uiharu/src/Lookup/WebLookup.php

<?php
namespace Uiharu\Lookup;

use stdClass;
use DOMDocument;
use RuntimeException;
use Uiharu\Config;
use Uiharu\FFMPEG;
use Uiharu\MediaTypeExts;
use Uiharu\Url;
use Index\MediaType;

// TODO: Content-Disposition should be honoured for the filename (title).
final class WebLookup implements \Uiharu\ILookup {
    public function match(Url $url): bool {
        return $url->isWeb();
    }

    private static function reqCreate(string $url) {
        $curl = curl_init($url);
        curl_setopt_array($curl, [
            CURLOPT_AUTOREFERER => true,
            CURLOPT_CERTINFO => false,
            CURLOPT_FAILONERROR => false,
            CURLOPT_FOLLOWLOCATION => true,
            CURLOPT_MAXREDIRS => 5,
            CURLOPT_PATH_AS_IS => true,
            CURLOPT_RETURNTRANSFER => true,
            CURLOPT_TCP_FASTOPEN => true,
            CURLOPT_CONNECTTIMEOUT => 2,
            CURLOPT_PROTOCOLS => CURLPROTO_HTTP | CURLPROTO_HTTPS,
            CURLOPT_REDIR_PROTOCOLS => CURLPROTO_HTTP | CURLPROTO_HTTPS,
            CURLOPT_TIMEOUT => 5,
            CURLOPT_DEFAULT_PROTOCOL => 'https',
            CURLOPT_USERAGENT => 'Mozilla/5.0 (compatible) Uiharu/' . UIH_VERSION,
            CURLOPT_HTTPHEADER => [
                'Accept: text/html,application/xhtml+xml',
            ],
        ]);
        return $curl;
    }

    private static function reqHead($curl): array|null {
        curl_setopt_array($curl, [
            CURLOPT_NOBODY => true,
            CURLOPT_HEADER => true,
        ]);

        $headers = curl_exec($curl);
        if($headers === false)
            return null;

        $headers = explode("\r\n", trim($headers));
        $status = 200;
        $lines = [];

        foreach($headers as $header) {
            if(empty($header))
                continue;

            if(strpos($header, ':') === false) {
                $headParts = explode(' ', $header);
                if(isset($headParts[1]) && is_numeric($headParts[1]))
                    $status = (int)$headParts[1];
                $lines = [];
                continue;
            }

            $parts = explode(':', $header, 2);
            $parts[0] = mb_strtolower($parts[0]);
            if(isset($lines[$parts[0]]))
                $lines[$parts[0]] .= ', ' . trim($parts[1] ?? '');
            else
                $lines[$parts[0]] = trim($parts[1] ?? '');
        }

        return compact('status', 'lines');
    }

    private static function reqBody($curl): string|false {
        curl_setopt_array($curl, [
            CURLOPT_NOBODY => false,
            CURLOPT_HEADER => false,
        ]);

        return curl_exec($curl);
    }

    private static function reqError($curl): string {
        return curl_error($curl);
    }

    private static function reqClose($curl): void {
        curl_close($curl);
    }

    public function lookup(Url $url): WebLookupResult {
        $req = self::reqCreate($url);
        $head = self::reqHead($req);

        if($head === null)
            throw new RuntimeException('Web request timed out: ' . self::reqError($req));

        try {
            $mediaType = MediaType::parse($head['lines']['content-type'] ?? '');
        } catch(InvalidArgumentException $ex) {
            $mediaType = MediaType::parse('application/octet-stream');
        }

        $isXHTML = $mediaType->equals('application/xhtml+xml');
        if($isXHTML || $mediaType->equals('text/html'))
            return $this->lookupSite($url, $req, $mediaType, $isXHTML);

        self::reqClose($req);

        if(MediaTypeExts::isMedia($mediaType))
            return $this->lookupMedia($url, $mediaType);

        return new WebLookupFallbackResult($url, $mediaType, $url->getHost() . ': ' . basename($url->getPath()));
    }

    private function lookupSite(Url $url, $req, MediaType $mediaType, bool $isXHTML): WebLookupResult {
        $body = self::reqBody($req);
        self::reqClose($req);

        $document = new DOMDocument;
        if($isXHTML) {
            $document->loadXML($body, LIBXML_NOERROR | LIBXML_NONET | LIBXML_NOWARNING);
        } else {
            $document->loadHTML($body, LIBXML_NOERROR | LIBXML_NONET | LIBXML_NOWARNING);
            foreach($document->childNodes as $child)
                if($child->nodeType === XML_PI_NODE) {
                    $document->removeChild($child);
                    break;
                }

            $document->encoding = $mediaType->getCharset();
        }

        $charSet = $document->encoding;

        $siteInfo = new stdClass;
        $siteInfo->title = '';
        $siteInfo->metaTitle = '';
        $siteInfo->desc = '';
        $siteInfo->siteName = '';
        $siteInfo->image = '';
        $siteInfo->colour = '';
        $siteInfo->type = 'website';

        $titleTag = $document->getElementsByTagName('title');
        foreach($titleTag as $tag) {
            $siteInfo->title = trim(mb_convert_encoding($tag->textContent, 'utf-8', $charSet));
            break;
        }

        $metaTags = $document->getElementsByTagName('meta');
        foreach($metaTags as $tag) {
            $nameAttr = $tag->hasAttribute('name') ? $tag->getAttribute('name') : (
                $tag->hasAttribute('property') ? $tag->getAttribute('property') : ''
            );
            $valueAttr = $tag->hasAttribute('value') ? $tag->getAttribute('value') : (
                $tag->hasAttribute('content') ? $tag->getAttribute('content') : ''
            );
            $nameAttr  = trim(mb_convert_encoding($nameAttr,  'utf-8', $charSet));
            $valueAttr = trim(mb_convert_encoding($valueAttr, 'utf-8', $charSet));
            if(empty($nameAttr) || empty($valueAttr))
                continue;

            switch($nameAttr) {
                case 'og:title':
                case 'twitter:title':
                    $siteInfo->metaTitle = $valueAttr;
                    break;

                case 'description':
                case 'og:description':
                case 'twitter:description':
                    if(empty($siteInfo->desc))
                        $siteInfo->desc = $valueAttr;
                    break;

                case 'og:site_name':
                    $siteInfo->siteName = $valueAttr;
                    break;

                case 'og:image':
                case 'twitter:image':
                    $siteInfo->image = $valueAttr;
                    break;

                case 'theme-color':
                    $siteInfo->colour = $valueAttr;
                    break;

                case 'og:type':
                    $siteInfo->type = 'website:' . $valueAttr;
                    break;
            }
        }

        return new WebLookupSiteResult($url, $mediaType, $siteInfo);
    }

    private function lookupMedia(Url $url, MediaType $mediaType): WebLookupResult {
        $mediaInfo = FFMPEG::cleanProbe($url);
        return new WebLookupMediaResult($url, $mediaType, $mediaInfo);
    }
}