public static function main()
{
Log::open_file(Carbon::now()->format("F-Y-g-i-A") . ".log");
Database::connect();
Crawler::crawl();
//echo htmlentities( \utf8_decode( "(AGIELLE) - Galvagni (Cisl): Milano è una pentola a pressione, reagire subito - RPT" ) );
/* $query = "SELECT ID as 'identification', post_title FROM wp_posts";
$exec = Database::exec( $query );
while( $res = $exec->fetch_assoc() ){
//Log::log( $res[ 'post_title' ] );
$old = $res[ 'post_title' ];
$new = htmlentities( $old );
$query = "UPDATE wp_posts SET post_title = '".Database::escape( @!!$new ? $new : $old )."' WHERE id = ".$res[ 'identification' ];
Log::log( "<red>".$query );
Database::exec( $query );
//exit();
Log::log( "Updated post #".$res[ 'identification' ] );
}*/
/*$post = new Post( Carbon::now(), "Wassup bruh, dis is the contnt", "Taitol", "wassup-bruh1", [ "New Cat 1", "New Cat 2" ] );
$post->save();
$post = new Post( Carbon::now(), "Wassup bruh, dis is the contnt", "Taitol", "wassup-bruh2", [ "New Cat 2", "New Cat 3" ] );
$post->save();*/
Log::close_file();
}
public static function filter($root, $parent, $nodes, $verbose = true)
{
$eligible = [];
$filtered = [];
$ncount = count($nodes);
$invalid = 0;
foreach (self::$identifiers as $name => $id) {
foreach ($nodes as $node_key => $node) {
$identifier = call_user_func($id['class'] . '::' . $id['method'], $node);
if (@(!!$identifier)) {
$valid = true;
foreach (self::$filters as $filter) {
if ($filter['for'] == $id['name'] || $filter['for'] == '*') {
$valid = call_user_func($filter['class'] . "::" . $filter['method'], $node) ? $valid : false;
// Add filtered info into node
}
}
if (@(!!$valid)) {
$found = false;
foreach ($filtered as $filtered_key => $filtered_value) {
if ($filtered_value['pattern'] == $identifier) {
$found = $filtered_key;
}
}
if ($found !== false) {
$filtered[$found]['nodes'][] = $node;
} else {
$filtered[] = ['pattern' => $identifier, 'id_by' => $id['method'], "id" => ['name' => $id['name'], 'value' => $id['prefix'] . $identifier . $id['postfix']], 'nodes' => [$node]];
}
} else {
$invalid++;
}
}
}
}
foreach ($filtered as $filtered_value) {
$nodes = $filtered_value['nodes'];
$nodes_count = count($nodes);
$pattern = $filtered_value['pattern'];
$id = $filtered_value['id'];
if ($nodes_count > 3) {
// needs to be changed
if ($verbose) {
Log::log("\tThere are " . $nodes_count . " nodes with " . strtoupper($id['name']) . " pattern <" . $id['value'] . "> eligible.");
}
$eligible[] = ['count' => $nodes_count, 'pattern' => $pattern, 'id_by' => $filtered_value['id_by'], 'nodes' => $nodes];
}
}
if ($verbose && $invalid > 0) {
//Log::log( "\t".$invalid."/".$ncount." nodes invalidated by filters." ); // Needs to be corrected, because $invalid is incremented by 1 each $filter on each $identifier
}
if (count($eligible) > 0) {
return true;
} else {
return false;
}
}
public static function crawl($source)
{
$crawler = new DomCrawler($source);
Log::log("Generating tree..");
$node_tree = Parser::parse($crawler, 0, "", false);
Log::log("Calculating components..");
$augmented_tree = Parser::augment($node_tree[0]);
Log::log("Calculating path..");
$marked_tree = Parser::identify_parent_path($augmented_tree);
file_put_contents("a-tree", print_r($marked_tree, true));
Log::log("Scanning the tree for a pattern..");
Scanner::scan($marked_tree);
// Test di funzionamento delle funzioni get_node_from_path e get_parent_from_path
//Log::log( Tree::get_node_from_path( $marked_tree, "0,0" )[ 'tag' ] );
//Log::log( Tree::get_parent_from_path( $marked_tree, "0,0" )[ 'tag' ] );
}
public static function connect()
{
$verbose = Config::$env == "dev" ? true : false;
if ($verbose) {
Log::log("<blue>Connecting to MySQL " . Config::$db_user . "@" . Config::$db_host . ".. ");
}
$mysqli = @new \mysqli(Config::$db_host, Config::$db_user, Config::$db_pass, Config::$db_name);
if ($mysqli->connect_errno) {
Log::log("<red>[" . $mysqli->connect_errno . "] " . $mysqli->connect_error);
exit;
} else {
self::$conn = $mysqli;
if ($verbose) {
Log::log("<green>Connected.");
}
}
}
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 scan($node)
{
// The tweaker has to be applied BEFORE the scan
Tweaker::apply();
$priority = Scanner::scan_for_pattern($node);
Log::log("Tweaking results..");
//$priority = Tweaker::apply( $priority );
// Readable test
$prio = [];
foreach ($priority as $node) {
unset($node['children']);
unset($node['node']);
$prio[] = $node;
}
$priority = $prio;
// End readable test
// Test
Log::log("Printing results.");
file_put_contents("priority", print_r($priority, true));
}
public static function get($url)
{
$client = new Guzzle();
$res = null;
$tries = 0;
$error = null;
do {
try {
$res = $client->request('GET', $url, ['timeout' => 10, 'http_errors' => false]);
} catch (\GuzzleHttp\Exception\ConnectException $e) {
$error = "ConnectException: " . $e->getMessage();
} catch (\GuzzleHttp\Exception\RequestException $e) {
$error = "RequestException: " . $e->getMessage();
}
$tries++;
} while ((!$res || $res->getStatusCode() == 200) && $tries < 10);
if ($res == null) {
Log::log($error);
}
return $res != null ? $res->getBody() : false;
}
public static function deletion($priority)
{
for ($i = 0; $i < count($priority); $i++) {
$node = $priority[$i];
$unset = false;
foreach (self::$blacklist['parent'] as $parent) {
if ($node['tag'] == $parent) {
unset($priority[$i]);
$unset = true;
}
}
foreach (self::$blacklist['child'] as $child) {
$old_children = $node['children'];
$old_count = count($old_children);
for ($a = 0; $a < count($old_children); $a++) {
$node_child = $old_children[$a];
if (@(!!$node_child['tag']) && $node_child['tag'] == $child) {
unset($old_children[$a]);
}
}
if (!$unset) {
$new_children = $old_children;
if (count($new_children) > 0) {
if (count($new_children) != $old_count) {
Log::log("\tUnset " . ($old_count - count($new_children)) . " children from [" . $node['tag'] . "][" . ($i - 1) . "]");
}
unset($priority[$i]['children']);
$priority[$i]['children'] = $new_children;
} else {
Log::log("\tCompletely unset [" . $node['tag'] . "][" . ($i - 1) . "]");
unset($priority[$i]);
}
}
}
}
return $priority;
}
public function save_id()
{
$verbose = Config::$env == "dev" ? true : false;
$id = null;
if ($verbose) {
Log::log("<blue>Creating the category `" . $this->name . "`..");
}
$slug = Database::escape(strtolower(App::clean($this->name)));
$query = "INSERT INTO `" . Config::$db_name . "`.`" . Config::$wp_prefix . "terms`( `name`, `slug`, `term_group` ) VALUE( '" . Database::escape($this->name) . "', '" . $slug . "', 0 )";
$exec = Database::exec($query);
if ($exec === false) {
Log::log("<red>Cannot save the category:");
Log::log("<red>[" . Database::$conn->errno . "] " . Database::$conn->error);
Log::log("<red>" . $query);
exit;
} else {
$term_id = Database::$conn->insert_id;
if ($verbose) {
Log::log("<green>Created the category `" . $this->name . "` with the slug `" . $slug . "` and ID '" . $term_id . "'.");
}
}
$query = "INSERT INTO `" . Config::$db_name . "`.`" . Config::$wp_prefix . "term_taxonomy`( `term_id`, `taxonomy`, `description`, `parent`, `count`) VALUE( '" . $term_id . "', 'category', '', 0, 0 );";
$exec = Database::exec($query);
if ($exec === false) {
Log::log("<red>Cannot save the taxonomy:");
Log::log("<red>[" . Database::$conn->errno . "] " . Database::$conn->error);
Log::log("<red>" . $query);
exit;
} else {
if ($verbose) {
Log::log("<green>Taxonomy created.");
}
$tax_id = Database::$conn->insert_id;
}
return [$term_id, $tax_id];
}
public static function get_archive($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::$post_end) {
Log::log("<red>Ended.");
exit;
}
}
}
}
public static function dump($obj, $depth = 0, $exp = "")
{
if ($depth === true) {
$exp = $depth;
$depth = 0;
}
if (is_bool($exp) && $exp === true) {
$exp = "";
}
/*if( !$exp )
usleep( 10000 );*/
foreach ($obj as $key => $value) {
$string = str_repeat(" ", $depth * 3);
if (is_object($obj)) {
$string .= "[" . $key . ":" . get_class($obj) . "] => ";
} else {
$string .= "[" . $key . "] => ";
}
if (is_object($value)) {
$string .= get_class($value);
if (!is_string($exp)) {
$string = "<darkYellow>" . $string;
}
if (!is_string($exp)) {
Log::log($string);
} else {
$exp .= PHP_EOL . $string;
}
$exp = self::dump($value, $depth + 1, $exp);
} else {
if (is_array($value)) {
$string .= "Array";
if (@(!$value)) {
$string .= "()";
if (!$exp) {
$string = "<red>" . $string;
}
} else {
if (!$exp) {
$string = "<blue>" . $string;
}
}
if (!is_string($exp)) {
Log::log($string);
} else {
$exp .= PHP_EOL . $string;
}
$exp = self::dump($value, $depth + 1, $exp);
} else {
if (is_bool($value)) {
if ($value) {
if (!is_string($exp)) {
$string .= "<darkGreen>bool(true)";
} else {
$string .= "bool(true)";
}
} else {
if (!is_string($exp)) {
$string .= "<darkGreen>bool(false)";
} else {
$string .= "bool(false)";
}
}
} else {
if (is_string($value)) {
$string .= "string(" . $value . ")";
}
}
if (!is_string($exp)) {
Log::log($string);
} else {
$exp .= PHP_EOL . $string;
}
}
}
}
return $exp;
}
<?php
use Scraper\Middleware\Log;
use Scraper\Controllers\Main;
ini_set("max_execution_time", "900000");
require 'vendor/autoload.php';
Log::log("<red>Starting main controller.");
Main::main();
public function save()
{
$verbose = Config::$env == "dev" ? true : false;
$post_save_query = $this->get_query();
$post_saved = Database::exec($post_save_query);
if ($post_saved === false) {
Log::log("<red>Cannot save the article:");
Log::log("<red>[" . Database::$conn->errno . "] " . Database::$conn->error);
Log::log("<red>" . $post_save_query);
exit;
}
$this->post_id = Database::$conn->insert_id;
if ($verbose) {
Log::log("<green>Article with ID [" . $this->post_id . "] saved.");
}
foreach ($this->category as $cat_key => $cat) {
$cat->check_exists();
$relation_query = "INSERT INTO `" . Config::$db_name . "`.`" . Config::$wp_prefix . "term_relationships` ( `object_id`, `term_taxonomy_id`, `term_order` ) VALUE ( " . $this->post_id . ", " . $cat->tax_id . ", 0 )";
$relation_exec = Database::exec($relation_query);
if ($relation_exec === false) {
Log::log("<red>Cannot save the link between the article and the taxonomy:");
Log::log("<red>[" . Database::$conn->errno . "] " . Database::$conn->error);
Log::log("<red>" . $relation_query);
exit;
}
$count_query = "UPDATE `" . Config::$db_name . "`.`" . Config::$wp_prefix . "term_taxonomy` SET count = count +1 WHERE `term_taxonomy_id` = " . $cat->tax_id . ";";
$count_exec = Database::exec($count_query);
if ($count_exec === false) {
Log::log("<red>Cannot add the article to the counter:");
Log::log("<red>[" . Database::$conn->errno() . "] " . Database::$conn->error);
Log::log("<red>" . $count_query);
exit;
}
}
Log::log("<green>Article #" . $this->post_id . " saved with its categories.");
}
public static function parse($node, $deep = 0, $xPath = "", $verbose = false)
{
$ret_obj = [];
$spaces = str_repeat(" ", $deep * 5);
$count = count($node);
$elements_count = [];
if ($count > 0) {
foreach ($node as $domElement) {
$node_obj = [];
$nodeName = $domElement->nodeName;
$nodeValue = $domElement->nodeValue;
$tagName = isset($domElement->tagName) ? $domElement->tagName : false;
$parentNode = $domElement->parentNode;
$was_verbose = $verbose;
if (Parser::get_id($domElement) == "g-items-atf") {
//$verbose = true;
}
// Generation of an absolute xPath
$myPath = "";
if (isset($tagName)) {
if (!isset($elements_count[$tagName])) {
$elements_count[$tagName] = 1;
} else {
$elements_count[$tagName]++;
}
$myPath = $xPath . ($xPath == "" ? "descendant-or-self::" : "/") . $tagName . (Parser::tags_count($node, $tagName) > 1 ? "[position()=" . $elements_count[$tagName] . "]" : "");
}
// Logging node by name with a tree structure
$log_string = "[" . $deep . "]" . (strlen($deep . "") < 2 ? " " : "");
if ($deep == 0) {
$log_string .= $nodeName;
} else {
if ($deep > 1) {
$tmp_string = "|";
$tdeep = 0;
while ($tdeep < $deep - 1) {
$tmp_string .= str_repeat(" ", 4) . "|";
$tdeep++;
}
$log_string .= $tmp_string;
} else {
if ($deep == 1) {
$log_string .= "|";
}
}
$log_string .= str_repeat("-", 4) . $nodeName;
}
if (get_class($domElement) == "DOMElement" && @(!!$domElement->getAttribute('id'))) {
$log_string .= "[#" . $domElement->getAttribute('id') . "]";
}
//$log_string .= "[".get_class( $domElement )."]";
if ($verbose) {
Log::log($log_string);
}
$node_obj = ['name' => $nodeName, 'value' => "", 'parent' => "", 'node' => $domElement];
if ($tagName !== false) {
$node_obj['tag'] = $tagName;
}
$children = Parser::parse($domElement->childNodes, $deep + 1, $myPath, $verbose);
if (!!$children) {
$node_obj['children'] = $children;
}
if ($myPath != "") {
$node_obj['xPath'] = $myPath;
}
$ret_obj[] = $node_obj;
}
return $ret_obj;
} else {
return false;
}
}
