X-Git-Url: https://git.ao2.it/tweeper.git/blobdiff_plain/b187bb677361d17a468abf749332d081a194b4bd..66097964b2c241e20eae5af87fb4a14fd2c1d889:/src/Tweeper.php?ds=inline diff --git a/src/Tweeper.php b/src/Tweeper.php index 73cbe81..7b2654d 100644 --- a/src/Tweeper.php +++ b/src/Tweeper.php @@ -6,7 +6,7 @@ namespace Tweeper; * @file * Tweeper - a Twitter to RSS web scraper. * - * Copyright (C) 2013-2015 Antonio Ospite + * Copyright (C) 2013-2020 Antonio Ospite * * This program is free software: you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -25,8 +25,6 @@ namespace Tweeper; use DOMDocument; use XSLTProcessor; -require_once 'Symfony/Component/Serializer/autoload.php'; - use Symfony\Component\Serializer\Serializer; use Symfony\Component\Serializer\Encoder\XmlEncoder; use Symfony\Component\Serializer\Normalizer\ObjectNormalizer; @@ -38,13 +36,31 @@ date_default_timezone_set('UTC'); */ class Tweeper { - private static $userAgent = "Mozilla/5.0 (Windows NT 6.1; rv:22.0) Gecko/20130405 Firefox/22.0"; + private static $userAgent = "Mozilla/5.0 (Linux; U; Android 4.3; en-us; SM-N900T Build/JSS15J)"; + private static $maxConnectionTimeout = 5; + private static $maxConnectionRetries = 5; /** - * Constructor sets up {@link $generate_enclosure}. + * Create a new Tweeper object controlling optional settings. + * + * @param bool $generate_enclosure + * Enables the creation of elements (disabled by default). + * @param bool $show_usernames + * Enables showing the username in front of the content for multi-user + * sites (enabled by default). Only some stylesheets supports this + * functionality (twitter, instagram, pump.io). + * @param bool $show_multimedia + * Enables showing multimedia content (images, videos) directly in the + * item description (enabled by default). Only some stylesheets supports + * this functionality (twitter, instagram, dilbert). + * @param bool $verbose_output + * Enables showing non-fatal errors like XML parsing errors. */ - public function __construct($generate_enclosure = FALSE) { + public function __construct($generate_enclosure = FALSE, $show_usernames = TRUE, $show_multimedia = TRUE, $verbose_output = TRUE) { $this->generate_enclosure = $generate_enclosure; + $this->show_usernames = $show_usernames; + $this->show_multimedia = $show_multimedia; + $this->verbose_output = $verbose_output; } /** @@ -71,6 +87,46 @@ class Tweeper { } /** + * Convert Twitter mobile date to the date format expected in a RSS document. + */ + public static function twitterToRssDate($date) { + // Twitter uses relative timestamps in minutes for recent tweets. + if (preg_match('/^(\d+)m$/', $date, $matches)) { + $timestamp = strtotime("+" . $matches[1] . " min", time()); + if (FALSE === $timestamp) { + $timestamp = 0; + } + } + else { + /* + * In case the time is specified put it after the date, + * to make it recognized by strptime(). + */ + if (preg_match('/(.*) - (.*)/', $date, $matches)) { + $date = $matches[2] . " " . $matches[1]; + } + + $timestamp = strtotime($date); + if (FALSE === $timestamp) { + $timestamp = 0; + } + + /* + * The twitter mobile UI usually only specifies the month and the day, so + * strtotime($date) may interpret the date as being in the future. + * + * If the date is in the future it is probably in the same day but in the + * previous year. + */ + if ($timestamp > time()) { + $timestamp = strtotime('-1 years', $timestamp); + } + } + + return Tweeper::epochToRssDate($timestamp); + } + + /** * Convert string to UpperCamelCase. */ public static function toUpperCamelCase($str, $delim = ' ') { @@ -80,24 +136,40 @@ class Tweeper { } /** + * Perform a cURL session multiple times when it fails with a timeout. + * + * @param resource $ch + * a cURL session handle. + */ + private static function curlExec($ch) { + $ret = FALSE; + $attempt = 0; + do { + $ret = curl_exec($ch); + if (FALSE === $ret) { + trigger_error(curl_error($ch), E_USER_WARNING); + } + } while (curl_errno($ch) == CURLE_OPERATION_TIMEDOUT && ++$attempt < Tweeper::$maxConnectionRetries); + + return $ret; + } + + /** * Get the contents from a URL. */ private static function getUrlContents($url) { $ch = curl_init($url); curl_setopt_array($ch, array( CURLOPT_HEADER => FALSE, + CURLOPT_CONNECTTIMEOUT => Tweeper::$maxConnectionTimeout, // Follow http redirects to get the real URL. CURLOPT_FOLLOWLOCATION => TRUE, + CURLOPT_COOKIEFILE => "", CURLOPT_RETURNTRANSFER => TRUE, - CURLOPT_SSL_VERIFYHOST => FALSE, - CURLOPT_SSL_VERIFYPEER => FALSE, CURLOPT_HTTPHEADER => array('Accept-language: en'), CURLOPT_USERAGENT => Tweeper::$userAgent, )); - $contents = curl_exec($ch); - if (FALSE === $contents) { - trigger_error(curl_error($ch)); - } + $contents = Tweeper::curlExec($ch); curl_close($ch); return $contents; @@ -111,17 +183,22 @@ class Tweeper { curl_setopt_array($ch, array( CURLOPT_HEADER => TRUE, CURLOPT_NOBODY => TRUE, + CURLOPT_CONNECTTIMEOUT => Tweeper::$maxConnectionTimeout, // Follow http redirects to get the real URL. CURLOPT_FOLLOWLOCATION => TRUE, CURLOPT_RETURNTRANSFER => TRUE, - CURLOPT_SSL_VERIFYHOST => FALSE, - CURLOPT_SSL_VERIFYPEER => FALSE, CURLOPT_USERAGENT => Tweeper::$userAgent, )); - curl_exec($ch); + + $ret = Tweeper::curlExec($ch); + if (FALSE === $ret) { + curl_close($ch); + return FALSE; + } + $url_info = curl_getinfo($ch); if (FALSE === $url_info) { - trigger_error(curl_error($ch)); + trigger_error(curl_error($ch), E_USER_WARNING); } curl_close($ch); @@ -154,10 +231,14 @@ class Tweeper { ); $url_info = Tweeper::getUrlInfo($url); + if (FALSE === $url_info) { + trigger_error("Failed to retrieve info for URL: " . $url, E_USER_WARNING); + return ''; + } $supported = in_array($url_info['content_type'], $supported_content_types); if (!$supported) { - error_log("Unsupported enclosure content type \"" . $url_info['content_type'] . "\" for URL: " . $url_info['url']); + trigger_error("Unsupported enclosure content type \"" . $url_info['content_type'] . "\" for URL: " . $url_info['url'], E_USER_WARNING); return ''; } @@ -165,10 +246,17 @@ class Tweeper { // See http://sourceforge.net/p/feedvalidator/bugs/72/ $http_url = preg_replace("/^https/", "http", $url_info['url']); + // When the server does not provide a Content-Length header, + // curl_getinfo() would return a negative value for + // "download_content_length", however RSS recommends to use 0 when the + // enclosure's size cannot be determined. + // See: https://www.feedvalidator.org/docs/error/UseZeroForUnknown.html + $length = max($url_info['download_content_length'], 0); + $dom = new DOMDocument(); $enc = $dom->createElement('enclosure'); $enc->setAttribute('url', $http_url); - $enc->setAttribute('length', $url_info['download_content_length']); + $enc->setAttribute('length', $length); $enc->setAttribute('type', $url_info['content_type']); return $enc; @@ -205,7 +293,7 @@ class Tweeper { $output .= " line $error->line"; - error_log($output); + trigger_error($output, E_USER_WARNING); } /** @@ -232,7 +320,7 @@ class Tweeper { $xml_data = $serializer->serialize($data, 'xml', $serializer_options); if (!$xml_data) { - trigger_error("Cannot serialize data", E_USER_ERROR); + trigger_error("Cannot serialize data", E_USER_WARNING); return NULL; } @@ -247,11 +335,23 @@ class Tweeper { $json_match_expr = '/window._sharedData = (.*);/'; $ret = preg_match($json_match_expr, $html, $matches); if ($ret !== 1) { - trigger_error("Cannot match expression: $json_match_expr\n", E_USER_ERROR); + trigger_error("Cannot match expression: $json_match_expr\n", E_USER_WARNING); return NULL; } - return Tweeper::jsonToXml($matches[1], 'instagram'); + $data = json_decode($matches[1], $assoc = TRUE); + + // The "qe" object contains elements which will result in invalid XML + // element names, so remove it. + unset($data["qe"]); + + // The "knobs" object contains elements with undefined namespaces, so + // remove it to silence an error message. + unset($data["knobs"]); + + $json = json_encode($data); + + return Tweeper::jsonToXml($json, 'instagram'); } /** @@ -282,8 +382,10 @@ class Tweeper { $xmlDoc->loadHTML($html); } - foreach (libxml_get_errors() as $xml_error) { - Tweeper::logXmlError($xml_error); + if ($this->verbose_output) { + foreach (libxml_get_errors() as $xml_error) { + Tweeper::logXmlError($xml_error); + } } libxml_clear_errors(); libxml_use_internal_errors($xml_errors_value); @@ -297,11 +399,15 @@ class Tweeper { private function loadStylesheet($host) { $stylesheet = "file://" . __DIR__ . "/rss_converter_" . $host . ".xsl"; if (FALSE === file_exists($stylesheet)) { - trigger_error("Conversion to RSS not supported for $host ($stylesheet not found)", E_USER_ERROR); + trigger_error("Conversion to RSS not supported for $host ($stylesheet not found)", E_USER_WARNING); return NULL; } $stylesheet_contents = Tweeper::getUrlContents($stylesheet); + if (FALSE === $stylesheet_contents) { + trigger_error("Cannot open $stylesheet", E_USER_WARNING); + return NULL; + } $xslDoc = new DOMDocument(); $xslDoc->loadXML($stylesheet_contents); @@ -309,6 +415,8 @@ class Tweeper { $xsltProcessor = new XSLTProcessor(); $xsltProcessor->registerPHPFunctions(); $xsltProcessor->setParameter('', 'generate-enclosure', $this->generate_enclosure); + $xsltProcessor->setParameter('', 'show-usernames', $this->show_usernames); + $xsltProcessor->setParameter('', 'show-multimedia', $this->show_multimedia); $xsltProcessor->importStylesheet($xslDoc); return $xsltProcessor; @@ -317,21 +425,30 @@ class Tweeper { /** * Convert the site content to RSS. */ - public function tweep($src_url) { + public function tweep($src_url, $host = NULL, $validate_scheme = TRUE) { $url = parse_url($src_url); - if (FALSE === $url || empty($url["host"])) { - trigger_error("Invalid URL: $src_url", E_USER_ERROR); + if (FALSE === $url) { + trigger_error("Invalid URL: $src_url", E_USER_WARNING); return NULL; } - $scheme = $url["scheme"]; - if (!in_array($scheme, array("http", "https"))) { - trigger_error("unsupported scheme: $scheme", E_USER_ERROR); - return NULL; + if (TRUE === $validate_scheme) { + $scheme = $url["scheme"]; + if (!in_array($scheme, array("http", "https"))) { + trigger_error("unsupported scheme: $scheme", E_USER_WARNING); + return NULL; + } } - // Strip the leading www. to be more forgiving on input URLs. - $host = preg_replace('/^www\./', '', $url["host"]); + // If the host is not given derive it from the URL. + if (NULL === $host) { + if (empty($url["host"])) { + trigger_error("Invalid host in URL: $src_url", E_USER_WARNING); + return NULL; + } + // Strip the leading www. to be more forgiving on input URLs. + $host = preg_replace('/^www\./', '', $url["host"]); + } $xsltProcessor = $this->loadStylesheet($host); if (NULL === $xsltProcessor) { @@ -340,6 +457,7 @@ class Tweeper { $html = Tweeper::getUrlContents($src_url); if (FALSE === $html) { + trigger_error("Failed to retrieve $src_url", E_USER_WARNING); return NULL; } @@ -354,11 +472,11 @@ class Tweeper { } $output = $xsltProcessor->transformToXML($xmlDoc); - if (FALSE === $output) { - trigger_error('XSL transformation failed.', E_USER_ERROR); + trigger_error('XSL transformation failed.', E_USER_WARNING); return NULL; } + return $output; }