Files
yellowjacket/cmd/indexbuild/main.go
T
yonluandClaude Opus 5 b3737d30af feat(explore): carry multi-artist credits in the catalog
A track credited to more than one artist has exactly one navigable
artist in this app and the rest are punctuation. `primaryArtist()`
string-parses the credit, strips a " feat. " clause and discards the
guest; it deliberately does not split on "&", "with" or "," because
those live inside real artist names.

Measured on a real 26,069-file library plus an 80+80 MusicBrainz
sample: 13% of recordings are multi-artist upstream, while only 0.86%
of files carry any structured multi-artist tag — mp3 carries zero
files with multiple MUSICBRAINZ_ARTISTID across 19,840. Of 1,286 files
saying "feat.", 90% have nothing structured behind it, and a sample of
80 such files was multi-artist in MB 80 times out of 80.

CLAUDE.md justified plan 013's removal of the credit tables with "3
credits of 2,823 listed more than one artist". That measured our own
*writer* — cachedLinkArtist was called once per credit, so a
collaboration could never have been recorded. Dropping the join table
was still right on cost; the evidence for "multi-artist is rare" was
not.

A credit is ordered parts and the credit string is derived from them,
so join phrases are assembly instructions, not disassembly ones.
Nothing here reconstructs a credit by searching a name inside a credit
string: the stored text may come from tags while the parts come from
the catalog, and those disagree for ~1 in 3 multi-artist credits.

Where it comes from, after two dead ends: the canonical dump CI
already streams has no join phrases and no as-credited names, and the
JSON dumps cover 153,691 recordings of ~35M with *zero* overlap
against a real library. So mbdump.tar.bz2 — 7.1 GB, ~13.7 min in
pure-Go bzip2, whose members are alphabetical, which is what lets one
pass resolve an entity's credit without buffering 35M recordings.

- artist_credit_part / artist_credit_ref, multi-artist credits only:
  a single-artist credit is already explore_index's own artist_name.
- Column layouts verified against the real 20260815 export;
  ErrDumpShape makes a wrong guess a failed build, not a wrong catalog.
- The pass runs on every mode, not just a build. The job picks its mode
  from the index's own state, and a complete import means "refresh",
  which never enters the importer — so credits could otherwise only
  arrive via a rebuild that re-downloads ~205 GB. It reports whether it
  populated anything, which is what flips `changed` and republishes.
- The importer asks whether an artifact carries the tables, on the
  writer where `core` is attached, so the artifact already published
  still imports.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AfVYUVExXsx1nSWrXN8mAh
2026-08-17 08:25:36 -04:00

384 lines
10 KiB
Go

//go:build indexbuild
// Command indexbuild maintains the explore search index outside the
// desktop app, so the catalog can be built once centrally instead of by
// every install.
//
// It decides what to do from the index's own state rather than needing
// the caller to know:
//
// no completed import → build (first run, or resume a partial one)
// import older than 3mo → rebuild (re-import from the newest dump)
// otherwise → refresh (fold in new incremental listens)
//
// A full build streams ~89GB from the ListenBrainz spark dump — far
// more than one CI job should attempt — so builds are budgeted and
// resumable: the importer checkpoints its absolute stream offset, and
// each run continues where the last stopped. A refresh is cheap
// (~250MB incremental dumps) and finishes in one run.
//
// Usage:
//
// YJ_HOME=/cache indexbuild -budget 3h
//
// Exit codes:
//
// 0 up to date — nothing left to do
// 3 build incomplete — schedule another run to resume
// 1 error
//
// When GITHUB_OUTPUT is set, `complete` and `changed` are appended to it
// so a workflow can decide whether to publish a new artifact.
package main
import (
"context"
"errors"
"flag"
"fmt"
"log/slog"
"os"
"os/signal"
"path/filepath"
"strconv"
"syscall"
"time"
"yellowjacket/backend/database"
"yellowjacket/backend/explore"
"yellowjacket/backend/system"
)
// exitIncomplete tells the caller the build made progress but has not
// finished, so another run should follow. Distinct from a failure: the
// checkpoint is valid and resuming is the correct action.
const exitIncomplete = 3
// mode is what this run decided to do.
type mode string
const (
modeAuto mode = "auto"
modeBuild mode = "build"
modeRefresh mode = "refresh"
modeRebuild mode = "rebuild"
)
// errNoHome is returned when YJ_HOME is unset. The default per-user data
// directory is deliberately not used: a build host should always write
// to an explicit, persistent location.
var errNoHome = errors.New(
"YJ_HOME must be set to a persistent directory on real disk",
)
// errIncomplete signals a clean stop with work remaining.
var errIncomplete = errors.New("build incomplete")
var errBadMode = errors.New("unknown mode")
type opts struct {
budget time.Duration
mode mode
rebuildAfter time.Duration
refreshAfter time.Duration
verbose bool
}
func main() {
var (
budget = flag.Duration("budget", 3*time.Hour,
"stop and checkpoint a build after this long (0 = no limit)")
modeFlag = flag.String("mode", string(modeAuto),
"auto | build | refresh | rebuild")
rebuildAfter = flag.Duration("rebuild-after", 180*24*time.Hour,
"re-import from a fresh dump once the last import is older than this")
refreshAfter = flag.Duration("refresh-after", 7*24*time.Hour,
"minimum gap between incremental refreshes (0 = always)")
verbose = flag.Bool("v", false, "debug logging")
)
flag.Parse()
err := run(opts{
budget: *budget,
mode: mode(*modeFlag),
rebuildAfter: *rebuildAfter,
refreshAfter: *refreshAfter,
verbose: *verbose,
})
if err != nil {
if errors.Is(err, errIncomplete) {
os.Exit(exitIncomplete)
}
fmt.Fprintln(os.Stderr, "indexbuild:", err)
os.Exit(1)
}
}
func run(o opts) error {
logger := newLogger(o.verbose)
if os.Getenv("YJ_HOME") == "" {
return errNoHome
}
dataDir, err := system.GetUserDataDirPath()
if err != nil {
return fmt.Errorf("resolve data dir: %w", err)
}
// Before the schema is applied, not after: applying it over a table
// whose shape has since changed is what fails, and this database's
// non-catalog half is disposable. See retireLibraryTables.
if err := retireLibraryTables(context.Background(), logger); err != nil {
return fmt.Errorf("retire stale tables: %w", err)
}
db, err := database.NewDB(logger)
if err != nil {
return fmt.Errorf("open database: %w", err)
}
// NewExploreService is reused rather than reconstructing its
// dependency graph here, so the headless path cannot drift from what
// the app does. SetContext is never called: it installs the Wails
// runtime context, and emitting Wails events against a non-Wails
// context terminates the process.
svc := explore.NewExploreService(logger.WithGroup("explore"), db)
chosen, why := decide(o, svc)
logger.Info("index maintenance",
"mode", string(chosen),
"reason", why,
"dataDir", dataDir,
"staging", filepath.Join(dataDir, "explore-staging"),
"lastImported", stamp(svc.IndexLastImported()),
"baselineSeries", svc.IndexBaselineSeries(),
)
seriesBefore := svc.IndexBaselineSeries()
switch chosen {
case modeRefresh:
err = doRefresh(logger, svc, o.refreshAfter)
case modeRebuild:
svc.PrepareIndexRebuild()
err = doBuild(logger, svc, o.budget)
case modeBuild:
err = doBuild(logger, svc, o.budget)
case modeAuto:
return fmt.Errorf("%w: auto should have resolved", errBadMode)
default:
return fmt.Errorf("%w: %q", errBadMode, chosen)
}
complete := svc.IndexImportComplete() && !errors.Is(err, errIncomplete)
// Credits are maintenance, not part of any one mode. They come from
// a different dump, they are keyed on entities the catalog already
// holds, and a catalog built before the pass existed would otherwise
// only gain them from a rebuild — which re-downloads ~205 GB to
// re-derive rows it already has. Skipped when the import is not
// complete, because there is nothing to key them against yet.
creditsAdded := false
if complete {
creditsAdded = svc.EnsureArtistCredits(context.Background())
}
// "Changed" means there is something new worth publishing, so it is
// only ever true for a finished import: a build stamps the listens
// series early, long before its rows are assembled, and reporting a
// change off that would be a lie about a half-built index.
changed := complete &&
(svc.IndexBaselineSeries() != seriesBefore ||
chosen != modeRefresh ||
creditsAdded)
report(logger, svc, chosen, complete, changed)
if writeErr := writeOutputs(complete, changed); writeErr != nil {
logger.Warn("could not write workflow outputs", "err", writeErr)
}
return err
}
func newLogger(verbose bool) *slog.Logger {
level := slog.LevelInfo
if verbose {
level = slog.LevelDebug
}
return slog.New(slog.NewTextHandler(os.Stdout, &slog.HandlerOptions{
Level: level,
}))
}
// indexState is the slice of the index that the mode decision depends
// on. Narrowing it to an interface keeps the decision table testable
// without standing up a database.
type indexState interface {
IndexImportComplete() bool
IndexLastImported() time.Time
}
// decide picks the mode from the index's own state, so callers (a cron,
// a push hook, a human) need no knowledge of where the index stands.
func decide(o opts, svc *explore.Service) (mode, string) {
return decideFrom(o, svc)
}
func decideFrom(o opts, state indexState) (mode, string) {
if o.mode != modeAuto {
return o.mode, "explicitly requested"
}
if !state.IndexImportComplete() {
return modeBuild, "no completed import yet"
}
last := state.IndexLastImported()
if last.IsZero() {
// Marker present but unparseable — treat as due rather than
// letting a malformed timestamp wedge the rebuild cadence.
return modeRebuild, "import timestamp unreadable"
}
if age := time.Since(last); age >= o.rebuildAfter {
return modeRebuild, fmt.Sprintf("last import %s ago (>= %s)",
age.Round(time.Hour), o.rebuildAfter)
}
return modeRefresh, "import current, folding in new listens"
}
func doBuild(
logger *slog.Logger, svc *explore.Service, budget time.Duration,
) error {
stop := make(chan os.Signal, 1)
signal.Notify(stop, os.Interrupt, syscall.SIGTERM)
defer signal.Stop(stop)
var timer <-chan time.Time
if budget > 0 {
t := time.NewTimer(budget)
defer t.Stop()
timer = t.C
}
// finished is closed by this function alone, so the watchdog only
// reads it — no double close, and it exits whether the build ended
// on its own or was stopped.
finished := make(chan struct{})
go func() {
select {
case <-finished:
case sig := <-stop:
logger.Info("build: signal received, checkpointing",
"signal", sig.String())
svc.StopIndexBuild()
case <-timer:
logger.Info("build: budget reached, checkpointing")
svc.StopIndexBuild()
}
}()
start := time.Now()
svc.StartIndexBuild()
svc.WaitForIndexIdle()
close(finished)
logger.Info("build stopped", "elapsed", time.Since(start).Round(time.Second))
if !svc.IndexImportComplete() {
return errIncomplete
}
return nil
}
func doRefresh(
logger *slog.Logger, svc *explore.Service, minInterval time.Duration,
) error {
start := time.Now()
svc.RefreshIndexNow(minInterval)
logger.Info("refresh finished",
"elapsed", time.Since(start).Round(time.Second))
return nil
}
func report(
logger *slog.Logger, svc *explore.Service,
chosen mode, complete, changed bool,
) {
status := svc.GetIndexStatus()
logger.Info("index state",
"mode", string(chosen),
"complete", complete,
"changed", changed,
"artists", status.Artists,
"releaseGroups", status.ReleaseGroups,
"recordings", status.Recordings,
"totalRows", status.TotalRows,
"baselineSeries", svc.IndexBaselineSeries(),
)
for _, tier := range status.Tiers {
logger.Info(" stage",
"name", tier.Name,
"state", tier.State,
"completed", tier.Completed,
"total", tier.Total,
"error", tier.Error,
)
}
if !complete {
logger.Info("build incomplete — rerun to resume from checkpoint")
}
}
// writeOutputs appends step outputs when running under a workflow, so
// the caller can publish only when something actually changed.
func writeOutputs(complete, changed bool) error {
path := os.Getenv("GITHUB_OUTPUT")
if path == "" {
return nil
}
f, err := os.OpenFile(path, os.O_APPEND|os.O_CREATE|os.O_WRONLY, 0o600)
if err != nil {
return fmt.Errorf("open outputs: %w", err)
}
defer func() { _ = f.Close() }()
_, err = fmt.Fprintf(f, "complete=%s\nchanged=%s\n",
strconv.FormatBool(complete), strconv.FormatBool(changed))
if err != nil {
return fmt.Errorf("write outputs: %w", err)
}
return nil
}
func stamp(t time.Time) string {
if t.IsZero() {
return "never"
}
return t.UTC().Format(time.RFC3339)
}