<?php
/**
 * Robots.txt additions, AI crawler blocking, llms.txt and staging protection.
 *
 * @package OrbitstemSEO
 */

namespace Orbitstem\SEO\Frontend;

use Orbitstem\SEO\Content\MetaRepository;
use Orbitstem\SEO\Content\TemplateEngine;
use Orbitstem\SEO\Content\TypeRegistry;
use Orbitstem\SEO\Core\Settings;

defined( 'ABSPATH' ) || exit;

/**
 * WordPress builds robots.txt on the fly when no robots.txt file exists;
 * this adds the lines from Settings, blocks the AI crawlers that were ticked,
 * and lists Orbitstem's extra sitemaps. A real robots.txt file in the site
 * folder replaces all of that, so Settings warns when one exists.
 *
 * Staging protection: on a site that looks like a copy (staging, development
 * or local), every page asks not to be indexed, so a test copy never competes
 * with the live site in search results.
 */
class Robots {

	/**
	 * Settings.
	 *
	 * @var Settings
	 */
	private $settings;

	/**
	 * Content types.
	 *
	 * @var TypeRegistry
	 */
	private $types;

	/**
	 * Resolver for llms.txt descriptions.
	 *
	 * @var Resolver
	 */
	private $resolver;

	/**
	 * Constructor.
	 *
	 * @param Settings     $settings Settings.
	 * @param TypeRegistry $types    Types.
	 * @param Resolver     $resolver Resolver.
	 */
	public function __construct( Settings $settings, TypeRegistry $types, Resolver $resolver ) {
		$this->settings = $settings;
		$this->types    = $types;
		$this->resolver = $resolver;
	}

	/**
	 * Registers hooks.
	 */
	public function register() {
		add_filter( 'robots_txt', array( $this, 'robots_txt' ), 20, 2 );
		add_action( 'init', array( $this, 'serve_llms' ), 2 );
		if ( $this->staging_guard_active() ) {
			add_filter( 'wp_robots', array( $this, 'staging_robots' ), 99 );
			add_action( 'send_headers', array( $this, 'staging_header' ) );
		}
	}

	/**
	 * Known AI crawlers: user agent => description.
	 *
	 * @return array<string,string>
	 */
	public static function ai_bots() {
		return array(
			'GPTBot'             => __( 'OpenAI: collects pages to train ChatGPT models', 'orbitstem-seo' ),
			'OAI-SearchBot'      => __( 'OpenAI: ChatGPT search (blocking it can keep you out of ChatGPT search answers)', 'orbitstem-seo' ),
			'ChatGPT-User'       => __( 'OpenAI: fetches a page when a ChatGPT user asks for it', 'orbitstem-seo' ),
			'ClaudeBot'          => __( 'Anthropic: collects pages to train Claude models', 'orbitstem-seo' ),
			'Claude-User'        => __( 'Anthropic: fetches a page when a Claude user asks for it', 'orbitstem-seo' ),
			'Google-Extended'    => __( 'Google: use of your pages for Gemini (does not affect Google Search)', 'orbitstem-seo' ),
			'Applebot-Extended'  => __( 'Apple: use of your pages for Apple Intelligence (does not affect Apple search)', 'orbitstem-seo' ),
			'PerplexityBot'      => __( 'Perplexity: AI answer engine', 'orbitstem-seo' ),
			'CCBot'              => __( 'Common Crawl: public dataset used to train many AI models', 'orbitstem-seo' ),
			'Bytespider'         => __( 'ByteDance: AI training', 'orbitstem-seo' ),
			'meta-externalagent' => __( 'Meta: AI training', 'orbitstem-seo' ),
			'Amazonbot'          => __( 'Amazon: Alexa and AI answers', 'orbitstem-seo' ),
		);
	}

	/**
	 * Whether a robots.txt file in the site folder hides these settings.
	 *
	 * @return bool
	 */
	public static function physical_file() {
		return file_exists( ABSPATH . 'robots.txt' );
	}

	/**
	 * Adds Orbitstem's lines to the virtual robots.txt.
	 *
	 * @param string $output    Robots.txt so far.
	 * @param string $is_public "1" when search engines may index the site.
	 * @return string
	 */
	public function robots_txt( $output, $is_public ) {
		$output = rtrim( (string) $output ) . "\n";
		if ( $this->staging_guard_active() ) {
			return $output . "\n# Orbitstem SEO: this looks like a staging or development copy, so crawlers are asked to stay out.\nUser-agent: *\nDisallow: /\n";
		}
		$extra = trim( (string) $this->settings->get( 'robots_extra' ) );
		if ( '' !== $extra ) {
			// Rules written without a "User-agent:" line are for every crawler, as the settings page explains.
			$first = strtolower( (string) strtok( preg_replace( '/^(#[^\n]*\n)+/', '', $extra . "\n" ), "\n" ) );
			if ( 0 !== strpos( $first, 'user-agent:' ) ) {
				$extra = "User-agent: *\n" . $extra;
			}
			$output .= "\n# Orbitstem SEO: extra rules from Settings\n" . $extra . "\n";
		}
		$blocked = (array) $this->settings->get( 'ai_bots_block' );
		if ( $blocked ) {
			$output .= "\n# Orbitstem SEO: AI crawlers blocked in Settings\n";
			foreach ( $blocked as $agent ) {
				$output .= "User-agent: {$agent}\nDisallow: /\n\n";
			}
		}
		if ( '0' !== (string) $is_public ) {
			foreach ( \Orbitstem\SEO\Sitemap\ExtraSitemaps::urls( $this->settings ) as $url ) {
				$output .= 'Sitemap: ' . $url . "\n";
			}
		}
		return $output;
	}

	/**
	 * Serves /llms.txt: a short Markdown guide to the site for AI tools.
	 */
	public function serve_llms() {
		if ( ! $this->settings->get( 'llms_txt' ) || empty( $_SERVER['REQUEST_URI'] ) || file_exists( ABSPATH . 'llms.txt' ) ) {
			return;
		}
		$path = (string) wp_parse_url( esc_url_raw( wp_unslash( $_SERVER['REQUEST_URI'] ) ), PHP_URL_PATH );
		$want = (string) wp_parse_url( home_url( '/llms.txt' ), PHP_URL_PATH );
		if ( $want !== $path ) {
			return;
		}
		status_header( 200 );
		header( 'Content-Type: text/plain; charset=utf-8' );
		header( 'X-Robots-Tag: noindex' );
		echo $this->llms_txt(); // phpcs:ignore WordPress.Security.EscapeOutput.OutputNotEscaped -- plain text built from cleaned titles and URLs.
		exit;
	}

	/**
	 * The llms.txt text: site name, summary, then key pages and recent posts.
	 *
	 * @return string
	 */
	public function llms_txt() {
		$clean   = function ( $text ) {
			return trim( preg_replace( '/[\r\n\[\]]+/', ' ', TemplateEngine::plain( (string) $text ) ) );
		};
		$out     = '# ' . $clean( get_bloginfo( 'name' ) ) . "\n\n";
		$tagline = $clean( get_bloginfo( 'description' ) );
		if ( '' !== $tagline ) {
			$out .= '> ' . $tagline . "\n\n";
		}
		$groups = array(
			__( 'Pages', 'orbitstem-seo' )        => array(
				'post_type'      => 'page',
				'orderby'        => 'menu_order title',
				'order'          => 'ASC',
				'posts_per_page' => 30,
			),
			__( 'Recent posts', 'orbitstem-seo' ) => array(
				'post_type'      => 'post',
				'posts_per_page' => 30,
			),
		);
		foreach ( $groups as $heading => $query ) {
			if ( ! $this->types->supports_post_type( $query['post_type'] ) ) {
				continue;
			}
			$lines = array();
			foreach ( get_posts(
				$query + array(
					'post_status'   => 'publish',
					'has_password'  => false,
					'no_found_rows' => true,
				)
			) as $post ) {
				$data = $this->resolver->for_post( $post );
				if ( ! $data || $data['noindex'] || in_array( (int) $post->ID, \Orbitstem\SEO\Schema\Woo::private_page_ids(), true ) ) {
					continue; // Hidden pages, and the shop's cart, checkout and account pages.
				}
				// A canonical on another site says this copy is not the one to read.
				$canonical = wp_parse_url( (string) $data['canonical'] );
				if ( is_array( $canonical ) && ! empty( $canonical['host'] ) && ! \Orbitstem\SEO\Links\Url::same_site( $canonical ) ) {
					continue;
				}
				$line = '- [' . $clean( $post->post_title ) . '](' . esc_url_raw( $data['url'] ) . ')';
				if ( '' !== $data['description'] ) {
					$line .= ': ' . $clean( $data['description'] );
				}
				$lines[] = $line;
			}
			if ( $lines ) {
				$out .= '## ' . $heading . "\n\n" . implode( "\n", $lines ) . "\n\n";
			}
		}
		return rtrim( $out ) . "\n";
	}

	/**
	 * Whether this site is certainly not a live site: WordPress was told so
	 * (WP_ENVIRONMENT_TYPE is staging, development or local), or the address
	 * cannot be reached from the internet (localhost, .test, .local and the
	 * other reserved names). Other addresses are never guessed from: real sites
	 * use names like dev.to, test.com or an IP address.
	 *
	 * @return bool
	 */
	public static function looks_like_staging() {
		// No translated text here: this runs on plugins_loaded, before translations may load.
		// The reason shown on screens is built later, by staging_reason().
		return '' !== self::staging_kind();
	}

	/**
	 * What makes this site a non-live copy: 'env', 'host' or ''.
	 *
	 * @return string
	 */
	private static function staging_kind() {
		if ( in_array( wp_get_environment_type(), array( 'staging', 'development', 'local' ), true ) ) {
			return 'env';
		}
		$host = strtolower( (string) wp_parse_url( home_url(), PHP_URL_HOST ) );
		if ( 'localhost' === $host || preg_match( '/\.(local|test|localhost|invalid|example)$/', $host ) ) {
			return 'host';
		}
		return '';
	}

	/**
	 * Why this site counts as a non-live copy ('' when it does not), for screens.
	 * Meant for screens, so register() never calls it.
	 *
	 * @return string
	 */
	public static function staging_reason() {
		// Translations may only load from init on; code that asks earlier gets plain English.
		$translate = did_action( 'init' ) > 0;
		switch ( self::staging_kind() ) {
			case 'env':
				/* translators: %s: environment type such as "staging" */
				return sprintf( $translate ? __( 'WordPress is set up as a "%s" site (WP_ENVIRONMENT_TYPE in wp-config.php).', 'orbitstem-seo' ) : 'WordPress is set up as a "%s" site (WP_ENVIRONMENT_TYPE in wp-config.php).', wp_get_environment_type() );
			case 'host':
				/* translators: %s: host name */
				return sprintf( $translate ? __( 'The address %s cannot be reached from the internet.', 'orbitstem-seo' ) : 'The address %s cannot be reached from the internet.', strtolower( (string) wp_parse_url( home_url(), PHP_URL_HOST ) ) );
		}
		return '';
	}

	/**
	 * A staging-like address that is not certain enough to act on by itself:
	 * a staging word as the first part of a subdomain of another domain, such
	 * as staging.example.com (but not dev.to or stage-door.co.uk). Only used to
	 * suggest setting WP_ENVIRONMENT_TYPE.
	 *
	 * @return string Host name, or '' when it does not look like staging.
	 */
	public static function staging_like_address() {
		$host   = strtolower( (string) wp_parse_url( home_url(), PHP_URL_HOST ) );
		$labels = explode( '.', $host );
		if ( count( $labels ) < 3 || filter_var( $host, FILTER_VALIDATE_IP ) ) {
			return '';
		}
		if ( ! preg_match( '/^(staging|stage|stg|dev|develop|development|test|testing|qa|uat|preview|sandbox)\d*$/', $labels[0] ) ) {
			return '';
		}
		// "dev.co.uk": what follows is only a public suffix such as co.uk, so dev.co.uk is itself the main domain.
		if ( 3 === count( $labels ) && strlen( $labels[1] ) <= 3 && 2 === strlen( $labels[2] ) ) {
			return '';
		}
		return $host;
	}

	/**
	 * Whether staging protection is on and applies here.
	 *
	 * @param Settings $settings Settings.
	 * @return bool
	 */
	public static function guard_applies( Settings $settings ) {
		/**
		 * Filters whether staging protection applies. Return true to protect a
		 * staging copy WordPress does not know about, or false to switch it off.
		 *
		 * @param bool $active Whether it applies.
		 */
		return (bool) apply_filters( 'orbitstem_seo_staging_guard', $settings->get( 'staging_guard' ) && self::looks_like_staging() );
	}

	/**
	 * Whether staging protection is on and applies here.
	 *
	 * @return bool
	 */
	public function staging_guard_active() {
		return self::guard_applies( $this->settings );
	}

	/**
	 * Adds noindex, nofollow to every page of a staging copy.
	 *
	 * @param array $robots Robots directives.
	 * @return array
	 */
	public function staging_robots( array $robots ) {
		$robots['noindex']  = true;
		$robots['nofollow'] = true;
		unset( $robots['index'], $robots['follow'], $robots['max-image-preview'] );
		return $robots;
	}

	/**
	 * Also tells crawlers in the HTTP header (covers files and feeds).
	 */
	public function staging_header() {
		if ( ! headers_sent() ) {
			header( 'X-Robots-Tag: noindex, nofollow', false );
		}
	}
}