import wordpress page that crawl from googlecache to database

worpress_page_parser.php

<?php
ini_set('display_errors',true);
error_reporting(E_ALL ^ E_NOTICE);

$obj = new HtmlImporter();
$obj->run();


class HtmlImporter {

    function run(){
        $this->parsePage();
    }

    function getDbInstance(){
        require __DIR__."/wp-config.php";
        $dsn = sprintf('mysql:host=%s;port=3306;dbname=%s;charset=UTF8;',DB_HOST,DB_NAME);
        $db = new PDO($dsn,DB_USER,DB_PASSWORD);
        $db->query("SET NAMES utf8;");
        return $db;
    }

    function getPages(){
        $ret = array();
        $d = dir(".");
        while (false !== ($entry = $d->read())) {
            if (preg_match('/p_\d+\.html/', $entry)){
                $ret[] = $entry;
            }
        }
        $d->close();
        return $ret;
    }

    function parsePage(){
        $db = $this->getDbInstance();
        $db->exec('TRUNCATE wp_posts');
        $db->exec('TRUNCATE wp_term_taxonomy');
        $db->exec('TRUNCATE wp_terms');
        $db->exec('TRUNCATE wp_term_relationships');

        $pages = $this->getPages();
        foreach ($pages as $page) {
            echo "importing $page \n";
            try {
                $pageobj = new PageParser($page);
                $pagedb['id'] = $pageobj->getId();
                $pagedb['title'] = $pageobj->getTitle();
                $pagedb['content'] = $pageobj->getContent();
                $pagedb['pubdate'] = $pageobj->getPubdate();
                $cats = $pageobj->getCategory();    
            } catch (Exception $e) {
                echo $e->getMessage(). "\n";
            }
            $this->savePage($pagedb);
            $this->saveCat($pagedb['id'],$cats);
        }
    }

    function savePage($pagedb){
        $db = $this->getDbInstance();
        $sql =  sprintf("INSERT INTO `wp_posts` (`id`,`post_title`,`post_content`,`post_date`,`post_excerpt`,`to_ping`,`pinged`,`post_content_filtered`) values ('%s','%s','%s','%s','','','','')",$pagedb['id'],$pagedb['title'],$pagedb['content'],$pagedb['pubdate']);
        //echo $sql;
        $db->exec($sql) or die(print_r($db->errorInfo(), true));
        //todo:fix gmt date
    }

    function saveCat($page_id,$cats) {
        $db = $this->getDbInstance();
        foreach ($cats as $catdb) {
            $sql =  sprintf("INSERT INTO `wp_terms` (`term_id`,`name`,`slug`) values ('%s','%s','%s')",$catdb['term_id'],$catdb['name'],$catdb['name']);           
            $db->exec($sql);

            $sql = sprintf("INSERT INTO `wp_term_taxonomy` (`term_id`,`taxonomy`,`description`) values ('%s','%s','')",$catdb['term_id'],'category');
            if( $db->exec($sql) ){
                //映射关系查询表，getTermTaxonomyId要用
                $this->termTaxonomyR[$catdb['term_id']] = $db->lastInsertId();
            }

            $this->updatePageCat($page_id,$catdb['term_id']);

        }
    }

    function updatePageCat($page_id,$term_id){
        $db = $this->getDbInstance();
        $term_taxonomy_id = $this->getTermTaxonomyId($term_id);
        $sql = sprintf("INSERT INTO `wp_term_relationships` (`object_id`,`term_taxonomy_id`) values ('%s','%s')",$page_id,$term_taxonomy_id);
        $db->exec($sql);
    }

    function getTermTaxonomyId($term_id){
        return $this->termTaxonomyR[$term_id];
    }
}

class PageParser {
    function __construct($page){
        #$page = "p_640.html";
        $this->filename = $page;
        try{
            $this->content = file_get_contents($page);
        } catch (Exception $e){
            throw new Exception("Error To Read $page", 1);
        }

        $regex = '/<header class="entry-header">([\s\S]+)<h1 class="entry-title">([\s\S]+)<\/h1>([\S\s]+)<div class="entry-content">([\s\S]+)<\/div>([\S\s]+)class="entry-meta">([\s\S]+)<\/footer>/';

        if(!preg_match($regex,$this->content,$match)){
            throw new Exception("Regex Not Match ", 1);
        }

        $this->title = $match[2];
        $this->content = $match[4];
        $this->content_meta = $match[6];
    }

    public function getId(){
        if(!preg_match('/p_(\d+)\.html/',$this->filename,$match)){
            throw new Exception('No Id found');
            return;
        }
        return $match[1];
    }

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

    public function getContent(){
        return $this->content;
    }

    public function getPubdate(){
        if(preg_match('/datetime="([\s\S]+)">.*<\/time>/',$this->content_meta,$match)){
            $date = $match[1];
            return date("Y-m-d H:i:s",strtotime($date) - 8 * 3600);   
        }
    }

    public function getCategory(){
        $cats = array();
        $regex = '/cat=(\d+)" rel="category">([^<]*)<\/a>/';
        if(preg_match_all($regex,$this->content_meta,$matchs)){
            $ids = $matchs[1];
            $names = $matchs[2];
            for($i=0;$i<count($ids);$i++) {
                $cat['term_id'] = $ids[$i];
                $cat['name']  = $names[$i];
                $cats[] = $cat;
            }
        }
        return $cats;
    }
}