Improve compilation album parsing to prevent albums fragmentation (#279)

Fixes a bug where compilation albums are parsed as separate albums
instead of creating a single junction album with various artists.

Fixes #267

Author: kushalpandya

Managers/Database/DMCategoryQueries.swift

@@ -68,6 +68,9 @@ extension DatabaseManager {
     func getAlbumEntities() -> [AlbumEntity] {
         do {
             return try dbQueue.read { db in
+                // Prefer the album's primary artist from the album_artists junction
+                // (which findOrCreateAlbum populates, including "Various Artists" for compilations)
+                // before falling back to per-track tag aggregates.
                 let sql = """
                     SELECT
                         albums.id,
@@ -76,7 +79,17 @@ extension DatabaseManager {
                         albums.artwork_data,
                         albums.release_year,
                         COALESCE(SUM(tracks.duration), 0) as totalDuration,
-                        COALESCE(NULLIF(MAX(tracks.album_artist), ''), MAX(tracks.artist)) as artistName,
+                        COALESCE(
+                            (SELECT artists.name
+                             FROM album_artists
+                             JOIN artists ON artists.id = album_artists.artist_id
+                             WHERE album_artists.album_id = albums.id
+                               AND album_artists.role = 'primary'
+                             ORDER BY album_artists.position
+                             LIMIT 1),
+                            NULLIF(MAX(tracks.album_artist), ''),
+                            MAX(tracks.artist)
+                        ) as artistName,
                         albums.created_at
                     FROM albums
                     LEFT JOIN tracks ON albums.id = tracks.album_id AND tracks.is_duplicate = 0

Managers/Database/DMCleanup.swift

@@ -57,15 +57,24 @@ extension DatabaseManager {
 
             // 2. Now clean up main tables using GRDB where possible
 
-            // Get all artist IDs that are still referenced in track_artists
+            // Get all artist IDs that are still referenced in track_artists OR album_artists.
+            // Album-only artists (e.g. the canonical "Various Artists" used to anchor
+            // compilations) have no track_artists rows but must not be considered orphaned.
             let artistsWithTracks = try TrackArtist
                 .select(TrackArtist.Columns.artistId, as: Int64.self)
                 .distinct()
                 .fetchSet(db)
-            
-            // Delete artists that have NO tracks
+
+            let artistsWithAlbums = try AlbumArtist
+                .select(AlbumArtist.Columns.artistId, as: Int64.self)
+                .distinct()
+                .fetchSet(db)
+
+            let referencedArtists = artistsWithTracks.union(artistsWithAlbums)
+
+            // Delete artists that have NO references in either junction
             let allArtistIds = try Artist.select(Artist.Columns.id, as: Int64.self).fetchSet(db)
-            let artistsToDelete = allArtistIds.subtracting(artistsWithTracks)
+            let artistsToDelete = allArtistIds.subtracting(referencedArtists)
 
             if !artistsToDelete.isEmpty {
                 let orphanedArtists = try Artist

Managers/Database/DMNormalization.swift

@@ -191,10 +191,14 @@ extension DatabaseManager {
     func processTrackAlbum(_ track: inout FullTrack, in db: Database, cache: ScanLookupCache? = nil) throws {
         guard !track.album.isEmpty && track.album != "Unknown Album" else { return }
 
-        // Determine the album artist (prefer albumArtist field, fallback to first artist from multi-artist string)
+        // Determine the album artist (prefer albumArtist field, then compilation flag, then first artist from multi-artist string)
         let albumArtistName: String?
         if let albumArtist = track.albumArtist, !albumArtist.isEmpty {
             albumArtistName = albumArtist
+        } else if track.compilation {
+            // Compilation flag set but no album artist tag; group all tracks under a canonical name
+            // so a Various Artists comp doesn't fragment into one album per track artist.
+            albumArtistName = "Various Artists"
         } else if !track.artist.isEmpty && track.artist != "Unknown Artist" {
             // Parse the artist field and use the first artist as the album artist
             let artists = ArtistParser.parse(track.artist)