miniflux-legacy/vendor/PicoFeed/Parser.php

505 lines
13 KiB
PHP
Raw Normal View History

2013-02-18 03:48:21 +01:00
<?php
namespace PicoFeed;
2014-05-20 20:20:27 +02:00
use DateTime;
use DateTimeZone;
use DOMXPath;
use SimpleXMLElement;
use PicoFeed\Config;
use PicoFeed\Encoding;
use PicoFeed\Filter;
use PicoFeed\Grabber;
use PicoFeed\Logging;
use PicoFeed\XmlParser;
2013-02-18 03:48:21 +01:00
2014-03-17 02:35:57 +01:00
/**
* Base parser class
*
* @author Frederic Guillot
* @package parser
*/
2013-02-18 03:48:21 +01:00
abstract class Parser
{
2014-05-20 20:20:27 +02:00
/**
* Config object
*
* @access private
* @var \PicoFeed\Config
*/
private $config = null;
2014-03-17 02:35:57 +01:00
/**
* Hash algorithm used to generate item id, any value supported by PHP, see hash_algos()
*
2014-05-20 20:20:27 +02:00
* @access private
2014-03-17 02:35:57 +01:00
* @var string
*/
2014-05-20 20:20:27 +02:00
private $hash_algo = 'crc32b'; // crc32b seems to be faster and shorter than other hash algorithms
/**
* Timezone used to parse feed dates
*
* @access private
* @var string
*/
private $timezone = 'UTC';
2014-03-17 02:35:57 +01:00
/**
* Feed content (XML data)
*
* @access protected
* @var string
*/
2013-02-18 03:48:21 +01:00
protected $content = '';
2014-03-17 02:35:57 +01:00
/**
2014-05-20 20:20:27 +02:00
* XML namespaces
2014-03-17 02:35:57 +01:00
*
2014-05-20 20:20:27 +02:00
* @access protected
* @var array
2014-03-17 02:35:57 +01:00
*/
2014-05-20 20:20:27 +02:00
protected $namespaces = array();
2014-03-17 02:35:57 +01:00
/**
2014-05-20 20:20:27 +02:00
* Enable the content grabber
2014-03-17 02:35:57 +01:00
*
2014-05-20 20:20:27 +02:00
* @access private
* @var bool
2014-03-17 02:35:57 +01:00
*/
2014-05-20 20:20:27 +02:00
public $enable_grabber = false;
2013-02-18 03:48:21 +01:00
2014-03-17 02:35:57 +01:00
/**
2014-05-20 20:20:27 +02:00
* Ignore those urls for the content scraper
2014-03-17 02:35:57 +01:00
*
2014-05-20 20:20:27 +02:00
* @access private
* @var array
2014-03-17 02:35:57 +01:00
*/
2014-05-20 20:20:27 +02:00
private $grabber_ignore_urls = array();
2013-02-18 03:48:21 +01:00
2014-03-17 02:35:57 +01:00
/**
* Constructor
*
* @access public
* @param string $content Feed content
* @param string $http_encoding HTTP encoding (headers)
*/
2013-10-04 05:14:39 +02:00
public function __construct($content, $http_encoding = '')
2013-02-18 03:48:21 +01:00
{
2014-05-25 14:47:03 +02:00
$xml_encoding = XmlParser::getEncodingFromXmlTag($content);
2013-10-04 05:14:39 +02:00
// Strip XML tag to avoid multiple encoding/decoding in the next XML processing
$this->content = Filter::stripXmlTag($content);
2013-06-01 20:57:17 +02:00
// Encode everything in UTF-8
2014-05-25 14:47:03 +02:00
Logging::setMessage(get_called_class().': HTTP Encoding "'.$http_encoding.'" ; XML Encoding "'.$xml_encoding.'"');
$this->content = Encoding::convert($this->content, $xml_encoding ?: $http_encoding);
2013-06-01 20:57:17 +02:00
// Workarounds
$this->content = $this->normalizeData($this->content);
2013-02-18 03:48:21 +01:00
}
2014-05-20 20:20:27 +02:00
/**
* Parse the document
*
* @access public
* @return mixed \PicoFeed\Feed instance or false
*/
public function execute()
{
Logging::setMessage(get_called_class().': begin parsing');
$xml = XmlParser::getSimpleXml($this->content);
if ($xml === false) {
Logging::setMessage(get_called_class().': XML parsing error');
Logging::setMessage(XmlParser::getErrors());
return false;
}
$this->namespaces = $xml->getNamespaces(true);
$feed = new Feed;
$this->findFeedUrl($xml, $feed);
$this->findFeedTitle($xml, $feed);
$this->findFeedLanguage($xml, $feed);
$this->findFeedId($xml, $feed);
$this->findFeedDate($xml, $feed);
foreach ($this->getItemsTree($xml) as $entry) {
$item = new Item;
$this->findItemAuthor($xml, $entry, $item);
$this->findItemUrl($entry, $item);
$this->findItemTitle($entry, $item);
$this->findItemId($entry, $item, $feed);
$this->findItemDate($entry, $item);
$this->findItemContent($entry, $item);
$this->findItemEnclosure($entry, $item, $feed);
$this->findItemLanguage($entry, $item, $feed);
$feed->items[] = $item;
}
Logging::setMessage(get_called_class().PHP_EOL.$feed);
return $feed;
}
2014-03-17 02:35:57 +01:00
/**
* Filter HTML for entry content
*
* @access public
* @param string $item_content Item content
* @param string $item_url Item URL
* @return string Filtered content
*/
public function filterHtml($item_content, $item_url)
2013-02-18 03:48:21 +01:00
{
$content = '';
2014-05-20 20:20:27 +02:00
// Setup the content scraper
if ($this->enable_grabber && ! in_array($item_url, $this->grabber_ignore_urls)) {
$grabber = new Grabber($item_url);
2014-05-20 20:20:27 +02:00
$grabber->setConfig($this->config);
$grabber->download();
if ($grabber->parse()) {
$item_content = $grabber->getContent();
}
}
2013-02-18 03:48:21 +01:00
2014-05-20 20:20:27 +02:00
// Content filtering
if ($item_content) {
2013-02-18 03:48:21 +01:00
2014-05-20 20:20:27 +02:00
if ($this->config !== null) {
2013-02-18 03:48:21 +01:00
2014-05-20 20:20:27 +02:00
$callback = $this->config->getContentFilteringCallback();
2014-05-20 20:20:27 +02:00
if (is_callable($callback)) {
$content = $callback($item_content, $item_url);
}
}
2013-02-18 03:48:21 +01:00
2014-05-20 20:20:27 +02:00
if (! $content) {
$filter = new Filter($item_content, $item_url);
$filter->setConfig($this->config);
$content = $filter->execute();
}
2013-02-18 03:48:21 +01:00
}
2014-05-20 20:20:27 +02:00
return $content;
2013-02-18 03:48:21 +01:00
}
2014-03-17 02:35:57 +01:00
/**
* Dirty quickfixes before XML parsing
*
* @access public
* @param string $data Raw data
* @return string Normalized data
*/
2013-04-05 05:34:07 +02:00
public function normalizeData($data)
2013-02-18 03:48:21 +01:00
{
$invalid_chars = array(
"\x10",
"\xc3\x20",
"&#x1F;",
);
foreach ($invalid_chars as $needle) {
$data = str_replace($needle, '', $data);
}
2014-02-01 20:52:33 +01:00
$data = $this->replaceEntityAttribute($data);
return $data;
}
2014-03-17 02:35:57 +01:00
/**
* Replace & by &amp; for each href attribute (Fix broken feeds)
*
* @access public
* @param string $content Raw data
* @return string Normalized data
*/
2014-02-01 20:52:33 +01:00
public function replaceEntityAttribute($content)
{
$content = preg_replace_callback('/href="[^"]+"/', function(array $matches) {
return htmlspecialchars($matches[0], ENT_NOQUOTES, 'UTF-8', false);
}, $content);
return $content;
2013-02-18 03:48:21 +01:00
}
2013-06-01 20:57:17 +02:00
2014-03-17 02:35:57 +01:00
/**
* Trim whitespace from the begining, the end and inside a string and don't break utf-8 string
*
* @access public
* @param string $value Raw data
* @return string Normalized data
*/
public function stripWhiteSpace($value)
{
$value = str_replace("\r", "", $value);
$value = str_replace("\t", "", $value);
$value = str_replace("\n", "", $value);
return trim($value);
}
2014-03-17 02:35:57 +01:00
/**
* Generate a unique id for an entry (hash all arguments)
*
* @access public
* @param string $args Pieces of data to hash
* @return string Id
*/
public function generateId()
{
2014-05-20 20:20:27 +02:00
return hash($this->hash_algo, implode(func_get_args()));
}
2013-10-24 00:39:21 +02:00
2014-03-17 02:35:57 +01:00
/**
* Try to parse all date format for broken feeds
*
* @access public
* @param string $value Original date format
* @return integer Timestamp
*/
2013-10-24 00:39:21 +02:00
public function parseDate($value)
{
// Format => truncate to this length if not null
$formats = array(
DATE_ATOM => null,
DATE_RSS => null,
DATE_COOKIE => null,
DATE_ISO8601 => null,
DATE_RFC822 => null,
DATE_RFC850 => null,
DATE_RFC1036 => null,
DATE_RFC1123 => null,
DATE_RFC2822 => null,
DATE_RFC3339 => null,
'D, d M Y H:i:s' => 25,
'D, d M Y h:i:s' => 25,
'D M d Y H:i:s' => 24,
'Y-m-d H:i:s' => 19,
'Y-m-d\TH:i:s' => 19,
'd/m/Y H:i:s' => 19,
'D, d M Y' => 16,
'Y-m-d' => 10,
'd-m-Y' => 10,
'm-d-Y' => 10,
'd.m.Y' => 10,
'm.d.Y' => 10,
'd/m/Y' => 10,
'm/d/Y' => 10,
);
$value = trim($value);
foreach ($formats as $format => $length) {
$truncated_value = $value;
if ($length !== null) {
$truncated_value = substr($truncated_value, 0, $length);
}
$timestamp = $this->getValidDate($format, $truncated_value);
if ($timestamp > 0) {
return $timestamp;
}
2013-10-24 00:39:21 +02:00
}
2014-05-20 20:20:27 +02:00
$date = new DateTime('now', new DateTimeZone($this->timezone));
return $date->getTimestamp();
2013-10-24 00:39:21 +02:00
}
2014-03-17 02:35:57 +01:00
/**
* Get a valid date from a given format
*
* @access public
* @param string $format Date format
* @param string $value Original date value
* @return integer Timestamp
*/
2013-10-24 00:39:21 +02:00
public function getValidDate($format, $value)
{
2014-05-20 20:20:27 +02:00
$date = DateTime::createFromFormat($format, $value, new DateTimeZone($this->timezone));
2013-10-24 00:39:21 +02:00
if ($date !== false) {
2014-05-20 20:20:27 +02:00
$errors = DateTime::getLastErrors();
if ($errors['error_count'] === 0 && $errors['warning_count'] === 0) {
return $date->getTimestamp();
}
2013-10-24 00:39:21 +02:00
}
return 0;
}
2013-12-18 02:56:53 +01:00
2014-03-17 02:35:57 +01:00
/**
* Hardcoded list of hostname/token to exclude from id generation
*
* @access public
* @param string $url URL
* @return boolean
*/
2013-12-18 02:56:53 +01:00
public function isExcludedFromId($url)
{
2014-02-01 20:52:33 +01:00
$exclude_list = array('ap.org', 'jacksonville.com');
2013-12-18 02:56:53 +01:00
foreach ($exclude_list as $token) {
if (strpos($url, $token) !== false) return true;
}
return false;
}
2014-03-17 02:35:57 +01:00
/**
* Get xml:lang value
*
* @access public
* @param string $xml XML string
* @return string Language
*/
public function getXmlLang($xml)
{
2014-05-20 20:20:27 +02:00
$dom = XmlParser::getDomDocument($this->content);
2014-03-17 02:35:57 +01:00
2014-05-20 20:20:27 +02:00
if ($dom === false) {
return '';
}
$xpath = new DOMXPath($dom);
2014-03-17 02:35:57 +01:00
return $xpath->evaluate('string(//@xml:lang[1])') ?: '';
}
/**
* Return true if the given language is "Right to Left"
*
* @static
* @access public
* @param string $language Language: fr-FR, en-US
* @return bool
*/
public static function isLanguageRTL($language)
{
$language = strtolower($language);
2014-05-20 20:20:27 +02:00
$rtl_languages = array(
'ar', // Arabic (ar-**)
'fa', // Farsi (fa-**)
'ur', // Urdu (ur-**)
'ps', // Pashtu (ps-**)
'syr', // Syriac (syr-**)
'dv', // Divehi (dv-**)
'he', // Hebrew (he-**)
'yi', // Yiddish (yi-**)
);
2014-03-17 02:35:57 +01:00
2014-05-20 20:20:27 +02:00
foreach ($rtl_languages as $prefix) {
if (strpos($language, $prefix) === 0) {
return true;
}
}
2014-03-17 02:35:57 +01:00
2014-05-20 20:20:27 +02:00
return false;
}
2014-03-17 02:35:57 +01:00
2014-05-20 20:20:27 +02:00
/**
* Set Hash algorithm used for id generation
*
* @access public
* @param string $algo Algorithm name
* @return \PicoFeed\Parser
*/
public function setHashAlgo($algo)
{
$this->hash_algo = $algo ?: $this->hash_algo;
return $this;
}
2014-03-17 02:35:57 +01:00
2014-05-20 20:20:27 +02:00
/**
* Set a different timezone
*
* @see http://php.net/manual/en/timezones.php
* @access public
* @param string $timezone Timezone
* @return \PicoFeed\Parser
*/
public function setTimezone($timezone)
{
$this->timezone = $timezone ?: $this->timezone;
return $this;
}
2014-03-17 02:35:57 +01:00
2014-05-20 20:20:27 +02:00
/**
* Set config object
*
* @access public
* @param \PicoFeed\Config $config Config instance
* @return \PicoFeed\Parser
*/
public function setConfig($config)
{
$this->config = $config;
return $this;
}
2014-03-17 02:35:57 +01:00
2014-05-20 20:20:27 +02:00
/**
* Enable the content grabber
*
* @access public
* @return \PicoFeed\Parser
*/
public function enableContentGrabber()
{
$this->enable_grabber = true;
}
2014-03-17 02:35:57 +01:00
2014-05-20 20:20:27 +02:00
/**
* Set ignored URLs for the content grabber
*
* @access public
* @param array $urls URLs
* @return \PicoFeed\Parser
*/
public function setGrabberIgnoreUrls(array $urls)
{
$this->grabber_ignore_urls = $urls;
}
2014-03-17 02:35:57 +01:00
2014-05-20 20:20:27 +02:00
/**
* Get a value from a XML namespace
*
* @access public
* @param SimpleXMLElement $xml XML element
* @param array $namespaces XML namespaces
* @param string $property XML tag name
* @param string $attribute XML attribute name
2014-05-20 20:20:27 +02:00
* @return string
*/
public function getNamespaceValue(SimpleXMLElement $xml, array $namespaces, $property, $attribute = '')
2014-05-20 20:20:27 +02:00
{
foreach ($namespaces as $name => $url) {
$namespace = $xml->children($namespaces[$name]);
if ($namespace->$property->count() > 0) {
if ($attribute) {
foreach ($namespace->$property->attributes() as $xml_attribute => $xml_value) {
if ($xml_attribute === $attribute && $xml_value) {
return (string) $xml_value;
}
}
}
2014-05-20 20:20:27 +02:00
return (string) $namespace->$property;
}
}
return '';
2014-03-17 02:35:57 +01:00
}
2013-02-18 03:48:21 +01:00
}