navidrome/core/artwork/artwork.go
Deluan Quintão c26f6f9e98
feat(artwork): store the resolution trace so artwork explain works offline (#5980)
* feat(artwork): record the resolution trace so explain works without --live

The worker never attached a ChainTrace, so `artwork explain` had to re-walk the
priority chain at CLI time. That reconstruction could disagree with what actually
happened, and without --live it could not report the external tier at all.

The worker now traces every acquisition and stores it. `explain` reads the stored
trace by default and reports when it was recorded; --live re-walks and calls the
agents. Disc artwork keeps no row, so it always walks live.

A chain trace alone would have explained almost nothing about failures: six of the
seven ways an item can fail happen after the chain has already picked a winner. The
trace now covers those stages too, and has somewhere to live when they fail: the
retrying queue row carries the last failure, and the state row keeps it in
last_failure once the retry budget is spent and the queue row is deleted.

Measured on a copy of a 682MB / 43.6k-item library: +9.7MB (+1.4%). No row crosses
the WITHOUT ROWID overflow threshold, so list hydration is unchanged; only full
scans of item_artwork, which no request performs, read more pages.

* test(artwork): pin the give-up ordering that keeps a failure for unresolved items

recordGiveUp updates an existing row, and for a kind with a recheck path that row is
only created moments earlier by the absent settle. Recording before the settle would
lose the failure for every item that never resolved, with nothing to catch it.

* refactor(artwork): tighten the trace code after review

Four fixes worth taking:

The doc comments on ChainTrace and chainState.trace still said the worker never
attaches a trace and resolution stays allocation-free — the exact invariant this
branch reverses.

explain's report field meant both "the chain shown was walked just now" and "go out
for real", and was being passed to loadPluginAgents, which --live documents as the
only thing that may open external connections. Renamed to `walked` and restored
explainLive as the sole input to that decision.

A stored Detail is an error string on the failure paths, with no bound. The measured
"no row reaches the WITHOUT ROWID overflow limit" only holds while it is bounded, so
cap it at 200 runes.

offlineGate was a factory returning a constant closure; make it a plain gateFunc like
its sibling passthroughGate. Collapse five copies of the age-a-queue-row loop in the
worker tests into one helper.

* refactor(artwork): drop the offline explain walk, now that traces are stored

`artwork explain` reported the external tier without calling it, so a diagnostic
could not add load to a provider already rate-limiting us. Reading the stored trace
answers that better: it reports what the agents actually returned, not what would
be tried.

Nothing could reach the offline gate any more. It was installed only for a walk
with --live unset, which now happens for disc artwork alone, and disc rejects the
external candidate before any gate call. That made the gate, its sentinel error,
the would-try outcome and two of explain's verdicts unreachable.

Removes offlineGate, errOfflineSkipped, OutcomeWouldTry, the NewTracingResolver
live parameter and the CreateArtworkResolver argument threaded through wire.

Verified against a copy of a real library: disc artwork with "external" first in
DiscArtPriority and external services enabled still records the skip and issues no
agent call.

* fix(artwork): make explain's no-network guarantee structural, not incidental

Serving falls back disc -> album and track -> disc -> album. The resolver layer
explain uses has no such fallback today, so dropping the offline gate did not leak.
But the guarantee rested on which chains happen to lack an external tier, and the
serving layer already shows the fallback shape someone could mirror.

Without --live the tracing resolver is now built with no agents at all, so no chain
and no fallback added later can reach a provider. That is stronger than the gate it
replaces, which only intercepted the call.

The test pins it against exactly that regression: with the guard removed and the
serving fallback mirrored into resolveDisc, it fails.

* refactor(artwork): trim the trace plumbing

EncodeTrace was exported for nobody: only this package writes traces, and cmd reads
them. It becomes a ChainTrace method, which also drops the copy Steps made for a
caller that only wanted to serialize.

explain's report carried queuedSteps and failureSteps, both pure functions of the
queue and state rows already in the struct, which let a test set the two out of step
with each other. formatExplain derives them, as it already does for every other
display value.

The trace row format and its tabwriter empty-cell rule lived in two places, and the
"nothing was ever recorded" predicate in three.

* fix(artwork): clear the queue trace on a fresh re-enqueue

Enqueue's conflict clause reset attempts to 0 but left the new trace
column, so after a scan or refresh re-enqueued a previously-failed item
artwork explain showed "Attempts: 0" next to the prior lifecycle's
"Last attempt failed" trace. Clear trace in Enqueue (a fresh lifecycle
has no last attempt); EnqueuePreservingBackoff still keeps it.

* fix(artwork): treat a processing-stage error as indeterminate in explain

A read/hash/decode/store failure records an OutcomeError step and writes an
absent row, but explainResult only mapped external errors and unreadable
candidates to indeterminate, so the default verdict read "not resolved" —
presenting a processing failure as a definitive miss. The worker retries
these exactly as it retries an unreadable candidate, so classify any
OutcomeError as indeterminate too.

* fix(artwork): record a trace step when a chainless resolver faults

Playlist and radio resolvers walk no priority chain, so a fault (unreadable
upload/sidecar, or an m3u fetch error with no grid) returned localError/extError
without recording any trace step. The attempt then encoded [], leaving artwork
explain with an empty "Last attempt failed" and "Gave up after". Record a
fallback step in the faulted-no-image branch when nothing else did, and carry
the source label through resolveLocalFile so the step can name it.

* fix(artwork): trace the m3u failure at its source, not via the empty guard

A playlist's grid sampling records album-chain steps into the shared trace, so
the processor's empty-trace fallback no longer fires when the m3u remote image
fetch failed — the error that forced the retry was omitted from explain. Record
it where it happens, in resolvePlaylist's external step, as external:m3u.

* test(artwork): skip the chainless-fault spec on Windows

The spec provokes an open fault with a non-directory parent, but Windows maps
that to a not-exist error, so localError is never set and the item resolves
absent instead of failed. The sibling failed-on-unreadable-upload spec skips
Windows for the same class of reason.

* fix(artwork): don't label an absent empty-chain row as pre-tracing

explain reported "resolved before traces were recorded" for any stored row
with an empty chain, but an empty CoverArtPriority records a real, empty [] chain
and resolves absent. A recorded resolution that finds an image always records its
winning candidate, so only a row with a hash and no chain predates tracing; split
on the hash and report an absent empty chain plainly instead.

* fix(db): retimestamp the artwork trace migration after rebase

master merged a 2026-08-18 migration, so the original 2026-08-16 timestamp is now
older than the newest on the base branch and Goose would silently skip it on an
already-upgraded database. Bumped past it; the SQL is unchanged.

* fix(artwork): keep the m3u error detail in the trace

The m3u trace step recorded OutcomeError with no detail because resolveExternalStep
collapsed the gate's error to a bool, so explain showed only "external:m3u error -"
and could not tell a timeout from an HTTP error or an open breaker. Return the error
(normalizing not-found to nil so it stays a definitive miss, not a failure) and store
its message as the step detail; encodeSteps already bounds it.

* docs(artwork): note the give-up write relies on serial draining

recordGiveUp writes last_failure unconditionally; that is only correct because
the drain resolves each item serially, so no concurrent success can store artwork
between the write and the queue delete. Record the invariant at the call site.
2026-08-21 10:24:01 -04:00

442 lines
16 KiB
Go

package artwork
import (
"bytes"
"context"
"errors"
"fmt"
"io"
"os"
"time"
"github.com/navidrome/navidrome/conf"
"github.com/navidrome/navidrome/consts"
"github.com/navidrome/navidrome/core/agents"
"github.com/navidrome/navidrome/core/ffmpeg"
"github.com/navidrome/navidrome/log"
"github.com/navidrome/navidrome/model"
"github.com/navidrome/navidrome/resources"
"github.com/navidrome/navidrome/utils/cache"
)
var ErrUnavailable = errors.New("artwork unavailable")
// errStaleSource means the backing file's mtime no longer matches RefMtime, so the stored hash may be stale.
var errStaleSource = errors.New("artwork: source file changed since resolution")
// Image is one servable artwork response.
type Image struct {
io.ReadCloser
Hash string // pixel identity; "" for placeholders
ETag string // representation validator; "" means Hash applies (full-size original)
LastUpdated time.Time
Placeholder bool
}
// representationTag varies with dimensions and encode settings, so a config change invalidates
// a revalidating client's cache even though the pixel hash is unchanged.
func representationTag(hash string, size int, square bool) string {
return fmt.Sprintf("%s.%d.%v.%s", hash, size, square, formatQualityTag())
}
type Artwork interface {
// Get returns ErrUnavailable when there is nothing to serve and model.ErrNotFound when
// the id resolves to nothing, so the caller can pick placeholder vs 404.
Get(ctx context.Context, artID model.ArtworkID, size int, square bool) (*Image, error)
// GetOrPlaceholder accepts an artwork token or a raw entity id, falling back to the
// kind's placeholder image (never resized, Placeholder=true).
GetOrPlaceholder(ctx context.Context, id string, size int, square bool) (*Image, error)
}
func NewArtwork(ds model.DataStore, cache cache.FileCache, store *ImageStore, ffm ffmpeg.FFmpeg) Artwork {
return &service{ds: ds, cache: cache, store: store, ffmpeg: ffm}
}
// entityExists reports whether the entity an artwork id points at is still there: state rows
// outlive a deleted entity until the next prune, so a servable row is not evidence of its owner.
func entityExists(ctx context.Context, ds model.DataStore, artID model.ArtworkID) bool {
var found bool
var err error
switch artID.Kind {
case model.KindArtistArtwork:
found, err = ds.Artist(ctx).Exists(artID.ID)
case model.KindAlbumArtwork:
found, err = ds.Album(ctx).Exists(artID.ID)
case model.KindMediaFileArtwork:
found, err = ds.MediaFile(ctx).Exists(artID.ID)
case model.KindPlaylistArtwork:
found, err = ds.Playlist(ctx).Exists(artID.ID)
case model.KindRadioArtwork:
found, err = ds.Radio(ctx).Exists(artID.ID)
case model.KindDiscArtwork:
albumID, _, perr := model.ParseDiscArtworkID(artID.ID)
if perr != nil {
return false
}
found, err = ds.Album(ctx).Exists(albumID)
default:
return false
}
return err == nil && found
}
type service struct {
ds model.DataStore
cache cache.FileCache
store *ImageStore
ffmpeg ffmpeg.FFmpeg
}
func (s *service) GetOrPlaceholder(ctx context.Context, id string, size int, square bool) (*Image, error) {
artID, err := s.parseArtworkID(ctx, id)
var img *Image
if err == nil {
img, err = s.Get(ctx, artID, size, square)
}
// Only a resolvable entity with no art gets a placeholder; an unknown id must stay
// ErrNotFound so callers can still answer 404 / Subsonic error 70.
if errors.Is(err, ErrUnavailable) {
return placeholderImage(artID.Kind), nil
}
return img, err
}
func (s *service) Get(ctx context.Context, artID model.ArtworkID, size int, square bool) (*Image, error) {
if artID.ID == "" {
return nil, ErrUnavailable
}
if size < 0 {
size = 0 // a negative size means full-size, not a giant (OOM) resize rectangle
}
switch artID.Kind {
case model.KindDiscArtwork:
return s.serveDisc(ctx, artID, size, square)
case model.KindMediaFileArtwork:
return s.serveMediaFile(ctx, artID, size, square)
default:
return s.serveEntity(ctx, artID, size, square)
}
}
// requestRecheckAge throttles view-triggered rechecks so reopening a genuinely-absent page can't
// hammer external services; below StaleAbsentAge to catch younger absences.
const requestRecheckAge = time.Hour
func (s *service) serveEntity(ctx context.Context, artID model.ArtworkID, size int, square bool) (*Image, error) {
ia, err := s.ds.Artwork(ctx).GetItemArtwork(artID.Kind, artID.ID, model.ImageTypePrimary)
switch {
case errors.Is(err, model.ErrNotFound):
return s.provisional(ctx, artID, size, square)
case err != nil:
return nil, err
case ia.Hash == "":
// Inserts an immediately-eligible recheck for a settled absent row.
if time.Since(ia.AttemptedAt) > requestRecheckAge {
s.enqueue(ctx, artID, model.ArtworkPriorityBump)
}
return nil, ErrUnavailable
default:
return s.serveHash(ctx, artID, ia, size, square)
}
}
// serveSource is the one place bytes become an Image. hash is the pixel identity ("" for disc art)
// and doubles as the full-size validator, so an ETag is only needed when resized or hash is "".
func (s *service) serveSource(ctx context.Context, key, hash string, lastUpdate time.Time,
size int, square bool, open func() (io.ReadCloser, error),
) (*Image, error) {
if size == 0 && !square {
rc, err := open()
if err != nil {
return nil, err
}
if rc == nil {
return nil, ErrUnavailable
}
img := &Image{ReadCloser: rc, Hash: hash, LastUpdated: lastUpdate}
if hash == "" {
img.ETag = representationTag(key, size, square)
}
return img, nil
}
stream, err := s.cache.Get(ctx, &resizedItem{
hash: key, size: size, square: square, ffmpeg: s.ffmpeg, open: open,
})
if err != nil {
return nil, err
}
return &Image{ReadCloser: stream, Hash: hash, ETag: representationTag(key, size, square), LastUpdated: lastUpdate}, nil
}
// serveHash serves the bytes of a found state row. A mismatch/open error is dangling, but a
// cancelled request is not: it must not enqueue a re-resolution.
func (s *service) serveHash(ctx context.Context, artID model.ArtworkID, ia *model.ItemArtwork, size int, square bool) (*Image, error) {
// Only this path can hand back a deleted entity's bytes; the others load their entity anyway.
if !entityExists(ctx, s.ds, artID) {
return nil, ErrUnavailable
}
art, err := s.ds.Artwork(ctx).GetImage(ia.Hash)
if err != nil {
if errors.Is(err, model.ErrNotFound) {
return s.dangling(ctx, artID)
}
return nil, err
}
img, err := s.serveSource(ctx, ia.Hash, ia.Hash, ia.UpdatedAt, size, square,
func() (io.ReadCloser, error) { return openOriginal(ia, art.Mime, s.store) })
if err != nil {
if errors.Is(err, context.Canceled) {
return nil, err
}
log.Warn(ctx, "Artwork: Could not serve image", "artID", artID, "size", size, err)
return s.dangling(ctx, artID)
}
return img, nil
}
// openOriginal enforces the mtime invariant: bytes are never served under a hash they no longer match.
func openOriginal(ia *model.ItemArtwork, mime string, store *ImageStore) (io.ReadCloser, error) {
if isFileBacked(ia.Source) {
f, err := os.Open(ia.SourcePath)
if err != nil {
return nil, err
}
info, err := f.Stat()
if err != nil {
f.Close()
return nil, err
}
if ia.RefMtime != 0 && info.ModTime().UnixNano() != ia.RefMtime {
f.Close()
log.Debug("Artwork: Backing file changed since resolution", "path", ia.SourcePath,
"hash", ia.Hash, "resolvedMtime", ia.RefMtime, "currentMtime", info.ModTime().UnixNano())
return nil, errStaleSource
}
return f, nil
}
// Store-backed bytes still carry the source's mtime, to detect edits to embedded art.
if ia.SourcePath != "" && ia.RefMtime != 0 {
info, err := os.Stat(ia.SourcePath)
if err != nil {
return nil, err
}
if info.ModTime().UnixNano() != ia.RefMtime {
log.Debug("Artwork: Source file changed since resolution", "path", ia.SourcePath,
"hash", ia.Hash, "resolvedMtime", ia.RefMtime, "currentMtime", info.ModTime().UnixNano())
return nil, errStaleSource
}
}
return store.Open(ia.Hash, mime)
}
// provisional serves local bytes for an entity with no state row, enqueuing the worker but
// never writing a state row itself.
func (s *service) provisional(ctx context.Context, artID model.ArtworkID, size int, square bool) (*Image, error) {
item := model.ArtworkQueueItem{ItemKind: artID.Kind.Prefix(), ItemID: artID.ID, ImageType: model.ImageTypePrimary}
res, err := newLocalResolver(s.ds, s.ffmpeg).resolve(ctx, item)
if err != nil {
return nil, err
}
s.enqueue(ctx, artID, model.ArtworkPriorityBump)
log.Debug(ctx, "Artwork: Provisional read-through, no state row yet", "artID", artID,
"source", res.source, "hit", res.reader != nil)
return s.serveResolution(ctx, res, size, square)
}
// serveResolution turns a local resolution's bytes into a servable Image (byte-hash only, no decode).
func (s *service) serveResolution(ctx context.Context, res resolution, size int, square bool) (*Image, error) {
if res.reader == nil {
return nil, ErrUnavailable
}
defer res.reader.Close()
data, err := readCapped(res.reader)
if err != nil {
return nil, ErrUnavailable
}
hash, err := hashImage(bytes.NewReader(data))
if err != nil {
return nil, ErrUnavailable
}
// Keyed by the byte-hash, so the entry lines up with the worker's eventual store entry.
return s.serveSource(ctx, hash, hash, unixMtime(res.refMtime), size, square,
func() (io.ReadCloser, error) { return io.NopCloser(bytes.NewReader(data)), nil })
}
func (s *service) serveMediaFile(ctx context.Context, artID model.ArtworkID, size int, square bool) (*Image, error) {
// The setting is not in the config fingerprint, so honor it at serve time: a direct mf- URL
// must fall back to disc/album instead of serving stale persisted embedded art.
if !conf.Server.EnableMediaFileCoverArt {
mf, err := s.ds.MediaFile(ctx).Get(artID.ID)
if err != nil {
return nil, err
}
return s.Get(ctx, mf.DiscCoverArtID(), size, square)
}
ia, err := s.ds.Artwork(ctx).GetItemArtwork(model.KindMediaFileArtwork, artID.ID, model.ImageTypePrimary)
switch {
case err == nil && ia.Hash != "":
return s.serveHash(ctx, artID, ia, size, square)
case err == nil:
// absent row: fall through
case errors.Is(err, model.ErrNotFound):
// no row: fall through
default:
return nil, err
}
noRow := errors.Is(err, model.ErrNotFound)
mf, err := s.ds.MediaFile(ctx).Get(artID.ID)
if err != nil {
return nil, err
}
if noRow && conf.Server.EnableMediaFileCoverArt && mf.HasCoverArt {
return s.provisionalEmbedded(ctx, artID, *mf, size, square)
}
// Mirror MediaFile.CoverArtID: a track defers to its disc art, which falls back to the album.
return s.Get(ctx, mf.DiscCoverArtID(), size, square)
}
// provisionalEmbedded serves a track's embedded art immediately, leaving the state row to the worker.
func (s *service) provisionalEmbedded(ctx context.Context, artID model.ArtworkID, mf model.MediaFile, size int, square bool) (*Image, error) {
lib, err := loadLibraryView(ctx, s.ds, mf.LibraryID)
if err != nil {
return nil, err
}
res, ok := resolveEmbedded(ctx, lib, s.ffmpeg, mf.Path)
s.enqueue(ctx, artID, model.ArtworkPriorityBump)
if !ok {
// Eligible but unextractable: fall back the way CoverArtID does, not to a placeholder.
return s.Get(ctx, mf.DiscCoverArtID(), size, square)
}
return s.serveResolution(ctx, res, size, square)
}
// serveDisc reads disc art through with no state row and no enqueue, falling back to the album cover.
func (s *service) serveDisc(ctx context.Context, artID model.ArtworkID, size int, square bool) (*Image, error) {
dr, err := newDiscArtworkReader(ctx, s.ds, artID)
if err != nil {
return nil, err
}
// Single-disc albums run the chain too: a disc can carry art distinct from the album cover.
selectImage := func() (io.ReadCloser, error) {
res, err := dr.selectImage(ctx, s.ffmpeg, conf.Server.DiscArtPriority, &chainState{})
return res.reader, err
}
albumArtID := model.ArtworkID{Kind: model.KindAlbumArtwork, ID: dr.album.ID}
// Disc art has no state row, hence no content hash: keying on id, album mtime and
// DiscArtPriority lets a warm cache answer without running the chain or touching the disk.
key := fmt.Sprintf("%s|%d|%s", artID.ID, dr.cacheTime().UnixNano(), conf.Server.DiscArtPriority)
img, err := s.serveSource(ctx, key, "", dr.cacheTime(), size, square, selectImage)
if err != nil {
if errors.Is(err, context.Canceled) {
return nil, err
}
return s.Get(ctx, albumArtID, size, square)
}
return img, nil
}
// dangling enqueues a re-resolution and reports unavailable, leaving the state row untouched.
func (s *service) dangling(ctx context.Context, artID model.ArtworkID) (*Image, error) {
log.Debug(ctx, "Artwork: State row points at bytes we cannot serve, re-resolving", "artID", artID)
s.enqueue(ctx, artID, model.ArtworkPriorityScan)
return nil, ErrUnavailable
}
func (s *service) enqueue(ctx context.Context, artID model.ArtworkID, priority int) {
err := s.ds.ArtworkQueue(ctx).EnqueuePreservingBackoff(model.ArtworkQueueItem{
ItemKind: artID.Kind.Prefix(),
ItemID: artID.ID,
ImageType: model.ImageTypePrimary,
Priority: priority,
})
if err != nil {
log.Warn(ctx, "Artwork: Could not enqueue re-resolution", "artID", artID, err)
}
}
func placeholderImage(kind model.Kind) *Image {
path := consts.PlaceholderAlbumArt
if kind == model.KindArtistArtwork {
path = consts.PlaceholderArtistArt
}
r, _ := resources.FS().Open(path)
return &Image{ReadCloser: r, Placeholder: true}
}
type coverArtIDGetter interface {
CoverArtID() model.ArtworkID
}
// parseArtworkID accepts an artwork token or a raw entity id, resolving the latter to its CoverArtID.
func (s *service) parseArtworkID(ctx context.Context, id string) (model.ArtworkID, error) {
if id == "" {
return model.ArtworkID{}, ErrUnavailable
}
if artID, err := model.ParseArtworkID(id); err == nil {
return artID, nil
}
entity, err := model.GetEntityByID(ctx, s.ds, id)
if err != nil {
return model.ArtworkID{}, err
}
if e, ok := entity.(coverArtIDGetter); ok {
return e.CoverArtID(), nil
}
return model.ArtworkID{}, model.ErrNotFound
}
// TracingResolver is the CLI's read-only view of resolution: it walks the priority chain, records
// the walk and reports the winning source, without ever writing artwork state.
type TracingResolver struct {
inner *resolver
trace *ChainTrace
}
// NewTracingResolver builds a TracingResolver that records its priority-chain walk. Without live
// it gets no agents at all, so neither a chain nor any fallback added later can reach a provider;
// with it, one item is at most one call per agent, so the rate limiter and breaker are bypassed.
func NewTracingResolver(ds model.DataStore, ag *agents.Agents, ffm ffmpeg.FFmpeg, t *ChainTrace, live bool) *TracingResolver {
inner := newLocalResolver(ds, ffm)
if live {
inner = newResolver(ds, ag, ffm, passthroughGate)
}
return &TracingResolver{inner: inner, trace: t}
}
// Resolve walks kind's sources for id, recording the walk, and reports the winning source
// ("" when none produced an image).
func (r *TracingResolver) Resolve(ctx context.Context, kind model.Kind, id string) (string, error) {
switch kind {
case model.KindArtistArtwork:
return r.explain(ctx, r.inner.resolveArtist, id)
case model.KindAlbumArtwork:
return r.explain(ctx, r.inner.resolveAlbum, id)
case model.KindDiscArtwork:
return r.explain(ctx, r.inner.resolveDisc, id)
case model.KindMediaFileArtwork:
return r.explain(ctx, r.inner.resolveMediaFile, id)
}
return "", fmt.Errorf("artwork: %s artwork has no chain to explain", kind)
}
// explain discards the bytes: nothing downstream persists this resolution, so nothing else
// would close the reader either.
func (r *TracingResolver) explain(ctx context.Context, resolve func(context.Context, string) (resolution, error), id string) (string, error) {
res, err := resolve(withTrace(ctx, r.trace), id)
if err != nil {
return "", err
}
if res.reader != nil {
_ = res.reader.Close()
}
return res.source, nil
}
func unixMtime(mtime int64) time.Time {
if mtime <= 0 {
return time.Time{}
}
return time.Unix(0, mtime) // RefMtime is unix-nanoseconds
}