<?php
/**
 * WebFinger class file.
 *
 * @package Activitypub
 */

namespace Activitypub;

use Activitypub\Activity\Actor;
use Activitypub\Collection\Actors;
use Activitypub\Collection\Remote_Actors;

/**
 * ActivityPub WebFinger Class.
 *
 * @author Matthias Pfefferle
 *
 * @see https://webfinger.net/
 */
class Webfinger {
	/**
	 * Check whether a value looks like an `acct` identifier.
	 *
	 * Accepts any of:
	 *
	 * - `user@host` — bare WebFinger handle.
	 * - `@user@host` — Mastodon display form with a leading `@`.
	 * - `acct:user@host` — full RFC 7565 URI form.
	 *
	 * The host/local-part pattern follows `ACTIVITYPUB_USERNAME_REGEXP`.
	 *
	 * @since 8.3.0
	 *
	 * @param mixed $value The candidate value.
	 * @return bool True if the value matches the acct identifier pattern.
	 */
	public static function is_acct( $value ) {
		if ( ! \is_string( $value ) || '' === $value ) {
			return false;
		}

		return (bool) \preg_match( '/^(?:acct:)?@?' . ACTIVITYPUB_USERNAME_REGEXP . '$/iu', $value );
	}

	/**
	 * Returns a users WebFinger "resource".
	 *
	 * @param int $user_id The WordPress user id.
	 *
	 * @return string The user-resource.
	 */
	public static function get_user_resource( $user_id ) {
		$user = Actors::get_by_id( $user_id );
		if ( ! $user || \is_wp_error( $user ) ) {
			return '';
		}

		return $user->get_webfinger();
	}

	/**
	 * Get the `acct:` URI for a handle.
	 *
	 * An internationalized handle gets the form RFC 7565 asks for: the userpart percent-encoded
	 * and the host as A-label.
	 *
	 * @since 9.4.0
	 *
	 * @param string $handle The handle, `user@host`.
	 *
	 * @return string The `acct:` URI.
	 */
	public static function get_acct( $handle ) {
		// With the scheme in front, a username that starts with `@` keeps it.
		$identifier_and_host = self::get_identifier_and_host( 'acct:' . $handle );

		if ( \is_wp_error( $identifier_and_host ) ) {
			return 'acct:' . $handle;
		}

		return $identifier_and_host[0];
	}

	/**
	 * Resolve a WebFinger resource.
	 *
	 * @param string $uri The WebFinger Resource.
	 *
	 * @return string|\WP_Error The URL or WP_Error.
	 */
	public static function resolve( $uri ) {
		$data = self::get_data( $uri );

		if ( \is_wp_error( $data ) ) {
			return $data;
		}

		if ( ! \is_array( $data ) || empty( $data['links'] ) || ! \is_array( $data['links'] ) ) {
			return new \WP_Error(
				'webfinger_missing_links',
				\__( 'No valid Link elements found.', 'activitypub' ),
				array(
					'status' => 400,
					'data'   => $data,
				)
			);
		}

		foreach ( $data['links'] as $link ) {
			if (
				isset( $link['rel'], $link['href'], $link['type'] ) &&
				'self' === $link['rel'] &&
				\is_string( $link['href'] ) &&
				(
					'application/activity+json' === $link['type'] ||
					'application/ld+json; profile="https://www.w3.org/ns/activitystreams"' === $link['type']
				)
			) {
				return $link['href'];
			}
		}

		return new \WP_Error(
			'webfinger_url_no_activitypub',
			\__( 'The Site supports WebFinger but not ActivityPub', 'activitypub' ),
			array(
				'status' => 400,
				'data'   => $data,
			)
		);
	}

	/**
	 * Transform a URI to an acct <identifier>@<host>.
	 *
	 * @see https://swicg.github.io/activitypub-webfinger/#reverse-discovery
	 *
	 * @param string $uri The URI (acct:, mailto:, http:, https:).
	 *
	 * @return string|\WP_Error Error or acct URI.
	 */
	public static function uri_to_acct( $uri ) {
		$data = self::get_data( $uri );

		if ( \is_wp_error( $data ) ) {
			return $data;
		}

		// Check if subject is an acct URI.
		if (
			isset( $data['subject'] ) &&
			\is_string( $data['subject'] ) &&
			\str_starts_with( $data['subject'], 'acct:' )
		) {
			return $data['subject'];
		}

		// Search for an acct URI in the aliases.
		if ( isset( $data['aliases'] ) && \is_array( $data['aliases'] ) ) {
			foreach ( $data['aliases'] as $alias ) {
				if ( \is_string( $alias ) && \str_starts_with( $alias, 'acct:' ) ) {
					return $alias;
				}
			}
		}

		return new \WP_Error(
			'webfinger_url_no_acct',
			\__( 'No acct URI found.', 'activitypub' ),
			array(
				'status' => 400,
				'data'   => $data,
			)
		);
	}

	/**
	 * Convert a URI string to an identifier and its host.
	 * Automatically adds acct: if it's missing.
	 *
	 * @param string $url The URI (acct:, mailto:, http:, https:).
	 *
	 * @return \WP_Error|array Error reaction or array with identifier and host as values.
	 */
	public static function get_identifier_and_host( $url ) {
		if ( ! $url ) {
			return new \WP_Error(
				'webfinger_invalid_identifier',
				\__( 'Invalid Identifier', 'activitypub' ),
				array(
					'status' => 400,
					'data'   => $url,
				)
			);
		}

		// Remove leading @.
		$url = \ltrim( $url, '@' );

		if ( \str_starts_with( $url, '//' ) ) {
			/*
			 * A scheme-relative URL is a URL reference, not a handle: treating it as one would
			 * read the host off the first `@` in its path.
			 */
			$identifier = $url;
			$scheme     = '';
		} elseif (
			// Scheme grammar per RFC 3986: a digit or hyphen in it must not read as a handle.
			! \preg_match( '/^([a-zA-Z][a-zA-Z0-9+.\-]*):/', $url, $match )
		) {
			$identifier = 'acct:' . $url;
			$scheme     = 'acct';
		} else {
			$identifier = $url;

			// Schemes are case-insensitive, and the switch below compares them exactly.
			$scheme = \strtolower( $match[1] );
		}

		$host = null;

		switch ( $scheme ) {
			case 'acct':
			case 'mailto':
			case 'xmpp':
				// Split on the last `@`: a local part may contain one, the host may not.
				if ( \strrpos( $identifier, '@' ) !== false ) {
					$host = \substr( $identifier, \strrpos( $identifier, '@' ) + 1 );

					/*
					 * Cut anything a path, query or fragment starts. `wp_parse_url()` does this for
					 * the URL forms below, but this branch takes the host off the string itself, so
					 * `acct:user@example.com#x` would otherwise carry `#x` into the host and miss
					 * a domain block on `example.com`, and `/x` would end up in the lookup URL's path.
					 */
					$host = \substr( $host, 0, \strcspn( $host, '/?#' ) );

					/*
					 * An internationalized handle is looked up in the form RFC 7565 asks for: the
					 * userpart percent-encoded and the host as A-label. ASCII is left as it is.
					 */
					if ( 'acct' === $scheme && \preg_match( '/[^\x00-\x7F]/', $identifier ) ) {
						$userpart   = \substr( $identifier, 0, \strrpos( $identifier, '@' ) );
						$userpart   = \preg_replace_callback(
							'/[^\x00-\x7F]+/',
							static function ( $non_ascii ) {
								return \rawurlencode( $non_ascii[0] );
							},
							$userpart
						);
						$identifier = $userpart . '@' . to_ascii_host( $host );
					}
				}
				break;
			default:
				$host = \wp_parse_url( $identifier, PHP_URL_HOST );
				break;
		}

		if ( empty( $host ) ) {
			return new \WP_Error(
				'webfinger_invalid_identifier',
				\__( 'Invalid Identifier', 'activitypub' ),
				array(
					'status' => 400,
					'data'   => $url,
				)
			);
		}

		return array( $identifier, \strtolower( to_ascii_host( $host ) ) );
	}

	/**
	 * Get the host of an identifier.
	 *
	 * The host half of {@see self::get_identifier_and_host()}, for callers that only need that.
	 * Prefer this over parsing a host out directly when you need an identifier's host: a handle
	 * has none to parse, and a URI in a scheme other than `http` can carry an `@` in its path
	 * that would be read as one.
	 *
	 * Not a drop-in for `is_same_host()`, which deliberately fails closed on an identifier with
	 * no parsable host. Giving an `acct:` keyId a host there would re-admit what 9.2.1 rejected.
	 *
	 * @since 9.3.0
	 *
	 * @param string $uri The identifier, a URL or a handle.
	 *
	 * @return string The host, lowercased, or an empty string when there is none.
	 */
	public static function get_host( $uri ) {
		$identifier_and_host = self::get_identifier_and_host( (string) $uri );

		if ( \is_wp_error( $identifier_and_host ) ) {
			return '';
		}

		/*
		 * Folded here rather than in `get_identifier_and_host()`, whose host is also used to build
		 * WebFinger and intent URLs: an IPv6 authority needs its brackets to stay a valid URL.
		 */
		return fold_host( $identifier_and_host[1] );
	}

	/**
	 * Get the WebFinger data for a given URI.
	 *
	 * @param string $uri The Identifier: <identifier>@<host> or URI.
	 *
	 * @return \WP_Error|mixed Error reaction, or the decoded document. The remote server picks the
	 *                         body, so callers check the shape before indexing it.
	 */
	public static function get_data( $uri ) {
		$identifier_and_host = self::get_identifier_and_host( $uri );

		if ( \is_wp_error( $identifier_and_host ) ) {
			return $identifier_and_host;
		}

		list( $identifier, $host ) = $identifier_and_host;

		$webfinger_url = \sprintf(
			'https://%s/.well-known/webfinger?resource=%s',
			$host,
			\rawurlencode( $identifier )
		);

		// Use Http::get() which handles all caching (success and errors).
		$response = Http::get(
			$webfinger_url,
			array( 'headers' => array( 'Accept' => 'application/jrd+json' ) ),
			WEEK_IN_SECONDS
		);

		if ( \is_wp_error( $response ) ) {
			return $response;
		}

		$body = \wp_remote_retrieve_body( $response );

		return \json_decode( $body, true );
	}

	/**
	 * Get the Remote-Follow endpoint for a given URI.
	 *
	 * @param string $uri The WebFinger Resource URI.
	 *
	 * @return string|\WP_Error Error or the Remote-Follow endpoint URI.
	 */
	public static function get_remote_follow_endpoint( $uri ) {
		return self::get_intent_endpoint( $uri, 'follow', true );
	}

	/**
	 * Generate a cache key for a given URI.
	 *
	 * @param string $uri A WebFinger Resource URI.
	 *
	 * @return string The cache key.
	 */
	public static function generate_cache_key( $uri ) {
		$uri = \ltrim( $uri, '@' );

		if ( \filter_var( $uri, FILTER_VALIDATE_EMAIL ) ) {
			$uri = 'acct:' . $uri;
		}

		return 'webfinger_' . \md5( $uri );
	}

	/**
	 * Infer a shortname from the Actor ID or URL. Used only for fallbacks,
	 * we will try to use what's supplied.
	 *
	 * @param Actor|string $actor_or_uri The Actor or URI.
	 *
	 * @return string Hopefully the name of the Follower.
	 */
	public static function guess( $actor_or_uri ) {
		if ( ! $actor_or_uri instanceof Actor ) {
			$actor = Remote_Actors::fetch_by_uri( $actor_or_uri );
			if ( \is_wp_error( $actor ) ) {
				return extract_name_from_uri( $actor_or_uri ) . '@' . \wp_parse_url( $actor_or_uri, PHP_URL_HOST );
			}

			$actor_or_uri = $actor;
		}

		if ( $actor_or_uri->get_preferred_username() ) {
			return $actor_or_uri->get_preferred_username() . '@' . \wp_parse_url( $actor_or_uri->get_id(), PHP_URL_HOST );
		}

		return extract_name_from_uri( $actor_or_uri->get_id() ) . '@' . \wp_parse_url( $actor_or_uri->get_id(), PHP_URL_HOST );
	}

	/**
	 * Get the Intent endpoint for a given URI and intent.
	 *
	 * @since 8.0.0
	 *
	 * @see https://codeberg.org/fediverse/fep/src/branch/main/fep/3b86/fep-3b86.md
	 *
	 * @param string $uri      The WebFinger Resource URI.
	 * @param string $intent   The intent to look for.
	 * @param bool   $fallback Whether to fallback to the Remote-Follow endpoint.
	 *
	 * @return string|\WP_Error Error or the Intent endpoint URI (may contain `{uri}` placeholder).
	 */
	public static function get_intent_endpoint( $uri, $intent, $fallback = false ) {
		$data = self::get_data( $uri );

		if ( \is_wp_error( $data ) ) {
			return $data;
		}

		if ( empty( $data['links'] ) || ! \is_array( $data['links'] ) ) {
			return new \WP_Error(
				'webfinger_missing_links',
				\__( 'No valid Link elements found.', 'activitypub' ),
				array(
					'status' => 400,
					'data'   => $data,
				)
			);
		}

		// Normalize the links with $rel as key.
		$links = array();

		foreach ( $data['links'] as $link ) {
			if ( ! isset( $link['rel'], $link['template'] ) || ! \is_string( $link['rel'] ) || ! \is_string( $link['template'] ) ) {
				continue;
			}

			$template = \trim( $link['template'] );

			/*
			 * A scheme check alone would pass `//host`, `/path`, `https:///path` and the empty
			 * string: with no colon there is nothing for `wp_kses_bad_protocol()` to strip, so the
			 * comparison below always matches. An accepted junk template is also stored under its
			 * rel, which stops the OStatus and FEP-3b86 fallbacks from ever being reached.
			 */
			if ( ! get_url_authority( $template ) ) {
				continue;
			}

			// The list is explicit: the `wp_allowed_protocols()` default is wider and filterable.
			$allowed = \wp_kses_bad_protocol( $template, array( 'http', 'https' ) );

			// Lowercased both sides like `wp_http_validate_url()`, or `HTTPS://` reads as a rewrite.
			if ( \strtolower( $allowed ) !== \strtolower( $template ) ) {
				continue;
			}

			$links[ \strtolower( $link['rel'] ) ] = $template;
		}

		$intent = \sanitize_text_field( $intent );
		$intent = \strtolower( $intent );

		if ( ! \filter_var( $intent, FILTER_VALIDATE_URL ) ) {
			$intent = 'https://w3id.org/fep/3b86/' . $intent;
		}

		if ( isset( $links[ $intent ] ) ) {
			return $links[ $intent ];
		}

		if ( ! $fallback ) {
			return new \WP_Error(
				'webfinger_missing_intent_endpoint',
				\__( 'No valid Intent endpoint found.', 'activitypub' ),
				array(
					'status' => 400,
					'data'   => $data,
				)
			);
		}

		/*
		 * OStatus subscribe URL (deprecated but still widely supported)
		 *
		 * @see https://ostatus.github.io/spec/OStatus%201.0%20Draft%202.html#anchor10
		 */
		if ( isset( $links['http://ostatus.org/schema/1.0/subscribe'] ) ) {
			return $links['http://ostatus.org/schema/1.0/subscribe'];
		}

		/*
		 * FEP-3b86 Object Intent — the generic "open this object on my home
		 * server" link, equivalent to pasting the URL into the home server's
		 * search box. Useful when no verb-specific intent is advertised.
		 *
		 * @see https://codeberg.org/fediverse/fep/src/branch/main/fep/3b86/fep-3b86.md#5-1-object-intent
		 */
		if ( isset( $links['https://w3id.org/fep/3b86/object'] ) ) {
			return $links['https://w3id.org/fep/3b86/object'];
		}

		// Last-resort: construct a Mastodon-compatible authorize_interaction URL.
		$identifier_and_host = self::get_identifier_and_host( $uri );

		if ( \is_wp_error( $identifier_and_host ) ) {
			return new \WP_Error(
				'webfinger_missing_intent_endpoint',
				\__( 'No valid Intent endpoint found.', 'activitypub' ),
				array(
					'status' => 400,
					'data'   => $data,
				)
			);
		}

		return 'https://' . $identifier_and_host[1] . '/authorize_interaction?uri={uri}';
	}
}