<?php
/**
 * URL normalization, exclusion rules and robots.txt matching.
 *
 * @package SiteAuditHub
 */

namespace SiteAuditHub;

defined( 'ABSPATH' ) || exit;

/**
 * URL helpers.
 */
class Url_Tools {

	/**
	 * Extensions that are never crawled as pages.
	 *
	 * @var array
	 */
	private static $non_html = array( 'jpg', 'jpeg', 'png', 'gif', 'webp', 'avif', 'svg', 'ico', 'bmp', 'tif', 'tiff', 'pdf', 'zip', 'gz', 'rar', '7z', 'tar', 'mp3', 'mp4', 'm4a', 'mov', 'avi', 'webm', 'ogg', 'wav', 'css', 'js', 'json', 'xml', 'txt', 'csv', 'doc', 'docx', 'xls', 'xlsx', 'ppt', 'pptx', 'woff', 'woff2', 'ttf', 'eot', 'otf', 'exe', 'dmg', 'apk', 'map' );

	/**
	 * Normalize a URL: drop fragment, lower-case scheme/host, strip configured params.
	 *
	 * @param string $url    URL.
	 * @param array  $config Scan config.
	 * @return string|false
	 */
	public static function normalize( $url, array $config ) {
		$url   = preg_replace( '/#.*$/s', '', trim( (string) $url ) );
		$parts = wp_parse_url( $url );
		if ( ! is_array( $parts ) || empty( $parts['scheme'] ) || empty( $parts['host'] ) ) {
			return false;
		}
		$scheme = strtolower( $parts['scheme'] );
		if ( 'http' !== $scheme && 'https' !== $scheme ) {
			return false;
		}
		$out = $scheme . '://' . strtolower( $parts['host'] );
		if ( ! empty( $parts['port'] ) && ! ( ( 'http' === $scheme && 80 === (int) $parts['port'] ) || ( 'https' === $scheme && 443 === (int) $parts['port'] ) ) ) {
			$out .= ':' . (int) $parts['port'];
		}
		$out .= isset( $parts['path'] ) && '' !== $parts['path'] ? $parts['path'] : '/';

		if ( ! empty( $parts['query'] ) ) {
			$strip = Settings::sanitize_param_list( isset( $config['strip_params'] ) ? $config['strip_params'] : '' );
			$keep  = array();
			foreach ( explode( '&', $parts['query'] ) as $pair ) {
				if ( '' === $pair ) {
					continue;
				}
				$name = strtolower( urldecode( strtok( $pair, '=' ) ) );
				if ( ! self::matches_any( $name, $strip ) ) {
					$keep[] = $pair;
				}
			}
			if ( $keep ) {
				$out .= '?' . implode( '&', $keep );
			}
		}
		return $out;
	}

	/**
	 * Wildcard match against a list (only * is special).
	 *
	 * @param string $value    Value.
	 * @param array  $patterns Patterns.
	 * @return bool
	 */
	public static function matches_any( $value, array $patterns ) {
		foreach ( $patterns as $p ) {
			$regex = '#^' . str_replace( '\*', '.*', preg_quote( $p, '#' ) ) . '$#i';
			if ( preg_match( $regex, $value ) ) {
				return true;
			}
		}
		return false;
	}

	/**
	 * Whether an internal URL may be crawled as a page under the config.
	 *
	 * @param string $url    Normalized URL.
	 * @param array  $config Config.
	 * @return bool
	 */
	public static function is_crawlable( $url, array $config ) {
		if ( ! Security::is_internal( $url ) ) {
			return false;
		}
		$path  = (string) wp_parse_url( $url, PHP_URL_PATH );
		$query = (string) wp_parse_url( $url, PHP_URL_QUERY );
		$ext   = strtolower( pathinfo( $path, PATHINFO_EXTENSION ) );
		if ( $ext && in_array( $ext, self::$non_html, true ) ) {
			return false;
		}
		// Never crawl admin, login, REST, XML-RPC or cron endpoints.
		$admin_path = (string) wp_parse_url( admin_url(), PHP_URL_PATH );
		if ( ( $admin_path && 0 === strpos( $path, $admin_path ) ) || preg_match( '#/(wp-admin/|wp-login\.php|xmlrpc\.php|wp-cron\.php|wp-json/|wp-content/|wp-includes/)#', $path ) || false !== strpos( $query, 'rest_route=' ) ) {
			return false;
		}
		if ( Login_Guard::active() && trim( strtolower( $path ), '/' ) === trim( (string) wp_parse_url( Login_Guard::login_url(), PHP_URL_PATH ), '/' ) ) {
			return false; // Never crawl (and so never reveal) the custom login page.
		}
		$patterns = array_filter( preg_split( '/[\r\n]+/', isset( $config['exclude_paths'] ) ? (string) $config['exclude_paths'] : '' ) );
		if ( $patterns ) {
			$target = $path . ( '' !== $query ? '?' . $query : '' );
			foreach ( $patterns as $p ) {
				$p = trim( $p );
				if ( '' === $p ) {
					continue;
				}
				$subject = preg_match( '#^https?://#i', $p ) ? $url : $target;
				if ( self::matches_any( $subject, array( $p ) ) ) {
					return false;
				}
			}
		}
		return true;
	}

	/**
	 * Parse robots.txt rules that apply to "User-agent: *".
	 *
	 * @param string $body robots.txt content.
	 * @return array {rules: [[allow(bool), pattern]], sitemaps: []}
	 */
	public static function parse_robots( $body ) {
		$rules    = array();
		$sitemaps = array();
		$applies  = false;
		$in_group = false;
		foreach ( preg_split( '/\r\n|\r|\n/', (string) $body ) as $line ) {
			$line = trim( preg_replace( '/#.*$/', '', $line ) );
			if ( '' === $line || false === strpos( $line, ':' ) ) {
				continue;
			}
			list( $field, $value ) = array_map( 'trim', explode( ':', $line, 2 ) );
			$field                 = strtolower( $field );
			if ( 'sitemap' === $field ) {
				if ( count( $sitemaps ) < 10 ) {
					$sitemaps[] = $value;
				}
				continue;
			}
			if ( 'user-agent' === $field ) {
				if ( ! $in_group ) {
					$applies = false;
				}
				$in_group = true;
				if ( '*' === $value ) {
					$applies = true;
				}
				continue;
			}
			$in_group = false;
			if ( $applies && ( 'disallow' === $field || 'allow' === $field ) && '' !== $value && count( $rules ) < 500 ) {
				$rules[] = array( 'allow' === $field, $value );
			}
		}
		return array(
			'rules'    => $rules,
			'sitemaps' => $sitemaps,
		);
	}

	/**
	 * Whether a path is disallowed (longest match wins; Allow wins ties).
	 *
	 * @param string $url   URL.
	 * @param array  $rules Rules from parse_robots().
	 * @return string|false Matching Disallow rule or false.
	 */
	public static function robots_disallowed( $url, array $rules ) {
		$path  = (string) wp_parse_url( $url, PHP_URL_PATH );
		$query = (string) wp_parse_url( $url, PHP_URL_QUERY );
		$path  = ( '' === $path ? '/' : $path ) . ( '' !== $query ? '?' . $query : '' );
		$best  = null;
		$len   = -1;
		foreach ( $rules as $rule ) {
			$pattern = $rule[1];
			$anchor  = '$' === substr( $pattern, -1 );
			$core    = $anchor ? substr( $pattern, 0, -1 ) : $pattern;
			$regex   = '#^' . str_replace( '\*', '.*', preg_quote( $core, '#' ) ) . ( $anchor ? '$' : '' ) . '#';
			if ( preg_match( $regex, $path ) ) {
				$l = strlen( $pattern );
				if ( $l > $len || ( $l === $len && $rule[0] ) ) {
					$len  = $l;
					$best = $rule;
				}
			}
		}
		return ( $best && ! $best[0] ) ? 'Disallow: ' . $best[1] : false;
	}
}