Nota — a free DAW for macOS, Windows and Linux. No subscription, no account — just open it and make music.
Data Model: Multi-Channel Input and Simultaneous Multi-Track Recording
Spec: spec.md | Research: research.md
Engine (C++, src/native/nota.engine/src/)
InputDevice channels (control thread)
| Field | Type | Notes |
|---|---|---|
channelCount |
int32 | 0 when no device is open; clipped to kMaxInputChannels = 64 |
channelName(i) |
string | device name or "" (UI falls back to "Input N") |
Read from AudioInput after start; refreshed on device change. Never read on the audio
or input thread.
Track input pick (Track, atomics as today)
| Field | Type | Rule |
|---|---|---|
inputFirst_ |
int32, 0-based | 0 ≤ first < kMaxInputChannels |
inputWidth_ |
int32 | 1 (mono) or 2 (stereo pair); width 2 ⇒ first is even |
recordInputSource_ |
int32 | unchanged: 0 = hardware, > 0 = another track |
midiInputSource_ |
int32 | -2 All inputs (default), ≥ 0 a hardware source index |
midiInputChannel_ |
int32 | 0 = all, 1–16 |
Available ⇔ first + width ≤ channelCount. An unavailable pick is kept, the track records
nothing (FR-011). Replaces inputChannels_ (kInputStereo/Mono1/Mono2); the old setter maps
0 → (0,2), 1 → (0,1), 2 → (1,1).
CaptureLayout (seqlock, engine/CaptureRing.h)
| Field | Type | Notes |
|---|---|---|
count |
int32 | number of device channels carried in the rings, ≤ 64 |
channel[64] |
int8 | device channel per ring column |
columnOf[64] |
int8 | device channel → column, −1 = not carried |
generation |
uint32 | bumps on every publish; every ring block carries it |
One layout for both rings; readers copy it (seqlock) and keep their copy on a torn read. The message thread keeps the last 16 in a history for the drain.
AudioTakeLane (message thread)
| Field | Type | Notes |
|---|---|---|
trackId |
int32 | |
column / width |
int32 | ring column of the pick's first channel; 1 or 2 |
internalSource |
int32 | 0 = hardware, else source track (internal ring column) |
buf |
vector | width channels interleaved; reserved in 30 s chunks |
writePos |
size_t | frame cursor |
passes |
vector | loop-punch passes (specs/067), per lane |
scene |
int32 | Session slot, −1 for Arrangement |
keepFrom |
int64 | first frame to keep (Session launch boundary / retry) |
stopAt |
int64 | −1 running; else the frame its slot stopped at |
AudioTake (message thread; replaces the single-take fields)
| Field | Notes |
|---|---|
lanes |
vector, ≤ 64 hardware + 16 internal |
startBeat, sampleRate, latencyFrames |
shared by every lane (FR-006) |
loop geometry, wrapped, origin |
shared |
retry markers / epochs |
shared (specs/078) |
State: Idle → CountIn → Rolling → (lane stopped)* → Stopped → materialised (one undo
group). Discard: Rolling → Aborted drops every lane.
MidiTakeLane (audio thread, preallocated kMaxMidiTakeLanes = 16)
trackId, pendingActive[128], recRoutedActive[128], clipIndex (message-thread side),
scene. recordTrackRt_ becomes the lane table's published count + ids.
MidiEvent (engine/EngineTypes.h)
Adds int8_t source (−1 = non-hardware: computer keyboard, gamepad, Remote, UI) and
int8_t channel (0–15). Filter match:
(track.src == All || track.src == e.source) && (track.ch == 0 || track.ch == e.channel + 1);
e.source == −1 matches only All.
Managed (C#)
TrackModel (Nota.Infrastructure/Persistence/ProjectModel.cs)
| Property | Type | Migration |
|---|---|---|
InputFirst |
int? (1-based) | null ⇒ from InputChannels: 0/null → 1, 1 → 1, 2 → 2 |
InputWidth |
int? | null ⇒ from InputChannels: 0/null → 2, else 1 |
InputChannels |
int? | still written when the pick is (1,2), (1,1) or (2,1), so 0.59.x opens it |
MidiInputUid |
string? | null = All inputs |
MidiInputChannel |
int? | null/0 = All |
InputPick (Nota.Application, replaces InputChannels enum in the API)
record struct InputPick(int First, int Width) (0-based first). IAudioEngine:
SetTrackInput(int trackId, InputPick), GetTrackInput(int trackId),
AudioInputChannelCount, AudioInputChannelName(int i),
SetTrackMidiInput(int trackId, string? uid, int channel), GetTrackMidiInput(...).
SetTrackInputChannels stays as an adapter until callers move.
InputSetup (preferences, per (audioSystem, deviceUid))
| Field | Type | Default |
|---|---|---|
Enabled[i] |
bool per channel | true |
Name[i] |
string? per channel | null (device name) |
PairName[k] |
string? per odd-aligned pair | null ("In 2k+1/2k+2" or both device names) |
Display name of a pick: user name → device name → "Input N" (mono); pair name → "In a/b". Menu lists: enabled mono channels; pairs whose both channels are enabled.