X-Git-Url: https://git.ao2.it/tweeper.git/blobdiff_plain/1e7a09621f810d54f9badc8bcf3f01a6c6ae2347..2c7f7e5da9efb2905d9c87cb5a9a2d6c057f576c:/src/Tweeper.php?ds=inline diff --git a/src/Tweeper.php b/src/Tweeper.php index e98623b..f79bb08 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) 2013-2018 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 @@ -36,7 +36,7 @@ date_default_timezone_set('UTC'); */ class Tweeper { - private static $userAgent = "Mozilla/5.0"; + private static $userAgent = "APIs-Google (+https://developers.google.com/webmasters/APIs-Google.html)"; private static $maxConnectionTimeout = 5; private static $maxConnectionRetries = 5; @@ -111,24 +111,35 @@ class Tweeper { } } while (curl_errno($ch) == CURLE_OPERATION_TIMEDOUT && ++$attempt < Tweeper::$maxConnectionRetries); + $response_code = curl_getinfo($ch, CURLINFO_RESPONSE_CODE); + if (FALSE === $response_code) { + trigger_error(curl_error($ch), E_USER_WARNING); + return FALSE; + } + + if ($response_code >= 400) { + trigger_error("HTTP reponse code $response_code", E_USER_WARNING); + return FALSE; + } + return $ret; } /** * Get the contents from a URL. */ - private static function getUrlContents($url) { + private static function getUrlContents($url, $user_agent = NULL) { $ch = curl_init($url); - curl_setopt_array($ch, array( + curl_setopt_array($ch, [ CURLOPT_HEADER => FALSE, CURLOPT_CONNECTTIMEOUT => Tweeper::$maxConnectionTimeout, // Follow http redirects to get the real URL. CURLOPT_FOLLOWLOCATION => TRUE, CURLOPT_COOKIEFILE => "", CURLOPT_RETURNTRANSFER => TRUE, - CURLOPT_HTTPHEADER => array('Accept-language: en'), - CURLOPT_USERAGENT => Tweeper::$userAgent, - )); + CURLOPT_HTTPHEADER => ['Accept-language: en'], + CURLOPT_USERAGENT => isset($user_agent) ? $user_agent : Tweeper::$userAgent, + ]); $contents = Tweeper::curlExec($ch); curl_close($ch); @@ -138,17 +149,17 @@ class Tweeper { /** * Get the headers from a URL. */ - private static function getUrlInfo($url) { + private static function getUrlInfo($url, $user_agent = NULL) { $ch = curl_init($url); - curl_setopt_array($ch, array( + curl_setopt_array($ch, [ CURLOPT_HEADER => TRUE, CURLOPT_NOBODY => TRUE, CURLOPT_CONNECTTIMEOUT => Tweeper::$maxConnectionTimeout, // Follow http redirects to get the real URL. CURLOPT_FOLLOWLOCATION => TRUE, CURLOPT_RETURNTRANSFER => TRUE, - CURLOPT_USERAGENT => Tweeper::$userAgent, - )); + CURLOPT_USERAGENT => isset($user_agent) ? $user_agent : Tweeper::$userAgent, + ]); $ret = Tweeper::curlExec($ch); if (FALSE === $ret) { @@ -169,7 +180,7 @@ class Tweeper { * Generate an RSS element. */ public static function generateEnclosure($url) { - $supported_content_types = array( + $supported_content_types = [ "application/octet-stream", "application/ogg", "application/pdf", @@ -188,7 +199,7 @@ class Tweeper { "video/mp4", "video/mpeg", "video/ogg", - ); + ]; $url_info = Tweeper::getUrlInfo($url); if (FALSE === $url_info) { @@ -270,13 +281,13 @@ class Tweeper { $encoder = new XmlEncoder(); $normalizer = new ObjectNormalizer(); - $serializer = new Serializer(array($normalizer), array($encoder)); + $serializer = new Serializer([$normalizer], [$encoder]); - $serializer_options = array( + $serializer_options = [ 'xml_encoding' => "UTF-8", 'xml_format_output' => TRUE, 'xml_root_node_name' => $root_node_name, - ); + ]; $xml_data = $serializer->serialize($data, 'xml', $serializer_options); if (!$xml_data) { @@ -301,13 +312,19 @@ class Tweeper { $data = json_decode($matches[1], $assoc = TRUE); - // The "qe" object contains elements which will result in invalid XML - // element names, so remove it. + // Remove items that may contain elements which can result in invalid XML + // element names (for example names starting with a number). unset($data["qe"]); - - // The "knobs" object contains elements with undefined namespaces, so - // remove it to silence an error message. - unset($data["knobs"]); + unset($data["knobx"]); + unset($data["to_cache"]); + + // Stop here in case Instagram redirected to the login page, this can + // happen when too many consecutive requests have been made from the same + // IP. + if (array_key_exists("LoginAndSignupPage", $data["entry_data"])) { + trigger_error("Cannot open instagram page: redirected to Login page.\n", E_USER_WARNING); + return NULL; + } $json = json_encode($data); @@ -335,7 +352,10 @@ class Tweeper { // If there is a host-specific method to get the XML data, use it! $get_xml_host_method = 'getXml' . Tweeper::toUpperCamelCase($host, '.'); if (method_exists($this, $get_xml_host_method)) { - $xml_data = call_user_func_array(array($this, $get_xml_host_method), array($html)); + $xml_data = call_user_func_array([$this, $get_xml_host_method], [$html]); + if (NULL === $xml_data) { + return NULL; + } $xmlDoc->loadXML($xml_data); } else { @@ -363,7 +383,7 @@ class Tweeper { return NULL; } - $stylesheet_contents = Tweeper::getUrlContents($stylesheet); + $stylesheet_contents = file_get_contents($stylesheet); if (FALSE === $stylesheet_contents) { trigger_error("Cannot open $stylesheet", E_USER_WARNING); return NULL; @@ -394,7 +414,7 @@ class Tweeper { if (TRUE === $validate_scheme) { $scheme = $url["scheme"]; - if (!in_array($scheme, array("http", "https"))) { + if (!in_array($scheme, ["http", "https"])) { trigger_error("unsupported scheme: $scheme", E_USER_WARNING); return NULL; } @@ -423,7 +443,7 @@ class Tweeper { $preprocess_html_host_method = 'preprocessHtml' . Tweeper::toUpperCamelCase($host, '.'); if (method_exists($this, $preprocess_html_host_method)) { - $html = call_user_func_array(array($this, $preprocess_html_host_method), array($html)); + $html = call_user_func_array([$this, $preprocess_html_host_method], [$html]); } $xmlDoc = $this->htmlToXml($html, $host);