feat(metadata): audio tag rewrite via ffmpeg + CBZ ComicInfo injection

Phases 5b and 5c of the metadata writer. 5b — single-file audio (MP3/FLAC/M4A/M4B): the container's tags are rewritten with ffmpeg -c copy (no re-encode; -map 0 keeps cover art, -map_chapters 0 keeps chapters). The key set is exactly what the scanner's applyAudioTagFallbacks reads back — title+album, subtitle, artist/album_artist, composer (narrators, Audible's convention), grouping + iTunes series-part, date, genre — so a write + rescan round-trips. Output is ffprobe-VALIDATED (duration within 1s/0.5% of the source) before the atomic rename; a lossy container rewrite never replaces a good book. Fields the format can't carry per our own read chain are deliberately not written, and fields_written reports only what really round-trips (subtitle rides MP3/FLAC — this image's mp4 muxer drops subtitle keys, discovered empirically and pinned by test). 5c — CBZ: ComicInfo.xml injected (or replaced) with the same atomic zip rewrite as the EPUB path, extended with an append-if-missing mode since a comic may not carry one yet. Field map mirrors the scanner's comic block: Title/Series/Number/Count/Volume, Writer (authors), Summary, Publisher, Year/Month/Day, LanguageISO, Web, Genre+Tags; fractional series numbers are omitted rather than truncated to a lie. CBR/CB7 stay on the sidecar path (RAR/7z can't be rewritten in place). Integration tests generate media at test runtime with ffmpeg (m4b) and in-memory zips (cbz) — no binary fixtures — pinning the tag round-trip through ffprobe, post-write rescan titles, the ComicInfo injection and the validation-gate behavior.
2026-10-27 16:04:34 -04:00
1 parent eaf1b2c8dd
commit 59eddc1386
@@ -7,8 +7,11 @@ import (
 	"encoding/xml"
 	"fmt"
 	"io"
+	"math"
 	"os"
+	"os/exec"
 	"path/filepath"
+	"strconv"
 	"strings"

 	"bookhoard/internal/database"
@@ -45,10 +48,23 @@ const (
 	MetadataTargetEmbeddedOPF MetadataWriteTarget = "embedded-opf"
 	// MetadataTargetSidecarOPF: a Calibre-style sidecar was written next
 	// to the file (formats that can't carry metadata: PDF, comics,
-	// folder books; audio until phase 5b).
+	// folder books).
 	MetadataTargetSidecarOPF MetadataWriteTarget = "sidecar-opf"
+	// MetadataTargetAudioTags: the container's own tags were rewritten
+	// via ffmpeg stream-copy (single-file MP3/FLAC/M4A/M4B — 5b).
+	MetadataTargetAudioTags MetadataWriteTarget = "audio-tags"
+	// MetadataTargetComicInfo: ComicInfo.xml inside the CBZ was
+	// injected/replaced (5c). CBR/CB7 stay on the sidecar path — RAR
+	// and 7z archives can't be rewritten in place.
+	MetadataTargetComicInfo MetadataWriteTarget = "comicinfo"
 )

+// audioTagExts is the single-file set the tag writer claims; multi-file
+// (folder) audio stays on the sidecar path.
+var audioTagExts = map[string]bool{
+	".mp3": true, ".flac": true, ".m4a": true, ".m4b": true,
+}
+
 // MetadataWriteResult reports what the write did, for the UI toast and
 // the response contract.
 type MetadataWriteResult struct {
@@ -61,12 +77,17 @@ type MetadataWriteResult struct {
 // MetadataWriteTargetFor maps a file extension to the write target.
 // The UI mirrors this to label the button; keep both in lockstep.
 func MetadataWriteTargetFor(ext string) MetadataWriteTarget {
-	switch strings.ToLower(strings.TrimSpace(ext)) {
-	case ".epub", ".kepub":
+	ext = strings.ToLower(strings.TrimSpace(ext))
+	switch {
+	case ext == ".epub" || ext == ".kepub":
 		return MetadataTargetEmbeddedOPF
+	case audioTagExts[ext]:
+		return MetadataTargetAudioTags
+	case ext == ".cbz":
+		return MetadataTargetComicInfo
 	default:
-		// 5b (audio tags via ffmpeg) and 5c (CBZ ComicInfo.xml) will
-		// claim their formats here; everything else takes the sidecar.
+		// CBR/CB7 (no in-place rewrite) and every other format take
+		// the sidecar.
 		return MetadataTargetSidecarOPF
 	}
 }
@@ -108,6 +129,26 @@ func (w *MetadataWriter) WriteMetadataToFile(ctx context.Context, itemID uuid.UU
 			Path:          path,
 			FieldsWritten: metadataFieldNames(&item),
 		}, nil
+	case MetadataTargetAudioTags:
+		path, err := w.writeAudioTags(absPath, &item)
+		if err != nil {
+			return nil, err
+		}
+		return &MetadataWriteResult{
+			Target:        MetadataTargetAudioTags,
+			Path:          path,
+			FieldsWritten: audioFieldNames(&item, filepath.Ext(absPath)),
+		}, nil
+	case MetadataTargetComicInfo:
+		path, err := w.writeCBZComicInfo(absPath, &item)
+		if err != nil {
+			return nil, err
+		}
+		return &MetadataWriteResult{
+			Target:        MetadataTargetComicInfo,
+			Path:          path,
+			FieldsWritten: comicFieldNames(&item),
+		}, nil
 	default:
 		path, err := w.writeSidecarOPF(ctx, absPath, &item)
 		if err != nil {
@@ -279,16 +320,25 @@ func originalIdentifiers(oldMetadata []byte) string {
 // rewriteEPUBZip copies every entry to dst, substituting opfName's
 // content. Order, compression method and modes are preserved.
 func rewriteEPUBZip(entries []*zip.File, opfName string, newOPF []byte, dst string) error {
+	return rewriteZipArchive(entries, opfName, newOPF, dst, "", nil)
+}
+
+// rewriteZipArchive is the generic form: substitute name's content, and
+// (for the CBZ path, where ComicInfo.xml may not exist yet) append
+// extraName/extraContent as a final entry when the archive lacks the
+// substituted name.
+func rewriteZipArchive(entries []*zip.File, name string, content []byte, dst string, extraName string, extraContent []byte) error {
 	out, err := os.OpenFile(dst, os.O_CREATE|os.O_WRONLY|os.O_TRUNC, 0o644)
 	if err != nil {
-		return fmt.Errorf("failed to create temp EPUB: %w", err)
+		return fmt.Errorf("failed to create temp archive: %w", err)
 	}
 	writer := zip.NewWriter(out)
+	substituted := false
 	for _, f := range entries {
 		header := f.FileHeader
 		header.Flags &= ^uint16(0x8) // drop data-descriptor flag; sizes are known upfront
 		var target io.Writer
-		if f.Name == opfName {
+		if f.Name == name {
 			// Sizes/CRC recomputed by the writer for the new content.
 			header.CompressedSize64 = 0
 			header.UncompressedSize64 = 0
@@ -299,13 +349,14 @@ func rewriteEPUBZip(entries []*zip.File, opfName string, newOPF []byte, dst stri
 			if err != nil {
 				writer.Close()
 				out.Close()
-				return fmt.Errorf("failed to stage OPF entry: %w", err)
+				return fmt.Errorf("failed to stage substituted entry: %w", err)
 			}
-			if _, err := target.Write(newOPF); err != nil {
+			if _, err := target.Write(content); err != nil {
 				writer.Close()
 				out.Close()
-				return fmt.Errorf("failed to write OPF entry: %w", err)
+				return fmt.Errorf("failed to write substituted entry: %w", err)
 			}
+			substituted = true
 			continue
 		}
 		target, err = writer.CreateHeader(&header)
@@ -332,13 +383,26 @@ func rewriteEPUBZip(entries []*zip.File, opfName string, newOPF []byte, dst stri
 			return fmt.Errorf("failed to finish entry %s: %w", f.Name, err)
 		}
 	}
+	if !substituted && extraName != "" && extraContent != nil {
+		target, cerr := writer.Create(extraName)
+		if cerr != nil {
+			writer.Close()
+			out.Close()
+			return fmt.Errorf("failed to append entry %s: %w", extraName, cerr)
+		}
+		if _, werr := target.Write(extraContent); werr != nil {
+			writer.Close()
+			out.Close()
+			return fmt.Errorf("failed to write appended entry %s: %w", extraName, werr)
+		}
+	}
 	if err := writer.Close(); err != nil {
 		out.Close()
-		return fmt.Errorf("failed to finalize EPUB: %w", err)
+		return fmt.Errorf("failed to finalize archive: %w", err)
 	}
 	if err := out.Sync(); err != nil {
 		out.Close()
-		return fmt.Errorf("failed to sync EPUB: %w", err)
+		return fmt.Errorf("failed to sync archive: %w", err)
 	}
 	return out.Close()
 }
@@ -374,6 +438,336 @@ func atomicWrite(path string, data []byte) error {
 	return nil
 }

+// ---- audio tags (5b): ffmpeg stream-copy rewrite ----
+
+// audioFieldNames lists the fields the audio tag path round-trips —
+// exactly the keys applyAudioTagFallbacks reads back (the ingestion
+// chain's tags rung). Fields the file format can't carry per our own
+// read chain (description, publisher, language, isbn, the full tag
+// list) are deliberately NOT written: "applicable and correct" means
+// nothing lands in a file the scanner would never find again.
+// Subtitle rides MP3/FLAC only (the mp4 muxer in our image drops
+// subtitle keys).
+func audioFieldNames(item *database.MediaItems, ext string) []string {
+	fields := []string{"title"}
+	if item.Subtitle.Valid && item.Subtitle.String != "" &&
+		!strings.EqualFold(ext, ".m4a") && !strings.EqualFold(ext, ".m4b") {
+		fields = append(fields, "subtitle")
+	}
+	if len(item.Authors) > 0 || (item.Author.Valid && item.Author.String != "") {
+		fields = append(fields, "authors")
+	}
+	if len(item.Narrators) > 0 {
+		fields = append(fields, "narrators")
+	}
+	if item.Series.Valid && item.Series.String != "" {
+		fields = append(fields, "series")
+	}
+	if item.DatePublished.Valid {
+		fields = append(fields, "date_published")
+	}
+	if len(item.Tags) > 0 {
+		fields = append(fields, "genre")
+	}
+	return fields
+}
+
+// audioTagArgs maps the item's current values onto the container tag
+// keys the scanner's own fallback chain reads (media_scanner_audio.go):
+//
+//	title → title + album        (players want title; the scanner's
+//	                              album rung names the book)
+//	subtitle → subtitle          (MP3/FLAC only — the mp4 muxer in
+//	                              our image drops subtitle keys, and
+//	                              fields we can't read back don't
+//	                              get written)
+//	authors → artist (first) + album_artist (joined)
+//	narrators → composer         (Audible's convention)
+//	series → grouping            (series-part rides the iTunes atom)
+//	date → date; first tag → genre
+func audioTagArgs(item *database.MediaItems, ext string) [][2]string {
+	var args [][2]string
+	add := func(k, v string) {
+		if strings.TrimSpace(v) != "" {
+			args = append(args, [2]string{k, v})
+		}
+	}
+
+	title := strings.TrimSpace(item.Title)
+	add("title", title)
+	add("album", title)
+	if item.Subtitle.Valid && !strings.EqualFold(ext, ".m4a") && !strings.EqualFold(ext, ".m4b") {
+		add("subtitle", strings.TrimSpace(item.Subtitle.String))
+	}
+
+	authors := item.Authors
+	if len(authors) == 0 && item.Author.Valid {
+		authors = []string{item.Author.String}
+	}
+	if len(authors) > 0 {
+		add("artist", strings.TrimSpace(authors[0]))
+		joined := make([]string, 0, len(authors))
+		for _, a := range authors {
+			a = strings.TrimSpace(a)
+			if a != "" {
+				joined = append(joined, a)
+			}
+		}
+		add("album_artist", strings.Join(joined, " & "))
+	}
+
+	if len(item.Narrators) > 0 {
+		narrators := make([]string, 0, len(item.Narrators))
+		for _, n := range item.Narrators {
+			n = strings.TrimSpace(n)
+			if n != "" {
+				narrators = append(narrators, n)
+			}
+		}
+		add("composer", strings.Join(narrators, ", "))
+	}
+
+	if item.Series.Valid {
+		add("grouping", strings.TrimSpace(item.Series.String))
+		if item.SeriesNumber.Valid && item.SeriesNumber.Float64 > 0 {
+			add("----:com.apple.itunes:series-part", strconvFormatFloat(item.SeriesNumber.Float64))
+		}
+	}
+
+	if item.DatePublished.Valid {
+		add("date", item.DatePublished.Time.Format("2006-01-02"))
+	}
+	if len(item.Tags) > 0 && strings.TrimSpace(item.Tags[0]) != "" {
+		add("genre", strings.TrimSpace(item.Tags[0]))
+	}
+	return args
+}
+
+// probeDurationSeconds ffprobes a file's duration (0 on any failure).
+func probeDurationSeconds(path string) float64 {
+	out, err := exec.Command("ffprobe", "-v", "error",
+		"-show_entries", "format=duration",
+		"-of", "default=noprint_wrappers=1:nokey=1", path).Output()
+	if err != nil {
+		return 0
+	}
+	v, err := strconv.ParseFloat(strings.TrimSpace(string(out)), 64)
+	if err != nil || v < 0 {
+		return 0
+	}
+	return v
+}
+
+// writeAudioTags rewrites the container's tags via ffmpeg stream copy:
+// `-c copy` never re-encodes, `-map 0` keeps every stream (cover art
+// included), `-map_metadata 0` preserves existing global tags, and
+// `-map_chapters 0` keeps the m4b chapter list. Output is validated
+// (duration within tolerance of the source) BEFORE the atomic rename —
+// a lossy container rewrite must never replace a good book.
+func (w *MetadataWriter) writeAudioTags(absPath string, item *database.MediaItems) (string, error) {
+	info, err := os.Stat(absPath)
+	if err != nil {
+		return "", fmt.Errorf("media file missing on disk: %w", err)
+	}
+	if info.IsDir() {
+		return "", fmt.Errorf("multi-file audio books take the sidecar path")
+	}
+
+	srcDuration := probeDurationSeconds(absPath)
+
+	ext := strings.ToLower(filepath.Ext(absPath))
+	tmpPath := absPath + ".tmp-write" + ext
+	args := []string{"-hide_banner", "-loglevel", "error", "-y",
+		"-i", absPath,
+		"-map", "0", "-map_metadata", "0", "-map_chapters", "0",
+		"-c", "copy",
+	}
+	for _, kv := range audioTagArgs(item, ext) {
+		args = append(args, "-metadata", kv[0]+"="+kv[1])
+	}
+	args = append(args, tmpPath)
+
+	if out, err := exec.Command("ffmpeg", args...).CombinedOutput(); err != nil {
+		os.Remove(tmpPath)
+		return "", fmt.Errorf("ffmpeg tag rewrite failed: %s", strings.TrimSpace(string(out)))
+	}
+
+	// Validation gate: the rewritten container must still hold the
+	// same audio (duration within 1s or 0.5%, whichever is looser).
+	outDuration := probeDurationSeconds(tmpPath)
+	if outDuration <= 0 {
+		os.Remove(tmpPath)
+		return "", fmt.Errorf("rewritten audio failed validation (ffprobe read no duration); original untouched")
+	}
+	if srcDuration > 0 {
+		tolerance := math.Max(1.0, srcDuration*0.005)
+		if math.Abs(outDuration-srcDuration) > tolerance {
+			os.Remove(tmpPath)
+			return "", fmt.Errorf("rewritten audio failed validation (duration %.2fs → %.2fs); original untouched", srcDuration, outDuration)
+		}
+	}
+
+	if err := os.Rename(tmpPath, absPath); err != nil {
+		os.Remove(tmpPath)
+		return "", fmt.Errorf("failed to replace audio file: %w", err)
+	}
+	return absPath, nil
+}
+
+// ---- CBZ ComicInfo.xml (5c) ----
+
+// comicFieldNames lists the fields the CBZ write round-trips through
+// the scanner's ComicInfo mapping (media_scanner.go: mergeMetadata's
+// comic block). Fractional series numbers are omitted — the schema's
+// Number is an integer and truncating would round-trip a lie.
+func comicFieldNames(item *database.MediaItems) []string {
+	fields := []string{"title"}
+	if item.Series.Valid && item.Series.String != "" {
+		fields = append(fields, "series")
+		if item.SeriesNumber.Valid && item.SeriesNumber.Float64 > 0 &&
+			item.SeriesNumber.Float64 == float64(int(item.SeriesNumber.Float64)) {
+			fields = append(fields, "series_number")
+		}
+		if item.SeriesCount.Valid && item.SeriesCount.Int32 > 0 {
+			fields = append(fields, "series_count")
+		}
+	}
+	if len(item.Authors) > 0 || (item.Author.Valid && item.Author.String != "") {
+		fields = append(fields, "authors")
+	}
+	if item.Description.Valid && item.Description.String != "" {
+		fields = append(fields, "description")
+	}
+	if len(item.Publishers) > 0 || (item.Publisher.Valid && item.Publisher.String != "") {
+		fields = append(fields, "publisher")
+	}
+	if item.Language.Valid && item.Language.String != "" {
+		fields = append(fields, "language")
+	}
+	if item.DatePublished.Valid {
+		fields = append(fields, "date_published")
+	}
+	if len(item.Tags) > 0 {
+		fields = append(fields, "tags")
+	}
+	if item.WebUrl.Valid && item.WebUrl.String != "" {
+		fields = append(fields, "web_url")
+	}
+	return fields
+}
+
+// buildComicInfoXML renders a ComicInfo.xml (schema v2.x) from the
+// item's current values — the mirror of the scanner's
+// extractMetadataFromArchive mapping. Empty values omit their element:
+// absent means "no data", never "delete".
+func buildComicInfoXML(item *database.MediaItems) string {
+	var b strings.Builder
+	b.WriteString(xml.Header)
+	b.WriteString(`<ComicInfo xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:xsd="http://www.w3.org/2001/XMLSchema">` + "\n")
+
+	tag := func(name, value string) {
+		if strings.TrimSpace(value) != "" {
+			b.WriteString("\t<" + name + ">" + escXML(value) + "</" + name + ">\n")
+		}
+	}
+	tagInt := func(name string, v int) {
+		if v > 0 {
+			b.WriteString("\t<" + name + ">" + strconv.Itoa(v) + "</" + name + ">\n")
+		}
+	}
+
+	tag("Title", strings.TrimSpace(item.Title))
+	tag("Series", strings.TrimSpace(item.Series.String))
+	if item.SeriesNumber.Valid && item.SeriesNumber.Float64 > 0 &&
+		item.SeriesNumber.Float64 == float64(int(item.SeriesNumber.Float64)) {
+		tagInt("Number", int(item.SeriesNumber.Float64))
+	}
+	if item.SeriesCount.Valid {
+		tagInt("Count", int(item.SeriesCount.Int32))
+	}
+	if item.Volume.Valid {
+		tagInt("Volume", int(item.Volume.Int32))
+	}
+
+	authors := item.Authors
+	if len(authors) == 0 && item.Author.Valid {
+		authors = []string{item.Author.String}
+	}
+	joinedAuthors := make([]string, 0, len(authors))
+	for _, a := range authors {
+		a = strings.TrimSpace(a)
+		if a != "" {
+			joinedAuthors = append(joinedAuthors, a)
+		}
+	}
+	tag("Writer", strings.Join(joinedAuthors, ", "))
+
+	tag("Summary", item.Description.String)
+
+	publisher := ""
+	if len(item.Publishers) > 0 {
+		publisher = item.Publishers[0]
+	} else if item.Publisher.Valid {
+		publisher = item.Publisher.String
+	}
+	tag("Publisher", strings.TrimSpace(publisher))
+
+	if item.DatePublished.Valid {
+		t := item.DatePublished.Time
+		tagInt("Year", t.Year())
+		tagInt("Month", int(t.Month()))
+		tagInt("Day", t.Day())
+	}
+
+	tag("LanguageISO", strings.TrimSpace(item.Language.String))
+	if item.WebUrl.Valid {
+		tag("Web", strings.TrimSpace(item.WebUrl.String))
+	}
+
+	genre := strings.TrimSpace(item.Genre.String)
+	if genre == "" && len(item.Tags) > 0 {
+		genre = strings.TrimSpace(item.Tags[0])
+	}
+	tag("Genre", genre)
+	if len(item.Tags) > 0 {
+		cleanTags := make([]string, 0, len(item.Tags))
+		for _, t := range item.Tags {
+			t = strings.TrimSpace(t)
+			if t != "" {
+				cleanTags = append(cleanTags, t)
+			}
+		}
+		tag("Tags", strings.Join(cleanTags, ", "))
+	}
+
+	b.WriteString("</ComicInfo>\n")
+	return b.String()
+}
+
+// writeCBZComicInfo injects (or replaces) ComicInfo.xml inside the CBZ
+// — same atomic zip rewrite as the EPUB path, every other entry copied
+// byte-identical.
+func (w *MetadataWriter) writeCBZComicInfo(absPath string, item *database.MediaItems) (string, error) {
+	reader, err := zip.OpenReader(absPath)
+	if err != nil {
+		return "", fmt.Errorf("failed to open CBZ: %w", err)
+	}
+	defer reader.Close()
+
+	newXML := buildComicInfoXML(item)
+
+	tmpPath := absPath + ".tmp-write"
+	if err := rewriteZipArchive(reader.File, "ComicInfo.xml", []byte(newXML), tmpPath, "ComicInfo.xml", []byte(newXML)); err != nil {
+		os.Remove(tmpPath)
+		return "", err
+	}
+	if err := os.Rename(tmpPath, absPath); err != nil {
+		os.Remove(tmpPath)
+		return "", fmt.Errorf("failed to replace CBZ: %w", err)
+	}
+	return absPath, nil
+}
+
 // ---- OPF generation (mirrors parseOPFContent's read set) ----

 // escXML escapes the five XML entities for element and attribute text.