class.media-extractor.php

<?php
/**
 * Class with methods to extract metadata from a post/page about videos, images, links, mentions embedded
 * in or attached to the post/page.
 *
 * @todo Additionally, have some filters on number of items in each field
 */
class Jetpack_Media_Meta_Extractor {

	// Some consts for what to extract
	const ALL = 255;
	const LINKS = 1;
	const MENTIONS = 2;
	const IMAGES = 4;
	const SHORTCODES = 8; // Only the keeper shortcodes below
	const EMBEDS = 16;
	const HASHTAGS = 32;

	// For these, we try to extract some data from the shortcode, rather than just recording its presence (which we do for all)
	// There should be a function get_{shortcode}_id( $atts ) or static method SomethingShortcode::get_{shortcode}_id( $atts ) for these.
	private static $KEEPER_SHORTCODES = array(
		'youtube',
		'vimeo',
		'hulu',
		'ted',
		'wpvideo',
		'audio',
	);

	/**
	 * Gets the specified media and meta info from the given post.
	 * NOTE: If you have the post's HTML content already and don't need image data, use extract_from_content() instead.
	 *
	 * @param $blog_id The ID of the blog
	 * @param $post_id The ID of the post
	 * @param $what_to_extract (int) A mask of things to extract, e.g. Jetpack_Media_Meta_Extractor::IMAGES | Jetpack_Media_Meta_Extractor::MENTIONS
	 * @returns a structure containing metadata about the embedded things, or empty array if nothing found, or WP_Error on error
	 */
	static public function extract( $blog_id, $post_id, $what_to_extract = self::ALL ) {

		// multisite?
		if ( function_exists( 'switch_to_blog') )
			switch_to_blog( $blog_id );

		$post = get_post( $post_id );
		$content = $post->post_title . "\n\n" . $post->post_content;
		$char_cnt = strlen( $content );

		//prevent running extraction on really huge amounts of content
		if ( $char_cnt > 100000 ) //about 20k English words
			$content = substr( $content, 0, 100000 );

		$extracted = array();

		// Get images first, we need the full post for that
		if ( self::IMAGES & $what_to_extract ) {
			$extracted = self::get_image_fields( $post );

			// Turn off images so we can safely call extract_from_content() below
			$what_to_extract = $what_to_extract - self::IMAGES;
		}

		if ( function_exists( 'switch_to_blog') )
			restore_current_blog();

		// All of the other things besides images can be extracted from just the content
		$extracted = self::extract_from_content( $content, $what_to_extract, $extracted );

		return $extracted;
	}

	/**
	 * Gets the specified meta info from the given post content.
	 * NOTE: If you want IMAGES, call extract( $blog_id, $post_id, ...) which will give you more/better image extraction
	 * This method will give you an error if you ask for IMAGES.
	 *
	 * @param $content The HTML post_content of a post
	 * @param $what_to_extract (int) A mask of things to extract, e.g. Jetpack_Media_Meta_Extractor::IMAGES | Jetpack_Media_Meta_Extractor::MENTIONS
	 * @param $already_extracted (array) Previously extracted things, e.g. images from extract(), which can be used for x-referencing here
	 * @returns a structure containing metadata about the embedded things, or empty array if nothing found, or WP_Error on error
	 */
	static public function extract_from_content( $content, $what_to_extract = self::ALL, $already_extracted = array() ) {
		$stripped_content = self::get_stripped_content( $content );

		// Maybe start with some previously extracted things (e.g. images from extract()
		$extracted = $already_extracted;

		// Embedded media objects will have already been converted to shortcodes by pre_kses hooks on save.

 		if ( self::IMAGES & $what_to_extract ) {
			$images = Jetpack_Media_Meta_Extractor::extract_images_from_content( $stripped_content, array() );
			$extracted = array_merge( $extracted, $images );
		}

		// ----------------------------------- MENTIONS ------------------------------

		if ( self::MENTIONS & $what_to_extract ) {
			if ( preg_match_all( '/(^|\s)@(\w+)/u', $stripped_content, $matches ) ) {
				$mentions = array_values( array_unique( $matches[2] ) ); //array_unique() retains the keys!
				$mentions = array_map( 'strtolower', $mentions );
				$extracted['mention'] = array( 'name' => $mentions );
				if ( !isset( $extracted['has'] ) )
					$extracted['has'] = array();
				$extracted['has']['mention'] = count( $mentions );
			}
		}

		// ----------------------------------- HASHTAGS ------------------------------
		/** Some hosts may not compile with --enable-unicode-properties and kick a warning:
		  *   Warning: preg_match_all() [function.preg-match-all]: Compilation failed: support for \P, \p, and \X has not been compiled
		  * Therefore, we only run this code block on wpcom, not in Jetpack.
		 */
		if ( ( defined( 'IS_WPCOM' ) && IS_WPCOM ) && ( self::HASHTAGS & $what_to_extract ) ) {
			//This regex does not exactly match Twitter's
			// if there are problems/complaints we should implement this:
			//   https://github.com/twitter/twitter-text/blob/master/java/src/com/twitter/Regex.java
			if ( preg_match_all( '/(?:^|\s)#(\w*\p{L}+\w*)/u', $stripped_content, $matches ) ) {
				$hashtags = array_values( array_unique( $matches[1] ) ); //array_unique() retains the keys!
				$hashtags = array_map( 'strtolower', $hashtags );
				$extracted['hashtag'] = array( 'name' => $hashtags );
				if ( !isset( $extracted['has'] ) )
					$extracted['has'] = array();
				$extracted['has']['hashtag'] = count( $hashtags );
			}
		}

		// ----------------------------------- SHORTCODES ------------------------------

		// Always look for shortcodes.
		// If we don't want them, we'll just remove them, so we don't grab them as links below
		$shortcode_pattern = '/' . get_shortcode_regex() . '/s';
 		if ( preg_match_all( $shortcode_pattern, $content, $matches ) ) {

			$shortcode_total_count = 0;
			$shortcode_type_counts = array();
			$shortcode_types = array();
			$shortcode_details = array();

			if ( self::SHORTCODES & $what_to_extract ) {

				foreach( $matches[2] as $key => $shortcode ) {
					//Elasticsearch (and probably other things) doesn't deal well with some chars as key names
					$shortcode_name = preg_replace( '/[.,*"\'\/\\\\#+ ]/', '_', $shortcode );

					$attr = shortcode_parse_atts( $matches[3][ $key ] );

					$shortcode_total_count++;
					if ( ! isset( $shortcode_type_counts[$shortcode_name] ) )
						$shortcode_type_counts[$shortcode_name] = 0;
					$shortcode_type_counts[$shortcode_name]++;

					// Store (uniquely) presence of all shortcode regardless of whether it's a keeper (for those, get ID below)
					// @todo Store number of occurrences?
					if ( ! in_array( $shortcode_name, $shortcode_types ) )
						$shortcode_types[] = $shortcode_name;

					// For keeper shortcodes, also store the id/url of the object (e.g. youtube video, TED talk, etc.)
					if ( in_array( $shortcode, self::$KEEPER_SHORTCODES ) ) {
						unset( $id ); // Clear shortcode ID data left from the last shortcode
						// We'll try to get the salient ID from the function jetpack_shortcode_get_xyz_id()
						// If the shortcode is a class, we'll call XyzShortcode::get_xyz_id()
						$shortcode_get_id_func = "jetpack_shortcode_get_{$shortcode}_id";
						$shortcode_class_name = ucfirst( $shortcode ) . 'Shortcode';
						$shortcode_get_id_method = "get_{$shortcode}_id";
						if ( function_exists( $shortcode_get_id_func ) ) {
							$id = call_user_func( $shortcode_get_id_func, $attr );
						} else if ( method_exists( $shortcode_class_name, $shortcode_get_id_method ) ) {
							$id = call_user_func( array( $shortcode_class_name, $shortcode_get_id_method ), $attr );
						}
						if ( ! empty( $id )
							&& ( ! isset( $shortcode_details[$shortcode_name] ) || ! in_array( $id, $shortcode_details[$shortcode_name] ) ) )
							$shortcode_details[$shortcode_name][] = $id;
					}
				}

				if ( $shortcode_total_count > 0 ) {
					// Add the shortcode info to the $extracted array
					if ( !isset( $extracted['has'] ) )
						$extracted['has'] = array();
					$extracted['has']['shortcode'] = $shortcode_total_count;
					$extracted['shortcode'] = array();
					foreach ( $shortcode_type_counts as $type => $count )
						$extracted['shortcode'][$type] = array( 'count' => $count );
					if ( ! empty( $shortcode_types ) )
						$extracted['shortcode_types'] = $shortcode_types;
					foreach ( $shortcode_details as $type => $id )
						$extracted['shortcode'][$type]['id'] = $id;
				}
			}

			// Remove the shortcodes form our copy of $content, so we don't count links in them as links below.
			$content = preg_replace( $shortcode_pattern, ' ', $content );
		}

		// ----------------------------------- LINKS ------------------------------

		if ( self::LINKS & $what_to_extract ) {

			// To hold the extracted stuff we find
			$links = array();

			// @todo Get the text inside the links?

			// Grab any links, whether in <a href="..." or not, but subtract those from shortcodes and images
			// (we treat embed links as just another link)
			if ( preg_match_all( '#(?:^|\s|"|\')(https?://([^\s()<>]+(?:\([\w\d]+\)|([^[:punct:]\s]|/))))#', $content, $matches ) ) {

				foreach ( $matches[1] as $link_raw ) {
					$url = parse_url( $link_raw );

					// Data URI links
					if ( isset( $url['scheme'] ) && 'data' === $url['scheme'] )
						continue;

					// Remove large (and likely invalid) links
					if ( 4096 < strlen( $link_raw ) )
						continue;

					// Build a simple form of the URL so we can compare it to ones we found in IMAGES or SHORTCODES and exclude those
					$simple_url = $url['scheme'] . '://' . $url['host'] . ( ! empty( $url['path'] ) ? $url['path'] : '' );
					if ( isset( $extracted['image']['url'] ) ) {
						if ( in_array( $simple_url, (array) $extracted['image']['url'] ) )
							continue;
					}

					list( $proto, $link_all_but_proto ) = explode( '://', $link_raw );

					// Build a reversed hostname
					$host_parts = array_reverse( explode( '.', $url['host'] ) );
					$host_reversed = '';
					foreach ( $host_parts as $part ) {
						$host_reversed .= ( ! empty( $host_reversed ) ? '.' : '' ) . $part;
					}

					$link_analyzed = '';
					if ( !empty( $url['path'] ) ) {
						// The whole path (no query args or fragments)
						$path = substr( $url['path'], 1 ); // strip the leading '/'
						$link_analyzed .= ( ! empty( $link_analyzed ) ? ' ' : '' ) . $path;

						// The path split by /
						$path_split = explode( '/', $path );
						if ( count( $path_split ) > 1 ) {
							$link_analyzed .= ' ' . implode( ' ', $path_split );
						}

						// The fragment
						if ( ! empty( $url['fragment'] ) )
							$link_analyzed .= ( ! empty( $link_analyzed ) ? ' ' : '' ) . $url['fragment'];
					}

					// @todo Check unique before adding
					$links[] = array(
						'url' => $link_all_but_proto,
						'host_reversed' => $host_reversed,
						'host' => $url['host'],
					);
				}

			}

			$link_count = count( $links );
			if ( $link_count ) {
				$extracted[ 'link' ] = $links;
				if ( !isset( $extracted['has'] ) )
					$extracted['has'] = array();
				$extracted['has']['link'] = $link_count;
			}
		}

		// ----------------------------------- EMBEDS ------------------------------

		//Embeds are just individual links on their own line
		if ( self::EMBEDS & $what_to_extract ) {

			if ( !function_exists( '_wp_oembed_get_object' ) )
				include( ABSPATH . WPINC . '/class-oembed.php' );

			// get an oembed object
			$oembed = _wp_oembed_get_object();

			// Grab any links on their own lines that may be embeds
			if ( preg_match_all( '|^\s*(https?://[^\s"]+)\s*$|im', $content, $matches ) ) {

				// To hold the extracted stuff we find
				$embeds = array();

				foreach ( $matches[1] as $link_raw ) {
					$url = parse_url( $link_raw );

					list( $proto, $link_all_but_proto ) = explode( '://', $link_raw );

					// Check whether this "link" is really an embed.
					foreach ( $oembed->providers as $matchmask => $data ) {
						list( $providerurl, $regex ) = $data;

						// Turn the asterisk-type provider URLs into regex
						if ( !$regex ) {
							$matchmask = '#' . str_replace( '___wildcard___', '(.+)', preg_quote( str_replace( '*', '___wildcard___', $matchmask ), '#' ) ) . '#i';
							$matchmask = preg_replace( '|^#http\\\://|', '#https?\://', $matchmask );
						}

						if ( preg_match( $matchmask, $link_raw ) ) {
							$provider = str_replace( '{format}', 'json', $providerurl ); // JSON is easier to deal with than XML
							$embeds[] = $link_all_but_proto; // @todo Check unique before adding

							// @todo Try to get ID's for the ones we care about (shortcode_keepers)
							break;
						}
					}
				}

				if ( ! empty( $embeds ) ) {
					if ( !isset( $extracted['has'] ) )
						$extracted['has'] = array();
					$extracted['has']['embed'] = count( $embeds );
					$extracted['embed'] = array( 'url' => array() );
					foreach ( $embeds as $e )
						$extracted['embed']['url'][] = $e;
				}
			}
		}

		return $extracted;
	}

	/**
	 * @param $post A post object
	 * @param $args (array) Optional args, see defaults list for details
	 * @returns array Returns an array of all images meeting the specified criteria in $args
	 *
	 * Uses Jetpack Post Images
	 */
	private static function get_image_fields( $post, $args = array() ) {

		$defaults = array(
			'width'               => 200, // Required minimum width (if possible to determine)
			'height'              => 200, // Required minimum height (if possible to determine)
		);

		$args = wp_parse_args( $args, $defaults );

		$image_list = array();
		$image_booleans = array();
		$image_booleans['gallery'] = 0;

		$from_featured_image = Jetpack_PostImages::from_thumbnail( $post->ID, $args['width'], $args['height'] );
		if ( !empty( $from_featured_image ) ) {
			$srcs = wp_list_pluck( $from_featured_image, 'src' );
			$image_list = array_merge( $image_list, $srcs );
		}

		$from_slideshow = Jetpack_PostImages::from_slideshow( $post->ID, $args['width'], $args['height'] );
		if ( !empty( $from_slideshow ) ) {
			$srcs = wp_list_pluck( $from_slideshow, 'src' );
			$image_list = array_merge( $image_list, $srcs );
		}

		$from_gallery = Jetpack_PostImages::from_gallery( $post->ID );
		if ( !empty( $from_gallery ) ) {
			$srcs = wp_list_pluck( $from_gallery, 'src' );
			$image_list = array_merge( $image_list, $srcs );
			$image_booleans['gallery']++; // @todo This count isn't correct, will only every count 1
		}

		// @todo Can we check width/height of these efficiently?  Could maybe use query args at least, before we strip them out
		$image_list = Jetpack_Media_Meta_Extractor::get_images_from_html( $post->post_content, $image_list );

		return Jetpack_Media_Meta_Extractor::build_image_struct( $image_list );
	}

	public static function extract_images_from_content( $content, $image_list ) {
		$image_list = Jetpack_Media_Meta_Extractor::get_images_from_html( $content, $image_list );
		return Jetpack_Media_Meta_Extractor::build_image_struct( $image_list );
	}

	public static function build_image_struct( $image_list ) {
		if ( ! empty( $image_list ) ) {
			$retval = array( 'image' => array() );
			$unique_imgs = array_unique( $image_list );
			foreach ( $image_list as $img ) {
				$retval['image'][] = array( 'url' => $img );
			}
			$image_booleans['image'] = count( $retval['image'] );
			if ( ! empty( $image_booleans ) )
				$retval['has'] = $image_booleans;
			return $retval;
		} else {
			return array();
		}
	}

	/**
	 *
	 * @param string $html Some markup, possibly containing image tags
	 * @param array $images_already_extracted (just an array of image URLs without query strings, no special structure), used for de-duplication
	 * @return array Image URLs extracted from the HTML, stripped of query params and de-duped
	 */
	public static function get_images_from_html( $html, $images_already_extracted ) {
		$image_list = $images_already_extracted;
		$from_html = Jetpack_PostImages::from_html( $html );
		if ( !empty( $from_html ) ) {
			$srcs = wp_list_pluck( $from_html, 'src' );
			foreach( $srcs as $image_url ) {
				if ( ( $src = parse_url( $image_url ) ) && isset( $src['scheme'], $src['host'], $src['path'] ) ) {
					// Rebuild the URL without the query string
					$queryless = $src['scheme'] . '://' . $src['host'] . $src['path'];
				} elseif ( $length = strpos( $image_url, '?' ) ) {
					// If parse_url() didn't work, strip off the query string the old fashioned way
					$queryless = substr( $image_url, 0, $length );
				} else {
					// Failing that, there was no spoon! Err ... query string!
					$queryless = $image_url;
				}

				// Discard URLs that are longer then 4KB, these are likely data URIs or malformed HTML.
				if ( 4096 < strlen( $queryless ) ) {
					continue;
				}

				if ( ! in_array( $queryless, $image_list ) ) {
					$image_list[] = $queryless;
				}
			}
		}
		return $image_list;
	}

	private static function get_stripped_content( $content ) {
		$clean_content = strip_tags( $content );
		$clean_content = html_entity_decode( $clean_content );
		//completely strip shortcodes and any content they enclose
		$clean_content = strip_shortcodes( $clean_content );
		return $clean_content;
	}
}