因?yàn)殡y點(diǎn)在于如何去識(shí)別并保留網(wǎng)頁中的文章部分,而且刪除其它無用的信息,并且要做到通用化,不能像火車頭那樣根據(jù)目標(biāo)站來制定采集規(guī)則,因?yàn)樗阉饕娼Y(jié)果中有各種的網(wǎng)頁。
抓回一個(gè)頁面的數(shù)據(jù),如何匹配出正文部分,鄭曉在下班路上想了個(gè)思路是:
1. 提取出body標(biāo)簽部分–>剔除所有鏈接–>剔除所有script、注釋–>剔除所有空白標(biāo)簽(包括標(biāo)簽內(nèi)不含中文的)–>獲取結(jié)果。
2. 直接匹配出非鏈接的、 符合在div、p、h標(biāo)簽中的中文部分???
還是會(huì)有不少其它多余信息啊,比如底部信息等。。 如何搞?不知道大家有木有什么思路或建議?
這個(gè)類是從網(wǎng)上找到的一個(gè)php實(shí)現(xiàn)的提取網(wǎng)頁正文部分的算法,鄭曉在本地也測(cè)試了下,準(zhǔn)確率非常高。
代碼如下:
<?php
class Readability {
// 保存判定結(jié)果的標(biāo)記位名稱
const ATTR_CONTENT_SCORE = "contentScore";
// DOM 解析類目前只支持 UTF-8 編碼
const DOM_DEFAULT_CHARSET = "utf-8";
// 當(dāng)判定失敗時(shí)顯示的內(nèi)容
const MESSAGE_CAN_NOT_GET = "Readability was unable to parse this page for content.";
// DOM 解析類(PHP5 已內(nèi)置)
protected $DOM = null;
// 需要解析的源代碼
protected $source = "";
// 章節(jié)的父元素列表
private $parentNodes = array();
// 需要?jiǎng)h除的標(biāo)簽
// Note: added extra tags from
private $junkTags = Array("style", "form", "iframe", "script", "button", "input", "textarea",
"noscript", "select", "option", "object", "applet", "basefont",
"bgsound", "blink", "canvas", "command", "menu", "nav", "datalist",
"embed", "frame", "frameset", "keygen", "label", "marquee", "link");
// 需要?jiǎng)h除的屬性
private $junkAttrs = Array("style", "class", "onclick", "onmouseover", "align", "border", "margin");
/**
* 構(gòu)造函數(shù)
* @param $input_char 字符串的編碼。默認(rèn) utf-8,可以省略
*/
function __construct($source, $input_char = "utf-8") {
$this->source = $source;
// DOM 解析類只能處理 UTF-8 格式的字符
$source = mb_convert_encoding($source, 'HTML-ENTITIES', $input_char);
// 預(yù)處理 HTML 標(biāo)簽,剔除冗余的標(biāo)簽等
$source = $this->preparSource($source);
// 生成 DOM 解析類
$this->DOM = new DOMDocument('1.0', $input_char);
try {
//libxml_use_internal_errors(true);
// 會(huì)有些錯(cuò)誤信息,不過不要緊 :^)
if ( encoding="'.Readability::DOM_DEFAULT_CHARSET.'">'.$source)) {
throw new Exception("Parse HTML Error!");
}
foreach ($this->DOM->childNodes as $item) {
if ($item->nodeType == XML_PI_NODE) {
$this->DOM->removeChild($item); // remove hack
}
}
// insert proper
$this->DOM->encoding = Readability::DOM_DEFAULT_CHARSET;
} catch (Exception $e) {
// ...
}
}
/**
* 預(yù)處理 HTML 標(biāo)簽,使其能夠準(zhǔn)確被 DOM 解析類處理
*
* @return String
*/
private function preparSource($string) {
// 剔除多余的 HTML 編碼標(biāo)記,避免解析出錯(cuò)
preg_match("/charset=([\w|\-]+);?/", $string, $match);
if (isset($match[1])) {
$string = preg_replace("/charset=([\w|\-]+);?/", "", $string, 1);
}
// Replace all doubled-up <BR> tags with <P> tags, and remove fonts.
$string = preg_replace("/<br\/?>[ \r\n\s]*<br\/?>/i", "</p><p>", $string);
$string = preg_replace("/<\/?font[^>]*>/i", "", $string);
// @see
// - from
$string = preg_replace("#<script(.*?)>(.*?)</script>#is", "", $string);
return trim($string);
}
/**
* 刪除 DOM 元素中所有的 $TagName 標(biāo)簽
*
* @return DOMDocument
*/
private function removeJunkTag($RootNode, $TagName) {
$Tags = $RootNode->getElementsByTagName($TagName);
//Note: always index 0, because removing a tag removes it from the results as well.
while($Tag = $Tags->item(0)){
$parentNode = $Tag->parentNode;
$parentNode->removeChild($Tag);
}
return $RootNode;
}
/**
* 刪除元素中所有不需要的屬性
*/
private function removeJunkAttr($RootNode, $Attr) {
$Tags = $RootNode->getElementsByTagName("*");
$i = 0;
while($Tag = $Tags->item($i++)) {
$Tag->removeAttribute($Attr);
}
return $RootNode;
}
/**
* 根據(jù)評(píng)分獲取頁面主要內(nèi)容的盒模型
* 判定算法來自:
* 這里由鄭曉博客轉(zhuǎn)發(fā)
* @return DOMNode
*/
private function getTopBox() {
// 獲得頁面所有的章節(jié)
$allParagraphs = $this->DOM->getElementsByTagName("p");
// Study all the paragraphs and find the chunk that has the best score.
// A score is determined by things like: Number of <p>'s, commas, special classes, etc.
$i = 0;
while($paragraph = $allParagraphs->item($i++)) {
$parentNode = $paragraph->parentNode;
$contentScore = intval($parentNode->getAttribute(Readability::ATTR_CONTENT_SCORE));
$className = $parentNode->getAttribute("class");
$id = $parentNode->getAttribute("id");
// Look for a special classname
if (preg_match("/(comment|meta|footer|footnote)/i", $className)) {
$contentScore -= 50;
} else if(preg_match(
"/((^|\\s)(post|hentry|entry[-]?(content|text|body)?|article[-]?(content|text|body)?)(\\s|$))/i",
$className)) {
$contentScore += 25;
}
// Look for a special ID
if (preg_match("/(comment|meta|footer|footnote)/i", $id)) {
$contentScore -= 50;
} else if (preg_match(
"/^(post|hentry|entry[-]?(content|text|body)?|article[-]?(content|text|body)?)$/i",
$id)) {
$contentScore += 25;
}
// Add a point for the paragraph found
// Add points for any commas within this paragraph
if (strlen($paragraph->nodeValue) > 10) {
$contentScore += strlen($paragraph->nodeValue);
}
// 保存父元素的判定得分
$parentNode->setAttribute(Readability::ATTR_CONTENT_SCORE, $contentScore);
// 保存章節(jié)的父元素,以便下次快速獲取
array_push($this->parentNodes, $parentNode);
}
$topBox = null;
// Assignment from index for performance.
// See
for ($i = 0, $len = sizeof($this->parentNodes); $i < $len; $i++) {
$parentNode = $this->parentNodes[$i];
$contentScore = intval($parentNode->getAttribute(Readability::ATTR_CONTENT_SCORE));
$orgContentScore = intval($topBox ? $topBox->getAttribute(Readability::ATTR_CONTENT_SCORE) : 0);
if ($contentScore && $contentScore > $orgContentScore) {
$topBox = $parentNode;
}
}
// 此時(shí),$topBox 應(yīng)為已經(jīng)判定后的頁面內(nèi)容主元素
return $topBox;
}
/**
* 獲取 HTML 頁面標(biāo)題
*
* @return String
*/
public function getTitle() {
$split_point = ' - ';
$titleNodes = $this->DOM->getElementsByTagName("title");
if ($titleNodes->length
&& $titleNode = $titleNodes->item(0)) {
// @see
$title = trim($titleNode->nodeValue);
$result = array_map('strrev', explode($split_point, strrev($title)));
return sizeof($result) > 1 ? array_pop($result) : $title;
}
return null;
}
/**
* Get Leading Image Url
*
* @return String
*/
public function getLeadImageUrl($node) {
$images = $node->getElementsByTagName("img");
if ($images->length && $leadImage = $images->item(0)) {
return $leadImage->getAttribute("src");
}
return null;
}
/**
* 獲取頁面的主要內(nèi)容(Readability 以后的內(nèi)容)
*
* @return Array
*/
public function getContent() {
if (!$this->DOM) return false;
// 獲取頁面標(biāo)題
$ContentTitle = $this->getTitle();
// 獲取頁面主內(nèi)容
$ContentBox = $this->getTopBox();
//Check if we found a suitable top-box.
if($ContentBox === null)
throw new RuntimeException(Readability::MESSAGE_CAN_NOT_GET);
// 復(fù)制內(nèi)容到新的 DOMDocument
$Target = new DOMDocument;
$Target->appendChild($Target->importNode($ContentBox, true));
// 刪除不需要的標(biāo)簽
foreach ($this->junkTags as $tag) {
$Target = $this->removeJunkTag($Target, $tag);
}
// 刪除不需要的屬性
foreach ($this->junkAttrs as $attr) {
$Target = $this->removeJunkAttr($Target, $attr);
}
$content = mb_convert_encoding($Target->saveHTML(), Readability::DOM_DEFAULT_CHARSET, "HTML-ENTITIES");
// 多個(gè)數(shù)據(jù),以數(shù)組的形式返回
return Array(
'lead_image_url' => $this->getLeadImageUrl($Target),
'word_count' => mb_strlen(strip_tags($content), Readability::DOM_DEFAULT_CHARSET),
'title' => $ContentTitle ? $ContentTitle : null,
'content' => $content
);
}
function __destruct() { }
}
使用起來也非常簡(jiǎn)單,實(shí)例化時(shí)傳入網(wǎng)頁的html源碼和相應(yīng)的編碼,然后直接調(diào)用其getContent方法即可返回提取到的正文部分,提取出的文章中可能還會(huì)含有少部分鏈接,可以自己后期再修改
更多信息請(qǐng)查看IT技術(shù)專欄