X-Git-Url: https://git.ao2.it/tweeper.git/blobdiff_plain/4c4771f61019b68bdf3d6896995fe9fb24b5fd45..346aa866098b16fb986ba65aa5fa298acad3e7b8:/tweeper.php
diff --git a/tweeper.php b/tweeper.php
index 37b73f5..5379be7 100644
--- a/tweeper.php
+++ b/tweeper.php
@@ -1,6 +1,7 @@
*
@@ -18,35 +19,60 @@
* along with this program. If not, see .
*/
-require_once 'XML/Serializer.php';
+require_once 'Symfony/Component/Serializer/autoload.php';
+
+use Symfony\Component\Serializer\Serializer;
+use Symfony\Component\Serializer\Encoder\XmlEncoder;
+use Symfony\Component\Serializer\Normalizer\ObjectNormalizer;
date_default_timezone_set('UTC');
+/**
+ * Scrape supported websites and perform conversion to RSS.
+ */
class Tweeper {
private static $USER_AGENT = "Mozilla/5.0 (Windows NT 6.1; rv:22.0) Gecko/20130405 Firefox/22.0";
+ /**
+ * Constructor sets up {@link $generate_enclosure}.
+ */
public function __construct($generate_enclosure = FALSE) {
$this->generate_enclosure = $generate_enclosure;
}
- public static function epoch_to_gmdate($timestamp)
- {
+ /**
+ * Convert numeric Epoch to the date format expected in a RSS document.
+ */
+ public static function epoch_to_gmdate($timestamp) {
+ if (!is_numeric($timestamp) || is_nan($timestamp)) {
+ $timestamp = 0;
+ }
+
return gmdate('D, d M Y H:i:s', $timestamp) . ' GMT';
}
- public static function str_to_gmdate($date)
- {
+ /**
+ * Convert generic date string to the date format expected in a RSS document.
+ */
+ public static function str_to_gmdate($date) {
$timestamp = strtotime($date);
+ if (FALSE === $timestamp) {
+ $timestamp = 0;
+ }
+
return Tweeper::epoch_to_gmdate($timestamp);
}
- private static function get_contents($url)
- {
+ /**
+ * Get the contents from a URL.
+ */
+ private static function get_contents($url) {
$ch = curl_init($url);
curl_setopt_array($ch, array(
CURLOPT_HEADER => FALSE,
- CURLOPT_FOLLOWLOCATION => TRUE, // follow http redirects to get the real URL
+ // Follow http redirects to get the real URL.
+ CURLOPT_FOLLOWLOCATION => TRUE,
CURLOPT_RETURNTRANSFER => TRUE,
CURLOPT_SSL_VERIFYHOST => FALSE,
CURLOPT_SSL_VERIFYPEER => FALSE,
@@ -59,13 +85,16 @@ class Tweeper {
return $contents;
}
- private static function get_info($url)
- {
+ /**
+ * Get the headers from a URL.
+ */
+ 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
+ // Follow http redirects to get the real URL.
+ CURLOPT_FOLLOWLOCATION => TRUE,
CURLOPT_RETURNTRANSFER => TRUE,
CURLOPT_SSL_VERIFYHOST => FALSE,
CURLOPT_SSL_VERIFYPEER => FALSE,
@@ -78,8 +107,10 @@ class Tweeper {
return $url_info;
}
- public static function generate_enclosure($url)
- {
+ /**
+ * Generate an RSS element.
+ */
+ public static function generate_enclosure($url) {
$supported_content_types = array(
"application/ogg",
"audio/aac",
@@ -121,39 +152,47 @@ class Tweeper {
return $dom->saveXML($enc);
}
- /* Mimic the message from libxml.c::php_libxml_ctx_error_level() */
+ /**
+ * 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;
+ 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;
}
$output .= trim($error->message);
if ($error->file) {
$output .= " in $error->file";
- } else {
+ }
+ else {
$output .= " in Entity,";
}
- $output .=" line $error->line";
+ $output .= " line $error->line";
error_log($output);
}
+ /**
+ * Load a stylesheet if the web site is supported.
+ */
private function load_stylesheet($host) {
$stylesheet = "file://" . __DIR__ . "/rss_converter_" . $host . ".xsl";
if (FALSE === file_exists($stylesheet)) {
- trigger_error("Conversion to RSS not supported: $host", E_USER_ERROR);
+ trigger_error("Conversion to RSS not supported for $host ($stylesheet not found)", E_USER_ERROR);
return NULL;
}
@@ -170,53 +209,77 @@ class Tweeper {
return $xsltProcessor;
}
- private function json_to_xml($html, $json_match_expr, $rootName) {
- // pre-process, convert json to XML
- $ret = preg_match($json_match_expr, $html, $matches);
- if ($ret !== 1) {
- trigger_error("Cannot match expression: $json_match_expr\n", E_USER_ERROR);
- return NULL;
- }
-
- $data = json_decode($matches[1]);
+ /**
+ * Convert json to xml.
+ */
+ private function json_to_xml($json, $root_node_name) {
+ // Apparenty the ObjectNormalizer used afterwards is not able to handle
+ // the stdClass object created by json_decode() with the default setting
+ // $assoc = false; so use $assoc = true.
+ $data = json_decode($json, $assoc = TRUE);
if (!$data) {
return NULL;
}
- $serializer_options = array (
- 'addDecl' => TRUE,
- 'encoding' => "UTF-8",
- 'indent' => ' ',
- 'rootName' => $rootName,
- );
+ $encoder = new XmlEncoder();
+ $normalizer = new ObjectNormalizer();
+ $serializer = new Serializer(array($normalizer), array($encoder));
- $serializer = new XML_Serializer($serializer_options);
+ $serializer_options = array(
+ 'xml_encoding' => "UTF-8",
+ 'xml_format_output' => TRUE,
+ 'xml_root_node_name' => $root_node_name,
+ );
- $status = $serializer->serialize($data);
- if (PEAR::isError($status)) {
- trigger_error($status->getMessage(), E_USER_ERROR);
+ $xml_data = $serializer->serialize($data, 'xml', $serializer_options);
+ if (!$xml_data) {
+ trigger_error("Cannot serialize data", E_USER_ERROR);
return NULL;
}
- return $serializer->getSerializedData();
+ return $xml_data;
}
+ /**
+ * Convert the Instagram content to XML.
+ */
private function get_xml_instagram_com($html) {
- return $this->json_to_xml($html, '/window._sharedData = (.*);/', 'instagram');
+ // Extract the json data from the html code.
+ $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);
+ return NULL;
+ }
+
+ return $this->json_to_xml($matches[1], 'instagram');
+ }
+
+ /**
+ * Make the Facebook HTML processable.
+ */
+ private function preprocess_html_facebook_com($html) {
+ $html = str_replace('', '', $html);
+ return $html;
}
+ /**
+ * Convert the HTML retrieved from the site to XML.
+ */
private function html_to_xml($html, $host) {
$xmlDoc = new DOMDocument();
// Handle warnings and errors when loading invalid HTML.
- $xml_errors_value = libxml_use_internal_errors(true);
+ $xml_errors_value = libxml_use_internal_errors(TRUE);
// If there is a host-specific method to get the xml data, use it!
$get_xml_host_method = 'get_xml_' . str_replace(".", "_", $host);
if (method_exists($this, $get_xml_host_method)) {
$xml_data = call_user_func_array(array($this, $get_xml_host_method), array($html));
$xmlDoc->loadXML($xml_data);
- } else {
+ }
+ else {
$xmlDoc->loadHTML($html);
}
@@ -229,6 +292,9 @@ class Tweeper {
return $xmlDoc;
}
+ /**
+ * Convert the site content to RSS.
+ */
public function tweep($src_url) {
$url = parse_url($src_url);
if (FALSE === $url || empty($url["host"])) {
@@ -236,7 +302,10 @@ class Tweeper {
return NULL;
}
- $xsltProcessor = $this->load_stylesheet($url["host"]);
+ // Strip the leading www. to be more forgiving on input URLs.
+ $host = preg_replace('/^www\./', '', $url["host"]);
+
+ $xsltProcessor = $this->load_stylesheet($host);
if (NULL === $xsltProcessor) {
return NULL;
}
@@ -246,7 +315,12 @@ class Tweeper {
return NULL;
}
- $xmlDoc = $this->html_to_xml($html, $url["host"]);
+ $preprocess_html_host_method = 'preprocess_html_' . str_replace(".", "_", $host);
+ if (method_exists($this, $preprocess_html_host_method)) {
+ $html = call_user_func_array(array($this, $preprocess_html_host_method), array($html));
+ }
+
+ $xmlDoc = $this->html_to_xml($html, $host);
if (NULL === $xmlDoc) {
return NULL;
}
@@ -259,46 +333,57 @@ class Tweeper {
}
return $output;
}
+
}
-function is_cli()
-{
- return (substr(php_sapi_name(), 0, 3) == 'cli');
+/**
+ * Check if the script is being run from the command line.
+ */
+function is_cli() {
+ return (php_sapi_name() === "cli");
}
-function usage($argv)
-{
+/**
+ * Show the script usage.
+ */
+function usage($argv) {
if (is_cli()) {
$usage = "{$argv[0]} [-e|-h|--help] \n";
- } else {
+ }
+ else {
$usage = htmlentities("{$_SERVER['SCRIPT_NAME']}?src_url=&generate_enclosure=<0|1>");
}
return "usage: $usage";
}
-function parse_options_cli($argv, $argc)
-{
+/**
+ * Parse command line options.
+ */
+function parse_options_cli($argv, $argc) {
$options = array(
- 'generate_enclosure' => FALSE
+ 'generate_enclosure' => FALSE,
);
- if ($argc < 2)
+ if ($argc < 2) {
return $options;
+ }
$cli_options = getopt("eh", array("help"));
foreach ($cli_options as $opt => $val) {
switch ($opt) {
- case 'e':
- $options['generate_enclosure'] = TRUE;
- break;
- case 'h':
- case 'help':
- echo usage($argv);
- exit(0);
- default:
- fwrite(STDERR, usage($argv));
- exit(1);
+ case 'e':
+ $options['generate_enclosure'] = TRUE;
+ break;
+
+ case 'h':
+ case 'help':
+ echo usage($argv);
+ exit(0);
+
+ default:
+ fwrite(STDERR, usage($argv));
+ exit(1);
}
}
@@ -307,32 +392,36 @@ function parse_options_cli($argv, $argc)
return $options;
}
-function parse_options_query_string()
-{
+/**
+ * Parse options passed from a query string.
+ */
+function parse_options_query_string() {
$options = array(
- 'generate_enclosure' => FALSE
+ 'generate_enclosure' => FALSE,
);
- if (isset($_GET['src_url']))
+ if (isset($_GET['src_url'])) {
$options['src_url'] = $_GET['src_url'];
+ }
- if (isset($_GET['generate_enclosure']))
+ if (isset($_GET['generate_enclosure'])) {
$options['generate_enclosure'] = $_GET['generate_enclosure'] == 1;
+ }
return $options;
}
-
if (is_cli()) {
$options = parse_options_cli($argv, $argc);
$ERROR_STREAM = fopen('php://stderr', 'w');
-} else {
+}
+else {
$options = parse_options_query_string();
$ERROR_STREAM = fopen('php://output', 'w');
}
if (!isset($options['src_url'])) {
- fwrite($ERROR_STREAM, usage($argv));
+ fwrite($ERROR_STREAM, usage(is_cli() ? $argv : NULL));
exit(1);
}