indicate what it is back to the user
// in this event probably?
return c.noErr;
}
if (recorder.buf_list.*.mNumberBuffers == 1) {
recorder.readFn(recorder.user_data, @as([*]u8, @ptrCast(recorder.buf_list.*.mBuffers[0].mData.?))[0..new_len]);
} else {
@panic("TODO: convert planar to interleaved");
// for (recorder.channels, 0..) |*ch, i| {
// ch.ptr = @as([*]u8, @ptrCast(recorder.buf_list.*.mBuffers[i].mData.?));
// }
}
return c.noErr;
}
pub fn deinit(recorder: *Recorder) void {
_ = c.AudioOutputUnitStop(recorder.audio_unit);
_ = c.AudioUnitUninitialize(recorder.audio_unit);
_ = c.AudioComponentInstanceDispose(recorder.audio_unit);
recorder.allocator.destroy(recorder.buf_list);
recorder.allocator.destroy(recorder);
}
pub fn start(recorder: *Recorder) !void {
try recorder.record();
}
pub fn record(recorder: *Recorder) !void {
if (c.AudioOutputUnitStart(recorder.audio_unit) != c.noErr) {
return error.CannotRecord;
}
recorder.is_paused = false;
}
pub fn pause(recorder: *Recorder) !void {
if (c.AudioOutputUnitStop(recorder.audio_unit) != c.noErr) {
return error.CannotPause;
}
recorder.is_paused = true;
}
pub fn paused(recorder: *Recorder) bool {
return recorder.is_paused;
}
pub fn setVolume(recorder: *Recorder, vol: f32) !void {
if (c.AudioUnitSetParameter(
recorder.audio_unit,
c.kHALOutputParam_Volume,
c.kAudioUnitScope_Global,
0,
vol,
0,
) != c.noErr) {
if (is_darling) return;
return error.CannotSetVolume;
}
}
pub fn volume(recorder: *Recorder) !f32 {
var vol: f32 = 0;
if (c.AudioUnitGetParameter(
recorder.audio_unit,
c.kHALOutputParam_Volume,
c.kAudioUnitScope_Global,
0,
&vol,
) != c.noErr) {
if (is_darling) return 1;
return error.CannotGetVolume;
}
return vol;
}
};
fn freeDevice(allocator: std.mem.Allocator, device: main.Device) void {
allocator.free(device.id);
allocator.free(device.name);
allocator.free(device.channels);
}
fn createStreamDesc(format: main.Format, sample_rate: u24, ch_count: usize) !c.AudioStreamBasicDescription {
var desc = c.AudioStreamBasicDescription{
.mSampleRate = @as(f64, @floatFromInt(sample_rate)),
.mFormatID = c.kAudioFormatLinearPCM,
.mFormatFlags = switch (format) {
.i16 => c.kAudioFormatFlagIsSignedInteger,
.i24 => c.kAudioFormatFlagIsSignedInteger,
.i32 => c.kAudioFormatFlagIsSignedInteger,
.f32 => c.kAudioFormatFlagIsFloat,
.u8 => return error.IncompatibleDevice,
},
.mBytesPerPacket = format.frameSize(@intCast(ch_count)),
.mFramesPerPacket = 1,
.mBytesPerFrame = format.frameSize(@intCast(ch_count)),
.mChannelsPerFrame = @intCast(ch_count),
.mBitsPerChannel = switch (format) {
.i16 => 16,
.i24 => 24,
.i32 => 32,
.f32 => 32,
.u8 => return error.IncompatibleDevice,
},
.mReserved = 0,
};
if (native_endian == .big) {
desc.mFormatFlags |= c.kAudioFormatFlagIsBigEndian;
}
return desc;
}