欢迎您访问程序员文章站本站旨在为大家提供分享程序员计算机编程知识！
您现在的位置是： 首页  >  后端开发

解析HTML标签，并实现快速查找节点，获取节点信息

程序员文章站 2023-12-30 12:20:46
...
详细介绍和使用请点击源码出处。
  1. /**
  2. * html标签解析包
  3. *
  4. * @category TagParse
  5. * @package TagParse
  6. * @author kun
  7. * @copyright 2014 kun
  8. * @license http://www.php.net/license/3_01.txt PHP License 3.01
  9. * @version 1.0
  10. * @link http://www.blogkun.com
  11. * @since 1.0
  12. */
  13. namespace TagParse;
  15. /**
  16. * TagDomRoot
  17. *
  18. * @category TagParse
  19. * @package TagParse
  20. * @author kun
  21. * @copyright 2014 kun
  22. * @license http://www.php.net/license/3_01.txt PHP License 3.01
  23. * @version 1.0
  24. * @link http://www.blogkun.com
  25. * @since 1.0
  26. */
  27. class TagDomRoot
  28. {
  29. public $tag = 'root';
  30. public $plaintext;
  31. public $child = array();
  32. public $level = 0;
  33. public static $TagParseError = false;
  34. protected static $TagSet = array();
  35. protected static $FoundNode = array();
  36. public static $ErrorTag = array();
  38. /**
  39. * initProperty
  40. *
  41. * @access public
  42. *
  43. * @return null
  44. */
  45. public function initProperty()
  46. {
  47. $TagParseError = false;
  48. $TagSet = array();
  49. $FoundNode = array();
  50. $DumpScriptCode = array();
  51. $ErrorTag = array();
  52. }
  54. /**
  55. * __construct
  56. *
  57. * @param string $str The tag string to be parse.
  58. *
  59. * @access public
  60. *
  61. * @return TagDomRoot
  62. */
  63. public function __construct($str)
  64. {
  65. $this->_removeNoise($str);
  66. if ($str === null) {
  67. self::$TagParseError = true;
  68. } else {
  69. $l = strpos($str, ' if ($l !== false) {
  70. $this->plaintext = substr($str, 0, $l);
  71. }
  72. $res = preg_match_all('~>(.*?) if ($res !== false && $res > 0) {
  73. $this->plaintext .= implode($matches[1]);
  74. }
  75. $r = strrpos($str, '>');
  76. if ($r !== false) {
  77. $this->plaintext .= substr($str, $r+1);
  78. }
  80. $tagCollect = array();
  81. $attrCollect = array();
  82. $innerContentCollect = array();
  84. if ($this->parseTag($str, $tagCollect, $attrCollect, $innerContentCollect) === false) {
  85. self::$TagParseError = true;
  86. }
  88. foreach ($tagCollect as $index => $tag) {
  89. $this->child[] = new TagDomNode($tag, $this, $attrCollect[$index], $innerContentCollect[$index], $this->level+1);
  90. }
  91. }
  92. }
  94. /**
  95. * parseTag
  96. *
  97. * @param mixed $str Description.
  98. * @param mixed &$tagCollect Description.
  99. * @param mixed &$attrCollect Description.
  100. * @param mixed &$innerContentCollect Description.
  101. *
  102. * @access protected
  103. *
  104. * @return boolean Value.
  105. */
  106. protected function parseTag($str, array &$tagCollect, array &$attrCollect, array &$innerContentCollect)
  107. {
  108. $selfClosingTags = array('img' => 1, 'br' => 1, 'input' => 1, 'meta' => 1, 'link' => 1, 'hr' => 1, 'base' => 1, 'embed' => 1, 'spacer' => 1);
  109. $end = -2;
  110. $close = 0;
  111. $error = false;
  112. $tag = '';
  113. while (true) {
  114. $l = strpos($str, ' if ($l === false) {//parse end
  115. break;
  116. }
  117. if (strpos(substr($str, $l, 2), '/') !== false) {//surplus closing tag,discard
  118. $error = true;
  119. $end = $l+strlen($tag);
  120. self::$ErrorTag[] = substr($str, $l, strpos($str, '>', $l)-$l+1);
  121. continue;
  122. }
  124. $r = strpos($str, '>', $l);
  125. $tag = substr($str, $l+1, $r-$l-1);
  126. if (!ctype_alpha($tag[0]) || strpos($tag, ' $end = $r + 1;
  127. continue;
  128. }
  129. $tag = preg_replace("~\n+~", ' ', $tag);
  130. $space = strpos($tag, ' ');
  131. if ($space !== false) {
  132. $attrCollect[] = substr($tag, $space+1);
  133. $tag = substr($tag, 0, $space);
  134. } else {
  135. $attrCollect[] = '';
  136. }
  137. $tagCollect[] = $tag;
  138. if (isset($selfClosingTags[$tag])) {
  139. $innerContentCollect[] = '';
  140. $end = $r-strlen($tag)-2;
  141. $close = $r+1;
  142. continue;
  143. }
  145. $countOpen = -1;
  146. $open = strpos($str, ' $close = strpos($str, ''.$tag.'>', $open);
  147. if ($close === false) {//surplus opening tag
  148. $innerContentCollect[] = substr($str, $r+1);
  149. $error = true;
  150. self::$ErrorTag[] = '';
  151. break;
  152. }
  153. $start = $open;
  154. while ($open $countOpen++;
  155. $open = strpos($str, ' }
  157. while ($countOpen > 0 && $close !== false) {
  158. $open = strpos($str, ' $close = strpos($str, ''.$tag.'>', $close+strlen($tag)+3);
  159. if ($close === false) {
  160. break;
  161. }
  162. $countOpen--;
  163. while ($open $open = strpos($str, ' $countOpen++;
  164. }
  165. }
  166. if ($close === false) {//标签闭合不配对
  167. $innerContentCollect[] = substr($str, $r+1);
  168. $error = true;
  169. break;
  170. }
  171. $end = $close;
  172. $r = strpos($str, '>', $start);
  173. $innerContentCollect[] = substr($str, $r+1, $end - $r - 1);
  174. }
  176. return !$error;
  177. }
  179. /**
  180. * _removeNoise
  181. *
  182. * @param string &$str The tag string to be parse.
  183. *
  184. * @access private
  185. *
  186. * @return string
  187. */
  188. private function _removeNoise(&$str)
  189. {
  190. $str = preg_replace('~~is', '', $str);
  191. $str = preg_replace('~~is', '', $str);
  192. $str = preg_replace('~*?>~is', '', $str);
  193. }
  195. /**
  196. * parseSelectors
  197. *
  198. * @param string $selectors user's select condition.
  199. * @param array &$selectorsTag tags
  200. * @param array &$selectorsAttr attributes
  201. *
  202. * @access protected
  203. *
  204. * @return null
  205. */
  206. protected function parseSelectors($selectors, array &$selectorsTag, array &$selectorsAttr)
  207. {
  208. preg_match_all('~([\w\d]+)(\[[\w\d -="._/]+\])?~', $selectors, $matches);
  209. $selectorsTag = $matches[1];
  210. foreach ($matches[2] as $key => $value) {
  211. $selectorsAttr[$key] = array();
  212. if ($value !== '') {
  213. preg_match_all('~([\w\d-]+)="([\w\d-. _/]+)"~', $value, $matches);
  214. foreach ($matches[1] as $index => $attr) {
  215. $selectorsAttr[$key][$attr] = $matches[2][$index];
  216. }
  217. }
  218. }
  219. }
  221. /**
  222. * find
  223. *
  224. * @param mixed $selectors user's select condition.
  225. * @param array $selectorsTag tags.
  226. * @param array $selectorsAttr attributes.
  227. *
  228. * @access public
  229. *
  230. * @return array
  231. */
  232. public function find($selectors, $selectorsTag = array(), $selectorsAttr = array())
  233. {
  234. if ($selectors !== null) {
  235. $this->parseSelectors($selectors, $selectorsTag, $selectorsAttr);
  236. }
  237. var_dump($selectorsTag, $selectorsAttr);exit();
  238. if (!empty($selectorsTag)) {
  239. $this->seek($selectorsTag, $selectorsAttr);
  240. foreach ($this->child as $key => $node) {
  241. $node->find(null, $selectorsTag, $selectorsAttr);
  242. }
  243. }
  245. if ($selectors !== null) {
  246. $res = self::$FoundNode;
  247. self::$FoundNode = array();
  249. return $res;
  250. }
  251. }
  253. /**
  254. * findGlobal
  255. *
  256. * @param string $selectors user's select condition.
  257. *
  258. * @access public
  259. *
  260. * @return array
  261. */
  262. public function findGlobal($selectors)
  263. {
  264. $space = strpos($selectors, ' ', strpos($selectors, ']'));
  265. if ($space === false) {
  266. return $this->findOneGlobal($selectors);
  267. } else {
  268. $selectorsAttr = array();
  269. $selectorsTag = array();
  270. $this->findOneGlobal(substr($selectors, 0, $space), false);
  271. $this->parseSelectors(substr($selectors, $space + 1), $selectorsTag, $selectorsAttr);
  272. if (!empty(self::$FoundNode) && !empty($selectorsTag)) {
  273. $nodes = self::$FoundNode;
  274. self::$FoundNode = array();
  275. foreach ($nodes as $key => $node) {
  276. $node->seek($selectorsTag, $selectorsAttr);
  277. }
  278. }
  281. }
  283. $res = self::$FoundNode;
  284. self::$FoundNode = array();
  286. return $res;
  287. }
  289. /**
  290. * seek
  291. *
  292. * @param array $selectorsTag tags.
  293. * @param array $selectorsAttr attributes.
  294. *
  295. * @access protected
  296. *
  297. * @return null
  298. */
  299. protected function seek($selectorsTag, $selectorsAttr)
  300. {
  301. foreach ($this->child as $key => $node) {
  302. $isFind = true;
  303. if ($node->tag === $selectorsTag[0]) {
  304. foreach ($selectorsAttr[0] as $attrName => $value) {
  306. if (isset($node->attr[$attrName])
  307. && (preg_match('~.*? '.$value.' .*?~', $node->attr[$attrName]) > 0
  308. || preg_match('~^'.$value.'$~', $node->attr[$attrName]) > 0
  309. || preg_match('~^'.$value.' ~', $node->attr[$attrName]) > 0
  310. || preg_match('~ '.$value.'$~', $node->attr[$attrName]) > 0)
  311. ) {
  312. continue;
  313. } else {
  314. $isFind = false;
  315. break;
  316. }
  317. }
  318. } else {
  319. $isFind = false;
  320. }
  321. if ($isFind) {
  322. if (count($selectorsTag) === 1) {
  323. self::$FoundNode[] = $node;
  324. } else {
  325. $node->seek(
  326. array_slice($selectorsTag, 1),
  327. array_slice($selectorsAttr, 1)
  328. );
  329. }
  330. }
  331. }
  332. }
  334. /**
  335. * findOneGlobal
  336. *
  337. * @param string $selector user's select condition.
  338. * @param bool $isReturn weather return value.
  339. *
  340. * @access public
  341. *
  342. * @return array
  343. */
  344. public function findOneGlobal($selector, $isReturn = true)
  345. {
  346. preg_match('~([\w\d]+)(\[[\w\d -="._/]+\])?~', $selector, $matches);
  347. $tag = $matches[1];
  348. $attr = array();
  349. if (isset($matches[2])) {
  350. preg_match_all('~([\w\d-]+)="([\w\d-. _/]+)"~', $matches[2], $matches);
  351. foreach ($matches[1] as $key => $value) {
  352. $attr[$value] = $matches[2][$key];
  353. }
  354. }
  355. if (isset(self::$TagSet[$tag])) {
  356. foreach (self::$TagSet[$tag] as $attrValue => $nodeArray) {
  357. $isFind = true;
  358. foreach ($attr as $attrName => $value) {
  359. if (preg_match('~'.$attrName.'=".*? '.$value.' .*?"~', $attrValue)
  360. || preg_match('~'.$attrName.'="'.$value.' .*?"~', $attrValue)
  361. || preg_match('~'.$attrName.'=".*? '.$value.'"~', $attrValue)
  362. || preg_match('~'.$attrName.'="'.$value.'"~', $attrValue)
  363. ) {
  364. continue;
  365. } else {
  366. $isFind = false;
  367. break;
  368. }
  369. }
  370. if ($isFind) {
  371. foreach ($nodeArray as $key => $node) {
  372. self::$FoundNode[] = $node;
  373. }
  374. }
  375. }
  376. }
  377. if ($isReturn) {
  378. $res = self::$FoundNode;
  379. self::$FoundNode = array();
  381. return $res;
  382. }
  383. }
  384. }
  386. /**
  387. * TagDomNode
  388. *
  389. * @uses TagDomRoot
  390. *
  391. * @category TagParse
  392. * @package TagParse
  393. * @author kun
  394. * @copyright 2014 kun
  395. * @license http://www.php.net/license/3_01.txt PHP License 3.01
  396. * @version 1.0
  397. * @link http://www.blogkun.com
  398. * @since 1.0
  399. */
  400. class TagDomNode extends TagDomRoot
  401. {
  402. public $attr = array();
  403. public $parent = null;
  405. /**
  406. * __construct
  407. *
  408. * @param mixed $tag tag.
  409. * @param mixed $parent parent node.
  410. * @param mixed $attr attribute.
  411. * @param mixed $innerContent tag content.
  412. * @param mixed $level node level.
  413. *
  414. * @access public
  415. *
  416. * @return TagDomNode
  417. */
  418. public function __construct($tag, $parent, $attr, $innerContent, $level)
  419. {
  420. $this->tag = $tag;
  421. $this->parent = $parent;
  422. $this->_parseAttr($attr);
  423. $this->level = $level;
  425. $l = strpos($innerContent, ' if ($l !== false) {
  426. $this->plaintext = substr($innerContent, 0, $l);
  427. }
  428. $res = preg_match_all('~>(.*?) if ($res !== false && $res > 0) {
  429. $this->plaintext .= implode($matches[1]);
  430. } else {
  431. $this->plaintext .= $innerContent;
  432. }
  433. $r = strrpos($innerContent, '>');
  434. if ($r !== false) {
  435. $this->plaintext .= substr($innerContent, $r+1);
  436. }
  438. $tagCollect = array();
  439. $attrCollect = array();
  440. $innerContentCollect = array();
  442. if ($this->parseTag($innerContent, $tagCollect, $attrCollect, $innerContentCollect) === false) {
  443. self::$TagParseError = true;
  444. }
  446. foreach ($tagCollect as $index => $tag) {
  447. $this->child[] = new TagDomNode($tag, $this, $attrCollect[$index], $innerContentCollect[$index], $this->level+1);
  448. }
  450. if (!isset(self::$TagSet[$this->tag])) {
  451. self::$TagSet[$this->tag] = array();
  452. }
  453. if (!isset(self::$TagSet[$this->tag][$attr])) {
  454. self::$TagSet[$this->tag][$attr] = array();
  455. }
  456. self::$TagSet[$this->tag][$attr][] = &$this;
  457. }
  459. /**
  460. * _parseAttr
  461. *
  462. * @param string $str attribute string.
  463. *
  464. * @access public
  465. *
  466. * @return null
  467. */
  468. private function _parseAttr($str)
  469. {
  470. preg_match_all('~(?[\w-]+)="(?.*?)"~s', $str, $matches);
  471. foreach ($matches['attrName'] as $key => $value) {
  472. $this->attr[$value] = $matches['attrValue'][$key];
  473. }
  474. }
  475. }
复制代码
相关标签： 解析HTML标签，并实现快速查找节点，获取节点信息

上一篇：

下一篇：