<?php
namespace Yipresser\AdminOptimizer\Modules;

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

/** Safely manages Admin Optimizer's AI policy additions to robots.txt. */
final class AI_Bot_Policy {
	const POLICY_ALLOW = 'allow';
	const POLICY_BLOCK = 'block';
	const BLOCK_START = '# BEGIN Admin Optimizer AI Bot Policy';
	const BLOCK_END   = '# END Admin Optimizer AI Bot Policy';

	public static function default_policy(): array {
		return [ AI_Bot_Registry::CATEGORY_SEARCH => self::POLICY_ALLOW, AI_Bot_Registry::CATEGORY_AGENT => self::POLICY_ALLOW, AI_Bot_Registry::CATEGORY_TRAINING => self::POLICY_BLOCK ];
	}

	public static function sanitize_policy( array $raw ): array {
		$clean = [];
		foreach ( self::default_policy() as $category => $fallback ) {
			$value = $raw[ $category ] ?? $fallback;
			$clean[ $category ] = in_array( $value, [ self::POLICY_ALLOW, self::POLICY_BLOCK ], true ) ? $value : $fallback;
		}
		return $clean;
	}

	/** Return the canonical Content-Signal directive for User-agent: *. */
	public static function content_signal( array $policy ): string {
		$policy = self::sanitize_policy( $policy );
		$fields = [];
		foreach ( AI_Bot_Registry::CONTENT_SIGNAL_FIELD_MAP as $category => $field ) {
			$fields[] = $field . '=' . ( self::POLICY_ALLOW === $policy[ $category ] ? 'yes' : 'no' );
		}
		return 'Content-Signal: ' . implode( ', ', $fields );
	}

	/** Build only explicit fallback crawler rules; an all-allow policy needs none. */
	public static function build_block( array $policy ): string {
		$policy = self::sanitize_policy( $policy );
		$lines = [];
		foreach ( array_keys( AI_Bot_Registry::CONTENT_SIGNAL_FIELD_MAP ) as $category ) {
			if ( self::POLICY_BLOCK !== $policy[ $category ] ) {
				continue;
			}
			$bots = wp_list_pluck( AI_Bot_Registry::get_by_category( $category ), 'user_agent' );
			if ( empty( $bots ) ) {
				continue;
			}
			if ( empty( $lines ) ) {
				$lines[] = self::BLOCK_START;
				$lines[] = '# AI training crawlers that may not honor Content-Signal';
			} else {
				$lines[] = '';
			}
			foreach ( $bots as $agent ) {
				$lines[] = 'User-agent: ' . $agent;
			}
			$lines[] = 'Disallow: /';
		}
		if ( empty( $lines ) ) {
			return '';
		}
		$lines[] = self::BLOCK_END;
		return implode( "\n", $lines );
	}

	/** @return array{status: string, policy?: array<string, string>} */
	public static function detect_policy( string $content ): array {
		$groups = self::wildcard_groups( $content );
		if ( 1 !== count( $groups ) || 1 !== count( $groups[0]['signals'] ) || self::managed_block_count( $content ) > 1 ) {
			return [ 'status' => 'custom' ];
		}
		$values = self::parse_signal( $groups[0]['signals'][0] );
		if ( null === $values ) {
			return [ 'status' => 'custom' ];
		}
		return [ 'status' => 'detected', 'policy' => [
			AI_Bot_Registry::CATEGORY_SEARCH => 'yes' === $values['search'] ? self::POLICY_ALLOW : self::POLICY_BLOCK,
			AI_Bot_Registry::CATEGORY_AGENT => 'yes' === $values['ai-input'] ? self::POLICY_ALLOW : self::POLICY_BLOCK,
			AI_Bot_Registry::CATEGORY_TRAINING => 'yes' === $values['ai-train'] ? self::POLICY_ALLOW : self::POLICY_BLOCK,
		] ];
	}

	/** Merge only AO additions, preserving unrelated directives and ordering. */
	public static function apply_to_content( string $content, array $policy ): string {
		$policy = self::sanitize_policy( $policy );
		if ( count( self::wildcard_groups( $content ) ) > 1 ) {
			return $content; // Multiple wildcard groups are unsafe to normalize.
		}
		$content = self::remove_managed_blocks( $content );
		$groups = self::wildcard_groups( $content );
		$lines = preg_split( '/\r\n|\r|\n/', $content );
		$signal = self::content_signal( $policy );
		if ( empty( $groups ) ) {
			$lines = trim( $content ) === '' ? [ 'User-agent: *', $signal ] : array_merge( $lines, [ '', 'User-agent: *', $signal ] );
		} else {
			$group = $groups[0];
			foreach ( array_reverse( $group['signal_indexes'] ) as $index ) {
				array_splice( $lines, $index, 1 );
			}
			array_splice( $lines, $group['start'] + 1, 0, [ $signal ] );
		}
		$content = rtrim( implode( "\n", $lines ) );
		$block = self::build_block( $policy );
		return '' === $block ? $content : $content . "\n\n" . $block;
	}

	/** @return array<int, array{start: int, signals: array<int, string>, signal_indexes: array<int, int>}> */
	private static function wildcard_groups( string $content ): array {
		$lines = preg_split( '/\r\n|\r|\n/', $content );
		$groups = [];
		$current = null;
		foreach ( $lines as $index => $line ) {
			if ( preg_match( '/^\s*User-agent\s*:\s*\*\s*(?:#.*)?$/i', $line ) ) {
				$groups[] = [ 'start' => $index, 'signals' => [], 'signal_indexes' => [] ];
				$current = count( $groups ) - 1;
				continue;
			}
			if ( preg_match( '/^\s*User-agent\s*:/i', $line ) ) {
				$current = null;
				continue;
			}
			if ( null !== $current && preg_match( '/^\s*Content-Signal\s*:\s*(.*?)\s*$/i', $line, $matches ) ) {
				$groups[ $current ]['signals'][] = $matches[1];
				$groups[ $current ]['signal_indexes'][] = $index;
			}
		}
		return $groups;
	}

	/** @return array<string, string>|null */
	private static function parse_signal( string $signal ): ?array {
		$values = [];
		foreach ( explode( ',', $signal ) as $part ) {
			if ( ! preg_match( '/^\s*(search|ai-train|ai-input)\s*=\s*(yes|no)\s*$/i', $part, $matches ) || isset( $values[ strtolower( $matches[1] ) ] ) ) {
				return null;
			}
			$values[ strtolower( $matches[1] ) ] = strtolower( $matches[2] );
		}
		return 3 === count( $values ) && isset( $values['search'], $values['ai-train'], $values['ai-input'] ) ? $values : null;
	}

	private static function remove_managed_blocks( string $content ): string {
		$pattern = '/' . preg_quote( self::BLOCK_START, '/' ) . '.*?' . preg_quote( self::BLOCK_END, '/' ) . '\s*/s';
		return preg_replace( $pattern, '', $content );
	}

	private static function managed_block_count( string $content ): int {
		return preg_match_all( '/' . preg_quote( self::BLOCK_START, '/' ) . '.*?' . preg_quote( self::BLOCK_END, '/' ) . '/s', $content );
	}
}