<?php
/**
 * Pulls URLs out of HTML, block markup, Elementor data and plain values.
 *
 * @package LinkTender
 */

defined( 'ABSPATH' ) || exit;

/**
 * URL extraction.
 */
class LinkTender_Extractor {

	/**
	 * Tag => attributes that hold URLs.
	 *
	 * @var array<string,string[]>
	 */
	const TAG_ATTRS = array(
		'a'      => array( 'href' ),
		'img'    => array( 'src' ),
		'iframe' => array( 'src' ),
		'video'  => array( 'src', 'poster' ),
		'audio'  => array( 'src' ),
		'source' => array( 'src' ),
		'embed'  => array( 'src' ),
	);

	/**
	 * Base URL for resolving relative links.
	 *
	 * @var string
	 */
	private $base = '';

	/**
	 * Sets the base URL used to resolve relative URLs.
	 *
	 * @param string $base Absolute URL.
	 * @return $this
	 */
	public function with_base( $base ) {
		$this->base = $base ? $base : home_url( '/' );
		return $this;
	}

	/**
	 * Extracts from any post content (HTML with optional block comments).
	 *
	 * @param string $content Content.
	 * @param string $field   Field label stored with each instance.
	 * @return array[]
	 */
	public function from_content( $content, $field ) {
		$content = (string) $content;
		if ( '' === trim( $content ) ) {
			return array();
		}

		$found = $this->from_html( $content, $field );

		if ( function_exists( 'has_blocks' ) && has_blocks( $content ) ) {
			$found = array_merge( $found, $this->from_blocks( parse_blocks( $content ), $field ) );
		}

		if ( false !== strpos( $content, '[' ) ) {
			$found = array_merge( $found, $this->from_shortcodes( $content, $field ) );
		}

		// Bare URLs on their own line become oEmbeds.
		if ( preg_match_all( '#^\s*(https?://[^\s<>"\']+)\s*$#im', $content, $m ) ) {
			foreach ( $m[1] as $url ) {
				$found[] = $this->item( $url, $field, 'embed', '' );
			}
		}

		return $this->finish( $found, $field );
	}

	/**
	 * Extracts from an HTML fragment.
	 *
	 * @param string $html  HTML.
	 * @param string $field Field label.
	 * @return array[]
	 */
	public function from_html( $html, $field ) {
		$found = array();
		$html  = (string) $html;
		if ( false === strpos( $html, '<' ) ) {
			return $found;
		}

		if ( ! preg_match_all( '#<(a|img|iframe|video|audio|source|embed)\b[^>]*>#i', $html, $matches, PREG_OFFSET_CAPTURE ) ) {
			return $found;
		}

		foreach ( $matches[0] as $index => $match ) {
			$tag_html = $match[0];
			$offset   = $match[1];
			$tag      = strtolower( $matches[1][ $index ][0] );

			$processor = new WP_HTML_Tag_Processor( $tag_html );
			if ( ! $processor->next_tag() ) {
				continue;
			}

			foreach ( self::TAG_ATTRS[ $tag ] as $attr ) {
				$value = $processor->get_attribute( $attr );
				if ( ! is_string( $value ) || '' === trim( $value ) ) {
					continue;
				}
				$anchor = '';
				if ( 'a' === $tag ) {
					$end = stripos( $html, '</a>', $offset );
					if ( false !== $end ) {
						$inner  = substr( $html, $offset + strlen( $tag_html ), $end - $offset - strlen( $tag_html ) );
						$anchor = trim( preg_replace( '/\s+/', ' ', wp_strip_all_tags( $inner ) ) );
						if ( '' === $anchor && preg_match( '#<img\b#i', $inner ) ) {
							$anchor = __( '[image]', 'linktender' );
						}
					}
				} elseif ( 'img' === $tag ) {
					$alt    = $processor->get_attribute( 'alt' );
					$anchor = is_string( $alt ) ? $alt : '';
				}
				$element = 'a' === $tag ? 'a' : ( 'img' === $tag ? 'img' : $tag );
				$found[] = $this->item( $value, $field, $element, $anchor );
			}
		}

		return $found;
	}

	/**
	 * Extracts URL-like attributes from parsed blocks (embeds, navigation links, buttons).
	 *
	 * @param array  $blocks Parsed blocks.
	 * @param string $field  Field label.
	 * @return array[]
	 */
	public function from_blocks( array $blocks, $field ) {
		$found = array();
		foreach ( $blocks as $block ) {
			if ( ! empty( $block['attrs'] ) && is_array( $block['attrs'] ) ) {
				$this->walk_attrs( $block['attrs'], $field, $found );
			}
			if ( ! empty( $block['innerBlocks'] ) ) {
				$found = array_merge( $found, $this->from_blocks( $block['innerBlocks'], $field ) );
			}
		}
		return $found;
	}

	/**
	 * Walks block attributes for URL values.
	 *
	 * @param array  $attrs  Attributes.
	 * @param string $field  Field label.
	 * @param array  $found  Collected items (by reference).
	 * @return void
	 */
	private function walk_attrs( array $attrs, $field, array &$found ) {
		foreach ( $attrs as $key => $value ) {
			if ( is_array( $value ) ) {
				$this->walk_attrs( $value, $field, $found );
				continue;
			}
			if ( ! is_string( $value ) ) {
				continue;
			}
			$key_name = strtolower( (string) $key );
			if ( preg_match( '/^(url|href|src|link|linkurl|mediaurl|poster)$/', $key_name ) && preg_match( '#^(https?:)?//#i', $value ) ) {
				$found[] = $this->item( $value, $field, 'block', '' );
			}
		}
	}

	/**
	 * Reads URLs from shortcode attributes without running any shortcode.
	 *
	 * Page builders such as Divi, WPBakery, Avada and Beaver Builder's shortcode mode keep
	 * button, image and video URLs in attributes like url="", button_url="", src="" or
	 * link="url:...|title:...". The content is only parsed as text: no shortcode callback
	 * runs, so there are no side effects, no extra queries and no outgoing requests.
	 *
	 * @param string $content Content.
	 * @param string $field   Field label.
	 * @return array[]
	 */
	public function from_shortcodes( $content, $field ) {
		$found   = array();
		$content = (string) $content;
		if ( ! preg_match_all( '/(?<!\[)\[([a-zA-Z][\w-]*)(\s[^\[\]]*?)?\/?\](?!\])/', $content, $matches, PREG_SET_ORDER ) ) {
			return $found;
		}
		foreach ( $matches as $match ) {
			if ( empty( $match[2] ) || ( false === strpos( $match[2], '//' ) && false === stripos( $match[2], '%2F%2F' ) ) ) {
				continue;
			}
			$atts = shortcode_parse_atts( trim( $match[2], " \t\n\r/" ) );
			if ( ! is_array( $atts ) ) {
				continue;
			}
			foreach ( $atts as $key => $value ) {
				if ( ! is_string( $key ) || ! is_string( $value ) ) {
					continue;
				}
				$item = $this->shortcode_attr_item( strtolower( $key ), trim( $value ), $field );
				if ( null !== $item ) {
					$found[] = $item;
				}
			}
		}
		return $found;
	}

	/**
	 * One URL item from a shortcode attribute, or null.
	 *
	 * @param string $key   Attribute name (lower case).
	 * @param string $value Attribute value.
	 * @param string $field Field label.
	 * @return array|null
	 */
	private function shortcode_attr_item( $key, $value, $field ) {
		// WPBakery link fields: url:https%3A%2F%2Fexample.com|title:Text|target:_blank.
		if ( 0 === strpos( $value, 'url:' ) && preg_match( '/^url:([^|]+)/', $value, $m ) ) {
			$url = rawurldecode( $m[1] );
			return preg_match( '#^https?://#i', $url ) ? $this->item( $url, $field, 'shortcode', '' ) : null;
		}

		/**
		 * Filters the shortcode attribute names that hold a URL.
		 *
		 * @param string $pattern Regular expression matched against the lower-case attribute name.
		 */
		$pattern = (string) apply_filters( 'linktender_shortcode_url_attributes', '/(^|_)(url|href|src|link|image|img|poster|mp4|webm|ogg|mp3|logo|background_image|bg_image)$/' );
		if ( ! preg_match( $pattern, $key ) || ! preg_match( '#^(https?:)?//[^\s"\'<>]+$#i', $value ) ) {
			return null;
		}
		$is_image = preg_match( '/(image|img|logo|poster)$/', $key ) || preg_match( '#\.(png|jpe?g|gif|webp|avif|svg)(\?.*)?$#i', $value );
		$element  = $is_image ? 'img' : 'shortcode';
		return $this->item( $value, $field, $element, '' );
	}

	/**
	 * Extracts from Elementor's JSON page data.
	 *
	 * @param string|array $data  JSON string or decoded array.
	 * @param string       $field Field label.
	 * @return array[]
	 */
	public function from_elementor( $data, $field ) {
		if ( is_string( $data ) ) {
			$data = json_decode( $data, true );
		}
		if ( ! is_array( $data ) ) {
			return array();
		}
		$found = array();
		$this->walk_elementor( $data, $field, $found, '' );
		return $this->finish( $found, $field );
	}

	/**
	 * Recursive Elementor walker.
	 *
	 * @param array  $node   Node.
	 * @param string $field  Field label.
	 * @param array  $found  Collected items (by reference).
	 * @param string $parent_key Parent key.
	 * @return void
	 */
	private function walk_elementor( array $node, $field, array &$found, $parent_key ) {
		foreach ( $node as $key => $value ) {
			if ( is_array( $value ) ) {
				$this->walk_elementor( $value, $field, $found, (string) $key );
				continue;
			}
			if ( ! is_string( $value ) || '' === $value ) {
				continue;
			}
			if ( 'url' === $key && preg_match( '#^(https?:)?//#i', $value ) ) {
				$element = ( false !== strpos( $parent_key, 'image' ) ) ? 'img' : 'a';
				$found[] = $this->item( $value, $field, $element, '' );
			} elseif ( false !== strpos( $value, '<' ) ) {
				$found = array_merge( $found, $this->from_html( $value, $field ) );
			}
		}
	}

	/**
	 * Extracts from a generic stored value (custom fields, widget settings).
	 *
	 * @param mixed  $value Value (string or nested array).
	 * @param string $field Field label.
	 * @return array[]
	 */
	public function from_value( $value, $field ) {
		$found = array();
		$this->walk_value( $value, $field, $found, 0 );
		return $this->finish( $found, $field );
	}

	/**
	 * Recursive walker for generic values.
	 *
	 * @param mixed  $value Value.
	 * @param string $field Field label.
	 * @param array  $found Collected (by reference).
	 * @param int    $depth Recursion depth.
	 * @return void
	 */
	private function walk_value( $value, $field, array &$found, $depth ) {
		if ( $depth > 8 ) {
			return;
		}
		if ( is_array( $value ) ) {
			foreach ( $value as $child ) {
				$this->walk_value( $child, $field, $found, $depth + 1 );
			}
			return;
		}
		if ( ! is_string( $value ) || '' === $value ) {
			return;
		}
		$trimmed = trim( $value );
		if ( preg_match( '#^https?://[^\s<>"]+$#i', $trimmed ) ) {
			$found[] = $this->item( $trimmed, $field, 'field', '' );
			return;
		}
		if ( false !== strpos( $value, '<' ) ) {
			$found = array_merge( $found, $this->from_html( $value, $field ) );
			if ( has_blocks( $value ) ) {
				$found = array_merge( $found, $this->from_blocks( parse_blocks( $value ), $field ) );
			}
		}
	}

	/**
	 * Builds one raw item.
	 *
	 * @param string $raw     URL as written.
	 * @param string $field   Field label.
	 * @param string $element Element type.
	 * @param string $anchor  Anchor text.
	 * @return array
	 */
	private function item( $raw, $field, $element, $anchor ) {
		return array(
			'raw_url' => $raw,
			'field'   => $field,
			'element' => $element,
			'anchor'  => (string) $anchor,
		);
	}

	/**
	 * Resolves, filters and de-duplicates raw items.
	 *
	 * @param array[] $found Raw items.
	 * @param string  $field Field label.
	 * @return array[]
	 */
	public function finish( array $found, $field ) {
		$settings = LinkTender_Settings::all();
		$out      = array();
		$seen     = array();
		$home     = wp_parse_url( home_url(), PHP_URL_HOST );

		foreach ( $found as $item ) {
			$url = $this->resolve( $item['raw_url'] );
			if ( null === $url ) {
				continue;
			}
			if ( 'img' === $item['element'] && empty( $settings['check_images'] ) ) {
				continue;
			}
			if ( in_array( $item['element'], array( 'iframe', 'video', 'audio', 'source', 'embed' ), true ) && empty( $settings['check_media'] ) ) {
				continue;
			}
			$host = strtolower( (string) wp_parse_url( $url, PHP_URL_HOST ) );
			if ( empty( $settings['check_internal'] ) && $host === $home ) {
				continue;
			}
			if ( LinkTender_Links::is_admin_or_login( $url ) || LinkTender_Settings::is_url_excluded( $url ) ) {
				continue;
			}
			// Block and embed hits repeat what the HTML pass already found.
			$key = $url . '|' . $item['raw_url'];
			if ( isset( $seen[ $key ] ) ) {
				continue;
			}
			$seen[ $key ] = true;

			$item['url'] = $url;
			$out[]       = $item;
		}

		/**
		 * Filters URLs extracted from one field.
		 *
		 * @param array[] $out   Items with url, raw_url, field, element, anchor.
		 * @param string  $field Field label.
		 */
		return apply_filters( 'linktender_extracted_urls', $out, $field );
	}

	/**
	 * Makes a URL absolute; returns null for URLs that cannot be checked.
	 *
	 * @param string $raw URL as written.
	 * @return string|null
	 */
	public function resolve( $raw ) {
		$url = LinkTender_Links::normalize( $raw );
		if ( '' === $url ) {
			return null;
		}
		if ( preg_match( '#^([a-z][a-z0-9+.-]*):#i', $url, $m ) ) {
			$scheme = strtolower( $m[1] );
			if ( 'http' !== $scheme && 'https' !== $scheme ) {
				// Skip non-web schemes such as mail, phone, script and data links.
				return null;
			}
		} else {
			if ( false !== strpos( $url, '{' ) || false !== strpos( $url, '[' ) ) {
				return null; // Template placeholders and shortcodes.
			}
			$url = WP_Http::make_absolute_url( $url, $this->base ? $this->base : home_url( '/' ) );
		}
		if ( ! wp_http_validate_url( $url ) && ! preg_match( '#^https?://[^/\s]+#i', $url ) ) {
			return null;
		}
		$host = wp_parse_url( $url, PHP_URL_HOST );
		if ( empty( $host ) || strlen( $url ) > 2000 ) {
			return null;
		}
		return $url;
	}
}