<?php
/**
 * Public discoverability — robots.txt block + <head> link references.
 *
 * Advertises the active AI Discovery Files to automated tools that scan
 * robots.txt or the homepage <head>. Without this, the files serve correctly
 * but are invisible to crawlers that don't already know the canonical URLs.
 *
 * Only files the user has actually activated in plugin settings are emitted.
 *
 * @package AIDF
 * @since   1.3.9
 */

if ( ! defined( 'ABSPATH' ) ) {
	exit;
}

/**
 * Emits discoverability hints into robots.txt and the HTML head.
 *
 * Two hooks, both honour the active_files setting:
 *
 * - robots_txt filter (WP 3.0+) — appends a block listing every currently-
 *   active discovery file URL. URLs are emitted as comments (lines starting
 *   with "#") so they are safe with any robots.txt parser, per RFC 9309.
 *   Note this filter only fires when WordPress is generating robots.txt
 *   virtually; sites with a physical robots.txt at the webroot will need to
 *   copy the block in by hand, which the admin UI exposes.
 *
 * - wp_head action — emits a <link rel="alternate"> for every active file
 *   so tools that scan the head (the same way they scan for RSS feeds) can
 *   discover them.
 */
class AIDF_Discovery {

	/**
	 * Slug -> human-readable title used in <link title="..."> attribute.
	 *
	 * @var array<string, string>
	 */
	private static $head_titles = array(
		'llms-txt'         => 'AI Discovery (llms.txt)',
		'llm-txt'          => 'AI Discovery legacy alias (llm.txt)',
		'llms-html'        => 'AI Discovery, HTML (llms.html)',
		'ai-txt'           => 'AI Usage Policy (ai.txt)',
		'ai-json'          => 'AI Permissions (ai.json)',
		'identity-json'    => 'Business Identity (identity.json)',
		'brand-txt'        => 'Brand Terminology (brand.txt)',
		'faq-ai-txt'       => 'FAQs for AI (faq-ai.txt)',
		'developer-ai-txt' => 'Developer Context (developer-ai.txt)',
		'robots-ai-txt'    => 'AI Crawler Access (robots-ai.txt)',
	);

	/**
	 * Register hooks.
	 *
	 * @return void
	 */
	public static function init() {
		add_filter( 'robots_txt', array( __CLASS__, 'inject_discovery_refs' ), 20, 2 );
		add_action( 'wp_head', array( __CLASS__, 'inject_head_links' ), 5 );
	}

	/**
	 * Filter callback for robots_txt.
	 *
	 * Appends a labelled block listing every active AI Discovery File URL.
	 * If the site is set to "discourage search engines" ($public === false),
	 * we don't add anything — WordPress will already be returning
	 * Disallow: / and the discovery files shouldn't be advertised.
	 *
	 * @param  string $output Existing robots.txt body.
	 * @param  int    $public Whether the blog is public.
	 * @return string
	 */
	public static function inject_discovery_refs( $output, $public ) {
		if ( ! $public ) {
			return $output;
		}

		$block = self::get_robots_txt_block();

		if ( '' === $block ) {
			return $output;
		}

		// Make sure we land on a new line cleanly.
		$separator = ( '' === $output || "\n" === substr( $output, -1 ) ) ? '' : "\n";

		return $output . $separator . $block;
	}

	/**
	 * Action callback for wp_head.
	 *
	 * Emits one <link rel="alternate"> per active discovery file so
	 * crawlers that scan the head (the way they scan for feeds) can
	 * pick the URLs up. Skipped in admin and feed contexts.
	 *
	 * @return void
	 */
	public static function inject_head_links() {
		if ( is_admin() || is_feed() ) {
			return;
		}

		$entries = self::get_active_entries();

		if ( empty( $entries ) ) {
			return;
		}

		echo "\n<!-- AI Discovery Files (v" . esc_html( AIDF_VERSION ) . ") -->\n";

		foreach ( $entries as $entry ) {
			printf(
				'<link rel="alternate" type="%s" href="%s" title="%s">' . "\n",
				esc_attr( $entry['short_type'] ),
				esc_url( $entry['url'] ),
				esc_attr( $entry['title'] )
			);
		}

		echo "<!-- /AI Discovery Files -->\n\n";
	}

	/**
	 * Build the labelled robots.txt block for the active discovery files.
	 *
	 * Returned without a trailing newline. Returns an empty string when no
	 * files are active so callers can decide whether to emit anything.
	 *
	 * This method is public so the admin UI can show it to the user when
	 * a physical robots.txt is detected (and the filter therefore can't
	 * inject it automatically).
	 *
	 * @return string
	 */
	public static function get_robots_txt_block() {
		$entries = self::get_active_entries();

		if ( empty( $entries ) ) {
			return '';
		}

		$lines   = array();
		$lines[] = '# AI Discovery Files';
		$lines[] = '# Specification: https://www.ai-visibility.org.uk/specifications/';
		$lines[] = '# Generated by AI Discovery Files for WordPress v' . AIDF_VERSION;

		foreach ( $entries as $entry ) {
			$lines[] = '# ' . $entry['filename'] . ': ' . $entry['url'];
		}

		return implode( "\n", $lines );
	}

	/**
	 * Whether a physical robots.txt exists at the WordPress root.
	 *
	 * When true, WordPress will serve that file as-is and the robots_txt
	 * filter never fires. The admin UI uses this to warn the user that
	 * they need to copy the block into their robots.txt by hand.
	 *
	 * @return bool
	 */
	public static function has_physical_robots_txt() {
		$abspath = untrailingslashit( ABSPATH );
		return file_exists( $abspath . '/robots.txt' );
	}

	/**
	 * Resolve the list of active discovery files to displayable entries.
	 *
	 * Each entry has:
	 *   - slug       : the file slug (e.g. 'llms-txt')
	 *   - filename   : the on-disk filename (e.g. 'llms.txt')
	 *   - url        : the absolute URL where the plugin serves it
	 *   - short_type : the MIME type without parameters (e.g. 'text/plain')
	 *   - title      : the human-readable title used in <link title="...">
	 *
	 * Skips:
	 *   - Slugs that are not present in AIDF_Plugin::get_file_types().
	 *   - llm-txt (it is a 301 redirect to llms.txt — advertising both would
	 *     just send crawlers through an extra hop, so we list only llms-txt).
	 *
	 * @return array<int, array<string, string>>
	 */
	private static function get_active_entries() {
		$settings   = AIDF_Plugin::get_settings();
		$active     = isset( $settings['active_files'] ) ? (array) $settings['active_files'] : array();
		$file_types = AIDF_Plugin::get_file_types();

		$out = array();

		foreach ( $active as $slug ) {
			if ( 'llm-txt' === $slug ) {
				// llm.txt is a redirect to llms.txt, no point advertising it separately.
				continue;
			}

			if ( ! isset( $file_types[ $slug ] ) ) {
				continue;
			}

			$meta = $file_types[ $slug ];

			$out[] = array(
				'slug'       => $slug,
				'filename'   => $meta['filename'],
				'url'        => home_url( '/' . $meta['filename'] ),
				'short_type' => self::short_type( $meta['content_type'] ),
				'title'      => isset( self::$head_titles[ $slug ] ) ? self::$head_titles[ $slug ] : $meta['filename'],
			);
		}

		return $out;
	}

	/**
	 * Strip the charset parameter from a Content-Type so it can be used in
	 * a <link type="..."> attribute (which expects bare MIME types).
	 *
	 * 'text/plain; charset=utf-8' -> 'text/plain'
	 *
	 * @param  string $content_type Full Content-Type header value.
	 * @return string
	 */
	private static function short_type( $content_type ) {
		$parts = explode( ';', $content_type, 2 );
		return trim( $parts[0] );
	}
}