miniflux-legacy/vendor/PicoFeed/Grabber.php
2013-10-03 23:14:39 -04:00

271 lines
7.4 KiB
PHP

<?php
namespace PicoFeed;
require_once __DIR__.'/Client.php';
require_once __DIR__.'/Encoding.php';
require_once __DIR__.'/Logging.php';
require_once __DIR__.'/Filter.php';
class Grabber
{
public $content = '';
public $html = '';
public $encoding = '';
// Order is important, generic terms at the end
public $candidatesAttributes = array(
'articleBody',
'articlebody',
'article-body',
'articleContent',
'articlecontent',
'article-content',
'articlePage',
'post-content',
'post_content',
'entry-content',
'main-content',
'story_content',
'storycontent',
'entryBox',
'entrytext',
'comic',
'post',
'article',
'content',
'main',
);
public $stripAttributes = array(
'comment',
'share',
'links',
'toolbar',
'fb',
'footer',
'credit',
'bottom',
'nav',
'header',
'social',
'tag',
'metadata',
'entry-utility',
'related-posts',
'tweet',
'categories',
);
public $stripTags = array(
'script',
'style',
'nav',
'header',
'footer',
'aside',
'form',
);
public function __construct($url, $html = '', $encoding = 'utf-8')
{
$this->url = $url;
$this->html = $html;
$this->encoding = $encoding;
}
public function parse()
{
if ($this->html) {
Logging::log(\get_called_class().' Fix encoding');
Logging::log(\get_called_class().': HTTP Encoding "'.$this->encoding.'"');
$this->html = Filter::stripMetaTags($this->html);
if ($this->encoding == 'windows-1251') {
$this->html = Encoding::cp1251ToUtf8($this->html);
}
else {
$this->html = Encoding::toUTF8($this->html);
}
Logging::log(\get_called_class().' Try to find rules');
$rules = $this->getRules();
if (is_array($rules)) {
Logging::log(\get_called_class().' Parse content with rules');
$this->parseContentWithRules($rules);
}
else {
Logging::log(\get_called_class().' Parse content with candidates');
$this->parseContentWithCandidates();
}
}
else {
Logging::log(\get_called_class().' No content fetched');
}
Logging::log(\get_called_class().' Content length: '.strlen($this->content).' bytes');
Logging::log(\get_called_class().' Grabber done');
return $this->content !== '';
}
public function download($timeout = 5, $user_agent = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_8_4) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/29.0.1547.62 Safari/537.36')
{
$client = Client::create();
$client->url = $this->url;
$client->timeout = $timeout;
$client->user_agent = $user_agent;
$client->execute();
$this->html = $client->getContent();
return $this->html;
}
public function getRules()
{
$hostname = parse_url($this->url, PHP_URL_HOST);
$files = array($hostname);
if (substr($hostname, 0, 4) == 'www.') {
$files[] = substr($hostname, 4);
}
if (($pos = strpos($hostname, '.')) !== false) {
$files[] = substr($hostname, $pos);
$files[] = substr($hostname, 0, $pos);
}
foreach ($files as $file) {
$filename = __DIR__.'/Rules/'.$file.'.php';
if (file_exists($filename)) {
return include $filename;
}
}
return false;
}
public function parseContentWithRules(array $rules)
{
\libxml_use_internal_errors(true);
$dom = new \DOMDocument;
$dom->loadHTML('<?xml version="1.0" encoding="UTF-8">'.$this->html);
$xpath = new \DOMXPath($dom);
if (isset($rules['strip']) && is_array($rules['strip'])) {
foreach ($rules['strip'] as $pattern) {
$nodes = $xpath->query($pattern);
if ($nodes !== false && $nodes->length > 0) {
foreach ($nodes as $node) {
$node->parentNode->removeChild($node);
}
}
}
}
if (isset($rules['body']) && is_array($rules['body'])) {
foreach ($rules['body'] as $pattern) {
$nodes = $xpath->query($pattern);
if ($nodes !== false && $nodes->length > 0) {
foreach ($nodes as $node) {
$this->content .= $dom->saveXML($node);
}
}
}
}
}
public function parseContentWithCandidates()
{
\libxml_use_internal_errors(true);
$dom = new \DOMDocument;
$dom->loadHTML('<?xml version="1.0" encoding="UTF-8">'.$this->html);
$xpath = new \DOMXPath($dom);
// Try to lookup in each tag
foreach ($this->candidatesAttributes as $candidate) {
Logging::log(\get_called_class().' Try this candidate: "'.$candidate.'"');
$nodes = $xpath->query('//*[(contains(@class, "'.$candidate.'") or @id="'.$candidate.'") and not (contains(@class, "nav") or contains(@class, "page"))]');
if ($nodes !== false && $nodes->length > 0) {
$this->content = $dom->saveXML($nodes->item(0));
Logging::log(\get_called_class().' Find candidate "'.$candidate.'" ('.strlen($this->content).' bytes)');
break;
}
}
// Try to fetch <article/>
if (! $this->content) {
$nodes = $xpath->query('//article');
if ($nodes !== false && $nodes->length > 0) {
$this->content = $dom->saveXML($nodes->item(0));
Logging::log(\get_called_class().' Find <article/> tag ('.strlen($this->content).' bytes)');
}
}
if (strlen($this->content) < 50) {
Logging::log(\get_called_class().' No enought content fetched, get the full body');
$this->content = $dom->saveXML($dom->firstChild);
}
Logging::log(\get_called_class().' Strip garbage');
$this->stripGarbage();
}
public function stripGarbage()
{
\libxml_use_internal_errors(true);
$dom = new \DOMDocument;
$dom->loadXML($this->content);
$xpath = new \DOMXPath($dom);
foreach ($this->stripTags as $tag) {
$nodes = $xpath->query('//'.$tag);
if ($nodes !== false && $nodes->length > 0) {
Logging::log(\get_called_class().' Strip tag: "'.$tag.'"');
foreach ($nodes as $node) {
$node->parentNode->removeChild($node);
}
}
}
foreach ($this->stripAttributes as $attribute) {
$nodes = $xpath->query('//*[contains(@class, "'.$attribute.'") or contains(@id, "'.$attribute.'")]');
if ($nodes !== false && $nodes->length > 0) {
Logging::log(\get_called_class().' Strip attribute: "'.$tag.'"');
foreach ($nodes as $node) {
$node->parentNode->removeChild($node);
}
}
}
$this->content = $dom->saveXML($dom->documentElement);
}
}