ITPub博客

首页 > Linux操作系统 > Linux操作系统 > 十个超级有用的PHP代码片段

十个超级有用的PHP代码片段

原创 Linux操作系统 作者:sg2012 时间:2012-10-23 14:11:07 0 删除 编辑
1. 发送短信

调用 TextMagic API。

// Include the TextMagic PHP lib  require('textmagic-sms-api-php/TextMagicAPI.php');   // Set the username and password information  $username = 'myusername';  $password = 'mypassword';   // Create a new instance of TM  $router = new TextMagicAPI(array(      'username' => $username,      'password' => $password ));   // Send a text message to '999-123-4567'  $result = $router->send('Wake up!', array(9991234567), true);   // result:  Result is: Array ( [messages] => Array ( [19896128] => 9991234567 ) [sent_text] => Wake up! [parts_count] => 1 ) 

2. 根据IP查找地址

function detect_city($ip) {           $default = 'UNKNOWN';           if (!is_string($ip) || strlen($ip) < 1 || $ip == '127.0.0.1' || $ip == 'localhost')              $ip = '8.8.8.8';           $curlopt_useragent = 'Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.2) Gecko/20100115 Firefox/3.6 (.NET CLR 3.5.30729)';           $url = 'http://ipinfodb.com/ip_locator.php?ip=' . urlencode($ip);          $ch = curl_init();           $curl_opt = array(              CURLOPT_FOLLOWLOCATION  => 1,              CURLOPT_HEADER      => 0,              CURLOPT_RETURNTRANSFER  => 1,              CURLOPT_USERAGENT   => $curlopt_useragent,              CURLOPT_URL       => $url,              CURLOPT_TIMEOUT         => 1,              CURLOPT_REFERER         => 'http://' . $_SERVER['HTTP_HOST'],          );           curl_setopt_array($ch, $curl_opt);           $content = curl_exec($ch);           if (!is_null($curl_info)) {              $curl_info = curl_getinfo($ch);          }           curl_close($ch);           if ( preg_match('{City : ([^<]*)<>}i', $content, $regs) )  {              $city = $regs[1];          }          if ( preg_match('{State/Province : ([^<]*)<>}i', $content, $regs) )  {              $state = $regs[1];          }           if( $city!='' && $state!='' ){            $location = $city . ', ' . $state;            return $location;          }else{            return $default;          }       } <!--]*)<!--]*)

3. 显示网页的源代码

foreach ($lines as $line_num => $line) {      // loop thru each line and prepend line numbers      echo "Line #{$line_num} : " . htmlspecialchars($line) . "
n";  } 

4. 检查服务器是否使用HTTPS

if ($_SERVER['HTTPS'] != "on") {      echo "This is not HTTPS";  }else{      echo "This is HTTPS";  } 

5. 显示Facebook粉丝数量

function fb_fan_count($facebook_name){      // Example: https://graph.facebook.com/digimantra      $data = json_decode(file_get_contents("https://graph.facebook.com/".$facebook_name));      echo $data->likes;  } 

6. 检测图片的主要颜色

$i = imagecreatefromjpeg("image.jpg");   for ($x=0;$x        $rgb = imagecolorat($i,$x,$y);          $r   = ($rgb >> 16) & 0xFF;          $g   = ($rgb >>  & 0xFF;          $b   = $rgb & 0xFF;           $rTotal += $r;          $gTotal += $g;          $bTotal += $b;          $total++;      }  }   $rAverage = round($rTotal/$total);  $gAverage = round($gTotal/$total);  $bAverage = round($bTotal/$total); 

7. 获取内存使用信息

echo "Initial: ".memory_get_usage()." bytes n";  /* prints  Initial: 361400 bytes  */  // let's use up some memory  for ($i = 0; $i < 100000; $i++) {      $array []= md5($i);  }   // let's remove half of the array  for ($i = 0; $i < 100000; $i++) {      unset($array[$i]);  }   echo "Final: ".memory_get_usage()." bytes n";  /* prints  Final: 885912 bytes  */  echo "Peak: ".memory_get_peak_usage()." bytes n";  /* prints  Peak: 13687072 bytes  */ 

8. 使用 gzcompress() 压缩数据

$string =  "Lorem ipsum dolor sit amet, consectetur  adipiscing elit. Nunc ut elit id mi ultricies  adipiscing. Nulla facilisi. Praesent pulvinar,  sapien vel feugiat vestibulum, nulla dui pretium orci,  non ultricies elit lacus quis ante. Lorem ipsum dolor  sit amet, consectetur adipiscing elit. Aliquam  pretium ullamcorper urna quis iaculis. Etiam ac massa  sed turpis tempor luctus. Curabitur sed nibh eu elit  mollis congue. Praesent ipsum diam, consectetur vitae  ornare a, aliquam a nunc. In id magna pellentesque  tellus posuere adipiscing. Sed non mi metus, at lacinia  augue. Sed magna nisi, ornare in mollis in, mollis  sed nunc. Etiam at justo in leo congue mollis.  Nullam in neque eget metus hendrerit scelerisque  eu non enim. Ut malesuada lacus eu nulla bibendum  id euismod urna sodales. ";   $compressed = gzcompress($string);   echo "Original size: ". strlen($string)."n";  /* prints  Original size: 800  */   echo "Compressed size: ". strlen($compressed)."n";  /* prints  Compressed size: 418  */   // getting it back  $original = gzuncompress($compressed); 

9. 使用PHP做Whois检查

function whois_query($domain) {       // fix the domain name:      $domain = strtolower(trim($domain));      $domain = preg_replace('/^http:///i', '', $domain);      $domain = preg_replace('/^', '', $domain);      $domain = explode('/', $domain);      $domain = trim($domain[0]);       // split the TLD from domain name      $_domain = explode('.', $domain);      $lst = count($_domain)-1;      $ext = $_domain[$lst];       // You find resources and lists      // like these on wikipedia:      //      // http://de.wikipedia.org/wiki/Whois      //      $servers = array(          "biz" => "whois.neulevel.biz",          "com" => "whois.internic.net",          "us" => "whois.nic.us",          "coop" => "whois.nic.coop",          "info" => "whois.nic.info",          "name" => "whois.nic.name",          "net" => "whois.internic.net",          "gov" => "whois.nic.gov",          "edu" => "whois.internic.net",          "mil" => "rs.internic.net",          "int" => "whois.iana.org",          "ac" => "whois.nic.ac",          "ae" => "whois.uaenic.ae",          "at" => "whois.ripe.net",          "au" => "whois.aunic.net",          "be" => "whois.dns.be",          "bg" => "whois.ripe.net",          "br" => "whois.registro.br",          "bz" => "whois.belizenic.bz",          "ca" => "whois.cira.ca",          "cc" => "whois.nic.cc",          "ch" => "whois.nic.ch",          "cl" => "whois.nic.cl",          "cn" => "whois.cnnic.net.cn",          "cz" => "whois.nic.cz",          "de" => "whois.nic.de",          "fr" => "whois.nic.fr",          "hu" => "whois.nic.hu",          "ie" => "whois.domainregistry.ie",          "il" => "whois.isoc.org.il",          "in" => "whois.ncst.ernet.in",          "ir" => "whois.nic.ir",          "mc" => "whois.ripe.net",          "to" => "whois.tonic.to",          "tv" => "whois.tv",          "ru" => "whois.ripn.net",          "org" => "whois.pir.org",          "aero" => "whois.information.aero",          "nl" => "whois.domain-registry.nl"      );       if (!isset($servers[$ext])){          die('Error: No matching nic server found!');      }       $nic_server = $servers[$ext];       $output = '';       // connect to whois server:      if ($conn = fsockopen ($nic_server, 43)) {          fputs($conn, $domain."rn");          while(!feof($conn)) {              $output .= fgets($conn,128);          }          fclose($conn);      }      else { die('Error: Could not connect to ' . $nic_server . '!'); }       return $output;  } 

10. 通过Email发送PHP错误

// Our custom error handler  function nettuts_error_handler($number, $message, $file, $line, $vars){      $email = "          An error ($number) occurred on line          $line and in the file: $file.           $message ";       $email .= "" . print_r($vars, 1) . "";       $headers = 'Content-type: text/html; charset=iso-8859-1' . "rn";       // Email the error to someone...      error_log($email, 1, 'you@youremail.com', $headers);       // Make sure that you decide how to respond to errors (on the user's side)      // Either echo an error message, or kill the entire project. Up to you...      // The code below ensures that we only "die" if the error was more than      // just a NOTICE.      if ( ($number !== E_NOTICE) && ($number < 2048) ) {          die("There was an error. Please try again later.");      }  }   // We should use our custom function to handle errors.  set_error_handler('nettuts_error_handler');   // Trigger an error... (var doesn't exist)  echo $somevarthatdoesnotexist;
本文来自: 尚观知识库 http://www.uplook.cn/index-Index-show-view19985.html

来自 “ ITPUB博客 ” ,链接:https://blog.itpub.net/28299513/viewspace-747008/ ,如需转载,请注明出处,否则将追究法律责任。

上一篇: 没有了~
下一篇: 没有了~
请登录后发表评论 登录
全部评论

注册时间:2012-10-23

  • 博文量
    1
  • 访问量
    5962