DomFramework/src/Sitemap.php

<?php

/**
 * DomFramework
 * @package domframework
 * @author Dominique Fournier <dominique@fournier38.fr>
 * @license BSD
 */

namespace Domframework;

/**
 * This class allow to read the sitemaps files available in websites, and give
 * the available URL and parameters
 */
class Sitemap
{
    /**
     * Return an array containing the URL in sitemap associated with the
     * information of priority and changefreq (in seconds)
     * array (
     *   "urls" => array (
     *     [http://domain.tld/path...] => array (
     *       "changefreq" => XX,
     *       "priority" => ZZ),
     *     ),
     *   "sitemaps" => array (
     *     [http://domain.tld/sitemap2] => array ("lastmod" => XX)
     * ));
     * The lastmod is return in time() based on UTC
     * @param string $content The content file to analyze
     * @param string $url The website URL
     * @return array The content of the file if it is valid
     */
    public function analyze($content, $url)
    {
        $finfo = new \finfo(FILEINFO_MIME_TYPE);
        $type = $finfo->buffer($content);
        if ($type === "application/x-gzip") {
            // Uncompress GZIP
            $content = gzinflate(substr($content, 10));
        }

        $content = trim($content);
        $type = $finfo->buffer($content);
        if ($type === "text/plain" || $type === "application/x-empty") {
            return $this->analyzeText($content, $url);
        } elseif ($type === "application/xml" || $type === "text/xml") {
            return $this->analyzeXML($content, $url);
        } else {
            throw new \Exception("Type of sitemap '$type' unknown for '$url'", 406);
        }
    }

    /**
     * Return an array containing the URL in sitemap associated with an empty
     * array, as the Text format provide only URL, so return empty array.
     * New sitemaps files can not be defined in text format.
     * array (
     *   "urls" => array (
     *     [http://domain.tld/path...] => array (),
     *   ),
     *   "sitemaps" => array ()
     * );
     * @param string $content The content file to analyze
     * @param string $url The website URL
     * @return array The content of the file if it is valid
     */
    public function analyzeText($content, $url)
    {
        if (strlen($content) > 10000000) {
            trigger_error("Sitemap '$url' size is too big -> skip", E_USER_ERROR);
            return ["urls" => [], "sitemaps" => []];
        }
        $urls = preg_split('/\r\n|\r|\n/', trim($content));
        if ($urls === [""]) {
            $urls = [];
        }
        $urls = array_fill_keys($urls, []);
        return ["urls" => $urls, "sitemaps" => []];
    }

    /**
     * Return an array containing the URL in sitemap associated with the
     * information of priority and changefreq (in seconds)
     * array (
     *   "urls" => array (
     *     [http://domain.tld/path...] => array (
     *        "changefreq" => XX,
     *        "priority" => ZZ),
     *     ),
     *   "sitemaps" => array (
     *     [http://domain.tld/sitemap2] => array ("lastmod" => XX)
     * ));
     * @param string $content The content file to analyze
     * @param string $url The website URL
     * @return array The content of the file if it is valid
     */
    public function analyzeXML($content, $url)
    {
        if (strlen($content) > 10000000) {
            throw new \Exception("Sitemap '$url' size is too big -> skip", 406);
        }
        $xml = @simplexml_load_string($content);
        $res = ["urls" => [], "sitemaps" => []];
        foreach ($xml->sitemap as $s) {
            $tmp = [];
            if (! isset($s->loc)) {
                trigger_error("No location in sitemap '$url'", E_USER_WARNING);
                continue;
            }
            $loc = (string)$s->loc;
            if (isset($s->lastmod)) {
                $lastmod = (string)$s->lastmod;
                $ts = strptime($lastmod, "%Y-%m-%dT%H:%M:%S%z");
                if ($ts === false) {
                    $ts = strptime($lastmod, "%Y-%m-%d");
                }
                if ($ts === false) {
                    trigger_error(
                        "Invalid lastmod in '$url' ($lastmod)",
                        E_USER_WARNING
                    );
                } else {
                    $tmp["lastmod"] = gmmktime(
                        $ts['tm_hour'],
                        $ts['tm_min'],
                        $ts['tm_sec'],
                        $ts['tm_mon'] + 1,
                        $ts['tm_mday'],
                        ($ts['tm_year'] + 1900)
                    );
                }
            }
            $res["sitemaps"][$loc] = $tmp;
        }
        foreach ($xml->url as $u) {
            $tmp = [];
            if (! isset($u->loc)) {
                trigger_error(
                    "No location in sitemap '$url' for url",
                    E_USER_WARNING
                );
                continue;
            }
            $loc = (string)$u->loc;
            if (isset($u->changefreq)) {
                $changefreq = (string)$u->changefreq;
                switch (strtolower($changefreq)) {
                    case "always":
                    case "hourly":
                    case "daily":
                    case "weekly":
                    case "monthly":
                    case "yearly":
                    case "never":
                        $tmp["changefreq"] = strtolower($u->changefreq);
                        break;
                    default:
                        trigger_error(
                            "Invalid chagefreq in '$url' ($changefreq)",
                            E_USER_WARNING
                        );
                }
            }
            if (isset($u->priority)) {
                $priority = (float)$u->priority;
                if ($priority < 0.0 || $priority > 1.0) {
                    trigger_error(
                        "Invalid priority in '$url' ($priority)",
                        E_USER_WARNING
                    );
                } else {
                    $tmp["priority"] = $priority;
                }
            }
            if (isset($u->lastmod)) {
                $lastmod = (string)$u->lastmod;
                $ts = strptime($lastmod, "%Y-%m-%dT%H:%M:%S%z");
                if ($ts === false) {
                    $ts = strptime($lastmod, "%Y-%m-%d");
                }
                if ($ts === false) {
                    trigger_error(
                        "Invalid lastmod in '$url' ($lastmod)",
                        E_USER_WARNING
                    );
                } else {
                    $tmp["lastmod"] = gmmktime(
                        $ts['tm_hour'],
                        $ts['tm_min'],
                        $ts['tm_sec'],
                        $ts['tm_mon'] + 1,
                        $ts['tm_mday'],
                        ($ts['tm_year'] + 1900)
                    );
                }
            }
            $res["urls"][$loc] = $tmp;
        }
        return $res;
    }
}