radio/api/lyrics.php

<?php
/**
 * api/lyrics.php
 * --------------
 * Lyrics lookup, tried in order: LRCLIB (free, no key, has synced lyrics),
 * then dumb (if USE_DUMB is on), then scraping the public Genius song page
 * (if a Genius token is configured). Each tier only runs if the previous one
 * found nothing, so dumb and Genius can both be configured at once. Returns
 * JSON:
 *
 *   {
 *     "found":       true|false,
 *     "lyrics":      "plain text\nwith newlines",
 *     "synced":      [{ "time": 12.34, "text": "…" }, …],   // [] when unavailable
 *     "url":         "…",
 *     "source":      "lrclib" | "dumb" | "genius",
 *     "annotations": [{ "text": "…", "path": "…" }, …]      // [] unless source is "dumb"
 *   }
 *
 * Lyrics are returned as plain text (tags/timestamps stripped) so the
 * front-end can render them with textContent — no HTML injection from a
 * third-party page. `synced` (line-level LRC timestamps, seconds) is only
 * populated when LRCLIB has synced lyrics; the Genius fallback has no timing
 * data so it's always [] there. `url` is '' for LRCLIB (the front-end only
 * shows a "source" link for dumb/genius). `annotations` lists Genius
 * annotation links found within the lyrics (dumb only — the front-end
 * splices a 📝 marker into each matching span and fetches the annotation
 * body from api/annotation.php on click).
 *
 * Note: the Genius fallback scrapes HTML, which can change at any time. If
 * Genius restructures their page the extraction may need updating.
 */

require_once __DIR__ . '/_bootstrap.php';

$empty = ['found' => false, 'lyrics' => '', 'synced' => [], 'url' => '', 'annotations' => []];

if (!cfg('lyrics.enabled')) {
    json_out($empty);
}

[$artist, $title] = request_track();
if ($artist === '' && $title === '') {
    json_out($empty);
}

$timeout = 5;

/**
 * Parse an LRC string into [{time, text}, …] sorted by time. A line can carry
 * more than one leading "[mm:ss.xx]" tag (repeated text, e.g. a chorus reused
 * at several timestamps) — each tag becomes its own entry. Non-timestamp
 * metadata lines ("[ar:...]", "[ti:...]", …) don't match and are skipped.
 */
function parse_lrc($lrc) {
    $out = [];
    foreach (explode("\n", $lrc) as $line) {
        if (!preg_match_all('/\[(\d+):(\d+(?:\.\d+)?)\]/', $line, $tags, PREG_SET_ORDER)) continue;
        $text = trim((string) preg_replace('/\[\d+:\d+(?:\.\d+)?\]/', '', $line));
        if ($text === '') continue;
        foreach ($tags as $tag) {
            $out[] = ['time' => ((int) $tag[1]) * 60 + (float) $tag[2], 'text' => $text];
        }
    }
    usort($out, function ($a, $b) { return $a['time'] <=> $b['time']; });
    return $out;
}

/* -- LRCLIB (primary, no key required) -------------------------------------- */
function lrclib_lyrics($artist, $title, $timeout) {
    $url = 'https://lrclib.net/api/search?track_name=' . rawurlencode($title)
         . '&artist_name=' . rawurlencode($artist);
    $raw = http_get($url, $timeout);
    if ($raw === null) return null;

    $hits = json_decode($raw, true);
    if (!is_array($hits) || count($hits) === 0) return null;

    $synced = !empty($hits[0]['syncedLyrics']) ? parse_lrc((string) $hits[0]['syncedLyrics']) : [];
    $lyrics = trim((string) ($hits[0]['plainLyrics'] ?? ''));
    if ($lyrics === '' && !empty($synced)) {
        $lyrics = implode("\n", array_map(function ($l) { return $l['text']; }, $synced));
    }
    return $lyrics === '' ? null : ['lyrics' => $lyrics, 'synced' => $synced];
}

$lrc = lrclib_lyrics($artist, $title, $timeout);
if ($lrc !== null) {
    json_out(['found' => true, 'lyrics' => $lrc['lyrics'], 'synced' => $lrc['synced'], 'url' => '', 'source' => 'lrclib', 'annotations' => []]);
}

/* -- dumb (tried before Genius; falls through to it if nothing found) ------- */
if (cfg('dumb.enabled')) {
    $path = dumb_search($artist, $title, $timeout);
    if ($path !== '') {
        $song = dumb_song($path, $timeout);
        if ($song !== null && $song['lyrics'] !== '') {
            json_out([
                'found' => true, 'lyrics' => $song['lyrics'], 'synced' => [],
                'url' => cfg('dumb.url') . $path, 'source' => 'dumb',
                'annotations' => $song['annotations'],
            ]);
        }
    }
    // Nothing on dumb — fall through to the Genius API below.
}

/* -- Genius (fallback) ------------------------------------------------------- */
if (cfg('genius.token') === '') json_out($empty);

$token   = cfg('genius.token');
$headers = ['Authorization: Bearer ' . $token];

/* -- Find the song on Genius ----------------------------------------------- */
$query  = rawurlencode(trim($artist . ' - ' . $title, ' -'));
$search = http_get('https://api.genius.com/search?q=' . $query, $timeout, $headers);
if ($search === null) json_out($empty);

$json = json_decode($search, true);
$hits = dot_get($json, 'response.hits', []);
if (!is_array($hits) || count($hits) === 0) json_out($empty);

$pageUrl = (string) dot_get($hits, '0.result.url', '');

// Optionally prefer the first English-language match (a few extra API calls).
if (cfg('lyrics.prefer_english')) {
    $checked = 0;
    foreach ($hits as $h) {
        if ($checked++ >= 5) break;
        $id = dot_get($h, 'result.id', null);
        if (!$id) continue;
        $songRaw = http_get('https://api.genius.com/songs/' . rawurlencode((string) $id), $timeout, $headers);
        if ($songRaw === null) continue;
        if (dot_get(json_decode($songRaw, true), 'response.song.language', '') === 'en') {
            $pageUrl = (string) dot_get($h, 'result.url', $pageUrl);
            break;
        }
    }
}
if ($pageUrl === '') json_out($empty);

/* -- Scrape the lyric containers ------------------------------------------- */
$page = http_get($pageUrl, $timeout);
if ($page === null) {
    json_out(['found' => false, 'lyrics' => '', 'synced' => [], 'url' => $pageUrl, 'source' => 'genius', 'annotations' => []]);
}

$marker = 'data-lyrics-container="true"';
$pieces = [];
$offset = 0;
while (($mpos = strpos($page, $marker, $offset)) !== false) {
    // Backtrack to the start of this opening <div ...> tag.
    $lt = strrpos(substr($page, 0, $mpos), '<div');
    if ($lt === false) { $offset = $mpos + strlen($marker); continue; }
    $inner = inner_balanced_div($page, $lt);
    $text  = trim(container_to_text($inner));
    if ($text !== '') $pieces[] = $text;
    $offset = $mpos + strlen($marker);
}

$lyrics = trim(implode("\n\n", $pieces));
// Trim stray indentation left on each line, then collapse blank-line runs.
$lyrics = implode("\n", array_map('trim', explode("\n", $lyrics)));
$lyrics = preg_replace("/\n{3,}/", "\n\n", $lyrics);

if ($lyrics === '') {
    json_out(['found' => false, 'lyrics' => '', 'synced' => [], 'url' => $pageUrl, 'source' => 'genius', 'annotations' => []]);
}

json_out(['found' => true, 'lyrics' => $lyrics, 'synced' => [], 'url' => $pageUrl, 'source' => 'genius', 'annotations' => []]);