Files
yellowjacket/backend/explore/mix.go
T
yonluandClaude Opus 5 e7748f1fd5
CI / check (push) Successful in 3m7s
CI / e2e (push) Canceled after 1m45s
feat(database): shape the library like files, and shrink the catalog
Plans 013 and 014, the album page that prompted them, and the smaller
fixes they turned up. Changelog, largest first.

## The local library is shaped like files, not like MusicBrainz

`audio_files` carries its own tags and points at `albums` and
`artists`; `file_genres` is the one real many-to-many. `recordings`,
`release_group_recordings`, `artist_credit`, `artist_credit_artist`,
`recording_genres`, `release_groups` and `release_to_rg` are gone from
the local side, and with them a six-way join in every read, a
`MIN(release_group_id)` subquery in eleven queries and a
first-credited-artist subquery in nine. Measured on a real 25,966-file
library, every many-to-many that model expressed was 1:1 in the data.

- Ownership is a file. `GetFilePathsByRecordingMBIDs`,
  `LibraryMBIDIndex.CheckMBIDs`, `collectLibraryEntities` and
  `pruneStaleLocalCrossReferences` all join `audio_files`, so the 812
  orphaned recordings, 216 release groups and 260 artists that library
  carried are now structurally impossible.
- One projection: every track query selects from the `track_metadata`
  view, one row type, one mapper. Nine hand-rolled copies had drifted
  far enough to report different years on different screens.
- `library_id = 0` means every library, so each list query exists once
  instead of scoped and unscoped with a branch at every call site.
- No migration chain. `sql/schemas/` is the one description of the
  shape; `sql/migrations/`, `applyMigrations` and `schema_migrations`
  are squashed away, along with the drift between them that had sqlc
  generating against a stale schema.
- `database.InsertTestTrack` is the one test seeder; twenty test files
  had been assembling the old FK chain each in its own order.

## The catalog stores its ids as bytes

`explore_index`'s three 36-char MBID columns and its entity-type text
are 16 raw bytes and a small integer. The table and its six indexes go
780 MB to 405 MB on a real 2,052,200-row catalog, which is why a fresh
install is ~0.6 GB rather than ~1.0 GB.

- `backend/explore/mbid.go` is the only place the encoding is known;
  everything above it speaks dashed strings.
- `CHECK(length(mbid) = 16)` makes a stringly write fail at the insert
  rather than silently returning no rows, since SQLite does not coerce
  between TEXT and BLOB.
- The importer asks the artifact what encoding it carries and converts
  on the way in, so the artifact already published keeps working and no
  format bump is needed.
- `indexRowColumns`/`scanIndexRow` replace four copies of a 22-column
  list, and `TestStoredEncodingRoundTrips` sweeps every read path.

## An album page that says how much of the album is yours

- One question, asked once: is there a file. `filePaths` is filled by a
  single batched lookup when the tracklist settles, and the badge, the
  Play count, the dimmed rows and every menu item read it — replacing
  four claims of decreasing confidence that could show a green tick on
  an album whose every action did nothing.
- Play, Play 7 of 12, or no play button at all.
- `total_tracks` on `explore_index` (~2 bytes over 400,677 release
  groups) and on `audio_files` from tags that have always carried it:
  a complete MBID-matched album now makes no catalog call at all, where
  it used to spend the most expensive request the app makes.
- A merged cluster shows the running order the most releases agree on,
  and the version list marks the release you own rather than standing a
  synthetic entry in for it.
- `AlbumReleasesFailed`: a slow fetch is no longer reported as a failed
  one by a 12-second timer.
- Rows not in the library are dimmed in place (with `aria-disabled`)
  instead of the owned ones wearing a green tick and a legend.

## Caches and cover art get ceilings

- Only the three tiers of a cover are stored; the full-resolution copy
  nothing rendered was 1,134 MB of a 1.4 GB covers directory.
- One artist portrait is downloaded and the rest are remembered as
  URLs — 4.1 GB of a 5.3 GB cache was candidates no code path reads.
- `browsedArtBudget` and `httpCacheBudget` bound what an age cannot:
  the same install held art for 5,770 artists in a 1,301-artist
  library.
- `OrphanedArtistImagesJob` joined a bare MBID onto a sharded
  directory, so it deleted the rows that were the only record of the
  files it left behind. `explore.ArtistImageDir` is that layout's one
  definition now.

## The autotag queue asks whether there is work

`tagging_items` was a row per album folder, not a queue, and no query
read the `tag_status` column that held the answer. The four queue
queries ask the files, which matters most where it is least visible:
`startPrefetch` was scoring every album in a tagged library against
MusicBrainz.

## Phantom playlist tracks resolve in place

An M3U8 imported before its files leaves phantom rows; they now match
by path and fall back to position, keep their place in the playlist
when resolved, and pair best-first so two phantoms cannot claim the
same file.

## Playing a track plays the list it is in

Double-click, and Play on a single row's menu, queue the list as
displayed with `startIndex` on that row — the album page and the track
list used to queue one track and discard the album around it. A
multi-row selection still plays exactly itself.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AfVYUVExXsx1nSWrXN8mAh
2026-08-16 13:58:15 -04:00

304 lines
7.6 KiB
Go

package explore
import (
"context"
"database/sql"
"math/rand/v2"
)
// mixBatchSize is how many tracks one GenerateMix call returns.
const mixBatchSize = 30
// mixGenreBoost is added to a candidate's similarity weight for each
// genre it shares with the seed, biasing the pick toward tracks that
// match on both artist and tag rather than artist alone.
const mixGenreBoost = 0.5
// mixSimilarArtistsPerSeed caps how many similar artists are expanded
// per distinct seed artist, so a seed with an unusually long tail in
// similar_artist_map doesn't turn one fallback trigger into hundreds
// of queries.
const mixSimilarArtistsPerSeed = 15
// mixSession is a dynamic mix in progress: the seed it was built from
// (fixed for the life of the session, so successive batches don't
// drift away from what the session started as) and what it has
// already handed out, so a batch doesn't repeat a track that just
// played.
type mixSession struct {
seedPaths []string
played map[string]bool
}
// GenerateMix returns the next batch of tracks for a dynamic-mix queue
// fallback, built by expanding the seed's artists to their similar
// artists (weighted by how often each appears in the seed, sharpened
// by shared genre tags) and restricting candidates to what is actually
// in the library — a queue can only play files that exist.
//
// continuing extends the current mix session — regenerating from its
// original seed rather than seedPaths — instead of starting a fresh
// one. Pass false whenever the queue that just exhausted was not
// itself a mix batch (a real selection just ran out); pass true when
// it was (the mix keeps going indefinitely). label names the batch
// after its most-represented seed artist, for the "Playing from" UI.
func (e *Service) GenerateMix(
ctx context.Context,
seedPaths []string,
continuing bool,
) (paths []string, label string, err error) {
e.mixMu.Lock()
defer e.mixMu.Unlock()
if !continuing || e.mix == nil {
e.mix = &mixSession{seedPaths: seedPaths, played: map[string]bool{}}
}
seed := e.mix.seedPaths
if len(seed) == 0 {
return nil, "", nil
}
artistCounts, topArtistName, genres := e.mixSeedProfile(ctx, seed)
if len(artistCounts) == 0 {
return nil, "", nil
}
candidates := e.mixCandidates(ctx, artistCounts, genres, seed, e.mix.played)
// The session has played through everything this seed can offer —
// rather than dead-ending an "indefinite" mix, start handing out
// repeats.
if len(candidates) == 0 && len(e.mix.played) > 0 {
candidates = e.mixCandidates(ctx, artistCounts, genres, seed, nil)
}
if len(candidates) == 0 {
return nil, "", nil
}
picked := weightedSample(candidates, mixBatchSize)
for _, p := range picked {
e.mix.played[p] = true
}
if topArtistName != "" {
label = "a mix inspired by " + topArtistName
} else {
label = "a dynamic mix"
}
return picked, label, nil
}
// mixSeedProfile tallies the seed's artists by frequency, its genre
// tags, and names the most-represented artist for the UI label.
func (e *Service) mixSeedProfile(
ctx context.Context,
seedPaths []string,
) (artistCounts map[string]int, topArtistName string, genres map[string]bool) {
artistCounts = map[string]int{}
artistNames := map[string]string{}
genres = map[string]bool{}
for _, p := range seedPaths {
artist, err := e.db.ReadQueries.GetArtistByFilePath(ctx, p)
if err == nil && artist.ArtistMbid != "" {
artistCounts[artist.ArtistMbid]++
artistNames[artist.ArtistMbid] = artist.ArtistName
}
}
for _, names := range e.genresByPath(ctx, seedPaths) {
for _, g := range names {
genres[g] = true
}
}
var topCount int
for mbid, count := range artistCounts {
if count > topCount {
topCount = count
topArtistName = artistNames[mbid]
}
}
return artistCounts, topArtistName, genres
}
// mixCandidates builds the weighted pool of library tracks to draw a
// batch from: every owned track by a similar artist, weighted by that
// artist's similarity score times how often the seed artist it came
// from appears in the seed, boosted for a shared genre tag, excluding
// the seed itself and anything already excluded (typically what the
// mix has already played).
func (e *Service) mixCandidates(
ctx context.Context,
artistCounts map[string]int,
seedGenres map[string]bool,
seedPaths []string,
exclude map[string]bool,
) map[string]float64 {
excludeSeed := make(map[string]bool, len(seedPaths))
for _, p := range seedPaths {
excludeSeed[p] = true
}
candidates := map[string]float64{}
// Every candidate path's genres, in one query. This used to be a
// single-row lookup *per candidate* inside two nested loops -
// twenty seed artists by twenty similar artists by thirty paths is
// twelve thousand queries to assemble one mix.
pathGenres := e.genresByPath(ctx, e.similarArtistPaths(ctx, artistCounts))
for seedArtistMBID, count := range artistCounts {
similar, err := e.SimilarArtists(seedArtistMBID)
if err != nil {
continue
}
if len(similar) > mixSimilarArtistsPerSeed {
similar = similar[:mixSimilarArtistsPerSeed]
}
for _, s := range similar {
if s.ArtistMBID == "" {
continue
}
paths, err := e.db.ReadQueries.GetFilePathsByArtistMBID(
ctx,
sql.NullString{String: s.ArtistMBID, Valid: true},
)
if err != nil {
continue
}
weight := s.Score * float64(count)
for _, p := range paths {
if excludeSeed[p] || exclude[p] {
continue
}
for _, g := range pathGenres[p] {
if seedGenres[g] {
weight += mixGenreBoost
break
}
}
candidates[p] += weight
}
}
}
return candidates
}
// genresByPath returns the genres of many files in one query.
func (e *Service) genresByPath(
ctx context.Context, paths []string,
) map[string][]string {
out := make(map[string][]string, len(paths))
if len(paths) == 0 {
return out
}
rows, err := e.db.ReadQueries.GetGenreNamesByFilePaths(ctx, paths)
if err != nil {
return out
}
for _, row := range rows {
out[row.FilePath] = append(out[row.FilePath], row.Name)
}
return out
}
// similarArtistPaths collects every owned file by an artist similar to
// one of the seeds, so their genres can be fetched in one go.
func (e *Service) similarArtistPaths(
ctx context.Context, artistCounts map[string]int,
) []string {
var paths []string
for seedArtistMBID := range artistCounts {
similar, err := e.SimilarArtists(seedArtistMBID)
if err != nil {
continue
}
if len(similar) > mixSimilarArtistsPerSeed {
similar = similar[:mixSimilarArtistsPerSeed]
}
for _, sim := range similar {
if sim.ArtistMBID == "" {
continue
}
p, err := e.db.ReadQueries.GetFilePathsByArtistMBID(
ctx, sql.NullString{String: sim.ArtistMBID, Valid: true},
)
if err != nil {
continue
}
paths = append(paths, p...)
}
}
return paths
}
// weightedSample picks up to n distinct keys from weights without
// replacement, biased toward higher weight (roulette-wheel selection).
// A key with zero or negative weight is never picked.
func weightedSample(weights map[string]float64, n int) []string {
type entry struct {
key string
weight float64
}
pool := make([]entry, 0, len(weights))
var total float64
for k, w := range weights {
if w <= 0 {
continue
}
pool = append(pool, entry{k, w})
total += w
}
picked := make([]string, 0, min(n, len(pool)))
for len(picked) < n && len(pool) > 0 {
r := rand.Float64() * total
idx := 0
for i, e := range pool {
r -= e.weight
if r <= 0 {
idx = i
break
}
}
picked = append(picked, pool[idx].key)
total -= pool[idx].weight
pool = append(pool[:idx], pool[idx+1:]...)
}
return picked
}