public static function exec($url)
{
Log::log("Starting download of the url <" . $url . ">");
// Download the source
$source = Downloader::get($url);
if ($source == false) {
Log::log("No source found. Exiting now.");
exit;
}
Crawler::crawl((string) $source);
/*$cat = [];
if( @!!"" ){
echo "true";
}else{
echo "false";
}*/
}
public static function get_archive($arc_key)
{
$verbose = Config::$env == "dev" ? true : false;
Log::log("<blue>Downloading archive #" . $arc_key . " - " . self::$archives[$arc_key]['text'] . " - " . self::$archives[$arc_key]['href']);
$crawler;
$valid = false;
do {
$source = Downloader::get(self::$archives[$arc_key]['href']);
//$source = Downloader::get( "http://localhost:8000" );
$crawler = new DomCrawler((string) $source);
$crawler = $crawler->filter('.base_pagina_articoli > div > a');
$count = count($crawler);
Log::log("<blue>Download ended.");
if ($count == 0) {
Log::log("<red>Found 0 articles. Why? Retrying to download.");
} else {
$valid = true;
}
sleep(1);
} while (!$valid);
if ($verbose) {
Log::log("<green>The download was successfull.");
Log::log("<cyan>Found " . $count . " posts in the archive.");
}
foreach ($crawler as $art_key => $domElement) {
if ($art_key >= Config::$status_post) {
if ($verbose) {
Log::log("<cyan>Article ID #" . $art_key . " of archive #" . $arc_key . ".");
}
Config::$status_post = 0;
$href = $domElement->getAttribute('href');
$text = trim($domElement->nodeValue);
$text = explode("\n", $text);
$title = \utf8_decode(htmlentities(trim($text[0])));
$date = [0, 0, 0, 0, 0, 0];
preg_match("#(\\d+)\\/(\\d+)\\/(\\d+)\\s\\-\\s(\\d+):(\\d+)#is", $text[1], $date);
$date = Carbon::create($date[3], $date[2], $date[1], $date[4], $date[5]);
if ($verbose) {
Log::log("<blue>[" . $date . "] - " . $title . " - " . $href);
}
$valid = false;
$article = null;
$categories = [];
do {
$source = Downloader::get($href);
$crawl = new DomCrawler((string) $source);
preg_match("#articolo\\/(.+?)\\/\\d+#is", $href, $match);
$name = trim($match[1]);
if ($verbose) {
Log::log("<darkGreen>" . $name);
}
$article = $crawl->filter('.dettagli_articoli_riassunto > p');
if (count($article) == 0) {
Log::log("<red>Cannot get the article #" . $art_key . ". Retrying.");
} else {
foreach ($article as $art) {
$article = \utf8_decode(htmlentities(trim($art->nodeValue)));
//var_dump( mb_detect_encoding( $art->nodeValue ) );
if ($verbose) {
Log::log("<darkGreen>" . $article);
}
break;
}
}
$cats = $crawl->filter('.dettagli_articoli_famiglie > a');
$categories = [];
foreach ($cats as $cat) {
$cat = $cat->nodeValue;
if (@(!!trim($cat))) {
$categories[] = $cat;
}
}
Log::log("<darkGreen>Categories: " . (@(!!$categories) ? "none" : implode(", ", $categories)));
if (@(!!$article)) {
$valid = true;
}
} while (!$valid);
$post = new Post($date, $article, $title, $name, @(!!$categories) ? $categories : false);
$post->save();
self::$count++;
if ($verbose) {
Log::log("<red>Count: " . self::$count);
}
if (Config::$post_end !== false && self::$count >= Config::$post_end) {
Log::log("<red>Ended.");
exit;
}
}
}
}
