<?php
/**
 * Builds the content index for existing posts.
 *
 * @package Yangra\Conditional
 */

namespace Yangra\Conditional\Index;

defined( 'ABSPATH' ) || exit;

/**
 * Scans every post in batches with WP-Cron and indexes it, after install and
 * whenever the index format changes (VERSION), or when an administrator asks
 * for a rescan (e.g. after importing content).
 *
 * The format version is only saved when a scan finishes. While one runs, its
 * progress is kept in STATE_OPTION, which the Dashboard shows.
 */
class IndexScan {

	/**
	 * Index format version. Bump it when ContentIndex records something new,
	 * so existing sites are scanned again.
	 */
	const VERSION = 4;

	/**
	 * Option holding the format version of the last finished scan.
	 */
	const VERSION_OPTION = 'yangra_conditional_index_version';

	/**
	 * Option holding a running scan's progress.
	 */
	const STATE_OPTION = 'yangra_conditional_index_scan';

	/**
	 * Option holding when the last scan finished and how many posts it read.
	 */
	const LAST_OPTION = 'yangra_conditional_index_last_scan';

	/**
	 * Cron hook.
	 */
	const HOOK = 'yangra_conditional_index_scan';

	/**
	 * Posts read per query.
	 */
	const BATCH = 100;

	/**
	 * Seconds one cron run keeps scanning before it schedules the next run.
	 */
	const TIME_BUDGET = 15;

	/**
	 * Content index.
	 *
	 * @var ContentIndex
	 */
	private $index;

	/**
	 * Constructor.
	 *
	 * @param ContentIndex $index Content index.
	 */
	public function __construct( ContentIndex $index ) {
		$this->index = $index;
	}

	/**
	 * Registers hooks.
	 *
	 * @return void
	 */
	public function register(): void {
		add_action( self::HOOK, array( $this, 'run' ) );
		add_action( 'admin_init', array( $this, 'maybe_start' ) );
	}

	/**
	 * Starts a scan when the index is missing or in an older format. A scan
	 * that is already running is resumed instead, in case its cron event was lost.
	 *
	 * @return void
	 */
	public function maybe_start(): void {
		if ( null !== $this->progress() ) {
			$this->resume();
			return;
		}

		if ( (int) get_option( self::VERSION_OPTION, 0 ) >= self::VERSION ) {
			return;
		}

		$this->start();
	}

	/**
	 * Schedules the next batch of a running scan when no cron event is waiting
	 * for it. WP-Cron can lose an event when requests overlap, e.g. a cron run
	 * saving a cron array it read before the event was scheduled, which would
	 * leave the scan stuck.
	 *
	 * @return void
	 */
	public function resume(): void {
		if ( null === $this->progress() || false !== wp_next_scheduled( self::HOOK ) ) {
			return;
		}

		wp_schedule_single_event( time(), self::HOOK );
	}

	/**
	 * Starts a scan from the first post, replacing any scan that is running.
	 *
	 * @return void
	 */
	public function start(): void {
		update_option(
			self::STATE_OPTION,
			array(
				'last_id' => 0,
				'done'    => 0,
				'total'   => $this->count(),
			),
			false
		);

		wp_clear_scheduled_hook( self::HOOK );
		wp_schedule_single_event( time(), self::HOOK );
	}

	/**
	 * Indexes posts until the time budget is used, then schedules the next run,
	 * or finishes the scan.
	 *
	 * @param int $budget Seconds to keep going.
	 * @return void
	 */
	public function run( $budget = self::TIME_BUDGET ): void {
		$state = get_option( self::STATE_OPTION );
		if ( ! is_array( $state ) ) {
			return;
		}

		$until = microtime( true ) + (int) $budget;
		do {
			$rows = $this->batch( (int) $state['last_id'] );
			foreach ( $rows as $row ) {
				$this->index->update( (int) $row->ID, (string) $row->post_content );
				$state['last_id'] = (int) $row->ID;
				++$state['done'];
			}
			$more = self::BATCH === count( $rows );
		} while ( $more && microtime( true ) < $until );

		if ( ! $more ) {
			delete_option( self::STATE_OPTION );
			update_option( self::VERSION_OPTION, self::VERSION );
			update_option(
				self::LAST_OPTION,
				array(
					'finished' => time(),
					'scanned'  => (int) $state['done'],
				),
				false
			);

			return;
		}

		update_option( self::STATE_OPTION, $state, false );
		wp_schedule_single_event( time(), self::HOOK );
	}

	/**
	 * A running scan's progress, or null when none is running.
	 *
	 * @return array{done: int, total: int}|null
	 */
	public function progress(): ?array {
		$state = get_option( self::STATE_OPTION );
		if ( ! is_array( $state ) ) {
			return null;
		}

		$total = max( (int) ( $state['total'] ?? 0 ), (int) ( $state['done'] ?? 0 ) );

		return array(
			'done'  => min( (int) ( $state['done'] ?? 0 ), $total ),
			'total' => $total,
		);
	}

	/**
	 * What the index holds (see ContentIndex::summary()), for the Dashboard.
	 *
	 * @return array{posts: int, blocks: int, menu_items: int, schedules: int, shortcodes: int, unpublished: int}
	 */
	public function summary(): array {
		return $this->index->summary();
	}

	/**
	 * When the last scan finished and how many posts it read, or null when no
	 * scan has finished since this was first recorded.
	 *
	 * @return array{finished: int, scanned: int}|null
	 */
	public function last_scan(): ?array {
		$last = get_option( self::LAST_OPTION );
		if ( ! is_array( $last ) || empty( $last['finished'] ) ) {
			return null;
		}

		return array(
			'finished' => (int) $last['finished'],
			'scanned'  => max( 0, (int) ( $last['scanned'] ?? 0 ) ),
		);
	}

	/**
	 * Posts after an ID, with their content. Revisions (and autosaves, which are
	 * revisions) and auto-drafts are left out.
	 *
	 * @param int $last_id Last post ID scanned.
	 * @return array<int, object>
	 */
	private function batch( int $last_id ): array {
		global $wpdb;

		return $wpdb->get_results( // phpcs:ignore WordPress.DB.DirectDatabaseQuery
			$wpdb->prepare(
				"SELECT ID, post_content FROM {$wpdb->posts} WHERE ID > %d AND post_type <> 'revision' AND post_status <> 'auto-draft' ORDER BY ID ASC LIMIT %d",
				$last_id,
				self::BATCH
			)
		);
	}

	/**
	 * How many posts a scan goes through.
	 *
	 * @return int
	 */
	private function count(): int {
		global $wpdb;

		return (int) $wpdb->get_var( "SELECT COUNT(*) FROM {$wpdb->posts} WHERE post_type <> 'revision' AND post_status <> 'auto-draft'" ); // phpcs:ignore WordPress.DB.DirectDatabaseQuery
	}
}