trove/app/Actions/FetchPageAction.php

<?php

declare(strict_types=1);

namespace App\Actions;

use App\Enums\CrawlOutcomeEnum;
use App\Services\UrlService;
use App\ValueObjects\FetchResult;
use fivefilters\Readability\Configuration;
use fivefilters\Readability\Readability;
use GuzzleHttp\Exception\ConnectException;
use Illuminate\Http\Client\ConnectionException;
use Illuminate\Http\Client\Factory;
use Illuminate\Http\Client\Response;
use InvalidArgumentException;
use League\Uri\BaseUri;
use Symfony\Component\DomCrawler\Crawler;
use Throwable;

class FetchPageAction
{
    public function __construct(
        private Factory $http,
        private UrlService $urlService,
    ) {}

    public function __invoke(string $url): FetchResult
    {
        try {
            $response = $this->http
                ->timeout(config('crawler.timeout'))
                ->withHeaders([
                    'User-Agent' => config('crawler.user_agent'),
                    'Accept' => 'text/html',
                ])
                ->withOptions([
                    'allow_redirects' => ['max' => config('crawler.max_redirects')],
                ])
                ->get($url);

        } catch (ConnectionException|ConnectException $e) {
            return $this->failureResult($e);
        }

        [$outcome, $error] = $this->validateResponse($response);

        if ($outcome === CrawlOutcomeEnum::Success) {
            [$title, $extractedText, $links] = $this->extractTitleTextAndLinks($response->body(), $url);
            $wordCount = $extractedText !== '' ? count(preg_split('/\s+/u', trim($extractedText))) : 0;
        }

        return new FetchResult(
            outcome: $outcome,
            statusCode: $response->status(),
            finalUrl: $url,
            title: $title ?? null,
            extractedText: $extractedText ?? null,
            outboundLinks: $links ?? collect(),
            wordCount: $wordCount ?? null,
            errorMessage: $error ?? null,
        );
    }

    private function validateResponse(Response $response): array
    {
        $status = $response->status();

        if ($status >= 400 && $status < 500) {
            return [CrawlOutcomeEnum::Blocked4xx, "HTTP {$status}"];
        }

        if ($status >= 500) {
            return [CrawlOutcomeEnum::Blocked5xx, "HTTP {$status}"];
        }

        $contentType = $response->header('Content-Type');
        if (! str_starts_with(mb_strtolower($contentType), 'text/html')) {
            return [CrawlOutcomeEnum::Rejected, "Unsupported file type: {$contentType}"];
        }

        return [CrawlOutcomeEnum::Success, null];
    }

    private function failureResult(ConnectionException|ConnectException $e): FetchResult
    {
        $guzzleException = $e instanceof ConnectException
            ? $e
            : ($e->getPrevious() instanceof ConnectException
                ? $e->getPrevious()
                : null);

        $errno = $guzzleException?->getHandlerContext()['errno'] ?? null;

        $outcome = $errno === CURLE_OPERATION_TIMEDOUT
            ? CrawlOutcomeEnum::Timeout
            : CrawlOutcomeEnum::Failed;

        return new FetchResult(
            outcome: $outcome,
            statusCode: null,
            finalUrl: null,
            title: null,
            extractedText: null,
            outboundLinks: collect(),
            wordCount: null,
            errorMessage: $e->getMessage(),
        );
    }

    private function extractTitleTextAndLinks(string $body, string $url): array
    {
        $crawler = new Crawler($body);

        $title = $crawler->filter('title')->count() > 0
            ? trim($crawler->filter('title')->text())
            : null;

        $readability = new Readability(new Configuration);
        $readability->parse($body);
        $mainContent = $readability->getContent() ?? '';
        $extractedText = trim(strip_tags($mainContent));

        $links = collect();
        if ($mainContent !== '') {
            $linkCrawler = new Crawler($mainContent);
            if ($linkCrawler->filter('a[href]')->count() > 0) {
                $links = collect($linkCrawler->filter('a[href]')->extract(['href']));
            }
        }

        $linksResolved = $links
            ->map(fn (string $href) => $this->resolveAndValidateLink($href, $url))
            ->filter()
            ->unique()
            ->values();

        return [$title, $extractedText, $linksResolved];
    }

    private function resolveAndValidateLink(string $href, string $finalUrl): ?string
    {
        try {
            $resolved = (string) BaseUri::from($finalUrl)->resolve($href);
            $resolved = strstr($resolved, '#', true) ?: $resolved;
        } catch (Throwable) {
            return null;
        }

        if ($resolved === $finalUrl) {
            return null;
        }

        try {
            $this->urlService->host($resolved);
        } catch (InvalidArgumentException) {
            return null;
        }

        return $resolved;
    }
}