Better dumb search

api/_bootstrap.php

@@ -102,6 +102,46 @@ function request_track() {
     return [trim((string) $artist), trim((string) $title)];
 }
 
+/**
+ * Strip a featuring-artist credit ("(feat. X)", "(ft. X)", "(featuring X)",
+ * "(with X)", bracket variants, or an unparenthesized trailing "feat. X")
+ * before it goes into a Genius/dumb search query.
+ *
+ * Genius's canonical song title never includes the featuring credit (it's
+ * separate metadata), but translation/cover pages often spell it out
+ * verbatim in their title — e.g. searching "Havana (feat. Young Thug)"
+ * surfaces only German/French translation pages (titled "... ft. Young
+ * Thug ... (Deutsche Übersetzung)" etc.), while "Havana" alone correctly
+ * finds the original as the first hit. Leaves other qualifiers like
+ * "(Remix)"/"(Live)" alone — those genuinely change which page you want.
+ */
+function strip_feature_credit($title) {
+    $title = preg_replace('/[\(\[]\s*(?:feat\.?|ft\.?|featuring|with)\b[^\)\]]*[\)\]]/i', '', $title);
+    $title = preg_replace('/\s+(?:feat\.?|ft\.?|featuring)\b.+$/i', '', $title);
+    return trim(preg_replace('/\s{2,}/', ' ', $title));
+}
+
+/**
+ * Loose "is this search hit actually by the artist we want" check —
+ * case-insensitive, ignores punctuation/spacing and anything in brackets
+ * (so "Camila Cabello (Ft. Young Thug)" still matches "Camila Cabello").
+ * An empty side always matches (nothing to compare against). Used to skip
+ * past hits attributed to someone else entirely — most notably Genius's
+ * "Genius Deutsche Übersetzungen" / "Genius traductions françaises" / etc.
+ * pseudo-artist accounts that translation pages get filed under.
+ */
+function artist_loosely_matches($hitArtist, $wantArtist) {
+    $norm = function ($s) {
+        $s = preg_replace('/[\(\[][^\)\]]*[\)\]]/', '', (string) $s);
+        $s = preg_replace('/[^\p{L}\p{N}]+/u', '', $s);
+        return strtolower($s);
+    };
+    $a = $norm($hitArtist);
+    $b = $norm($wantArtist);
+    if ($a === '' || $b === '') return true;
+    return strpos($a, $b) !== false || strpos($b, $a) !== false;
+}
+
 /**
  * Return the inner HTML of a balanced <div> whose opening tag starts at $lt.
  * Correctly handles nested <div> elements. Shared by the Genius and dumb
@@ -154,15 +194,36 @@ function container_to_text($fragment) {
  * templates, these regexes may need updating.
  */
 
-/** Search dumb for a track; returns the first song result's page path (e.g. "/Artist-title-lyrics"), or '' if none. */
+/**
+ * Search dumb for a track; returns a song result's page path (e.g.
+ * "/Artist-title-lyrics"), or '' if none. Scans the first 5 song hits for
+ * one actually attributed to $artist (see artist_loosely_matches) and
+ * returns that; falls back to the top hit if none match rather than
+ * returning nothing, so an unusual artist-name spelling can't regress
+ * availability entirely.
+ */
 function dumb_search($artist, $title, $timeout) {
     $base = cfg('dumb.url');
     if ($base === '') return '';
-    $query = rawurlencode(trim($artist . ' - ' . $title, ' -'));
+    $query = rawurlencode(trim($artist . ' - ' . strip_feature_credit($title), ' -'));
     $html = http_get($base . '/search?q=' . $query, $timeout);
     if ($html === null) return '';
-    if (!preg_match('/<h2>\s*Songs\s*<\/h2>.*?<a[^>]+id="search-item"[^>]+href="([^"]+)"/is', $html, $m)) return '';
-    return html_entity_decode($m[1], ENT_QUOTES | ENT_HTML5, 'UTF-8');
+
+    if (!preg_match('/<h2>\s*Songs\s*<\/h2>(.*?)(?=<h2>|$)/is', $html, $section)) return '';
+    if (!preg_match_all('/<a[^>]+id="search-item"[^>]+href="([^"]+)"[^>]*>.*?<span>([^<]*)<\/span>/is', $section[1], $hits, PREG_SET_ORDER)) {
+        return '';
+    }
+
+    $fallback = '';
+    $checked = 0;
+    foreach ($hits as $h) {
+        $href = html_entity_decode($h[1], ENT_QUOTES | ENT_HTML5, 'UTF-8');
+        if ($fallback === '') $fallback = $href;
+        if ($checked++ >= 5) break;
+        $hitArtist = html_entity_decode($h[2], ENT_QUOTES | ENT_HTML5, 'UTF-8');
+        if (artist_loosely_matches($hitArtist, $artist)) return $href;
+    }
+    return $fallback;
 }
 
 /**