抓取“中國 IC 網”供應商程序

<?php 
/** 
* 抓取“中國 IC 網（http://www.ic37.com）”供應商主程序 
* author Lee. 
* Last modify $Date: 2012-2-9 9:32:21 $ 
* 注：本程序按照編碼 GB2312 執行，因為“中國 IC 網”網站是GB2312編碼，數據庫也得保持一致
*/ 
class ic37 { 
    private $key; // 型號  
    private $pageNum; // 頁碼  
 
    /**
     * 入口程序
     */ 
    public function go($key) { 
        $this->key = $key; 
        $this->pageNum = $this->getPageNum(); 
        $this->getInfo(); 
    } 
 
    /**
     * 獲取供應商 url 鏈接數組
     * @return ArrayObject
     */ 
    private function getInfo() { 
        if ($this->pageNum==1) { # 處理只有一頁的情況 
            $arr = $this->shopAddContact($this->shopUrlMatchReArr($this->getContent())); 
            $this->isAddSuccess($arr); 
        } elseif ($this->pageNum>1) { # 多頁 
            for ($i=1; $i<=$this->pageNum; $i++) { 
                $arr = $this->shopAddContact($this->shopUrlMatchReArr($this->getContent($i))); 
                $this->isAddSuccess($arr); 
            } 
        }    
    } 
     
    /**
     * 打印是否添加成功
     * @param ArrayObject $arr
     * @return string
     */ 
    private function isAddSuccess($arr) { 
        foreach ($arr as $k=>$v) { 
            if ($this->execAdd($this->getInfoByShopUrl($v))) { 
                echo 'Add Success!!'; 
            } else { 
                echo 'Add Faild!!'; 
            } 
        } 
    } 
 
    /**
     * 執行添加到數據庫
     * @param ArrayObject $infoArr
     * @return Number 受影響的行數
     */ 
    private function execAdd($infoArr) { 
        $mysqli = $this->getDb(); 
        if (!emptyempty($infoArr['company'])) { 
            if (!$this->isExists($mysqli, $infoArr)) { 
                $num = $mysqli->query("INSERT INTO ic37(company,person,phone,mobile,qq,msn,fax,email,address,country,region,zip,web,shopUrl) VALUES ('{$infoArr['company']}','{$infoArr['person']}','{$infoArr['phone']}','{$infoArr['mobile']}','{$infoArr['qq']}','{$infoArr['msn']}','{$infoArr['fax']}','{$infoArr['email']}','{$infoArr['address']}','{$infoArr['country']}','{$infoArr['region']}','{$infoArr['zip']}','{$infoArr['web']}','{$infoArr['shopUrl']}')"); 
                return $num; 
            } else { 
                return false; # 表示數據已經存在 
            } 
        } else { 
            return false; 
        } 
    } 
     
    private function formatStr($str) { 
        $str = trim($str); 
        $str = str_replace(' ', '', $str); 
        $str = str_replace('==聯系我們', '', $str); 
        return $str; 
    } 
 
    /**
     * 連接數據庫
     */ 
    private function getDb() { 
        $mysqli = new mysqli('localhost', 'root', '1715544', 'weiku'); 
        $mysqli->query('SET NAMES GB2312'); 
        return $mysqli; 
    } 
 
    /**
     * 檢查公司是否已經存在
     * @param Resource $mysqli
     * @param ArrayObject $infoArr
     * @return bool
     */ 
    private function isExists($mysqli, $infoArr) { 
        $mysqli->query("SELECT company FROM weiku WHERE company = '{$infoArr['company']}'"); 
        if ($mysqli->affected_rows) { 
            return true; 
        } else { 
            return false; 
        } 
    } 
 
    /**
     * 抓取信息
     * @param $url 
     * @return ArrayObject
     */ 
    private function getInfoByShopUrl($url) { 
        $re = preg_replace('/<a.+>(.*)<\/a>/', '\1', str_replace('</font>', '', str_replace('<font color="#000099">', '', $this->getUrlInfo($url)))); 
        preg_match_all('/<title>(.*)<\/title>/Usi', $re, $companyArr); 
        preg_match_all('/<strong>聯系人：<\/strong><\/td>\s*<td.*>(.*)<\/td>/Usi', $re, $personArr); 
        preg_match_all('/<strong>電話：<\/strong><\/td>\s*<td.*>(.*)<\/td>/Usi', $re, $phoneArr); 
        preg_match_all('/<strong>手機：<\/strong><\/td>\s*<td.*>(.*)<\/td>/Usi', $re, $mobileArr); 
        preg_match_all('/<strong>QQ：<\/strong><\/td>\s*<td.*>(.*)<\/td>/Usi', $re, $qqArr); 
        preg_match_all('/<strong>MSN：<\/strong><\/td>\s*<td.*>(.*)<\/td>/Usi', $re, $msnArr); 
        preg_match_all('/<strong>傳真：<\/strong><\/td>\s*<td.*>(.*)<\/td>/Usi', $re, $faxArr); 
        preg_match_all('/<strong>EMail：<\/strong><\/td>\s*<td.*>(.*)<\/td>/Usi', $re, $emailArr); 
        preg_match_all('/司地址[：]*[<\/strong>]*[<strong>]*[：]*[<\/strong>]*<\/td>\s*<td.*>(.*)<\/td>/Usi', $re, $addressArr); 
        preg_match_all('/<strong>國家[：]*<\/strong>[<strong>]*[：]*[<\/strong>]*<\/td>\s*<td.*>(.*)<\/td>/Usi', $re, $countryArr); 
        preg_match_all('/<strong>地區：<\/strong><\/td>\s*<td.*>(.*)<\/td>/Usi', $re, $regionArr); 
        preg_match_all('/<strong>郵政編碼：<\/strong><\/td>\s*<td.*>(.*)<\/td>/Usi', $re, $zipArr); 
        preg_match_all('/<strong>\s*網址[1]*：<\/strong><\/td>\s*<td.*>(.*)<\/td>/Usi', $re, $webArr); 
        $infoArr = array( 
            'company'=>$this->formatStr($companyArr[1][0]), 
            'person'=>$this->formatStr($personArr[1][0]), 
            'phone'=>$this->formatStr($phoneArr[1][0]), 
            'mobile'=>$this->formatStr($mobileArr[1][0]), 
            'qq'=>$this->formatStr($qqArr[1][0]), 
            'msn'=>$this->formatStr($msnArr[1][0]), 
            'fax'=>$this->formatStr($faxArr[1][0]), 
            'email'=>$this->formatStr($emailArr[1][0]), 
            'address'=>$this->formatStr($addressArr[1][0]), 
            'country'=>$this->formatStr($countryArr[1][0]), 
            'region'=>$this->formatStr($regionArr[1][0]), 
            'zip'=>$this->formatStr($zipArr[1][0]), 
            'web'=>$this->formatStr($webArr[1][0]), 
            'shopUrl'=>$url 
        ); 
        return $infoArr; 
    } 
 
    /**
     * 根據頁面獲取供應商 url 數組
     * @param string $re
     * @return ArrayObject
     */ 
    private function shopUrlMatchReArr($re) { 
        preg_match_all('/<p class="Company"><a.* href=\"(.+)\".*>[<font color="#FF0000">]*.*[<\/font>]*<\/a>\s*<\/p>/Usi', $re, $arr); 
        $arr = $this->formatUrlArr(array_unique($arr[1])); 
        return $arr; 
    } 
     
    /**
     * 格式化數組
     * @param Array $arr
     * @return ArrayObject
     */ 
    private function formatUrlArr($arr) { 
        $newArr = array(); 
        foreach ($arr as $key=>$value) { 
            if ($this->isExistsHttp($value)) { 
                $newArr[$key] = $value; 
            } 
        } 
        return $newArr; 
    } 
     
    /**
     * 格式化 QQ
     * @param string $str
     * @return string
     */ 
    private function formatQqMsn($str, $e='QQ') { 
        if (emptyempty($str)) return ''; 
        preg_match_all('/alt="'.$e.'\:(.+)"/Usi', $str, $arr); 
        if (count($arr[1])==1) return $arr[1][0]; 
        $newStr = null; 
        foreach ($arr[1] as $value) { 
            $newStr .= $value . ' '; 
        } 
        return rtrim($newStr, ' '); 
    } 
 
    /**
     * 供應商店鋪鏈接添加 contact.asp
     * @param array $arr
     * @return string    
     */ 
    private function shopAddContact($arr) { 
        foreach ($arr as $k=>$v) { 
            if (stristr($v, 'contact.asp')===FALSE)  
                $newArr[$k] = $this->addContact($v); 
            else 
                $newArr[$k] = $v; 
        } 
        return $newArr; 
    } 
     
    /**
     * 鏈接添加 contact.asp
     * @param string $str
     * @return string    
     */ 
    private function addContact($str) { 
        return $str . '/contact.asp'; 
    } 
 
    /**
     * 去掉網址的 A 標簽
     * @param string $site
     * @return string
     */ 
    private function stripATags($site) { 
        $site = preg_replace('/<a.+>(.+)<\/a>/', '\1', $site); 
        return $site; 
    } 
 
    /**
     * 檢查 url 是否有 http
     * @param string $url
     * @return bool
     */ 
    private function isExistsHttp($url) { 
        if (stristr($url, 'http://')) { 
            return true; 
        } else { 
            return false; 
        } 
    } 
     
    /**
     * 獲取頁面內容
     * @param Number $page
     * @return string
     */ 
    private function getContent($page=1) { 
        $re = file_get_contents($this->getUrl($this->key, $page)); 
        return $re; 
    } 
     
    /**
     * 獲取頁碼
     * @return Number
     */ 
    private function getPageNum() { 
        preg_match_all('/共.*條記錄分(.*)頁顯示/Usi', $this->getContent(), $arr); 
        return $arr[1][0]; 
    } 
 
    /**
     * 獲取 URL 鏈接
     * @param string $str
     * @param int $page 頁碼
     * @return string
     */ 
    private function getUrl($str, $page=1) { 
        return "http://www.ic37.com/sell/search.asp?keyword={$str}&x=86&y=22&page={$page}"; 
    } 
 
    /**
     * 獲取頁面內容
     * @param string $url
     * @return string
     */ 
    private function getUrlInfo($url) { 
        $re = file_get_contents($url); 
        return $re; 
    } 
} 
 
/*
程序運行思路：根據“中國 IC 網”的IC搜索功能，輸入型號進行搜索，然後抓取供應商信息
 
數據庫結構
CREATE TABLE `ic37` (
    `id` mediumint(8) unsigned NOT NULL auto_increment,
    `company` varchar(500) default NULL,
    `person` varchar(500) default NULL,
    `phone` varchar(500) default NULL,
    `mobile` varchar(500) default NULL,
    `qq` varchar(500) default NULL,
    `msn` varchar(500) default NULL,
    `fax` varchar(500) default NULL,
    `email` varchar(500) default NULL,
    `address` varchar(1000) default NULL,
    `country` varchar(500) default NULL,
    `region` varchar(500) default NULL,
    `zip` varchar(500) default NULL,
    `web` varchar(500) default NULL,
    `shopUrl` varchar(500) default NULL,
    PRIMARY KEY  (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=gb2312
*/ 
 
$k = new ic37(); 
$arr = array_unique(array('MAX3232', 'AML8613', 'MT6225A', 'OM8373PS/N3/A', 'PT7313', 'MAX8212ESA', 'TL431', 'S3C2440', 'TMS320F2812PGFA', 'PCM1704', 'AN6717', 'CA3162E', 'CA3161E', 'LM393N', 'DS18B20', 'SHT10', 'AML8613', 'AN6717', 'LM393N', 'CA3161E', 'CA3162E', 'PCM1704', 'STK392-040', 'K1667', 'MAX232', 'STM32F103', 'LM358', 'NE555', '78L05', 'LM324', 'TL431', 'PC817', '7805', 'LM339', 'LM317', '46A-3GRI', 'MODEL', '78L05', '93C46-3GRI', '8050', 'DS18B20', 'TDA2030', 'LM393', '74HC595', '6N137', 'SN75176BDR')); 
foreach ($arr as $v) { 
    $k->go($v); 
} 
?> 
<?php
/**
* 抓取“中國 IC 網（http://www.ic37.com）”供應商主程序
* author Lee.
* Last modify $Date: 2012-2-9 9:32:21 $
* 注：本程序按照編碼 GB2312 執行，因為“中國 IC 網”網站是GB2312編碼，數據庫也得保持一致
*/
class ic37 {
 private $key; // 型號
 private $pageNum; // 頁碼

 /**
  * 入口程序
  */
 public function go($key) {
  $this->key = $key;
  $this->pageNum = $this->getPageNum();
  $this->getInfo();
 }

 /**
  * 獲取供應商 url 鏈接數組
  * @return ArrayObject
  */
 private function getInfo() {
  if ($this->pageNum==1) { # 處理只有一頁的情況
   $arr = $this->shopAddContact($this->shopUrlMatchReArr($this->getContent()));
   $this->isAddSuccess($arr);
  } elseif ($this->pageNum>1) { # 多頁
   for ($i=1; $i<=$this->pageNum; $i++) {
    $arr = $this->shopAddContact($this->shopUrlMatchReArr($this->getContent($i)));
    $this->isAddSuccess($arr);
   }
  } 
 }
 
 /**
  * 打印是否添加成功
  * @param ArrayObject $arr
  * @return string
  */
 private function isAddSuccess($arr) {
  foreach ($arr as $k=>$v) {
   if ($this->execAdd($this->getInfoByShopUrl($v))) {
    echo 'Add Success!!';
   } else {
    echo 'Add Faild!!';
   }
  }
 }

 /**
  * 執行添加到數據庫
  * @param ArrayObject $infoArr
  * @return Number 受影響的行數
  */
 private function execAdd($infoArr) {
  $mysqli = $this->getDb();
  if (!empty($infoArr['company'])) {
   if (!$this->isExists($mysqli, $infoArr)) {
    $num = $mysqli->query("INSERT INTO ic37(company,person,phone,mobile,qq,msn,fax,email,address,country,region,zip,web,shopUrl) VALUES ('{$infoArr['company']}','{$infoArr['person']}','{$infoArr['phone']}','{$infoArr['mobile']}','{$infoArr['qq']}','{$infoArr['msn']}','{$infoArr['fax']}','{$infoArr['email']}','{$infoArr['address']}','{$infoArr['country']}','{$infoArr['region']}','{$infoArr['zip']}','{$infoArr['web']}','{$infoArr['shopUrl']}')");
    return $num;
   } else {
    return false; # 表示數據已經存在
   }
  } else {
   return false;
  }
 }
 
 private function formatStr($str) {
  $str = trim($str);
  $str = str_replace(' ', '', $str);
  $str = str_replace('==聯系我們', '', $str);
  return $str;
 }

 /**
  * 連接數據庫
  */
 private function getDb() {
  $mysqli = new mysqli('localhost', 'root', '1715544', 'weiku');
  $mysqli->query('SET NAMES GB2312');
  return $mysqli;
 }

 /**
  * 檢查公司是否已經存在
  * @param Resource $mysqli
  * @param ArrayObject $infoArr
  * @return bool
  */
 private function isExists($mysqli, $infoArr) {
  $mysqli->query("SELECT company FROM weiku WHERE company = '{$infoArr['company']}'");
  if ($mysqli->affected_rows) {
   return true;
  } else {
   return false;
  }
 }

 /**
  * 抓取信息
  * @param $url
  * @return ArrayObject
  */
 private function getInfoByShopUrl($url) {
  $re = preg_replace('/<a.+>(.*)<\/a>/', '\1', str_replace('</font>', '', str_replace('<font color="#000099">', '', $this->getUrlInfo($url))));
  preg_match_all('/<title>(.*)<\/title>/Usi', $re, $companyArr);
  preg_match_all('/<strong>聯系人：<\/strong><\/td>\s*<td.*>(.*)<\/td>/Usi', $re, $personArr);
  preg_match_all('/<strong>電話：<\/strong><\/td>\s*<td.*>(.*)<\/td>/Usi', $re, $phoneArr);
  preg_match_all('/<strong>手機：<\/strong><\/td>\s*<td.*>(.*)<\/td>/Usi', $re, $mobileArr);
  preg_match_all('/<strong>QQ：<\/strong><\/td>\s*<td.*>(.*)<\/td>/Usi', $re, $qqArr);
  preg_match_all('/<strong>MSN：<\/strong><\/td>\s*<td.*>(.*)<\/td>/Usi', $re, $msnArr);
  preg_match_all('/<strong>傳真：<\/strong><\/td>\s*<td.*>(.*)<\/td>/Usi', $re, $faxArr);
  preg_match_all('/<strong>EMail：<\/strong><\/td>\s*<td.*>(.*)<\/td>/Usi', $re, $emailArr);
  preg_match_all('/司地址[：]*[<\/strong>]*[<strong>]*[：]*[<\/strong>]*<\/td>\s*<td.*>(.*)<\/td>/Usi', $re, $addressArr);
  preg_match_all('/<strong>國家[：]*<\/strong>[<strong>]*[：]*[<\/strong>]*<\/td>\s*<td.*>(.*)<\/td>/Usi', $re, $countryArr);
  preg_match_all('/<strong>地區：<\/strong><\/td>\s*<td.*>(.*)<\/td>/Usi', $re, $regionArr);
  preg_match_all('/<strong>郵政編碼：<\/strong><\/td>\s*<td.*>(.*)<\/td>/Usi', $re, $zipArr);
  preg_match_all('/<strong>\s*網址[1]*：<\/strong><\/td>\s*<td.*>(.*)<\/td>/Usi', $re, $webArr);
  $infoArr = array(
   'company'=>$this->formatStr($companyArr[1][0]),
   'person'=>$this->formatStr($personArr[1][0]),
   'phone'=>$this->formatStr($phoneArr[1][0]),
   'mobile'=>$this->formatStr($mobileArr[1][0]),
   'qq'=>$this->formatStr($qqArr[1][0]),
   'msn'=>$this->formatStr($msnArr[1][0]),
   'fax'=>$this->formatStr($faxArr[1][0]),
   'email'=>$this->formatStr($emailArr[1][0]),
   'address'=>$this->formatStr($addressArr[1][0]),
   'country'=>$this->formatStr($countryArr[1][0]),
   'region'=>$this->formatStr($regionArr[1][0]),
   'zip'=>$this->formatStr($zipArr[1][0]),
   'web'=>$this->formatStr($webArr[1][0]),
   'shopUrl'=>$url
  );
  return $infoArr;
 }

 /**
  * 根據頁面獲取供應商 url 數組
  * @param string $re
  * @return ArrayObject
  */
 private function shopUrlMatchReArr($re) {
  preg_match_all('/<p class="Company"><a.* href=\"(.+)\".*>[<font color="#FF0000">]*.*[<\/font>]*<\/a>\s*<\/p>/Usi', $re, $arr);
  $arr = $this->formatUrlArr(array_unique($arr[1]));
  return $arr;
 }
 
 /**
  * 格式化數組
  * @param Array $arr
  * @return ArrayObject
  */
 private function formatUrlArr($arr) {
  $newArr = array();
  foreach ($arr as $key=>$value) {
   if ($this->isExistsHttp($value)) {
    $newArr[$key] = $value;
   }
  }
  return $newArr;
 }
 
 /**
  * 格式化 QQ
  * @param string $str
  * @return string
  */
 private function formatQqMsn($str, $e='QQ') {
  if (empty($str)) return '';
  preg_match_all('/alt="'.$e.'\:(.+)"/Usi', $str, $arr);
  if (count($arr[1])==1) return $arr[1][0];
  $newStr = null;
  foreach ($arr[1] as $value) {
   $newStr .= $value . ' ';
  }
  return rtrim($newStr, ' ');
 }

 /**
  * 供應商店鋪鏈接添加 contact.asp
  * @param array $arr
  * @return string 
  */
 private function shopAddContact($arr) {
  foreach ($arr as $k=>$v) {
   if (stristr($v, 'contact.asp')===FALSE)
    $newArr[$k] = $this->addContact($v);
   else
    $newArr[$k] = $v;
  }
  return $newArr;
 }
 
 /**
  * 鏈接添加 contact.asp
  * @param string $str
  * @return string 
  */
 private function addContact($str) {
  return $str . '/contact.asp';
 }

 /**
  * 去掉網址的 A 標簽
  * @param string $site
  * @return string
  */
 private function stripATags($site) {
  $site = preg_replace('/<a.+>(.+)<\/a>/', '\1', $site);
  return $site;
 }

 /**
  * 檢查 url 是否有 http
  * @param string $url
  * @return bool
  */
 private function isExistsHttp($url) {
  if (stristr($url, 'http://')) {
   return true;
  } else {
   return false;
  }
 }
 
 /**
  * 獲取頁面內容
  * @param Number $page
  * @return string
  */
 private function getContent($page=1) {
  $re = file_get_contents($this->getUrl($this->key, $page));
  return $re;
 }
 
 /**
  * 獲取頁碼
  * @return Number
  */
 private function getPageNum() {
  preg_match_all('/共.*條記錄分(.*)頁顯示/Usi', $this->getContent(), $arr);
  return $arr[1][0];
 }

 /**
  * 獲取 URL 鏈接
  * @param string $str
  * @param int $page 頁碼
  * @return string
  */
 private function getUrl($str, $page=1) {
  return "http://www.ic37.com/sell/search.asp?keyword={$str}&x=86&y=22&page={$page}";
 }

 /**
  * 獲取頁面內容
  * @param string $url
  * @return string
  */
 private function getUrlInfo($url) {
  $re = file_get_contents($url);
  return $re;
 }
}

/*
程序運行思路：根據“中國 IC 網”的IC搜索功能，輸入型號進行搜索，然後抓取供應商信息

數據庫結構
CREATE TABLE `ic37` (
 `id` mediumint(8) unsigned NOT NULL auto_increment,
 `company` varchar(500) default NULL,
 `person` varchar(500) default NULL,
 `phone` varchar(500) default NULL,
 `mobile` varchar(500) default NULL,
 `qq` varchar(500) default NULL,
 `msn` varchar(500) default NULL,
 `fax` varchar(500) default NULL,
 `email` varchar(500) default NULL,
 `address` varchar(1000) default NULL,
 `country` varchar(500) default NULL,
 `region` varchar(500) default NULL,
 `zip` varchar(500) default NULL,
 `web` varchar(500) default NULL,
 `shopUrl` varchar(500) default NULL,
 PRIMARY KEY  (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=gb2312
*/

$k = new ic37();
$arr = array_unique(array('MAX3232', 'AML8613', 'MT6225A', 'OM8373PS/N3/A', 'PT7313', 'MAX8212ESA', 'TL431', 'S3C2440', 'TMS320F2812PGFA', 'PCM1704', 'AN6717', 'CA3162E', 'CA3161E', 'LM393N', 'DS18B20', 'SHT10', 'AML8613', 'AN6717', 'LM393N', 'CA3161E', 'CA3162E', 'PCM1704', 'STK392-040', 'K1667', 'MAX232', 'STM32F103', 'LM358', 'NE555', '78L05', 'LM324', 'TL431', 'PC817', '7805', 'LM339', 'LM317', '46A-3GRI', 'MODEL', '78L05', '93C46-3GRI', '8050', 'DS18B20', 'TDA2030', 'LM393', '74HC595', '6N137', 'SN75176BDR'));
foreach ($arr as $v) {
 $k->go($v);
}
?>