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

<?php

namespace App\Services\Crawl;

use App\Models\CrawlAddress;
use Illuminate\Support\Collection;

class CrawlAddressSourceResolver
{
    /** @var array<string, Collection<int, CrawlAddress>> */
    protected array $cache = [];

    public function resolveByRequestUrl(?string $url, string $targetType = 'industry_news'): ?string
    {
        $normalized = $this->normalizeUrl($url);
        if ($normalized === '') {
            return null;
        }

        return $this->addresses($targetType)
            ->first(fn (CrawlAddress $address) => $this->normalizeUrl($address->request_url) === $normalized)
            ?->name;
    }

    public function resolveBySourceUrl(?string $sourceUrl, string $targetType = 'industry_news'): ?string
    {
        $host = strtolower((string) parse_url((string) $sourceUrl, PHP_URL_HOST));
        if ($host === '') {
            return null;
        }

        $matches = $this->addresses($targetType)
            ->filter(function (CrawlAddress $address) use ($host) {
                $addressHost = strtolower((string) parse_url($address->request_url, PHP_URL_HOST));

                return $addressHost !== '' && $addressHost === $host;
            })
            ->values();

        if ($matches->isEmpty()) {
            return null;
        }

        if ($matches->count() === 1) {
            return $matches->first()->name;
        }

        $path = (string) (parse_url((string) $sourceUrl, PHP_URL_PATH) ?: '/');
        $bestName = null;
        $bestLength = 0;

        foreach ($matches as $address) {
            $prefix = $this->listPathPrefix($address->request_url);
            if ($prefix !== '' && str_starts_with($path, $prefix) && strlen($prefix) > $bestLength) {
                $bestName = $address->name;
                $bestLength = strlen($prefix);
            }
        }

        return $bestName;
    }

    public function resolveForNews(?string $jobRequestUrl, ?string $articleSourceUrl, string $targetType = 'industry_news'): ?string
    {
        return $this->resolveByRequestUrl($jobRequestUrl, $targetType)
            ?? $this->resolveBySourceUrl($articleSourceUrl, $targetType);
    }

    /**
     * @return list<string>
     */
    public function genericAdapterSourceNames(): array
    {
        return [
            '通用资讯 HTML',
            '爬虫采集',
        ];
    }

    /**
     * @return Collection<int, CrawlAddress>
     */
    protected function addresses(string $targetType): Collection
    {
        if (! isset($this->cache[$targetType])) {
            $this->cache[$targetType] = CrawlAddress::query()
                ->where('target_type', $targetType)
                ->where('status', 1)
                ->orderBy('sort')
                ->orderBy('name')
                ->get();
        }

        return $this->cache[$targetType];
    }

    protected function normalizeUrl(?string $url): string
    {
        $trimmed = trim((string) $url);
        if ($trimmed === '') {
            return '';
        }

        if (! preg_match('#^https?://#i', $trimmed)) {
            $trimmed = 'https://'.$trimmed;
        }

        $parts = parse_url($trimmed);
        if (! is_array($parts) || empty($parts['host'])) {
            return rtrim($trimmed, '/');
        }

        $scheme = strtolower((string) ($parts['scheme'] ?? 'https'));
        $host = strtolower((string) $parts['host']);
        $path = $parts['path'] ?? '/';
        $path = $path === '' ? '/' : $path;
        $path = rtrim($path, '/') ?: '/';

        return $scheme.'://'.$host.$path;
    }

    protected function listPathPrefix(string $url): string
    {
        $path = (string) (parse_url($url, PHP_URL_PATH) ?: '/');
        if ($path === '/') {
            return '/';
        }

        $path = rtrim($path, '/');
        $lastSlash = strrpos($path, '/');
        if ($lastSlash === false) {
            return $path;
        }

        return substr($path, 0, $lastSlash + 1);
    }
}