manifest.py
python
sha256:18b983389ee1b55900fcd799bfbb496552d2e3ecded9d18cefbfef188947a12e
chore: remove blob-debug test marker file
Sonnet 4.6
1 day ago
| 1 | """Hierarchical chunk manifests for the Muse MIDI plugin. |
| 2 | |
| 3 | Evolves the flat ``{"files": {"track.mid": "<sha256>"}}`` snapshot beyond |
| 4 | a single content hash per file to a rich, per-bar, per-track manifest that |
| 5 | enables: |
| 6 | |
| 7 | - **Partial diff** — compare only the bars that changed. |
| 8 | - **Query acceleration** — answer note queries without reading full MIDI blobs. |
| 9 | - **Targeted merge** — attempt to merge only the bars with conflicts. |
| 10 | - **Historical analytics** — aggregate statistics over commit history without |
| 11 | re-parsing MIDI bytes on every query. |
| 12 | |
| 13 | Backward compatibility |
| 14 | ---------------------- |
| 15 | The ``MusicManifest.files`` field is identical to the standard |
| 16 | ``SnapshotManifest.files`` so that the core Muse engine and all existing |
| 17 | commands continue to work unchanged. The ``tracks`` field is additive |
| 18 | metadata stored as a sidecar under ``.muse/music_manifests/`` — never |
| 19 | replacing the canonical flat manifest. |
| 20 | |
| 21 | Storage layout:: |
| 22 | |
| 23 | .muse/music_manifests/ |
| 24 | <snapshot_id>.json — full MusicManifest for this snapshot |
| 25 | (rebuildable from history; add to .museignore [domain.midi] in CI) |
| 26 | |
| 27 | Public API |
| 28 | ---------- |
| 29 | - :class:`BarChunk` — per-bar chunk descriptor. |
| 30 | - :class:`TrackManifest` — rich metadata for one MIDI track. |
| 31 | - :class:`MusicManifest` — top-level sidecar manifest. |
| 32 | - :func:`build_bar_chunk` — build a :class:`BarChunk` from a bar's notes. |
| 33 | - :func:`build_track_manifest` — build a :class:`TrackManifest`. |
| 34 | - :func:`build_music_manifest` — build the full :class:`MusicManifest`. |
| 35 | - :func:`write_music_manifest` — persist to ``.muse/music_manifests/``. |
| 36 | - :func:`read_music_manifest` — load from the sidecar store. |
| 37 | """ |
| 38 | |
| 39 | import json |
| 40 | import logging |
| 41 | import pathlib |
| 42 | from typing import Literal, TypedDict |
| 43 | |
| 44 | from muse._version import __version__ |
| 45 | from muse.core.paths import music_manifests_dir |
| 46 | from muse.core.types import Manifest, blob_id, load_json_file |
| 47 | from muse.plugins.midi._query import ( |
| 48 | NoteInfo, |
| 49 | detect_chord, |
| 50 | key_signature_guess, |
| 51 | notes_by_bar, |
| 52 | ) |
| 53 | from muse.plugins.midi.midi_diff import extract_notes |
| 54 | |
| 55 | logger = logging.getLogger(__name__) |
| 56 | |
| 57 | type _BarMap = dict[str, "BarChunk"] |
| 58 | type _TrackMap = dict[str, "TrackManifest"] |
| 59 | type _ChangeMap = dict[str, list[int]] |
| 60 | |
| 61 | # --------------------------------------------------------------------------- |
| 62 | # Types |
| 63 | # --------------------------------------------------------------------------- |
| 64 | |
| 65 | class BarChunk(TypedDict): |
| 66 | """Descriptor for one bar's worth of note events in a MIDI track. |
| 67 | |
| 68 | ``bar`` 1-indexed bar number (assumes 4/4 time). |
| 69 | ``chunk_hash`` ``sha256:``-prefixed ID of the canonical JSON of all notes in this bar. |
| 70 | Used for per-bar change detection without re-parsing MIDI. |
| 71 | ``note_count`` Number of notes in this bar. |
| 72 | ``chord`` Best-guess chord name for this bar (e.g. ``"Cmaj"``). |
| 73 | ``pitch_range`` ``[min_pitch, max_pitch]`` MIDI pitch values in this bar. |
| 74 | """ |
| 75 | |
| 76 | bar: int |
| 77 | chunk_hash: str |
| 78 | note_count: int |
| 79 | chord: str |
| 80 | pitch_range: list[int] |
| 81 | |
| 82 | class TrackManifest(TypedDict): |
| 83 | """Rich metadata descriptor for one MIDI track at a specific snapshot. |
| 84 | |
| 85 | ``track_id`` Stable identifier for this track (SHA-256 of file path). |
| 86 | Stable across renames if you track by content; changes |
| 87 | on rename. Use entity IDs in the entity index for |
| 88 | true cross-rename continuity. |
| 89 | ``file_path`` Workspace-relative MIDI file path. |
| 90 | ``content_hash`` SHA-256 of the full MIDI file bytes (same as the flat |
| 91 | manifest entry — the canonical content address). |
| 92 | ``bars`` Mapping from ``str(bar_number)`` → :class:`BarChunk`. |
| 93 | JSON keys are always strings; callers convert to int. |
| 94 | ``ticks_per_beat`` MIDI ticks per quarter note for this file. |
| 95 | ``note_count`` Total note count across all bars. |
| 96 | ``key_guess`` Krumhansl-Schmuckler key estimate (e.g. ``"G major"``). |
| 97 | ``bar_count`` Number of bars with at least one note. |
| 98 | """ |
| 99 | |
| 100 | track_id: str |
| 101 | file_path: str |
| 102 | content_hash: str |
| 103 | bars: _BarMap |
| 104 | ticks_per_beat: int |
| 105 | note_count: int |
| 106 | key_guess: str |
| 107 | bar_count: int |
| 108 | |
| 109 | class MusicManifest(TypedDict): |
| 110 | """Top-level hierarchical manifest for a music snapshot. |
| 111 | |
| 112 | This is the sidecar companion to the standard :class:`~muse.domain.SnapshotManifest`. |
| 113 | The ``files`` field is identical to the flat manifest — the core engine |
| 114 | reads only ``files`` for content addressing. The ``tracks`` field is |
| 115 | additive richness for music-domain queries, diff, and merge. |
| 116 | |
| 117 | ``schema_version`` The Muse package version (read from ``muse._version``). |
| 118 | ``snapshot_id`` The snapshot this manifest belongs to. |
| 119 | ``files`` Standard flat ``{path: sha256}`` manifest (compat layer). |
| 120 | ``tracks`` ``{path: TrackManifest}`` for each MIDI file. |
| 121 | """ |
| 122 | |
| 123 | domain: Literal["midi"] |
| 124 | schema_version: str |
| 125 | snapshot_id: str |
| 126 | files: Manifest |
| 127 | tracks: _TrackMap |
| 128 | |
| 129 | # --------------------------------------------------------------------------- |
| 130 | # Builders |
| 131 | # --------------------------------------------------------------------------- |
| 132 | |
| 133 | def _bar_chunk_hash(notes: list[NoteInfo]) -> str: |
| 134 | """Return a ``sha256:``-prefixed ID of the canonical JSON of a bar's notes.""" |
| 135 | payload = json.dumps( |
| 136 | [ |
| 137 | { |
| 138 | "pitch": n.pitch, |
| 139 | "velocity": n.velocity, |
| 140 | "start_tick": n.start_tick, |
| 141 | "duration_ticks": n.duration_ticks, |
| 142 | "channel": n.channel, |
| 143 | } |
| 144 | for n in sorted(notes, key=lambda n: (n.start_tick, n.pitch)) |
| 145 | ], |
| 146 | sort_keys=True, |
| 147 | separators=(",", ":"), |
| 148 | ).encode() |
| 149 | return blob_id(payload) |
| 150 | |
| 151 | def build_bar_chunk(bar_num: int, notes: list[NoteInfo]) -> BarChunk: |
| 152 | """Build a :class:`BarChunk` descriptor for *bar_num*. |
| 153 | |
| 154 | Args: |
| 155 | bar_num: 1-indexed bar number. |
| 156 | notes: All notes in this bar. |
| 157 | |
| 158 | Returns: |
| 159 | A populated :class:`BarChunk`. |
| 160 | """ |
| 161 | pcs = frozenset(n.pitch_class for n in notes) |
| 162 | chord = detect_chord(pcs) |
| 163 | pitches = [n.pitch for n in notes] |
| 164 | pitch_range: list[int] = [min(pitches), max(pitches)] if pitches else [0, 0] |
| 165 | return BarChunk( |
| 166 | bar=bar_num, |
| 167 | chunk_hash=_bar_chunk_hash(notes), |
| 168 | note_count=len(notes), |
| 169 | chord=chord, |
| 170 | pitch_range=pitch_range, |
| 171 | ) |
| 172 | |
| 173 | def build_track_manifest( |
| 174 | notes: list[NoteInfo], |
| 175 | file_path: str, |
| 176 | content_hash: str, |
| 177 | ticks_per_beat: int, |
| 178 | ) -> TrackManifest: |
| 179 | """Build a :class:`TrackManifest` from a parsed note list. |
| 180 | |
| 181 | Args: |
| 182 | notes: All notes extracted from the MIDI file. |
| 183 | file_path: Workspace-relative MIDI file path. |
| 184 | content_hash: SHA-256 of the MIDI file bytes (from the flat manifest). |
| 185 | ticks_per_beat: MIDI timing resolution. |
| 186 | |
| 187 | Returns: |
| 188 | A populated :class:`TrackManifest`. |
| 189 | """ |
| 190 | track_id = blob_id(file_path.encode()) |
| 191 | bars_map = notes_by_bar(notes) |
| 192 | bars: _BarMap = {} |
| 193 | for bar_num, bar_notes in sorted(bars_map.items()): |
| 194 | bars[str(bar_num)] = build_bar_chunk(bar_num, bar_notes) |
| 195 | |
| 196 | key_guess = key_signature_guess(notes) |
| 197 | |
| 198 | return TrackManifest( |
| 199 | track_id=track_id, |
| 200 | file_path=file_path, |
| 201 | content_hash=content_hash, |
| 202 | bars=bars, |
| 203 | ticks_per_beat=ticks_per_beat, |
| 204 | note_count=len(notes), |
| 205 | key_guess=key_guess, |
| 206 | bar_count=len(bars), |
| 207 | ) |
| 208 | |
| 209 | # --------------------------------------------------------------------------- |
| 210 | # Persistence |
| 211 | # --------------------------------------------------------------------------- |
| 212 | |
| 213 | def _manifest_path(repo_root: pathlib.Path, snapshot_id: str) -> pathlib.Path: |
| 214 | return music_manifests_dir(repo_root) / f"{snapshot_id}.json" |
| 215 | |
| 216 | def write_music_manifest( |
| 217 | repo_root: pathlib.Path, |
| 218 | manifest: MusicManifest, |
| 219 | ) -> pathlib.Path: |
| 220 | """Persist *manifest* to ``.muse/music_manifests/<snapshot_id>.json``. |
| 221 | |
| 222 | Args: |
| 223 | repo_root: Repository root. |
| 224 | manifest: The manifest to write. |
| 225 | |
| 226 | Returns: |
| 227 | Path to the written file. |
| 228 | """ |
| 229 | snapshot_id = manifest.get("snapshot_id", "") |
| 230 | if not snapshot_id: |
| 231 | raise ValueError("MusicManifest.snapshot_id must be non-empty") |
| 232 | path = _manifest_path(repo_root, snapshot_id) |
| 233 | path.parent.mkdir(parents=True, exist_ok=True) |
| 234 | path.write_text(f"{json.dumps(manifest, indent=2)}\n") |
| 235 | logger.debug( |
| 236 | "✅ Music manifest written: %s (%d tracks)", |
| 237 | snapshot_id[:8], |
| 238 | len(manifest["tracks"]), |
| 239 | ) |
| 240 | return path |
| 241 | |
| 242 | def read_music_manifest( |
| 243 | repo_root: pathlib.Path, |
| 244 | snapshot_id: str, |
| 245 | ) -> MusicManifest | None: |
| 246 | """Load the music manifest for *snapshot_id*, or ``None`` if absent. |
| 247 | |
| 248 | Args: |
| 249 | repo_root: Repository root. |
| 250 | snapshot_id: Snapshot ID. |
| 251 | |
| 252 | Returns: |
| 253 | The :class:`MusicManifest`, or ``None`` when the sidecar file does |
| 254 | not exist. |
| 255 | """ |
| 256 | path = _manifest_path(repo_root, snapshot_id) |
| 257 | if not path.exists(): |
| 258 | return None |
| 259 | raw: MusicManifest | None = load_json_file(path) |
| 260 | if raw is None: |
| 261 | logger.warning("⚠️ Corrupt music manifest %s: unreadable or invalid JSON", path) |
| 262 | return None |
| 263 | return raw |
| 264 | |
| 265 | # --------------------------------------------------------------------------- |
| 266 | # Partial diff helper |
| 267 | # --------------------------------------------------------------------------- |
| 268 | |
| 269 | def diff_manifests_by_bar( |
| 270 | base: MusicManifest, |
| 271 | target: MusicManifest, |
| 272 | ) -> _ChangeMap: |
| 273 | """Return a per-track list of bars that changed between two manifests. |
| 274 | |
| 275 | Uses the per-bar ``chunk_hash`` values to detect changes without |
| 276 | loading any MIDI bytes. |
| 277 | |
| 278 | Args: |
| 279 | base: Ancestor manifest. |
| 280 | target: Newer manifest. |
| 281 | |
| 282 | Returns: |
| 283 | ``{track_path: [changed_bar_numbers]}`` for all tracks where at |
| 284 | least one bar differs. Tracks added or removed appear with ``[-1]`` |
| 285 | as a sentinel indicating the whole track changed. |
| 286 | """ |
| 287 | changed: _ChangeMap = {} |
| 288 | |
| 289 | all_tracks = set(base["tracks"]) | set(target["tracks"]) |
| 290 | |
| 291 | for track in sorted(all_tracks): |
| 292 | base_track = base["tracks"].get(track) |
| 293 | target_track = target["tracks"].get(track) |
| 294 | |
| 295 | if base_track is None or target_track is None: |
| 296 | changed[track] = [-1] |
| 297 | continue |
| 298 | |
| 299 | if base_track["content_hash"] == target_track["content_hash"]: |
| 300 | continue |
| 301 | |
| 302 | # Content changed — find which bars. |
| 303 | base_bars = base_track["bars"] |
| 304 | target_bars = target_track["bars"] |
| 305 | all_bar_keys = set(base_bars) | set(target_bars) |
| 306 | |
| 307 | changed_bars: list[int] = [] |
| 308 | for bar_key in sorted(all_bar_keys, key=lambda k: int(k)): |
| 309 | base_chunk = base_bars.get(bar_key) |
| 310 | target_chunk = target_bars.get(bar_key) |
| 311 | if base_chunk is None or target_chunk is None: |
| 312 | changed_bars.append(int(bar_key)) |
| 313 | elif base_chunk["chunk_hash"] != target_chunk["chunk_hash"]: |
| 314 | changed_bars.append(int(bar_key)) |
| 315 | |
| 316 | if changed_bars: |
| 317 | changed[track] = sorted(changed_bars) |
| 318 | |
| 319 | return changed |
File History
7 commits
sha256:18b983389ee1b55900fcd799bfbb496552d2e3ecded9d18cefbfef188947a12e
chore: remove blob-debug test marker file
Sonnet 4.6
1 day ago
sha256:e452ad9a6ace6ccc6d875a35e06caf9da5576a970c1c36133b69a891ce5fefa8
chore: prebuild timing test
Sonnet 4.6
9 days ago
sha256:0008ab6695e3e064b3e236b24fd19e538fef6a588eb0d211622f4466d919c0b1
merge: pull staging/dev — advance to 0.2.0rc12
Sonnet 4.6
patch
11 days ago
sha256:9c33d61749fff814c5226d5386aa2af7064c2c02788594a25fdd709358132eea
fix: _PROPOSAL_PREFIX_RESOLVE_LIMIT 200 → 100 to match hub …
Sonnet 4.6
22 days ago
sha256:36c3cb3e76619d4c30a6d9bf81b5ec4ff148e30dcfed913e3114ca7b43b81c7e
fix: rename objects→blobs in push client and all stale test…
Sonnet 4.6
patch
25 days ago
sha256:c06a9b9b9fee26c68ea725b44d54b2c0a171301ce9de746d5b656617b4463a9a
fix: repair four test failures from post-migration audit
Sonnet 4.6
patch
31 days ago
sha256:1900655993c83c4107067375548a7be823e471d2515830842f1a12cba4bd3cdf
fix: unified object store migration — idempotent writes, JS…
Sonnet 4.6
minor
⚠
31 days ago