核心類:FilterTools
<?php
class FilterTools {
public static $keyword = [];
/**
* 從檔案載入敏感詞字典(每行一個詞)
*/
public static function getBadWords(string $filename): array {
$handle = fopen($filename, 'r');
if (!$handle) return [];
while (($line = fgets($handle)) !== false) {
$word = trim($line);
if ($word !== '') self::$keyword[] = $word;
}
fclose($handle);
return self::$keyword;
}
/**
* 過濾內容
*
* @param string $content 待處理文字
* @param string $replace 替換字元(如 '*'、'[敏感詞]')
* @param string $filename 字典檔路徑
* @param array $memConfig Memcache 連線設定 ['host'=>'', 'port'=>11211, 'key'=>'bad_words']
* @return string
*/
public static function filterContent(string $content, string $replace, string $filename, array $memConfig): string {
$mem = new BadWordsMemcache($filename, $memConfig);
$keywords = $mem->getList();
if (empty($keywords)) {
$keywords = self::getBadWords($filename);
}
// 建立 詞 -> 替換字串 映射表
$replaceMap = array_combine($keywords, array_fill(0, count($keywords), $replace));
return strtr($content, $replaceMap);
}
}
Memcache 快取包裝類
<?php
class BadWordsMemcache {
private $mem;
private $key;
private $filename;
public function __construct(string $filename, array $config) {
$this->filename = $filename;
$this->key = $config['key'] ?? 'bad_words';
$this->mem = new Memcache();
$this->mem->connect($config['host'] ?? 'localhost', $config['port'] ?? 11211);
}
public function getList(): array {
$data = $this->mem->get($this->key);
if ($data !== false) return $data;
// 快取未命中,載入檔案並寫入快取(TTL 1 小時)
$words = FilterTools::getBadWords($this->filename);
if ($words) {
$this->mem->set($this->key, $words, 0, 3600);
}
return $words;
}
// 後台更新字典後呼叫此方法重載
public function reload(): void {
$this->mem->delete($this->key);
$this->getList();
}
}
使用範例
<?php
require_once 'FilterTools.php';
$memConfig = [
'host' => '127.0.0.1',
'port' => 11211,
'key' => 'app:bad_words',
];
$text = "你媽的我操一色狼雜種二山食物";
$clean = FilterTools::filterContent($text, '*', __DIR__ . '/config/word.txt', $memConfig);
echo $clean; // ****我*一**雜種**食物
?>
效能優化建議
- Trie 樹 (前綴樹):大量敏感詞(>1萬)時,
strtr 掃描效能下降,建議構建 Trie 樹單次遍歷匹配(參考 php-sensitive-word)。
- 非同步更新:後台修改字典後,發送訊息佇列通知所有 PHP-FPM 進程呼叫
BadWordsMemcache::reload()。
- 分級過濾:嚴重詞直接攔截/拒絕,輕微詞替換標記,輸出前再統一處理。