zenfolio-scraper/main.php

<?php

function getPaginatedUrl(int $photoSetId, int $offset, int $pageSize = 80) : string
{
    return sprintf(
        'https://api.zenfolio.com/api/private/1.7/zfapi.asmx/LoadPhotoSetPhotos'.
        '?startingIndex=%d&numberOfPhotos=%d&photosetId=%d',
        $offset,
        $pageSize,
        $photoSetId
    );
}

function getPhotosByPhotoSetId(int $photoSetId) : \Generator
{
    $page = 0;
    $pageSize = 1000;

    // Set up curl for fetching
    $curl = curl_init();
    curl_setopt($curl, CURLOPT_USERAGENT, 'curl/7.62.0');
    curl_setopt($curl, CURLOPT_RETURNTRANSFER, 1);

    while (true) {
        // Set URL to fetch
        curl_setopt($curl, CURLOPT_URL, getPaginatedUrl($photoSetId, $page * $pageSize, $pageSize));

        // Fetch url
        $result = curl_exec($curl);

        // Parse XML
        $sxe = simplexml_load_string($result);

        // If no photos, break the loop
        if (count($sxe->Photo) === 0) {
            break;
        }

        // Loop back photos
        foreach ($sxe->Photo as $photo) {
            // SimpleXML to stdClass
            yield json_decode(json_encode($photo));
        }

        // If less photos than pageSize, break the loop to not have to go to an empty page
        if (count($sxe->Photo) < $pageSize) {
            break;
        }

        $page++;
    }

    curl_close($curl);
}

function getPhotoGalleryMetadataByUrl(string $url, string $takenOn = null) : stdClass
{
    // Set up a dom document and fetch the url
    $doc = new DOMDocument();
    $doc->loadHTMLFile($url);

    // Assume that the page has a title element, grab the first one and trim it
    $pageTitle = $doc->getElementsByTagName('title')->item(0)->textContent;
    $pageTitle = trim(str_replace('Sportgardens fotogalleri |', '', $pageTitle));

    // Match out date from title
    preg_match('#(\d\d\d\d-?\d\d-?\d\d)#', $pageTitle, $matches);

    // We got gallery date
    $galleryDate = $matches[0];

    // Clean the date out of the title
    $galleryTitle = str_replace($galleryDate, '', $pageTitle);
    $galleryTitle = preg_replace('#\d.+\d#', '', $galleryTitle);
    $galleryTitle = preg_replace('#\s-#', '', $galleryTitle);
    $galleryTitle = preg_replace('#\s+#', ' ', $galleryTitle);

    // If dateformat is without dashes, reformat it
    if (strlen($galleryDate) === 8) {
        $galleryDate = sprintf(
            '%d-%02d-%02d',
            substr($galleryDate, 0, 4),
            substr($galleryDate, 4, 2),
            substr($galleryDate, 6, 2),
        );
    }

    if (empty($galleryDate) && $takenOn) {
        $galleryDate = date('Y-m-d', strtotime($takenOn));
    }

    // Put metadata together
    return (object) [
        'url' => $url,
        'date' => $galleryDate,
        'group' => trim($galleryTitle),
    ];
}

function getPhotoSetUrlsByPhotoSetId(int $photoSetId) : \Generator
{
    $photoSetUrls = [];

    foreach (getPhotosByPhotoSetId($photoSetId) as $photo) {
        $galleryUrl = preg_replace('#/\w+$#', '', $photo->PageUrl);

        if (!isset($photoSetUrls[$galleryUrl])) {
            $photoSetUrls[$galleryUrl] = getPhotoGalleryMetadataByUrl($galleryUrl, $photo->TakenOn);

            yield $photoSetUrls[$galleryUrl];
        }
    }
}

// Map years to photo gallery ids
$photoGalleryIds = [
    2016 => 3742586428944470221,
    2017 => 3742586429170821662,
    2018 => 3742586428889977272,
];

// Check arguments
if ($argv[1] === '--year' && isset($photoGalleryIds[$argv[2]])) {
    // Loop through all galleries
    foreach (getPhotoSetUrlsByPhotoSetId($photoGalleryIds[$argv[2]]) as $photoGallery) {
        echo $photoGallery->group.' - '.$photoGallery->date.' - '.$photoGallery->url.PHP_EOL;
    }

    return;
}

// Print help
echo 'Usage: php '.$argv[0].' --year [year]'.PHP_EOL;
echo 'Valid years: '.implode(', ', array_keys($photoGalleryIds)).PHP_EOL;