classes/feedparser.php

   1 <?php
   2 class FeedParser {
   3         private $doc;
   4         private $error;
   5         private $libxml_errors = array();
   6         private $items;
   7         private $link;
   8         private $title;
   9         private $type;
  10         private $xpath;
  11
  12         const FEED_RDF = 0;
  13         const FEED_RSS = 1;
  14         const FEED_ATOM = 2;
  15
  16         function normalize_encoding($data) {
  17                 if (preg_match('/^(<\?xml[\t\n\r ].*?encoding[\t\n\r ]*=[\t\n\r ]*["\'])(.+?)(["\'].*?\?>)/s', $data, $matches) === 1) {
  18                         $data = mb_convert_encoding($data, 'UTF-8', $matches[2]);
  19
  20                         $data = preg_replace('/^<\?xml[\t\n\r ].*?\?>/s', $matches[1] . "UTF-8" . $matches[3] , $data);
  21                 }
  22
  23                 return $data;
  24         }
  25
  26         function __construct($data) {
  27                 libxml_use_internal_errors(true);
  28                 libxml_clear_errors();
  29                 $this->doc = new DOMDocument();
  30                 $this->doc->loadXML($data);
  31
  32                 mb_substitute_character("none");
  33
  34                 $error = libxml_get_last_error();
  35
  36                 // libxml compiled without iconv?
  37                 if ($error && $error->code == 32) {
  38                         $data = $this->normalize_encoding($data);
  39
  40                         if ($data) {
  41                                 libxml_clear_errors();
  42
  43                                 $this->doc = new DOMDocument();
  44                                 $this->doc->loadXML($data);
  45
  46                                 $error = libxml_get_last_error();
  47                         }
  48                 }
  49
  50                 // some terrible invalid unicode entity?
  51                 if ($error) {
  52                         foreach (libxml_get_errors() as $err) {
  53                                 if ($err->code == 9) {
  54                                         // if the source feed is not in utf8, next conversion will fail
  55                                         $data = $this->normalize_encoding($data);
  56
  57                                         // remove dangling bytes
  58                                         $data = mb_convert_encoding($data, 'UTF-8', 'UTF-8');
  59
  60                                         // apparently not all UTF-8 characters are valid for XML
  61                                         $data = preg_replace('/[^\x{0009}\x{000a}\x{000d}\x{0020}-\x{D7FF}\x{E000}-\x{FFFD}]+/u', ' ', $data);
  62
  63                                         if ($data) {
  64                                                 libxml_clear_errors();
  65
  66                                                 $this->doc = new DOMDocument();
  67                                                 $this->doc->loadXML($data);
  68
  69                                                 $error = libxml_get_last_error();
  70                                         }
  71                                         break;
  72                                 }
  73                         }
  74                 }
  75
  76                 if ($error) {
  77                         foreach (libxml_get_errors() as $error) {
  78                                 if ($error->level == LIBXML_ERR_FATAL) {
  79                                         if(!isset($this->error)) //currently only the first error is reported
  80                                                 $this->error = $this->format_error($error);
  81                                         $this->libxml_errors [] = $this->format_error($error);
  82                                 }
  83                         }
  84                 }
  85                 libxml_clear_errors();
  86
  87                 $this->items = array();
  88         }
  89
  90         function init() {
  91                 $root = $this->doc->firstChild;
  92                 $xpath = new DOMXPath($this->doc);
  93                 $xpath->registerNamespace('atom', 'http://www.w3.org/2005/Atom');
  94                 $xpath->registerNamespace('atom03', 'http://purl.org/atom/ns#');
  95                 $xpath->registerNamespace('media', 'http://search.yahoo.com/mrss/');
  96                 $xpath->registerNamespace('rdf', 'http://www.w3.org/1999/02/22-rdf-syntax-ns#');
  97                 $xpath->registerNamespace('slash', 'http://purl.org/rss/1.0/modules/slash/');
  98                 $xpath->registerNamespace('dc', 'http://purl.org/dc/elements/1.1/');
  99                 $xpath->registerNamespace('content', 'http://purl.org/rss/1.0/modules/content/');
 100                 $xpath->registerNamespace('thread', 'http://purl.org/syndication/thread/1.0');
 101
 102                 $this->xpath = $xpath;
 103
 104                 $root = $xpath->query("(//atom03:feed|//atom:feed|//channel|//rdf:rdf|//rdf:RDF)");
 105
 106                 if ($root && $root->length > 0) {
 107                         $root = $root->item(0);
 108
 109                         if ($root) {
 110                                 switch (mb_strtolower($root->tagName)) {
 111                                 case "rdf:rdf":
 112                                         $this->type = $this::FEED_RDF;
 113                                         break;
 114                                 case "channel":
 115                                         $this->type = $this::FEED_RSS;
 116                                         break;
 117                                 case "feed":
 118                                         $this->type = $this::FEED_ATOM;
 119                                         break;
 120                                 default:
 121                                         if( !isset($this->error) ){
 122                                                 $this->error = "Unknown/unsupported feed type";
 123                                         }
 124                                         return;
 125                                 }
 126                         }
 127
 128                         switch ($this->type) {
 129                         case $this::FEED_ATOM:
 130
 131                                 $title = $xpath->query("//atom:feed/atom:title")->item(0);
 132
 133                                 if (!$title)
 134                                         $title = $xpath->query("//atom03:feed/atom03:title")->item(0);
 135
 136
 137                                 if ($title) {
 138                                         $this->title = $title->nodeValue;
 139                                 }
 140
 141                                 $link = $xpath->query("//atom:feed/atom:link[not(@rel)]")->item(0);
 142
 143                                 if (!$link)
 144                                         $link = $xpath->query("//atom03:feed/atom03:link[not(@rel)]")->item(0);
 145
 146
 147                                 if ($link && $link->hasAttributes()) {
 148                                         $this->link = $link->getAttribute("href");
 149                                 }
 150
 151                                 $articles = $xpath->query("//atom:entry");
 152
 153                                 if (!$articles || $articles->length == 0)
 154                                         $articles = $xpath->query("//atom03:entry");
 155
 156                                 foreach ($articles as $article) {
 157                                         array_push($this->items, new FeedItem_Atom($article, $this->doc, $this->xpath));
 158                                 }
 159
 160                                 break;
 161                         case $this::FEED_RSS:
 162                                 $title = $xpath->query("//channel/title")->item(0);
 163
 164                                 if ($title) {
 165                                         $this->title = $title->nodeValue;
 166                                 }
 167
 168                                 $link = $xpath->query("//channel/link")->item(0);
 169
 170                                 if ($link) {
 171                                         if ($link->getAttribute("href"))
 172                                                 $this->link = $link->getAttribute("href");
 173                                         else if ($link->nodeValue)
 174                                                 $this->link = $link->nodeValue;
 175                                 }
 176
 177                                 $articles = $xpath->query("//channel/item");
 178
 179                                 foreach ($articles as $article) {
 180                                         array_push($this->items, new FeedItem_RSS($article, $this->doc, $this->xpath));
 181                                 }
 182
 183                                 break;
 184                         case $this::FEED_RDF:
 185                                 $xpath->registerNamespace('rssfake', 'http://purl.org/rss/1.0/');
 186
 187                                 $title = $xpath->query("//rssfake:channel/rssfake:title")->item(0);
 188
 189                                 if ($title) {
 190                                         $this->title = $title->nodeValue;
 191                                 }
 192
 193                                 $link = $xpath->query("//rssfake:channel/rssfake:link")->item(0);
 194
 195                                 if ($link) {
 196                                         $this->link = $link->nodeValue;
 197                                 }
 198
 199                                 $articles = $xpath->query("//rssfake:item");
 200
 201                                 foreach ($articles as $article) {
 202                                         array_push($this->items, new FeedItem_RSS($article, $this->doc, $this->xpath));
 203                                 }
 204
 205                                 break;
 206
 207                         }
 208
 209                         if ($this->title) $this->title = trim($this->title);
 210                         if ($this->link) $this->link = trim($this->link);
 211
 212                 } else {
 213                         if( !isset($this->error) ){
 214                                 $this->error = "Unknown/unsupported feed type";
 215                         }
 216                         return;
 217                 }
 218         }
 219
 220         function format_error($error) {
 221                 if ($error) {
 222                         return sprintf("LibXML error %s at line %d (column %d): %s",
 223                                 $error->code, $error->line, $error->column,
 224                                 $error->message);
 225                 } else {
 226                         return "";
 227                 }
 228         }
 229
 230         function error() {
 231                 return $this->error;
 232         }
 233
 234         function errors() {
 235                 return $this->libxml_errors;
 236         }
 237
 238         function get_link() {
 239                 return $this->link;
 240         }
 241
 242         function get_title() {
 243                 return $this->title;
 244         }
 245
 246         function get_items() {
 247                 return $this->items;
 248         }
 249
 250         function get_links($rel) {
 251                 $rv = array();
 252
 253                 switch ($this->type) {
 254                 case $this::FEED_ATOM:
 255                         $links = $this->xpath->query("//atom:feed/atom:link");
 256
 257                         foreach ($links as $link) {
 258                                 if (!$rel || $link->hasAttribute('rel') && $link->getAttribute('rel') == $rel) {
 259                                         array_push($rv, trim($link->getAttribute('href')));
 260                                 }
 261                         }
 262                         break;
 263                 case $this::FEED_RSS:
 264                         $links = $this->xpath->query("//atom:link");
 265
 266                         foreach ($links as $link) {
 267                                 if (!$rel || $link->hasAttribute('rel') && $link->getAttribute('rel') == $rel) {
 268                                         array_push($rv, trim($link->getAttribute('href')));
 269                                 }
 270                         }
 271                         break;
 272                 }
 273
 274                 return $rv;
 275         }
 276 } ?>