// quickly bail out if there is no job for me
$workingQuery = work_getOneForMe();
if ($workingQuery == null) {
die("no jobs for me\n");
}
// configuration
const QUERY_HD_ONLY = true;
const QUERY_RESULTS_COUNT = 100;
const QUERY_SORTING_SEQUENCE = ['relevance', 'viewCount', 'rating'];
const MIN_VIDEO_VIEWS = 2000;
const FORCE_REINDEX = false;
define('VIDEOS_PROCESSED_SET', 'jam_videos_processed');
//define('VIDEOS_INDEXED_SET_NAME', 'jam_videos_indexed');
// create the global objects
require_once 'YTMachine.php';
$ytMachine = new \YTMachine();
//require_once 'IndexMachine_Algolia.php';
//$indexMachine = new \IndexMachine_Algolia(isset($_GET['index']) ? 'yt_' . $_GET['index'] : '');
require_once 'IndexMachine_ElasticSearch.php';
$indexMachine = new \IndexMachine_ElasticSearch(['173.230.144.120:9199']);
// loop until there's work
do {
// let's go with the current query
$outPrefix = 'Q: ' . $workingQuery . ': ';
echo $outPrefix . "started\n";
$someQueries = [$workingQuery];
// search youtube for N queries, for M (3) ordering criteria
/* @var $videoLeads YTVideo[] */
$videoLeads = [];
foreach ($someQueries as $query) {
foreach (QUERY_SORTING_SEQUENCE as $order) {
/**
* @return bool True if successful
*/
public function resolveCaptions()
{
// do it at most once
if ($this->resolvedCaptions) {
return $this->ytCC != null;
}
$this->resolvedCaptions = true;
$this->lastResolveCaptionsIssue = '';
// perform the resolution
// FIXME: SPEED BOTTLENECK (1/second)
$captionsList = YTMachine::getYoutube()->captions->listCaptions('snippet', $this->videoId);
// get all the SRT from this track that match the language
$ytCCs = [];
foreach ($captionsList->getItems() as $item) {
/* @var $item Google_Service_YouTube_Caption */
/* @var $cc Google_Service_YouTube_CaptionSnippet */
$cc = $item->getSnippet();
$ccId = $item->getId();
// Sanity: abort if the returned CC is for a different video
$ccVideoId = $cc->getVideoId();
if ($ccVideoId != $this->videoId) {
if (YT_VIOLENT) {
die('wrong video id, got ' . $ccVideoId . ' expecting ' . $this->videoId);
}
$this->lastResolveCaptionsIssue .= ' i';
continue;
}
// Sanity: check constant attributes, or stop if unexpected
$ccStatus = $cc->getStatus();
if ($ccStatus != 'serving') {
if (YT_VERBOSE) {
echo '{wrong cc status: ' . $ccStatus;
if ($ccStatus == 'failure') {
echo '(' . $cc->getFailureReason() . ')';
}
echo '}';
}
$this->lastResolveCaptionsIssue .= ' s';
continue;
}
// base fetching query
$fetchQuery = 'v=' . $ccVideoId;
// add kind
$ccKind = $cc->getTrackKind();
if ($ccKind == 'standard') {
// nothing to do here
} else {
if ($ccKind == 'ASR') {
// FILTER: TODO: we don't support the ASR format yet, at all. Always fails.
if (YT_VERBOSE) {
echo $ccVideoId . ', skipping ASR tracks (unsupported yet)' . "\n";
}
$this->lastResolveCaptionsIssue .= ' k-asr';
continue;
} else {
if (YT_VIOLENT) {
die('unknown track type ' . $ccKind);
}
$this->lastResolveCaptionsIssue .= ' k-' . $ccKind;
continue;
}
}
// add language
$ccLang = $cc->getLanguage();
if (!empty($ccLang)) {
// FILTER: skip if the language is not what we asked for
if ($ccLang != $this->language) {
// NOTE: in the future we could also stash other languages for later
if (YT_VERBOSE) {
echo $ccVideoId . ', skipping CC for different language: ' . $ccLang . "\n";
}
$this->lastResolveCaptionsIssue .= ' l-' . $ccLang;
continue;
}
$fetchQuery .= '&lang=' . $ccLang;
}
// add 'name'
$ccTrackName = $cc->getName();
if (!empty($ccTrackName)) {
$fetchQuery .= '&name=' . $ccTrackName;
}
// customize the output format. available formats:
// srv1: <text start="2.501" dur="3.671">
// srv2: <timedtext><window t="0" id="1" op="define" rc="15" cc="32" ap="7" ah="50" av="95"/><text w="2" t="5538" d="2536">RE</text>
// srv3: <timedtext format="3"><p t="2501" d="3671" w="2"><s>TH</s><s t="33">E</s>
// sbv: [SubViewer] 0:00:02.501,0:00:06.172 \n THE WASHINGTON CORRESPONDENT
// srt: [SubRip ] 1 \n 00:00:02,501 --> 00:00:06,172 \n THE WASHINGTON CORRESPONDENT
// ttml: [TTML ] <p begin="00:00:02.501" end="00:00:06.172" region="r3" style="s2"><span begin="00:00:00.000">TH</span>
// vtt: [WebVTT ] 00:00:02.501 --> 00:00:06.172 align:start position:0% line:7% \n THE WASHINGTON CORRESPONDENT
$fetchQuery .= '&fmt=srv1';
// Fetch the Caption from cache
$ccString = CacheMachine::retrieveValue('cc_' . $fetchQuery);
// Fetch the Caption (and expect a 200:OK code)
if ($ccString == null) {
try {
$response = YTMachine::getGuzzle()->get('https://www.youtube.com/api/timedtext?' . $fetchQuery);
if ($response->getStatusCode() != 200) {
if (YT_VIOLENT) {
die('wrong status code ' . $response->getStatusCode() . ' on ' . $fetchQuery);
}
$this->lastResolveCaptionsIssue .= ' h1-' . $response->getStatusCode();
continue;
}
$ccString = $response->getBody()->getContents();
CacheMachine::storeValue('cc_' . $fetchQuery, $ccString, null);
} catch (\GuzzleHttp\Exception\ClientException $exception) {
if (YT_VIOLENT) {
die('HTTP request failed: ' . $exception);
}
$this->lastResolveCaptionsIssue .= ' h2-"' . $exception->getMessage() . '"';
continue;
}
}
// FILTER: Size Heuristic (FIXME): reject semi-empty captions (usually with not much more than the title)
$ccStringSize = $ccString != null ? strlen($ccString) : 0;
if ($ccStringSize < self::SUB_OK_THRESHOLD) {
if (YT_VERBOSE) {
echo $ccVideoId . ', skipping for small size: ' . $ccStringSize . "\n";
}
$this->lastResolveCaptionsIssue .= ' S-' . $ccStringSize;
continue;
}
// FILTER: parse and validate XML
try {
$ccTranscript = new SimpleXMLElement($ccString);
if ($ccTranscript->getName() != 'transcript') {
if (YT_VIOLENT) {
die('expected a transcript root element, got a ' . $ccTranscript->getName() . ' instead ');
}
$this->lastResolveCaptionsIssue .= ' x-1';
continue;
}
} catch (Exception $e) {
if (YT_VERBOSE) {
echo 'skipping for xml parsing error' . $ccVideoId . "\n";
}
$this->lastResolveCaptionsIssue .= ' x-2';
continue;
}
// Break a SRT into individual Lines
$lines = [];
$maxLength = 0;
foreach ($ccTranscript->text as $line) {
$text = $this->fixSrv1Caption(strval($line));
// fix quoted strings
if (strlen($text) > YT_MIN_VALID_CHARS && substr($text, 0, 1) == '"' && substr($text, -1) == '"') {
$text = substr($text, 1, -1);
}
// skip lines with less than YT_MIN_VALID_CHARS chars
$textLength = strlen($text);
if ($textLength < YT_MIN_VALID_CHARS || $textLength > YT_MAX_VALID_CHARS) {
continue;
}
// FILTER: videos that have start but not duration are usually 1-liners
$attributes = $line->attributes();
$start = $attributes['start'];
$duration = $attributes['dur'];
if (empty($start) || empty($duration)) {
if (YT_VERBOSE) {
echo 'skipping cc line for start or duration empty on ' . $ccVideoId . " body: " . $ccString . "\n";
}
continue;
}
if (floatval($duration) > YT_MAX_VALID_DURATION) {
if (YT_VERBOSE) {
echo 'skipping cc line for duration ' . $ccVideoId . "\n";
}
continue;
}
// skip "( ... )" strings, since parenthesis are not real speech
if (substr($text, 0, 1) == '(' && substr($text, -1) == ')') {
if (YT_VERBOSE) {
echo 'skipping cc line for (parenthesis) ' . $ccVideoId . "\n";
}
continue;
}
// add the line
array_push($lines, ['t' => $text, 's' => floatval($start), 'd' => floatval($duration), 'e' => floatval($start) + floatval($duration)]);
if ($textLength > $maxLength) {
$maxLength = $textLength;
}
}
// FILTER: almost-empty docs, or docs with at most 3 letters per line
if ($maxLength < YT_MIN_VALID_CHARS || sizeof($lines) < YT_MIN_VALID_LINES) {
if (YT_VERBOSE) {
echo 'skipping for emptiness ' . sizeof($lines) . " lines and " . $maxLength . " max chars per line\n";
}
$this->lastResolveCaptionsIssue .= ' e';
continue;
}
// save the fully-fetched caption
array_push($ytCCs, new YTCC($ccId, $ccVideoId, $ccTrackName, $ccStringSize, $ccString, $lines, $cc->getLastUpdated()));
}
// use just best caption amongst those available, chosen by size
usort($ytCCs, function ($a, $b) {
return $b->ccSize - $a->ccSize;
});
// pick the best (if any), or null
$this->ytCC = empty($ytCCs) ? null : $ytCCs[0];
return $this->ytCC != null;
}
