slake-school-service/app/Services/Crawl/CrawlAuthorParser.php

<?php

namespace App\Services\Crawl;

class CrawlAuthorParser
{
    /**
     * @param  list<array{name:string,email?:?string,affiliation?:?string,university_name?:?string}>  $authorsParsed
     * @return array{name:string,email:?string,affiliation:?string,university_name:?string}|null
     */
    public static function leadAuthor(?string $authorsString, array $authorsParsed = []): ?array
    {
        if ($authorsParsed !== []) {
            $first = $authorsParsed[0];
            $name = trim((string) ($first['name'] ?? ''));
            if ($name !== '') {
                return [
                    'name' => $name,
                    'email' => self::normalizeEmail($first['email'] ?? null),
                    'affiliation' => self::cleanText($first['affiliation'] ?? null),
                    'university_name' => self::universityFromAffiliation($first['affiliation'] ?? null)
                        ?? self::cleanText($first['university_name'] ?? null),
                ];
            }
        }

        if (! $authorsString) {
            return null;
        }

        $parts = preg_split('/[;；,，]/u', $authorsString) ?: [];
        $name = trim((string) ($parts[0] ?? ''));
        if ($name === '') {
            return null;
        }

        return [
            'name' => $name,
            'email' => null,
            'affiliation' => null,
            'university_name' => null,
        ];
    }

    /**
     * @return list<array{name:string,email:?string,affiliation:?string,university_name:?string}>
     */
    public static function splitAuthorsString(?string $authors): array
    {
        if (! $authors) {
            return [];
        }

        $rows = [];
        foreach (preg_split('/[;；]/u', $authors) ?: [] as $chunk) {
            $name = trim($chunk);
            if ($name !== '') {
                $rows[] = ['name' => $name, 'email' => null, 'affiliation' => null, 'university_name' => null];
            }
        }

        return $rows;
    }

    public static function normalizeEmail(?string $email): ?string
    {
        $email = trim((string) $email);
        if ($email === '' || ! filter_var($email, FILTER_VALIDATE_EMAIL)) {
            return null;
        }

        return strtolower($email);
    }

    public static function cleanText(?string $text): ?string
    {
        $text = self::ensureUtf8((string) $text);
        if ($text === '') {
            return null;
        }

        $text = trim(preg_replace('/\s+/u', ' ', $text) ?? '');

        return $text === '' ? null : $text;
    }

    public static function ensureUtf8(string $text): string
    {
        if ($text === '') {
            return '';
        }

        if (! mb_check_encoding($text, 'UTF-8')) {
            $converted = @mb_convert_encoding($text, 'UTF-8', 'UTF-8,GBK,GB2312,BIG5,ISO-8859-1');
            $text = is_string($converted) ? $converted : $text;
        }

        if (function_exists('mb_scrub')) {
            $text = mb_scrub($text, 'UTF-8');
        } else {
            $text = iconv('UTF-8', 'UTF-8//IGNORE', $text) ?: '';
        }

        return $text;
    }

    /**
     * 判断文本是否像真实职称（过滤导航文案如「电气首页」）。
     */
    public static function looksLikeAcademicTitle(?string $title): bool
    {
        $title = trim((string) $title);
        if ($title === '' || mb_strlen($title) > 30) {
            return false;
        }

        if (preg_match('/首页|导航|菜单|返回|更多|搜索|登录|关于/u', $title)) {
            return false;
        }

        return (bool) preg_match(
            '/教授|副教授|讲师|助教|研究员|副研究员|助理研究员|实验师|工程师|院士|博导|导师|专家|学者|长聘|准聘|特聘|特任|教轨|兼职|访问|青年|副高|正高|中级|初级/u',
            $title,
        );
    }

    /**
     * 递归清洗数组/字符串，确保可安全 json_encode。
     *
     * @param  mixed  $value
     * @return mixed
     */
    public static function sanitizeUtf8Deep(mixed $value): mixed
    {
        if (is_string($value)) {
            return self::ensureUtf8($value);
        }

        if (is_array($value)) {
            $out = [];
            foreach ($value as $key => $item) {
                $safeKey = is_string($key) ? self::ensureUtf8($key) : $key;
                $out[$safeKey] = self::sanitizeUtf8Deep($item);
            }

            return $out;
        }

        return $value;
    }

    public static function universityFromAffiliation(?string $affiliation): ?string
    {
        $affiliation = self::cleanText($affiliation);
        if (! $affiliation) {
            return null;
        }

        if (preg_match('/^([^,，;；]+(?:大学|学院|研究院|研究所|University|College)[^,，;；]*)/iu', $affiliation, $m)) {
            return trim($m[1]);
        }

        return null;
    }
}