<?php
/**
 * Reads the "keep this page out of search" answer other plugins already store.
 *
 * @package AcidStudios\AEO
 */

declare(strict_types=1);

namespace AcidStudios\AEO\Detect;

defined( 'ABSPATH' ) || exit;

/**
 * Whether a post has been marked `noindex` by the SEO plugin running the site.
 *
 * A site owner who tells Yoast to keep a page out of search has answered the
 * question this plugin would otherwise ask them a second time. Advertising that
 * page in `llms.txt` or `ai-context.json` — files written specifically so that
 * machines read them — contradicts the answer they already gave, so the answer
 * is read rather than repeated.
 *
 * Every supported plugin stores it in post meta, except All in One SEO, which
 * keeps its per-post row in its own table. Each is read in the plugin's own
 * vocabulary: SEOPress says `yes` for "do not index", Rank Math keeps a list of
 * directives, Yoast and The SEO Framework store `1`.
 *
 * WooCommerce is here too, and not by analogy: it marks the cart, the checkout
 * and the account pages `noindex` at request time, through `wp_robots`, which
 * leaves nothing in the database to read. Those three pages are named directly.
 */
final class NoIndex {

	/**
	 * Meta keys whose value is compared as a string.
	 *
	 * @var array<string,string>
	 */
	private const EXACT = [
		// Yoast SEO: 1 = noindex, 2 = index, absent = follow the default.
		'_yoast_wpseo_meta-robots-noindex' => '1',
		// The SEO Framework: 1 = noindex, -1 = index, 0 = default.
		'_genesis_noindex'                 => '1',
		// SEOPress: the field asks "do not display in search results", so yes means noindex.
		'_seopress_robots_index'           => 'yes',
	];

	/**
	 * Meta keys holding a list of robots directives.
	 *
	 * @var array<int,string>
	 */
	private const LISTS = [
		'rank_math_robots',
	];

	/**
	 * Meta keys holding a settings array with a `noindex` member.
	 *
	 * @var array<int,string>
	 */
	private const MAPS = [
		'slim_seo',
	];

	/**
	 * Post IDs already answered for, by ID.
	 *
	 * @var array<int,bool>
	 */
	private static array $cache = [];

	/**
	 * Forgets the cached answers.
	 *
	 * @return void
	 */
	public static function reset(): void {
		self::$cache = [];
	}

	/**
	 * Whether this post is marked `noindex` by something else on the site.
	 *
	 * @param int $post_id Post ID.
	 * @return bool True when the post should not be advertised.
	 */
	public static function is( int $post_id ): bool {
		if ( $post_id <= 0 ) {
			return false;
		}

		if ( ! isset( self::$cache[ $post_id ] ) ) {
			self::$cache[ $post_id ] = self::detect( $post_id );
		}

		/**
		 * Filters whether a post counts as `noindex`.
		 *
		 * A site with an SEO plugin this does not know about, or with its own
		 * idea of what belongs in the generated files, answers here.
		 *
		 * @param bool $noindex Whether the post is marked `noindex`.
		 * @param int  $post_id Post ID.
		 */
		return (bool) apply_filters( 'acid_aeo_is_noindex', self::$cache[ $post_id ], $post_id );
	}

	/**
	 * The uncached answer.
	 *
	 * @param int $post_id Post ID.
	 * @return bool True when something on the site marked the post `noindex`.
	 */
	private static function detect( int $post_id ): bool {
		foreach ( self::EXACT as $key => $value ) {
			if ( self::meta( $post_id, $key ) === $value ) {
				return true;
			}
		}

		foreach ( self::LISTS as $key ) {
			$directives = get_post_meta( $post_id, $key, true );

			if ( is_array( $directives ) && in_array( 'noindex', array_map( 'strval', $directives ), true ) ) {
				return true;
			}
		}

		foreach ( self::MAPS as $key ) {
			$settings = get_post_meta( $post_id, $key, true );

			if ( is_array( $settings ) && ! empty( $settings['noindex'] ) ) {
				return true;
			}
		}

		return self::aioseo( $post_id ) || self::woocommerce_page( $post_id );
	}

	/**
	 * One meta value as a trimmed string.
	 *
	 * @param int    $post_id Post ID.
	 * @param string $key     Meta key.
	 * @return string The value, empty when it is not a scalar.
	 */
	private static function meta( int $post_id, string $key ): string {
		$value = get_post_meta( $post_id, $key, true );

		return is_scalar( $value ) ? trim( (string) $value ) : '';
	}

	/**
	 * All in One SEO's answer, which lives in its own table.
	 *
	 * Only asked when the plugin is actually active, and only for one post, so
	 * the query is cheap and the table is known to exist. `robots_default` means
	 * "whatever the site wide setting says", and only an explicit override
	 * counts as an answer about this page.
	 *
	 * @param int $post_id Post ID.
	 * @return bool True when All in One SEO marks the post `noindex`.
	 */
	private static function aioseo( int $post_id ): bool {
		if ( ! defined( 'AIOSEO_VERSION' ) ) {
			return false;
		}

		global $wpdb;

		if ( ! isset( $wpdb ) || ! is_object( $wpdb ) ) {
			return false;
		}

		$table = $wpdb->prefix . 'aioseo_posts';

		// phpcs:ignore WordPress.DB.DirectDatabaseQuery.DirectQuery, WordPress.DB.DirectDatabaseQuery.NoCaching -- Reading another plugin's per-post row; it has no API that answers without booting its own stack, and the answer is cached in this class for the request.
		$row = $wpdb->get_row(
			$wpdb->prepare( 'SELECT robots_default, robots_noindex FROM %i WHERE post_id = %d', $table, $post_id ),
			ARRAY_A
		);

		if ( ! is_array( $row ) ) {
			return false;
		}

		return empty( $row['robots_default'] ) && ! empty( $row['robots_noindex'] );
	}

	/**
	 * Whether this is one of WooCommerce's functional pages.
	 *
	 * The cart, the checkout and the account page are machinery, not content:
	 * WooCommerce itself sends them `noindex`, they read as an empty cart or a
	 * login form to anyone who fetches them, and an assistant citing them as a
	 * source for the business is citing a form. The shop page is left in — it is
	 * the catalogue, and it says something about what is sold.
	 *
	 * @param int $post_id Post ID.
	 * @return bool True for the cart, checkout or account page.
	 */
	private static function woocommerce_page( int $post_id ): bool {
		if ( ! function_exists( 'wc_get_page_id' ) ) {
			return false;
		}

		foreach ( [ 'cart', 'checkout', 'myaccount' ] as $page ) {
			if ( (int) wc_get_page_id( $page ) === $post_id ) {
				return true;
			}
		}

		return false;
	}
}