研学线路

app/Support/StudyTourDeclarationParser.php

@@ -139,6 +139,9 @@ class StudyTourDeclarationParser
 
                 continue;
             }
+            if ($key === 'implementation_html' && mb_strlen($plain) > 150) {
+                $score -= 80;
+            }
             $score += 12 + min(mb_strlen($plain), 120) / 20;
         }
 
@@ -159,7 +162,7 @@ class StudyTourDeclarationParser
 
         $introText = self::joinSectionLines($sections['intro'] ?? []);
         $feeText = self::joinSectionLines($sections['fee'] ?? []);
-        $implText = self::joinSectionLines($sections['impl'] ?? []);
+        $implText = self::joinSectionLines(self::filterImplementationLines($sections['impl'] ?? []));
 
         $routePlans = self::parseRouteSection($sections['route'] ?? []);
         $courses = self::parseCoursesSection($sections['courses'] ?? [], $routePlans);
@@ -815,6 +818,71 @@ class StudyTourDeclarationParser
         return (bool) preg_match('/(体验|探秘|制作|参观|采摘|仪式|晚宴|午餐|晚餐|早餐|午休|活动|课程|王国|有趣|小夜灯|缫丝|挂件|面包)/u', $line);
     }
 
+    /**
+     * @param  array<int, string>  $lines
+     * @return array<int, string>
+     */
+    private static function filterImplementationLines(array $lines): array
+    {
+        $kept = [];
+        foreach ($lines as $line) {
+            if ($line === '') {
+                continue;
+            }
+            if (! self::isValidImplementationLine($line)) {
+                break;
+            }
+            $kept[] = $line;
+        }
+
+        return $kept;
+    }
+
+    private static function isValidImplementationLine(string $line): bool
+    {
+        if (self::isGarbledLine($line) || self::isWordMetadataLine($line)) {
+            return false;
+        }
+
+        if (preg_match('/[￿]/u', $line)) {
+            return false;
+        }
+
+        if (preg_match('/(?:xmlns|http:|xml |ContentTypes|accent[0-9]|folHlink|theme\/|drawingml)/iu', $line)) {
+            return false;
+        }
+
+        if (preg_match('/^\d{1,2}月\d{1,2}日/u', $line)) {
+            return true;
+        }
+
+        if (! preg_match('/[月日场活动安排如下夏令营研学实施时间共]/u', $line)) {
+            return false;
+        }
+
+        $cjk = preg_match_all('/[\x{4e00}-\x{9fff}]/u', $line) ?: 0;
+
+        return $cjk >= 4 && ($cjk / max(mb_strlen($line), 1)) >= 0.45;
+    }
+
+    private static function isWordMetadataLine(string $line): bool
+    {
+        static $needles = [
+            '正文', '默认段落字体', '普通表格', '无列表', '页脚', '页眉', '网格型',
+            '微软用户', 'WordDocument', 'DocumentSummary', 'SummaryInformation',
+            'CompObj', 'ProductBuild', 'KSOProduct', 'Normal.dotm', 'Microsoft Office',
+            'xmlns', 'ContentTypes', 'theme/theme', 'themeManager', 'fontTable',
+            'Root Entry', 'SMWordDoc', 'Word.Document',
+        ];
+        foreach ($needles as $needle) {
+            if (str_contains($line, $needle)) {
+                return true;
+            }
+        }
+
+        return (bool) preg_match('/^(黑体|宋体|微软雅黑|等线(?: Light)?|Arial|Symbol|Tahoma|Times New Roman|Calibri|Cambria Math|Segoe UI Emoji|SimHei|SimSun|DengXian)$/u', $line);
+    }
+
     private static function isGarbledLine(string $line): bool
     {
         if ($line === '') {
@@ -829,10 +897,18 @@ class StudyTourDeclarationParser
             return false;
         }
 
+        if (self::isWordMetadataLine($line)) {
+            return true;
+        }
+
         if (preg_match('/[\x{0080}-\x{009F}]/u', $line)) {
             return true;
         }
 
+        if (preg_match('/[\x{0200}-\x{024F}\x{0400}-\x{04FF}\x{0500}-\x{052F}\x{0600}-\x{06FF}]/u', $line)) {
+            return true;
+        }
+
         if (preg_match('/[ᘀ-᛿]/u', $line)) {
             return true;
         }
@@ -841,17 +917,29 @@ class StudyTourDeclarationParser
             return true;
         }
 
+        if (preg_match('/[￿]/u', $line)) {
+            return true;
+        }
+
+        if (preg_match('/(?:xmlns|http:|ContentTypes|Normal\.dotm|Word\.Document)/iu', $line)) {
+            return true;
+        }
+
+        if (preg_match('/[^\x{4e00}-\x{9fff}0-9A-Za-z，。、：；（）""\'\-\s\/&“”]/u', $line) && mb_strlen($line) <= 8) {
+            return true;
+        }
+
         $cjk = preg_match_all('/[\x{4e00}-\x{9fff}]/u', $line) ?: 0;
         if ($cjk === 0 && mb_strlen($line) >= 3) {
             return true;
         }
 
         $len = mb_strlen($line);
-        if ($len < 6) {
-            return $cjk === 0;
+        if ($len <= 5) {
+            return $cjk < 2;
         }
 
-        $readable = preg_match_all('/[\x{4e00}-\x{9fff}0-9A-Za-z，。、：；（）\-\s\/&]/u', $line) ?: 0;
+        $readable = preg_match_all('/[\x{4e00}-\x{9fff}0-9A-Za-z，。、：；（）\-\s\/&“”]/u', $line) ?: 0;
 
         return ($readable / max($len, 1)) < 0.45;
     }