src/Tweeper.php: add a retry mechanism for cURL sessions

[tweeper.git] / src / Tweeper.php

diff --git a/src/Tweeper.php b/src/Tweeper.php
index 93ac9e0..d58b529 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) 2013-2018  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 +36,23 @@ 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 (Windows NT 6.1; WOW64; rv:60.0) Gecko/20100101 Firefox/60.0";
+  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).

    */
    */

-  public function __construct($generate_enclosure = FALSE) {
+  public function __construct($generate_enclosure = FALSE, $show_usernames = TRUE) {

     $this->generate_enclosure = $generate_enclosure;
     $this->generate_enclosure = $generate_enclosure;

+    $this->show_usernames = $show_usernames;

   }
 
   /**
   }
 
   /**


@@ -78,12 +88,32 @@ 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($ch, array(
       CURLOPT_HEADER => FALSE,
    * Get the contents from a URL.
    */
   private static function getUrlContents($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_RETURNTRANSFER => TRUE,
       // Follow http redirects to get the real URL.
       CURLOPT_FOLLOWLOCATION => TRUE,
       CURLOPT_RETURNTRANSFER => TRUE,


@@ -92,10 +122,7 @@ class Tweeper {
       CURLOPT_HTTPHEADER => array('Accept-language: en'),
       CURLOPT_USERAGENT => Tweeper::$userAgent,
     ));
       CURLOPT_HTTPHEADER => array('Accept-language: en'),
       CURLOPT_USERAGENT => Tweeper::$userAgent,
     ));

-    $contents = curl_exec($ch);
-    if (FALSE === $contents) {
-      trigger_error(curl_error($ch));
-    }
+    $contents = Tweeper::curlExec($ch);

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


@@ -109,6 +136,7 @@ class Tweeper {
     curl_setopt_array($ch, array(
       CURLOPT_HEADER => TRUE,
       CURLOPT_NOBODY => TRUE,
     curl_setopt_array($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,
       // Follow http redirects to get the real URL.
       CURLOPT_FOLLOWLOCATION => TRUE,
       CURLOPT_RETURNTRANSFER => TRUE,


@@ -116,10 +144,16 @@ class Tweeper {
       CURLOPT_SSL_VERIFYPEER => FALSE,
       CURLOPT_USERAGENT => Tweeper::$userAgent,
     ));
       CURLOPT_SSL_VERIFYPEER => FALSE,
       CURLOPT_USERAGENT => Tweeper::$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) {
     $url_info = curl_getinfo($ch);
     if (FALSE === $url_info) {

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

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


@@ -152,10 +186,14 @@ class Tweeper {
     );
 
     $url_info = Tweeper::getUrlInfo($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 '';
     }
 


@@ -203,7 +241,7 @@ class Tweeper {
 
     $output .= " line $error->line";
 
 
     $output .= " line $error->line";
 

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

   }
 
   /**
   }
 
   /**


@@ -230,7 +268,7 @@ class Tweeper {
 
     $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;
     }
 


@@ -245,11 +283,23 @@ 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;
     }
 

-    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');

   }
 
   /**
   }
 
   /**


@@ -295,11 +345,15 @@ 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;
     }
 
     $stylesheet_contents = Tweeper::getUrlContents($stylesheet);
       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);
 
     $xslDoc = new DOMDocument();
     $xslDoc->loadXML($stylesheet_contents);


@@ -307,6 +361,7 @@ 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);

     $xsltProcessor->importStylesheet($xslDoc);
 
     return $xsltProcessor;
     $xsltProcessor->importStylesheet($xslDoc);
 
     return $xsltProcessor;


@@ -315,21 +370,30 @@ class Tweeper {
   /**
    * Convert the site content to RSS.
    */
   /**
    * 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);
     $url = parse_url($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;
     }
 
       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) {
 
     $xsltProcessor = $this->loadStylesheet($host);
     if (NULL === $xsltProcessor) {


@@ -338,6 +402,7 @@ class Tweeper {
 
     $html = Tweeper::getUrlContents($src_url);
     if (FALSE === $html) {
 
     $html = Tweeper::getUrlContents($src_url);
     if (FALSE === $html) {

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

       return NULL;
     }
 
       return NULL;
     }
 


@@ -352,11 +417,11 @@ 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;
   }