math_cms/app/Services/AsyncMarkdownSplitter.php

<?php

namespace App\Services;

use Illuminate\Support\Facades\Log;

class AsyncMarkdownSplitter
{
    /**
     * 将 Markdown 切分为题目数组
     *
     * @param string $markdown 原始 Markdown 文本
     * @return array 题目数组，每个元素包含 index 和 raw_markdown
     */
    public function split(string $markdown): array
    {
        // 使用正则表达式识别题号作为切分点（只接受“数字 + 明确分隔符”）
        // 注意：不要用 “数字 + 空白” 作为切分点，会误切正文中的列表/步骤/年份等。
        $pattern = '/^\s*(\d{1,4})(?:[\\.．、\\)）\\]】])\\s*/m';

        // 找到所有匹配的位置
        preg_match_all($pattern, $markdown, $matches, PREG_OFFSET_CAPTURE);

        $candidates = [];

        if (empty($matches[0])) {
            // 没有找到题号，整个作为一块
            return [
                [
                    'index' => 1,
                    'raw_markdown' => trim($markdown)
                ]
            ];
        }

        // 构建分块
        $positions = [];
        foreach ($matches[0] as $match) {
            $positions[] = $match[1];
        }

        for ($i = 0; $i < count($positions); $i++) {
            $start = $positions[$i];
            $end = $i + 1 < count($positions) ? $positions[$i + 1] : strlen($markdown);

            $block = substr($markdown, $start, $end - $start);
            $block = trim($block);

            if (!empty($block)) {
                // 提取题号作为 index
                preg_match('/^\s*(\d+)/', $block, $indexMatch);
                $index = $indexMatch[1] ?? ($i + 1);

                $candidates[] = [
                    // sequence：文件内顺序，保证唯一，不会因为 index 重复而覆盖
                    'sequence' => $i + 1,
                    'index' => (int)$index,
                    'raw_markdown' => $block
                ];
            }
        }

        return $candidates;
    }

    /**
     * 验证切分结果
     *
     * @param array $candidates 切分结果
     * @return bool
     */
    public function validate(array $candidates): bool
    {
        // 题号重复在“多套试卷/多章节合并”场景是正常现象，不应判定为失败。
        // 仅做轻量日志，避免输出超长 indexes 列表刷屏。
        $indexes = array_map(fn($item) => $item['index'], $candidates);
        $uniqueCount = count(array_unique($indexes));
        $total = count($indexes);

        if ($total > 0 && $uniqueCount !== $total) {
            Log::warning('Duplicate question indexes detected', [
                'total' => $total,
                'unique' => $uniqueCount,
            ]);
        }

        // 检查每个候选是否有内容
        foreach ($candidates as $candidate) {
            if (empty($candidate['raw_markdown'])) {
                Log::warning('Empty markdown content detected', [
                    'index' => $candidate['index']
                ]);
                return false;
            }
        }

        return true;
    }

    /**
     * 获取切分统计信息
     *
     * @param array $candidates 切分结果
     * @return array
     */
    public function getStatistics(array $candidates): array
    {
        $total = count($candidates);
        $avgLength = 0;
        $maxLength = 0;
        $minLength = PHP_INT_MAX;

        foreach ($candidates as $candidate) {
            $length = strlen($candidate['raw_markdown']);
            $avgLength += $length;
            $maxLength = max($maxLength, $length);
            $minLength = min($minLength, $length);
        }

        if ($total > 0) {
            $avgLength = round($avgLength / $total, 2);
        }

        return [
            'total_candidates' => $total,
            'avg_length' => $avgLength,
            'max_length' => $maxLength,
            'min_length' => $minLength === PHP_INT_MAX ? 0 : $minLength
        ];
    }
}