[tt-rss.git] / classes / feedparser.php

<?php
class FeedParser {
	private $doc;
	private $error;
	private $libxml_errors = array();
	private $items;
	private $link;
	private $title;
	private $type;
	private $xpath;

	const FEED_RDF = 0;
	const FEED_RSS = 1;
	const FEED_ATOM = 2;

	function normalize_encoding($data) {
		if (preg_match('/^(<\?xml[\t\n\r ].*?encoding[\t\n\r ]*=[\t\n\r ]*["\'])(.+?)(["\'].*?\?>)/s', $data, $matches) === 1) {
			$data = mb_convert_encoding($data, 'UTF-8', $matches[2]);

			$data = preg_replace('/^<\?xml[\t\n\r ].*?\?>/s', $matches[1] . "UTF-8" . $matches[3] , $data);
		}

		return $data;
	}

	function __construct($data) {
		libxml_use_internal_errors(true);
		libxml_clear_errors();
		$this->doc = new DOMDocument();
		$this->doc->loadXML($data);

		mb_substitute_character("none");

		$error = libxml_get_last_error();

		// libxml compiled without iconv?
		if ($error && $error->code == 32) {
			$data = $this->normalize_encoding($data);

			if ($data) {
				libxml_clear_errors();

				$this->doc = new DOMDocument();
				$this->doc->loadXML($data);

				$error = libxml_get_last_error();
			}
		}

		// some terrible invalid unicode entity?
		if ($error) {
			foreach (libxml_get_errors() as $err) {
				if ($err->code == 9) {
					// if the source feed is not in utf8, next conversion will fail
					$data = $this->normalize_encoding($data);

					// remove dangling bytes
					$data = mb_convert_encoding($data, 'UTF-8', 'UTF-8');

					// apparently not all UTF-8 characters are valid for XML
					$data = preg_replace('/[^\x{0009}\x{000a}\x{000d}\x{0020}-\x{D7FF}\x{E000}-\x{FFFD}]+/u', ' ', $data);

					if ($data) {
						libxml_clear_errors();

						$this->doc = new DOMDocument();
						$this->doc->loadXML($data);

						$error = libxml_get_last_error();
					}
					break;
				}
			}
		}

		if ($error) {
			foreach (libxml_get_errors() as $error) {
				if ($error->level == LIBXML_ERR_FATAL) {
					if(!isset($this->error)) //currently only the first error is reported
						$this->error = $this->format_error($error);
					$this->libxml_errors [] = $this->format_error($error);
				}
			}
		}
		libxml_clear_errors();

		$this->items = array();
	}

	function init() {
		$root = $this->doc->firstChild;
		$xpath = new DOMXPath($this->doc);
		$xpath->registerNamespace('atom', 'http://www.w3.org/2005/Atom');
		$xpath->registerNamespace('atom03', 'http://purl.org/atom/ns#');
		$xpath->registerNamespace('media', 'http://search.yahoo.com/mrss/');
		$xpath->registerNamespace('rdf', 'http://www.w3.org/1999/02/22-rdf-syntax-ns#');
		$xpath->registerNamespace('slash', 'http://purl.org/rss/1.0/modules/slash/');
		$xpath->registerNamespace('dc', 'http://purl.org/dc/elements/1.1/');
		$xpath->registerNamespace('content', 'http://purl.org/rss/1.0/modules/content/');
		$xpath->registerNamespace('thread', 'http://purl.org/syndication/thread/1.0');

		$this->xpath = $xpath;

		$root = $xpath->query("(//atom03:feed|//atom:feed|//channel|//rdf:rdf|//rdf:RDF)");

		if ($root && $root->length > 0) {
			$root = $root->item(0);

			if ($root) {
				switch (mb_strtolower($root->tagName)) {
				case "rdf:rdf":
					$this->type = $this::FEED_RDF;
					break;
				case "channel":
					$this->type = $this::FEED_RSS;
					break;
				case "feed":
				case "atom:feed":
					$this->type = $this::FEED_ATOM;
					break;
				default:
					if( !isset($this->error) ){
						$this->error = "Unknown/unsupported feed type";
					}
					return;
				}
			}

			switch ($this->type) {
			case $this::FEED_ATOM:

				$title = $xpath->query("//atom:feed/atom:title")->item(0);

				if (!$title)
					$title = $xpath->query("//atom03:feed/atom03:title")->item(0);


				if ($title) {
					$this->title = $title->nodeValue;
				}

				$link = $xpath->query("//atom:feed/atom:link[not(@rel)]")->item(0);

				if (!$link)
					$link = $xpath->query("//atom:feed/atom:link[@rel='alternate']")->item(0);

				if (!$link)
					$link = $xpath->query("//atom03:feed/atom03:link[not(@rel)]")->item(0);

				if (!$link)
					$link = $xpath->query("//atom03:feed/atom03:link[@rel='alternate']")->item(0);

				if ($link && $link->hasAttributes()) {
					$this->link = $link->getAttribute("href");
				}

				$articles = $xpath->query("//atom:entry");

				if (!$articles || $articles->length == 0)
					$articles = $xpath->query("//atom03:entry");

				foreach ($articles as $article) {
					array_push($this->items, new FeedItem_Atom($article, $this->doc, $this->xpath));
				}

				break;
			case $this::FEED_RSS:
				$title = $xpath->query("//channel/title")->item(0);

				if ($title) {
					$this->title = $title->nodeValue;
				}

				$link = $xpath->query("//channel/link")->item(0);

				if ($link) {
					if ($link->getAttribute("href"))
						$this->link = $link->getAttribute("href");
					else if ($link->nodeValue)
						$this->link = $link->nodeValue;
				}

				$articles = $xpath->query("//channel/item");

				foreach ($articles as $article) {
					array_push($this->items, new FeedItem_RSS($article, $this->doc, $this->xpath));
				}

				break;
			case $this::FEED_RDF:
				$xpath->registerNamespace('rssfake', 'http://purl.org/rss/1.0/');

				$title = $xpath->query("//rssfake:channel/rssfake:title")->item(0);

				if ($title) {
					$this->title = $title->nodeValue;
				}

				$link = $xpath->query("//rssfake:channel/rssfake:link")->item(0);

				if ($link) {
					$this->link = $link->nodeValue;
				}

				$articles = $xpath->query("//rssfake:item");

				foreach ($articles as $article) {
					array_push($this->items, new FeedItem_RSS($article, $this->doc, $this->xpath));
				}

				break;

			}

			if ($this->title) $this->title = trim($this->title);
			if ($this->link) $this->link = trim($this->link);

		} else {
			if( !isset($this->error) ){
				$this->error = "Unknown/unsupported feed type";
			}
			return;
		}
	}

	function format_error($error) {
		if ($error) {
			return sprintf("LibXML error %s at line %d (column %d): %s",
				$error->code, $error->line, $error->column,
				$error->message);
		} else {
			return "";
		}
	}

	function error() {
		return $this->error;
	}

	function errors() {
		return $this->libxml_errors;
	}

	function get_link() {
		return $this->link;
	}

	function get_title() {
		return $this->title;
	}

	function get_items() {
		return $this->items;
	}

	function get_links($rel) {
		$rv = array();

		switch ($this->type) {
		case $this::FEED_ATOM:
			$links = $this->xpath->query("//atom:feed/atom:link");

			foreach ($links as $link) {
				if (!$rel || $link->hasAttribute('rel') && $link->getAttribute('rel') == $rel) {
					array_push($rv, trim($link->getAttribute('href')));
				}
			}
			break;
		case $this::FEED_RSS:
			$links = $this->xpath->query("//atom:link");

			foreach ($links as $link) {
				if (!$rel || $link->hasAttribute('rel') && $link->getAttribute('rel') == $rel) {
					array_push($rv, trim($link->getAttribute('href')));
				}
			}
			break;
		}

		return $rv;
	}
} ?>
Commit	Line	Data
cd07592c AD	1	<?php
	2	class FeedParser {
	3	private $doc;
	4	private $error;
54f23d38	5	private $libxml_errors = array();
cd07592c AD	6	private $items;
	7	private $link;
	8	private $title;
	9	private $type;
b9eee80e	10	private $xpath;
cd07592c AD	11
	12	const FEED_RDF = 0;
	13	const FEED_RSS = 1;
	14	const FEED_ATOM = 2;
	15
4ad04ee2 AD	16	function normalize_encoding($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 $data;
	24	}
	25
cd07592c AD	26	function __construct($data) {
	27	libxml_use_internal_errors(true);
	28	libxml_clear_errors();
	29	$this->doc = new DOMDocument();
	30	$this->doc->loadXML($data);
d1f3fa97	31
f612dbe8 AD	32	mb_substitute_character("none");
f612dbe8 AD	33
d1f3fa97 AD	34	$error = libxml_get_last_error();
d1f3fa97 AD	35
4f00f55c	36	// libxml compiled without iconv?
5fa36a64	37	if ($error && $error->code == 32) {
4ad04ee2	38	$data = $this->normalize_encoding($data);
4f00f55c	39
4ad04ee2 AD	40	if ($data) {
4ad04ee2 AD	41	libxml_clear_errors();
f612dbe8	42
4ad04ee2 AD	43	$this->doc = new DOMDocument();
4ad04ee2 AD	44	$this->doc->loadXML($data);
f612dbe8	45
4ad04ee2	46	$error = libxml_get_last_error();
5fa36a64	47	}
4f00f55c AD	48	}
	49
	50	// some terrible invalid unicode entity?
5fa36a64	51	if ($error) {
4d49863f	52	foreach (libxml_get_errors() as $err) {
5fa36a64	53	if ($err->code == 9) {
4ad04ee2 AD	54	// if the source feed is not in utf8, next conversion will fail
	55	$data = $this->normalize_encoding($data);
	56
5fa36a64	57	// remove dangling bytes
5fa36a64	58	$data = mb_convert_encoding($data, 'UTF-8', 'UTF-8');
4d49863f	59
5fa36a64	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	$error = libxml_get_last_error();
	70	}
	71	break;
	72	}
f612dbe8	73	}
d1f3fa97	74	}
4d49863f	75
4d49863f AD	76	if ($error) {
	77	foreach (libxml_get_errors() as $error) {
	78	if ($error->level == LIBXML_ERR_FATAL) {
54f23d38	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);
d3305ff8	82	}
4d49863f	83	}
d3305ff8	84	}
cd07592c AD	85	libxml_clear_errors();
	86
	87	$this->items = array();
	88	}
	89
	90	function init() {
	91	$root = $this->doc->firstChild;
04d2f9c8 AD	92	$xpath = new DOMXPath($this->doc);
04d2f9c8 AD	93	$xpath->registerNamespace('atom', 'http://www.w3.org/2005/Atom');
a3b9fd12	94	$xpath->registerNamespace('atom03', 'http://purl.org/atom/ns#');
4c00e15b	95	$xpath->registerNamespace('media', 'http://search.yahoo.com/mrss/');
852d4ac8	96	$xpath->registerNamespace('rdf', 'http://www.w3.org/1999/02/22-rdf-syntax-ns#');
d4992d6b AD	97	$xpath->registerNamespace('slash', 'http://purl.org/rss/1.0/modules/slash/');
d4992d6b AD	98	$xpath->registerNamespace('dc', 'http://purl.org/dc/elements/1.1/');
8a95d630	99	$xpath->registerNamespace('content', 'http://purl.org/rss/1.0/modules/content/');
d71ac5d3	100	$xpath->registerNamespace('thread', 'http://purl.org/syndication/thread/1.0');
d4992d6b	101
b9eee80e	102	$this->xpath = $xpath;
04d2f9c8	103
a9000b03	104	$root = $xpath->query("(//atom03:feed\|//atom:feed\|//channel\|//rdf:rdf\|//rdf:RDF)");
cd07592c	105
2c6f3c2e	106	if ($root && $root->length > 0) {
a9000b03 AD	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":
4834ad56	118	case "atom:feed":
a9000b03 AD	119	$this->type = $this::FEED_ATOM;
	120	break;
	121	default:
	122	if( !isset($this->error) ){
	123	$this->error = "Unknown/unsupported feed type";
	124	}
	125	return;
349c4229	126	}
cd07592c AD	127	}
cd07592c AD	128
cd07592c AD	129	switch ($this->type) {
cd07592c AD	130	case $this::FEED_ATOM:
cd07592c AD	131
	132	$title = $xpath->query("//atom:feed/atom:title")->item(0);
	133
a3b9fd12 AD	134	if (!$title)
	135	$title = $xpath->query("//atom03:feed/atom03:title")->item(0);
	136
	137
cd07592c AD	138	if ($title) {
	139	$this->title = $title->nodeValue;
	140	}
	141
	142	$link = $xpath->query("//atom:feed/atom:link[not(@rel)]")->item(0);
	143
e55a5ec6 AD	144	if (!$link)
	145	$link = $xpath->query("//atom:feed/atom:link[@rel='alternate']")->item(0);
	146
a3b9fd12 AD	147	if (!$link)
	148	$link = $xpath->query("//atom03:feed/atom03:link[not(@rel)]")->item(0);
	149
e55a5ec6 AD	150	if (!$link)
e55a5ec6 AD	151	$link = $xpath->query("//atom03:feed/atom03:link[@rel='alternate']")->item(0);
a3b9fd12	152
cd07592c AD	153	if ($link && $link->hasAttributes()) {
	154	$this->link = $link->getAttribute("href");
	155	}
	156
	157	$articles = $xpath->query("//atom:entry");
	158
a3b9fd12 AD	159	if (!$articles \|\| $articles->length == 0)
	160	$articles = $xpath->query("//atom03:entry");
	161
cd07592c	162	foreach ($articles as $article) {
4c00e15b	163	array_push($this->items, new FeedItem_Atom($article, $this->doc, $this->xpath));
cd07592c AD	164	}
cd07592c AD	165
cd07592c AD	166	break;
cd07592c AD	167	case $this::FEED_RSS:
04d2f9c8 AD	168	$title = $xpath->query("//channel/title")->item(0);
	169
	170	if ($title) {
	171	$this->title = $title->nodeValue;
	172	}
	173
	174	$link = $xpath->query("//channel/link")->item(0);
	175
1874c8d6 AD	176	if ($link) {
	177	if ($link->getAttribute("href"))
	178	$this->link = $link->getAttribute("href");
	179	else if ($link->nodeValue)
	180	$this->link = $link->nodeValue;
04d2f9c8 AD	181	}
	182
	183	$articles = $xpath->query("//channel/item");
	184
	185	foreach ($articles as $article) {
4c00e15b	186	array_push($this->items, new FeedItem_RSS($article, $this->doc, $this->xpath));
04d2f9c8 AD	187	}
04d2f9c8 AD	188
cd07592c	189	break;
852d4ac8 AD	190	case $this::FEED_RDF:
	191	$xpath->registerNamespace('rssfake', 'http://purl.org/rss/1.0/');
	192
	193	$title = $xpath->query("//rssfake:channel/rssfake:title")->item(0);
	194
	195	if ($title) {
	196	$this->title = $title->nodeValue;
	197	}
	198
	199	$link = $xpath->query("//rssfake:channel/rssfake:link")->item(0);
	200
	201	if ($link) {
	202	$this->link = $link->nodeValue;
	203	}
	204
	205	$articles = $xpath->query("//rssfake:item");
	206
	207	foreach ($articles as $article) {
	208	array_push($this->items, new FeedItem_RSS($article, $this->doc, $this->xpath));
	209	}
	210
	211	break;
	212
cd07592c	213	}
31bd6f76 AD	214
	215	if ($this->title) $this->title = trim($this->title);
	216	if ($this->link) $this->link = trim($this->link);
	217
852d4ac8	218	} else {
349c4229	219	if( !isset($this->error) ){
	220	$this->error = "Unknown/unsupported feed type";
	221	}
852d4ac8	222	return;
cd07592c AD	223	}
	224	}
	225
	226	function format_error($error) {
	227	if ($error) {
	228	return sprintf("LibXML error %s at line %d (column %d): %s",
	229	$error->code, $error->line, $error->column,
	230	$error->message);
	231	} else {
	232	return "";
	233	}
	234	}
	235
	236	function error() {
	237	return $this->error;
	238	}
	239
54f23d38	240	function errors() {
	241	return $this->libxml_errors;
	242	}
	243
cd07592c AD	244	function get_link() {
	245	return $this->link;
	246	}
	247
	248	function get_title() {
	249	return $this->title;
	250	}
	251
	252	function get_items() {
	253	return $this->items;
	254	}
	255
b9eee80e AD	256	function get_links($rel) {
	257	$rv = array();
	258
	259	switch ($this->type) {
	260	case $this::FEED_ATOM:
	261	$links = $this->xpath->query("//atom:feed/atom:link");
	262
	263	foreach ($links as $link) {
	264	if (!$rel \|\| $link->hasAttribute('rel') && $link->getAttribute('rel') == $rel) {
31bd6f76	265	array_push($rv, trim($link->getAttribute('href')));
b9eee80e AD	266	}
	267	}
	268	break;
	269	case $this::FEED_RSS:
f17c3ee2 AD	270	$links = $this->xpath->query("//atom:link");
f17c3ee2 AD	271
3c8060ac AD	272	foreach ($links as $link) {
3c8060ac AD	273	if (!$rel \|\| $link->hasAttribute('rel') && $link->getAttribute('rel') == $rel) {
31bd6f76	274	array_push($rv, trim($link->getAttribute('href')));
3c8060ac	275	}
b9eee80e AD	276	}
	277	break;
	278	}
	279
	280	return $rv;
	281	}
cd07592c	282	} ?>