fix: filter and repair of some records

src/analyse.php

@@ -225,6 +225,9 @@ function splitHtml($htmlFile) {
     }
     foreach ($result as &$row) {
         $row = trim($row);
+        $row = str_replace(''', '\'', $row);
+        $row = str_replace('&', '&', $row);
+        $row = str_replace('"', '"', $row);
     }
     return $result;
 }
@@ -260,7 +263,7 @@ function getHtmlType($str) { // 提取TLD类型
         case '(Restricted generic top-level domain)':
             return 'Restricted TLD';
         case '(Test top-level domain)':
-            return 'TLD for test';
+            return 'Test TLD';
         default:
             die('error analyse -> type');
     }
@@ -269,8 +272,8 @@ function getHtmlType($str) { // 提取TLD类型
 function getHtmlManager($str) { // 提取TLD所有者信息
     if ($str == '') {
         return array(
-            'manager' => '',
+            'name' => array(),
-            'manager_info' => ''
+            'addr' => array()
         );
     }
     $temp = explode('</b><br/>', $str);
@@ -278,17 +281,19 @@ function getHtmlManager($str) { // 提取TLD所有者信息
         die('error analyse -> manager');
     }
     $manager = trim($temp[0]);
+    preg_match('/\\\u[0-9a-f]{4}/', $manager, $match);
     $manager = substr($manager, 3 - strlen($manager));
     if ($manager === 'Not assigned') {
         return array(
-            'manager' => '',
+            'name' => array(),
-            'manager_info' => ''
+            'addr' => array()
         );
     }
+    $manager = explode('<br>', $manager);
     if ($temp[1] == '') {
         return array(
-            'manager' => $manager,
+            'name' => $manager,
-            'manager_info' => ''
+            'addr' => array()
         );
     }
     $temp = str_replace('<br>', '<br/>', trim($temp[1]));
@@ -303,8 +308,8 @@ function getHtmlManager($str) { // 提取TLD所有者信息
         die('error analyse -> manager');
     }
     return array(
-        'manager' => $manager,
+        'name' => $manager,
-        'manager_addr' => $manager_addr
+        'addr' => $manager_addr
     );
 }
 
@@ -370,7 +375,7 @@ function getHtmlContact($str) { // 提取联系人信息
     $flag = false;
     foreach ($temp as $line) {
         if (!$flag) {
-            $result['org'] = $line;
+            $result['org'] = preg_replace('/[\s]+/', ' ', $line);
             $flag = true;
             continue;
         }
@@ -398,15 +403,21 @@ function getHtmlContact($str) { // 提取联系人信息
     if ($result['email'] != '' && substr($result['email'], -5) !== '<br/>') {
         die('error analyse -> contact');
     }
+    if ($result['email'] !== '') {
         $result['email'] = substr($result['email'], 0, strlen($result['email']) - 5);
+    }
     if ($result['voice'] != '' && substr($result['voice'], -5) !== '<br/>') {
         die('error analyse -> contact');
     }
+    if ($result['voice'] !== '') {
         $result['voice'] = substr($result['voice'], 0, strlen($result['voice']) - 5);
+    }
     if ($result['fax'] != '' && substr($result['fax'], -5) !== '<br/>') {
         die('error analyse -> contact');
     }
+    if ($result['fax'] !== '') {
         $result['fax'] = substr($result['fax'], 0, strlen($result['fax']) - 5);
+    }
     if ($result['fax'] === 'n/a' || $result['fax'] === 'NA' || $result['fax'] === 'N/A' || $result['fax'] === '-') {
         $result['fax'] = '';
     }
@@ -476,12 +487,6 @@ function getHtmlNS($str) { // 提取TLD名称服务器
 }
 
 function getHtmlInfo($str) { // 获取官网/Whois服务器信息
-    // if ($str == '') {
-    //     return array(
-    //         'website' => '',
-    //         'whois' => ''
-    //     );
-    // }
     preg_match_all('/<p>[\s\S]+?<\/p>/', $str, $match);
     if (count($match) !== 1) {
         die('error analyse -> info');

src/load.php

@@ -8,7 +8,7 @@ function getTldsInfo($tldList, $htmlDir) { // 抓取各个TLD数据
             die('error analyse -> title');
         }
         $info['type'] = getHtmlType($html['type']);
-        $info += getHtmlManager($html['manager']);
+        $info['manager'] = getHtmlManager($html['manager']);
         $info['admin_contact'] = getHtmlContact($html['admin']);
         $info['tech_contact'] = getHtmlContact($html['tech']);
         $info['nameserver'] = getHtmlNS($html['ns']);

src/main.php

@@ -4,30 +4,60 @@ require_once './load.php';
 require_once './analyse.php';
 require_once './punycode.php';
 
-// function writeFile($filename, $data) {
+$html_path = '../html/';
-//     $file = fopen($filename, 'w');
+$release_path = '../release/';
-//     fwrite($file, $data);
+
-//     fclose($file);
+// Get TLD list from IANA website
-// }
+// main.html -> https://www.iana.org/domains/root/db
+$tlds = getIanaTlds($html_path . 'main.html');
+writeFile($release_path . 'tld-list.txt', implode(PHP_EOL, $tlds) . PHP_EOL);
+
+// Analyse all TLDs from html files
+$data = getTldsInfo($tlds, $html_path . 'tlds/');
+
+// $str = $data['.to']['admin_contact']['org'];
+// echo preg_replace('/[\s]+/', ' ', $str);
+// echo $str;
+// exit;
+
+// var_dump($data['.com']);
+// exit;
+foreach ($data as $index => $row) {
+    // if ($row['admin_contact']['voice'] === false) {
+        // echo $index . ' -> ' . implode(' | ', $row['manager']['addr']) . PHP_EOL;
+        // if (count($row['nameserver']) === 0) { continue; }
+        // echo $row['website'] . PHP_EOL;
+        // echo $row['whois'] . PHP_EOL;
+        // echo $row['last_updated'] . PHP_EOL;
+        // echo $row['regist_date'] . PHP_EOL;
+        // echo $index . ' -> ' . $row['tech_contact']['name'] . PHP_EOL;
+        // echo $index . ' -> ' . $row['tech_contact']['org'] . PHP_EOL;
+        // echo $index . ' -> ' . implode(' | ', $row['tech_contact']['addr']) . PHP_EOL;
+        // echo $index . ' -> ' . $row['tech_contact']['email'] . PHP_EOL;
+        // echo $index . ' -> ' . $row['tech_contact']['voice'] . PHP_EOL;
+        // echo $index . ' -> ' . $row['tech_contact']['fax'] . PHP_EOL;
 
-// $data = getIanaTlds('../html/main.html'); // https://www.iana.org/domains/root/db
-// // writeFile('tlds.txt', implode(PHP_EOL, $data) . PHP_EOL);
-// $urls = '';
-// foreach ($data as $tld) {
-//     $urls .= 'https://www.iana.org/domains/root/db/';
-//     $urls .= substr($tld, 1, strlen($tld) - 1);
-//     $urls .= '.html' . PHP_EOL;
     // }
-// writeFile('urls.txt', $urls);
+}
+
+// return;
 
-$tlds = getIanaTlds('../html/main.html');
+// Output data by json format
-$data = getTldsInfo($tlds, '../html/tlds/');
+writeFile($release_path . 'all-data.json', json_encode($data));
-// echo count($data);
 
+// Output whois server list by csv format
+$whoisStr = '';
 foreach ($data as $index => $row) {
     if ($row['whois'] !== '') {
-        echo $index . ' -> ' . $row['whois'] . PHP_EOL;
+        $whoisStr .= $index . ',' . $row['whois'] . PHP_EOL;
     }
 }
+writeFile('../release/whois-server.csv', $whoisStr);
+
+function writeFile($filename, $data) {
+    $file = fopen($filename, 'w');
+    fwrite($file, $data);
+    fclose($file);
+}
 
 ?>