Lyric fallback order: lrclib, dumb, genius, lastfm

api/_bootstrap.php

@@ -165,11 +165,33 @@ function dumb_search($artist, $title, $timeout) {
     return html_entity_decode($m[1], ENT_QUOTES | ENT_HTML5, 'UTF-8');
 }
 
+/**
+ * Extract Genius annotation ("referent") links out of a raw lyrics HTML
+ * fragment. Returns an ordered list of ['text' => plain annotated lyric
+ * text, 'path' => dumb path to fetch the annotation JSON from]. Matched
+ * structurally — href = "/{id}/{song-slug}/{verse-slug}" — rather than by
+ * CSS class, since Genius's class names are build-hashed (styled-components)
+ * and change on their end without notice.
+ */
+function dumb_extract_annotations($fragment) {
+    $out = [];
+    if (!preg_match_all('#<a\s+href="(/\d+/[^"/]+/[^"/]+)"[^>]*>(.*?)</a>#is', $fragment, $matches, PREG_SET_ORDER)) {
+        return $out;
+    }
+    foreach ($matches as $m) {
+        $text = trim(container_to_text($m[2]));
+        if ($text === '') continue;
+        $out[] = ['text' => $text, 'path' => $m[1] . '/annotations'];
+    }
+    return $out;
+}
+
 /**
  * Fetch + parse a dumb song page ($path from dumb_search()). Returns
- * ['lyrics' => plain text, 'album' => name, 'art' => absolute album art url,
- * 'song_art' => absolute song thumbnail url] (any field may be ''), or null
- * if the page couldn't be fetched.
+ * ['lyrics' => plain text, 'annotations' => [{text, path}, …] (see
+ * dumb_extract_annotations), 'album' => name, 'art' => absolute album art
+ * url, 'song_art' => absolute song thumbnail url] (any field may be ''/[]),
+ * or null if the page couldn't be fetched.
  */
 function dumb_song($path, $timeout) {
     $base = cfg('dumb.url');
@@ -177,14 +199,18 @@ function dumb_song($path, $timeout) {
     $html = http_get($base . $path, $timeout);
     if ($html === null) return null;
 
-    $out = ['lyrics' => '', 'album' => '', 'art' => '', 'song_art' => ''];
+    $out = ['lyrics' => '', 'annotations' => [], 'album' => '', 'art' => '', 'song_art' => ''];
 
     // <div id="lyrics"> wraps the raw Genius lyrics HTML that dumb already
     // extracted server-side (no data-lyrics-container markers left in it).
     $idPos = strpos($html, 'id="lyrics"');
     if ($idPos !== false) {
         $lt = strrpos(substr($html, 0, $idPos), '<div');
-        if ($lt !== false) $out['lyrics'] = trim(container_to_text(inner_balanced_div($html, $lt)));
+        if ($lt !== false) {
+            $fragment = inner_balanced_div($html, $lt);
+            $out['annotations'] = dumb_extract_annotations($fragment);
+            $out['lyrics'] = trim(container_to_text($fragment));
+        }
     }
 
     // Two <img id="album-artwork"> tags can appear: the song thumbnail