src/Tweeper.php: only override the User-Agent to a mobile one for twitter.com

[tweeper.git] / src / Tweeper.php

diff --git a/src/Tweeper.php b/src/Tweeper.php
index d60e43b..f1d579f 100644 (file)
--- a/src/Tweeper.php
+++ b/src/Tweeper.php
@@ -6,7 +6,7 @@ namespace Tweeper;
  * @file
  * Tweeper - a Twitter to RSS web scraper.
  *
  * @file
  * Tweeper - a Twitter to RSS web scraper.
  *

- * Copyright (C) 2013-2016  Antonio Ospite <ao2@ao2.it>
+ * Copyright (C) 2013-2020  Antonio Ospite <ao2@ao2.it>

  *
  * 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
  *
  * 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


@@ -36,13 +36,31 @@ date_default_timezone_set('UTC');
  */
 class Tweeper {
 
  */
 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";
+  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 <enclosure/> 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->generate_enclosure = $generate_enclosure;

+    $this->show_usernames = $show_usernames;
+    $this->show_multimedia = $show_multimedia;
+    $this->verbose_output = $verbose_output;

   }
 
   /**
   }
 
   /**


@@ -69,6 +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, $delim = ' ') {
    * Convert string to UpperCamelCase.
    */
   public static function toUpperCamelCase($str, $delim = ' ') {


@@ -78,24 +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.
    */
    * Get the contents from a URL.
    */

-  private static function getUrlContents($url) {
+  private static function getUrlContents($url, $user_agent = NULL) {

     $ch = curl_init($url);
     $ch = curl_init($url);

-    curl_setopt_array($ch, array(
+    curl_setopt_array($ch, [

       CURLOPT_HEADER => FALSE,
       CURLOPT_HEADER => FALSE,

+      CURLOPT_CONNECTTIMEOUT => Tweeper::$maxConnectionTimeout,

       // Follow http redirects to get the real URL.
       CURLOPT_FOLLOWLOCATION => TRUE,
       // Follow http redirects to get the real URL.
       CURLOPT_FOLLOWLOCATION => TRUE,

+      CURLOPT_COOKIEFILE => "",

       CURLOPT_RETURNTRANSFER => TRUE,
       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));
-    }
+      CURLOPT_HTTPHEADER => ['Accept-language: en'],
+      CURLOPT_USERAGENT => isset($user_agent) ? $user_agent : Tweeper::$userAgent,
+    ]);
+    $contents = Tweeper::curlExec($ch);

     curl_close($ch);
 
     return $contents;
     curl_close($ch);
 
     return $contents;


@@ -104,22 +178,27 @@ class Tweeper {
   /**
    * Get the headers from a URL.
    */
   /**
    * Get the headers from a URL.
    */

-  private static function getUrlInfo($url) {
+  private static function getUrlInfo($url, $user_agent = NULL) {

     $ch = curl_init($url);
     $ch = curl_init($url);

-    curl_setopt_array($ch, array(
+    curl_setopt_array($ch, [

       CURLOPT_HEADER => TRUE,
       CURLOPT_NOBODY => TRUE,
       CURLOPT_HEADER => TRUE,
       CURLOPT_NOBODY => TRUE,

+      CURLOPT_CONNECTTIMEOUT => Tweeper::$maxConnectionTimeout,

       // Follow http redirects to get the real URL.
       CURLOPT_FOLLOWLOCATION => TRUE,
       CURLOPT_RETURNTRANSFER => TRUE,
       // 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);
+      CURLOPT_USERAGENT => isset($user_agent) ? $user_agent : Tweeper::$userAgent,
+    ]);
+
+    $ret = Tweeper::curlExec($ch);
+    if (FALSE === $ret) {
+      curl_close($ch);
+      return FALSE;
+    }
+

     $url_info = curl_getinfo($ch);
     if (FALSE === $url_info) {
     $url_info = curl_getinfo($ch);
     if (FALSE === $url_info) {

-      trigger_error(curl_error($ch));
+      trigger_error(curl_error($ch), E_USER_WARNING);

     }
     curl_close($ch);
 
     }
     curl_close($ch);
 


@@ -130,7 +209,7 @@ class Tweeper {
    * Generate an RSS <enclosure/> element.
    */
   public static function generateEnclosure($url) {
    * Generate an RSS <enclosure/> element.
    */
   public static function generateEnclosure($url) {

-    $supported_content_types = array(
+    $supported_content_types = [

       "application/octet-stream",
       "application/ogg",
       "application/pdf",
       "application/octet-stream",
       "application/ogg",
       "application/pdf",


@@ -149,13 +228,17 @@ class Tweeper {
       "video/mp4",
       "video/mpeg",
       "video/ogg",
       "video/mp4",
       "video/mpeg",
       "video/ogg",

-    );
+    ];

 
     $url_info = Tweeper::getUrlInfo($url);
 
     $url_info = Tweeper::getUrlInfo($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) {
 
     $supported = in_array($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 '';
     }
 
       return '';
     }
 


@@ -163,10 +246,17 @@ class Tweeper {
     // See http://sourceforge.net/p/feedvalidator/bugs/72/
     $http_url = preg_replace("/^https/", "http", $url_info['url']);
 
     // 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);
     $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;
     $enc->setAttribute('type', $url_info['content_type']);
 
     return $enc;


@@ -203,7 +293,7 @@ class Tweeper {
 
     $output .= " line $error->line";
 
 
     $output .= " line $error->line";
 

-    error_log($output);
+    trigger_error($output, E_USER_WARNING);

   }
 
   /**
   }
 
   /**


@@ -220,17 +310,17 @@ class Tweeper {
 
     $encoder = new XmlEncoder();
     $normalizer = new ObjectNormalizer();
 
     $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_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) {
 
     $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;
     }
 
       return NULL;
     }
 


@@ -245,12 +335,21 @@ class Tweeper {
     $json_match_expr = '/window._sharedData = (.*);/';
     $ret = preg_match($json_match_expr, $html, $matches);
     if ($ret !== 1) {
     $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 NULL;
     }
 

-    // Filter elements which will result in invalid XML element names.
-    $json = str_replace('404_as_react', '_404_as_react', $matches[1]);
+    $data = json_decode($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');
   }
 
     return Tweeper::jsonToXml($json, 'instagram');
   }


@@ -276,15 +375,17 @@ 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)) {
     // 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]);

       $xmlDoc->loadXML($xml_data);
     }
     else {
       $xmlDoc->loadHTML($html);
     }
 
       $xmlDoc->loadXML($xml_data);
     }
     else {
       $xmlDoc->loadHTML($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($xml_errors_value);
     }
     libxml_clear_errors();
     libxml_use_internal_errors($xml_errors_value);


@@ -298,11 +399,15 @@ class Tweeper {
   private function loadStylesheet($host) {
     $stylesheet = "file://" . __DIR__ . "/rss_converter_" . $host . ".xsl";
     if (FALSE === file_exists($stylesheet)) {
   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;
     }
 
       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;
+    }

 
     $xslDoc = new DOMDocument();
     $xslDoc->loadXML($stylesheet_contents);
 
     $xslDoc = new DOMDocument();
     $xslDoc->loadXML($stylesheet_contents);


@@ -310,6 +415,8 @@ class Tweeper {
     $xsltProcessor = new XSLTProcessor();
     $xsltProcessor->registerPHPFunctions();
     $xsltProcessor->setParameter('', 'generate-enclosure', $this->generate_enclosure);
     $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;
     $xsltProcessor->importStylesheet($xslDoc);
 
     return $xsltProcessor;


@@ -318,25 +425,25 @@ class Tweeper {
   /**
    * Convert the site content to RSS.
    */
   /**
    * Convert the site content to RSS.
    */

-  public function tweep($src_url, $host=NULL, $validate_scheme=TRUE) {
+  public function tweep($src_url, $host = NULL, $validate_scheme = TRUE) {

     $url = parse_url($src_url);
     if (FALSE === $url) {
     $url = parse_url($src_url);
     if (FALSE === $url) {

-      trigger_error("Invalid URL: $src_url", E_USER_ERROR);
+      trigger_error("Invalid URL: $src_url", E_USER_WARNING);

       return NULL;
     }
 
     if (TRUE === $validate_scheme) {
       $scheme = $url["scheme"];
       return NULL;
     }
 
     if (TRUE === $validate_scheme) {
       $scheme = $url["scheme"];

-      if (!in_array($scheme, array("http", "https"))) {
-        trigger_error("unsupported scheme: $scheme", E_USER_ERROR);
+      if (!in_array($scheme, ["http", "https"])) {
+        trigger_error("unsupported scheme: $scheme", E_USER_WARNING);

         return NULL;
       }
     }
 
         return NULL;
       }
     }
 

-    // if the host is not given derive it from the URL
+    // If the host is not given derive it from the URL.

     if (NULL === $host) {
       if (empty($url["host"])) {
     if (NULL === $host) {
       if (empty($url["host"])) {

-        trigger_error("Invalid host in URL: $src_url", E_USER_ERROR);
+        trigger_error("Invalid host in URL: $src_url", E_USER_WARNING);

         return NULL;
       }
       // Strip the leading www. to be more forgiving on input URLs.
         return NULL;
       }
       // Strip the leading www. to be more forgiving on input URLs.


@@ -348,14 +455,23 @@ class Tweeper {
       return NULL;
     }
 
       return NULL;
     }
 

-    $html = Tweeper::getUrlContents($src_url);
+    // Override User-Agent for twitter.com to force it to serve the mobile UI.
+    if ($host == "twitter.com") {
+      $user_agent = "Mozilla/5.0 (Linux; U; Android 4.3; en-us; SM-N900T Build/JSS15J)";
+    }
+    else {
+      $user_agent = NULL;
+    }
+
+    $html = Tweeper::getUrlContents($src_url, $user_agent);

     if (FALSE === $html) {
     if (FALSE === $html) {

+      trigger_error("Failed to retrieve $src_url", E_USER_WARNING);

       return NULL;
     }
 
     $preprocess_html_host_method = 'preprocessHtml' . Tweeper::toUpperCamelCase($host, '.');
     if (method_exists($this, $preprocess_html_host_method)) {
       return NULL;
     }
 
     $preprocess_html_host_method = 'preprocessHtml' . Tweeper::toUpperCamelCase($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);
     }
 
     $xmlDoc = $this->htmlToXml($html, $host);


@@ -364,11 +480,11 @@ class Tweeper {
     }
 
     $output = $xsltProcessor->transformToXML($xmlDoc);
     }
 
     $output = $xsltProcessor->transformToXML($xmlDoc);

-

     if (FALSE === $output) {
     if (FALSE === $output) {

-      trigger_error('XSL transformation failed.', E_USER_ERROR);
+      trigger_error('XSL transformation failed.', E_USER_WARNING);

       return NULL;
     }
       return NULL;
     }

+

     return $output;
   }
 
     return $output;
   }