[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) {

			$encoding = strtolower($matches[2]);

			if (in_array($encoding, array_map('strtolower', mb_list_encodings())))
				$data = mb_convert_encoding($data, 'UTF-8', $encoding);

			$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) {
4ad04ee2 AD	17	if (preg_match('/^(<\?xml[\t\n\r ].?encoding[\t\n\r ]=[\t\n\r ]["\'])(.+?)(["\'].?\?>)/s', $data, $matches) === 1) {
cde82722 AD	18
	19	$encoding = strtolower($matches[2]);
	20
e4071d25	21	if (in_array($encoding, array_map('strtolower', mb_list_encodings())))
cde82722	22	$data = mb_convert_encoding($data, 'UTF-8', $encoding);
4ad04ee2 AD	23
	24	$data = preg_replace('/^<\?xml[\t\n\r ].*?\?>/s', $matches[1] . "UTF-8" . $matches[3] , $data);
	25	}
	26
	27	return $data;
	28	}
	29
cd07592c AD	30	function __construct($data) {
	31	libxml_use_internal_errors(true);
	32	libxml_clear_errors();
	33	$this->doc = new DOMDocument();
	34	$this->doc->loadXML($data);
d1f3fa97	35
f612dbe8 AD	36	mb_substitute_character("none");
f612dbe8 AD	37
d1f3fa97 AD	38	$error = libxml_get_last_error();
d1f3fa97 AD	39
4f00f55c	40	// libxml compiled without iconv?
5fa36a64	41	if ($error && $error->code == 32) {
4ad04ee2	42	$data = $this->normalize_encoding($data);
4f00f55c	43
4ad04ee2 AD	44	if ($data) {
4ad04ee2 AD	45	libxml_clear_errors();
f612dbe8	46
4ad04ee2 AD	47	$this->doc = new DOMDocument();
4ad04ee2 AD	48	$this->doc->loadXML($data);
f612dbe8	49
4ad04ee2	50	$error = libxml_get_last_error();
5fa36a64	51	}
4f00f55c AD	52	}
	53
	54	// some terrible invalid unicode entity?
5fa36a64	55	if ($error) {
4d49863f	56	foreach (libxml_get_errors() as $err) {
5fa36a64	57	if ($err->code == 9) {
4ad04ee2 AD	58	// if the source feed is not in utf8, next conversion will fail
	59	$data = $this->normalize_encoding($data);
	60
5fa36a64	61	// remove dangling bytes
5fa36a64	62	$data = mb_convert_encoding($data, 'UTF-8', 'UTF-8');
4d49863f	63
5fa36a64	64	// apparently not all UTF-8 characters are valid for XML
	65	$data = preg_replace('/[^\x{0009}\x{000a}\x{000d}\x{0020}-\x{D7FF}\x{E000}-\x{FFFD}]+/u', ' ', $data);
	66
	67	if ($data) {
	68	libxml_clear_errors();
	69
	70	$this->doc = new DOMDocument();
	71	$this->doc->loadXML($data);
	72
	73	$error = libxml_get_last_error();
	74	}
	75	break;
	76	}
f612dbe8	77	}
d1f3fa97	78	}
4d49863f	79
4d49863f AD	80	if ($error) {
	81	foreach (libxml_get_errors() as $error) {
	82	if ($error->level == LIBXML_ERR_FATAL) {
54f23d38	83	if(!isset($this->error)) //currently only the first error is reported
	84	$this->error = $this->format_error($error);
	85	$this->libxml_errors [] = $this->format_error($error);
d3305ff8	86	}
4d49863f	87	}
d3305ff8	88	}
cd07592c AD	89	libxml_clear_errors();
	90
	91	$this->items = array();
	92	}
	93
	94	function init() {
	95	$root = $this->doc->firstChild;
04d2f9c8 AD	96	$xpath = new DOMXPath($this->doc);
04d2f9c8 AD	97	$xpath->registerNamespace('atom', 'http://www.w3.org/2005/Atom');
a3b9fd12	98	$xpath->registerNamespace('atom03', 'http://purl.org/atom/ns#');
4c00e15b	99	$xpath->registerNamespace('media', 'http://search.yahoo.com/mrss/');
852d4ac8	100	$xpath->registerNamespace('rdf', 'http://www.w3.org/1999/02/22-rdf-syntax-ns#');
d4992d6b AD	101	$xpath->registerNamespace('slash', 'http://purl.org/rss/1.0/modules/slash/');
d4992d6b AD	102	$xpath->registerNamespace('dc', 'http://purl.org/dc/elements/1.1/');
8a95d630	103	$xpath->registerNamespace('content', 'http://purl.org/rss/1.0/modules/content/');
d71ac5d3	104	$xpath->registerNamespace('thread', 'http://purl.org/syndication/thread/1.0');
d4992d6b	105
b9eee80e	106	$this->xpath = $xpath;
04d2f9c8	107
a9000b03	108	$root = $xpath->query("(//atom03:feed\|//atom:feed\|//channel\|//rdf:rdf\|//rdf:RDF)");
cd07592c	109
2c6f3c2e	110	if ($root && $root->length > 0) {
a9000b03 AD	111	$root = $root->item(0);
	112
	113	if ($root) {
	114	switch (mb_strtolower($root->tagName)) {
	115	case "rdf:rdf":
	116	$this->type = $this::FEED_RDF;
	117	break;
	118	case "channel":
	119	$this->type = $this::FEED_RSS;
	120	break;
	121	case "feed":
4834ad56	122	case "atom:feed":
a9000b03 AD	123	$this->type = $this::FEED_ATOM;
	124	break;
	125	default:
	126	if( !isset($this->error) ){
	127	$this->error = "Unknown/unsupported feed type";
	128	}
	129	return;
349c4229	130	}
cd07592c AD	131	}
cd07592c AD	132
cd07592c AD	133	switch ($this->type) {
cd07592c AD	134	case $this::FEED_ATOM:
cd07592c AD	135
	136	$title = $xpath->query("//atom:feed/atom:title")->item(0);
	137
a3b9fd12 AD	138	if (!$title)
	139	$title = $xpath->query("//atom03:feed/atom03:title")->item(0);
	140
	141
cd07592c AD	142	if ($title) {
	143	$this->title = $title->nodeValue;
	144	}
	145
	146	$link = $xpath->query("//atom:feed/atom:link[not(@rel)]")->item(0);
	147
e55a5ec6 AD	148	if (!$link)
	149	$link = $xpath->query("//atom:feed/atom:link[@rel='alternate']")->item(0);
	150
a3b9fd12 AD	151	if (!$link)
	152	$link = $xpath->query("//atom03:feed/atom03:link[not(@rel)]")->item(0);
	153
e55a5ec6 AD	154	if (!$link)
e55a5ec6 AD	155	$link = $xpath->query("//atom03:feed/atom03:link[@rel='alternate']")->item(0);
a3b9fd12	156
cd07592c AD	157	if ($link && $link->hasAttributes()) {
	158	$this->link = $link->getAttribute("href");
	159	}
	160
	161	$articles = $xpath->query("//atom:entry");
	162
a3b9fd12 AD	163	if (!$articles \|\| $articles->length == 0)
	164	$articles = $xpath->query("//atom03:entry");
	165
cd07592c	166	foreach ($articles as $article) {
4c00e15b	167	array_push($this->items, new FeedItem_Atom($article, $this->doc, $this->xpath));
cd07592c AD	168	}
cd07592c AD	169
cd07592c AD	170	break;
cd07592c AD	171	case $this::FEED_RSS:
04d2f9c8 AD	172	$title = $xpath->query("//channel/title")->item(0);
	173
	174	if ($title) {
	175	$this->title = $title->nodeValue;
	176	}
	177
	178	$link = $xpath->query("//channel/link")->item(0);
	179
1874c8d6 AD	180	if ($link) {
	181	if ($link->getAttribute("href"))
	182	$this->link = $link->getAttribute("href");
	183	else if ($link->nodeValue)
	184	$this->link = $link->nodeValue;
04d2f9c8 AD	185	}
	186
	187	$articles = $xpath->query("//channel/item");
	188
	189	foreach ($articles as $article) {
4c00e15b	190	array_push($this->items, new FeedItem_RSS($article, $this->doc, $this->xpath));
04d2f9c8 AD	191	}
04d2f9c8 AD	192
cd07592c	193	break;
852d4ac8 AD	194	case $this::FEED_RDF:
	195	$xpath->registerNamespace('rssfake', 'http://purl.org/rss/1.0/');
	196
	197	$title = $xpath->query("//rssfake:channel/rssfake:title")->item(0);
	198
	199	if ($title) {
	200	$this->title = $title->nodeValue;
	201	}
	202
	203	$link = $xpath->query("//rssfake:channel/rssfake:link")->item(0);
	204
	205	if ($link) {
	206	$this->link = $link->nodeValue;
	207	}
	208
	209	$articles = $xpath->query("//rssfake:item");
	210
	211	foreach ($articles as $article) {
	212	array_push($this->items, new FeedItem_RSS($article, $this->doc, $this->xpath));
	213	}
	214
	215	break;
	216
cd07592c	217	}
31bd6f76 AD	218
	219	if ($this->title) $this->title = trim($this->title);
	220	if ($this->link) $this->link = trim($this->link);
	221
852d4ac8	222	} else {
349c4229	223	if( !isset($this->error) ){
	224	$this->error = "Unknown/unsupported feed type";
	225	}
852d4ac8	226	return;
cd07592c AD	227	}
	228	}
	229
	230	function format_error($error) {
	231	if ($error) {
	232	return sprintf("LibXML error %s at line %d (column %d): %s",
	233	$error->code, $error->line, $error->column,
	234	$error->message);
	235	} else {
	236	return "";
	237	}
	238	}
	239
	240	function error() {
	241	return $this->error;
	242	}
	243
54f23d38	244	function errors() {
	245	return $this->libxml_errors;
	246	}
	247
cd07592c AD	248	function get_link() {
	249	return $this->link;
	250	}
	251
	252	function get_title() {
	253	return $this->title;
	254	}
	255
	256	function get_items() {
	257	return $this->items;
	258	}
	259
b9eee80e AD	260	function get_links($rel) {
	261	$rv = array();
	262
	263	switch ($this->type) {
	264	case $this::FEED_ATOM:
	265	$links = $this->xpath->query("//atom:feed/atom:link");
	266
	267	foreach ($links as $link) {
	268	if (!$rel \|\| $link->hasAttribute('rel') && $link->getAttribute('rel') == $rel) {
31bd6f76	269	array_push($rv, trim($link->getAttribute('href')));
b9eee80e AD	270	}
	271	}
	272	break;
	273	case $this::FEED_RSS:
f17c3ee2 AD	274	$links = $this->xpath->query("//atom:link");
f17c3ee2 AD	275
3c8060ac AD	276	foreach ($links as $link) {
3c8060ac AD	277	if (!$rel \|\| $link->hasAttribute('rel') && $link->getAttribute('rel') == $rel) {
31bd6f76	278	array_push($rv, trim($link->getAttribute('href')));
3c8060ac	279	}
b9eee80e AD	280	}
	281	break;
	282	}
	283
	284	return $rv;
	285	}
22a866ed	286	}