X-Git-Url: https://git.ao2.it/tweeper.git/blobdiff_plain/d38cac35c8e0aed0a223533bb54e46958c693212..19d2c915502213e887ba7fd75c2e6b52400cd6d1:/tweeper.php?ds=sidebyside diff --git a/tweeper.php b/tweeper.php index 45bc786..4be5c7d 100644 --- a/tweeper.php +++ b/tweeper.php @@ -2,7 +2,7 @@ /* * tweeper - a Twitter to RSS web scraper * - * Copyright (C) 2013 Antonio Ospite + * Copyright (C) 2013-2014 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 @@ -20,107 +20,151 @@ date_default_timezone_set('UTC'); -define('USER_AGENT', 'Mozilla/5.0 (Windows NT 6.1; rv:22.0) Gecko/20130405 Firefox/22.0'); +class Tweeper { -function str_to_gmdate($date) -{ - $timestamp = strtotime($date); - return epoch_to_gmdate($timestamp); -} + private static $USER_AGENT = "Mozilla/5.0 (Windows NT 6.1; rv:22.0) Gecko/20130405 Firefox/22.0"; -function epoch_to_gmdate($timestamp) -{ - return gmdate('D, d M Y H:i:s', $timestamp) . ' GMT'; -} + public function __construct($stylesheet, $generate_enclosure = FALSE) { + $stylesheet_contents = $this->get_contents($stylesheet); -function get_url_info($url) -{ - $ch = curl_init($url); - curl_setopt_array($ch, array( + $xslDoc = new DOMDocument(); + $xslDoc->loadXML($stylesheet_contents); + + $this->xsltProcessor = new XSLTProcessor(); + $this->xsltProcessor->registerPHPFunctions(); + $this->xsltProcessor->setParameter('', 'generateEnclosure', $generate_enclosure); + $this->xsltProcessor->importStylesheet($xslDoc); + } + + public static function epoch_to_gmdate($timestamp) + { + return gmdate('D, d M Y H:i:s', $timestamp) . ' GMT'; + } + + public static function str_to_gmdate($date) + { + $timestamp = strtotime($date); + return Tweeper::epoch_to_gmdate($timestamp); + } + + private static function get_contents($url) + { + $ch = curl_init($url); + curl_setopt_array($ch, array( + CURLOPT_HEADER => FALSE, + CURLOPT_FOLLOWLOCATION => TRUE, // follow http redirects to get the real URL + CURLOPT_RETURNTRANSFER => TRUE, + CURLOPT_SSL_VERIFYHOST => FALSE, + CURLOPT_SSL_VERIFYPEER => FALSE, + CURLOPT_HTTPHEADER => array('Accept-language: en'), + CURLOPT_USERAGENT => Tweeper::$USER_AGENT, + )); + $contents = curl_exec($ch); + curl_close($ch); + + return $contents; + } + + private static function get_info($url) + { + $ch = curl_init($url); + curl_setopt_array($ch, array( CURLOPT_HEADER => TRUE, CURLOPT_NOBODY => TRUE, CURLOPT_FOLLOWLOCATION => TRUE, // follow http redirects to get the real URL CURLOPT_RETURNTRANSFER => TRUE, CURLOPT_SSL_VERIFYHOST => FALSE, CURLOPT_SSL_VERIFYPEER => FALSE, - CURLOPT_USERAGENT => USER_AGENT, - )); - curl_exec($ch); - $url_info = curl_getinfo($ch);; - curl_close($ch); - - return $url_info; -} + CURLOPT_USERAGENT => Tweeper::$USER_AGENT, + )); + curl_exec($ch); + $url_info = curl_getinfo($ch); + curl_close($ch); -function generate_enclosure($url) -{ - $url_info = get_url_info($url); - - switch($url_info['content_type']) { - case "video/avi": - case "image/gif": - case "image/jpeg": - case "audio/x-midi": - case "audio/mpeg": - case "video/mpeg": - case "audio/vorbis": - case "application/ogg": - case "audio/webm": - case "audio/aac": - case "audio/mp4": - case "audio/wav": - case "video/mp4": - case "video/ogg": - $dom = new DomDocument(); - $enc = $dom->createElement('enclosure'); - $enc->setAttribute( 'url', $url_info['url']); - $enc->setAttribute( 'length', $url_info['download_content_length']); - $enc->setAttribute( 'type', $url_info['content_type']); - - $dom->appendChild($enc); - - return $dom->saveXML($enc); - default: - return; + return $url_info; } -} + public static function generate_enclosure($url) + { + $supported_content_types = array( + "application/ogg", + "audio/aac", + "audio/mp4", + "audio/mpeg", + "audio/ogg", + "audio/vorbis", + "audio/wav", + "audio/webm", + "audio/x-midi", + "image/gif", + "image/jpeg", + "video/avi", + "video/mp4", + "video/mpeg", + "video/ogg", + ); -class Tweeper { + $url_info = Tweeper::get_info($url); - public function __construct($stylesheet) { - $stylesheet_contents = $this->get_contents($stylesheet); + $supported = in_array($url_info['content_type'], $supported_content_types); + if (!$supported) { + error_log("Unsupported enclosure content type \"" . $url_info['content_type'] . "\" for URL: " . $url_info['url']); + return ''; + } - $xslDoc = new DOMDocument(); - $xslDoc->loadXML($stylesheet_contents); + $dom = new DomDocument(); + $enc = $dom->createElement('enclosure'); + $enc->setAttribute('url', $url_info['url']); + $enc->setAttribute('length', $url_info['download_content_length']); + $enc->setAttribute('type', $url_info['content_type']); - $this->xsltProcessor = new XSLTProcessor(); - $this->xsltProcessor->registerPHPFunctions(); - $this->xsltProcessor->importStylesheet($xslDoc); + $dom->appendChild($enc); + + return $dom->saveXML($enc); } - private function get_contents($uri) { - # https://www.wjsams.com/c/docs/Wiki/Php-HowToSetUserAgentOnFileGetContents - $opts = array( - 'http' => array( - 'method' => "GET", - 'header' => join(array( - "Accept-language: en\r\n", - "User-Agent: {" + USER_AGENT + "}\r\n" - )) - ) - ); + /* Mimic the message from libxml.c::php_libxml_ctx_error_level() */ + private function log_xml_error($error) { + $output = ""; + + switch ($error->level) { + case LIBXML_ERR_WARNING: + $output .= "Warning $error->code: "; + break; + case LIBXML_ERR_ERROR: + $output .= "Error $error->code: "; + break; + case LIBXML_ERR_FATAL: + $output .= "Fatal Error $error->code: "; + break; + } - $context = stream_context_create($opts); - $contents = file_get_contents($uri, false, $context); - return $contents; + $output .= trim($error->message); + + if ($error->file) { + $output .= " in $error->file"; + } else { + $output .= " in Entity,"; + } + + $output .=" line $error->line"; + + error_log($output); } public function tweep($uri) { - $html = $this->get_contents($uri); + $html = Tweeper::get_contents($uri); $xmlDoc = new DOMDocument(); + + // Handle warnings and errors when loading invalid HTML. + $xml_errors_value = libxml_use_internal_errors(true); $xmlDoc->loadHTML($html); + foreach (libxml_get_errors() as $xml_error) { + $this->log_xml_error($xml_error); + } + libxml_clear_errors(); + libxml_use_internal_errors($xml_errors_value); $output = $this->xsltProcessor->transformToXML($xmlDoc); @@ -135,29 +179,35 @@ class Tweeper { function usage($argv) { if (php_sapi_name() != 'cli') - $usage = htmlentities("{$_SERVER['SCRIPT_NAME']}?src_url="); + $usage = htmlentities("{$_SERVER['SCRIPT_NAME']}?src_url=&generate_enclosure=<0|1>"); else - $usage = "{$argv[0]} [-h|--help] \n"; + $usage = "{$argv[0]} [-e|-h|--help] \n"; return "usage: $usage"; } function parse_options_cli($argv, $argc) { - $options = array(); + $options = array( + 'generate_enclosure' => FALSE + ); if ($argc < 2) return $options; - $cli_options = getopt("h", array("help")); + $cli_options = getopt("eh", array("help")); foreach ($cli_options as $opt => $val) { switch ($opt) { + case 'e': + $options['generate_enclosure'] = TRUE; + break; case 'h': case 'help': echo usage($argv); exit(0); default: - die(usage($argv)); + fwrite(STDERR, usage($argv)); + exit(1); } } @@ -168,30 +218,44 @@ function parse_options_cli($argv, $argc) function parse_options_query_string() { - $options = array(); + $options = array( + 'generate_enclosure' => FALSE + ); if (isset($_GET['src_url'])) $options['src_url'] = $_GET['src_url']; + if (isset($_GET['generate_enclosure'])) + $options['generate_enclosure'] = $_GET['generate_enclosure'] == 1; + return $options; } -if (php_sapi_name() != 'cli') +if (php_sapi_name() != 'cli') { $options = parse_options_query_string(); -else + $ERROR_STREAM = fopen('php://output', 'w'); +} else { $options = parse_options_cli($argv, $argc); + $ERROR_STREAM = fopen('php://stderr', 'w'); +} -if (!isset($options['src_url'])) - die(usage($argv)); +if (!isset($options['src_url'])) { + fwrite($ERROR_STREAM, usage($argv)); + exit(1); +} $url = parse_url($options['src_url']); -if (FALSE === $url || empty($url["host"])) - die("Invalid url: ${options['src_url']}\n"); +if (FALSE === $url || empty($url["host"])) { + fwrite($ERROR_STREAM, "Invalid url: ${options['src_url']}\n"); + exit(1); +} -$stylesheet = __DIR__ . "/rss_converter_" . $url["host"] . ".xsl"; -if (FALSE === file_exists($stylesheet)) - die("Conversion to RSS not supported: {$url["host"]}\n"); +$stylesheet = "file://" . __DIR__ . "/rss_converter_" . $url["host"] . ".xsl"; +if (FALSE === file_exists($stylesheet)) { + fwrite($ERROR_STREAM, "Conversion to RSS not supported: {$url["host"]}\n"); + exit(1); +} -$tweeper = new Tweeper($stylesheet); +$tweeper = new Tweeper($stylesheet, $options['generate_enclosure']); echo $tweeper->tweep($options['src_url']);