A PHP class for WordPress to import posts (or pages, cpt) as basic HTML, including importing images from content.

class-cli-sitemap-importer.php

<?php

// Prevent direct file access.
defined( 'ABSPATH' ) || die;

/**
 * WP-CLI to delete all content.
 *
wp post delete $(wp post list --post_type=post --format=ids) --force
wp post delete $(wp post list --post_type=page --format=ids) --force
wp post delete $(wp post list --post_type=attachment --format=ids) --force
wp term delete category $(wp term list category --field=term_id)
wp term delete post_tag $(wp term list post_tag --field=term_id)
 */

/**
 * Instantiate the class.
 *
 * @since 0.1.0
 *
 * @return void
 */
new Mai_CLI_Sitemap_Importer;

/**
 * Gets it started.
 *
 * @since 0.1.0
 *
 * @link https://docs.wpvip.com/how-tos/write-custom-wp-cli-commands/
 * @link https://webdevstudios.com/2019/10/08/making-wp-cli-commands/
 *
 * @return void
 */
add_action( 'cli_init', function() {
	WP_CLI::add_command( 'maisitemap', 'Mai_CLI_Sitemap_Importer' );
});

/**
 * Split testing class.
 *
 * @version 0.1.0
 */
class Mai_CLI_Sitemap_Importer {
	/**
	 * Gets environment.
	 *
	 * Usage: wp maisitemap get_environment
	 *
	 * @return void
	 */
	function get_environment() {
		WP_CLI::log( sprintf( 'Environment: %s', wp_get_environment_type() ) );
	}

	/**
	 * Import pages from sitemap.
	 *
	 * Usage:
	 * wp maisitemap import --sitemap_url=https://example.com/sitemap.xml
	 * wp maisitemap import --sitemap_url=https://www.example.com/sitemap.xml --xpath="(//div[@class='page-content'])[1]" --post_type=page
	 *
	 * @since 0.1.0
	 *
	 * @param array $args       Standard command args.
	 * @param array $assoc_args Keyed args like --search and --fields.
	 *
	 * @return void
	 */
	function import( $args, $assoc_args ) {
		// Parse args.
		$assoc_args = wp_parse_args(
			$assoc_args,
			[
				'sitemap_url' => '',       // Required. The sitemap URL.
				'xpath'       => '//main', // The XPath to the content.
				'post_type'   => 'post',
				'post_status' => 'publish',
			]
		);

		// Bail if no URL.
		if ( ! $assoc_args['sitemap_url'] ) {
			WP_CLI::error( 'Please provide a sitemap URL.' );
			return;
		}

		// Get sitemap data.
		$response = wp_remote_get( $assoc_args['sitemap_url'] );
		$code     = wp_remote_retrieve_response_code( $response );

		// Bail if error.
		if ( is_wp_error( $response ) ) {
			WP_CLI::line( $response->get_error_message() );
			return;
		}

		// Bail if error.
		if ( 200 !== $code ) {
			// Build error message.
			$message = $code;

			// Add error message if available.
			if ( isset( $body['error']['message'] ) ) {
				$message .= ' ' . $body['error']['message'];
			}

			WP_CLI::line( $message );
			return;
		}

		// Get body and parse XML.
		$body = wp_remote_retrieve_body( $response );
		$xml  = simplexml_load_string( $body );

		// Bail if no XML.
		if ( ! $xml ) {
			WP_CLI::error( 'Could not parse XML.' );
			return;
		}

		// Bail if no URLs.
		if ( ! $xml->url || ! count( $xml->url ) ) {
			WP_CLI::error( 'No URLs found in sitemap.' );
			return;
		}

		// Loop through URLs
		foreach ( $xml->url as $url ) {
			$url = (string) $url->loc;

			// Skip if no URL.
			if ( ! $url ) {
				continue;
			}

			// Fetch HTML content for each URL.
			$html_response = wp_remote_get( $url );

			// Skip if error.
			if ( is_wp_error( $html_response ) ) {
				WP_CLI::line( $response->get_error_message() );
				return;
			}

			// Get HTML.
			$html = wp_remote_retrieve_body( $html_response );

			// Create the new document.
			$dom = new DOMDocument();

			// Modify state.
			$libxml_previous_state = libxml_use_internal_errors( true );

			// Load the content in the document HTML.
			$dom->loadHTML( $html );

			// Setup XPath.
			$xpath = new DOMXPath( $dom );

			// Parse URL.
			$parse = wp_parse_url( $url );

			// Get the host.
			$host = $parse['host'];

			// Get slug.
			$slug = $parse['path'];
			$slug = trim( $slug, '/' );

			// Get first h1.
			$nodes = $xpath->query( '(//h1)[1]' );
			$title = $nodes->item(0)->nodeValue;

			// If no h1, get title from slug.
			if ( ! $title ) {
				$title = str_replace( '-', ' ', $slug );
				$title = ucwords( $title );
			}

			// Get main content.
			$nodes = $xpath->query( $assoc_args['xpath'] );

			// Skip if no nodes.
			if ( ! $nodes->length ) {
				continue;
			}

			// Save new HTML.
			$content = $dom->saveHTML( $nodes->item(0) );

			// Remove content of <script> and <style> tags
			$content = preg_replace( '/<script\b[^>]*>.*?<\/script>/is', '', $content );
			$content = preg_replace( '/<style\b[^>]*>.*?<\/style>/is', '', $content );

			// Get valid tags.
			$tags = [
				'p',
				'a',
				'ul',
				'ol',
				'li',
				'h1',
				'h2',
				'h3',
				'h4',
				'h5',
				'h6',
				'blockquote',
				'img',
				'figure',
				'figcaption',
				'iframe',
				'video',
				'audio',
				'source',
				'pre',
				'code',
				'br',
				'hr',
				'em',
				'strong',
			];

			// Strip tags.
			$content = strip_tags( $content, $tags );

			// Loop through tags.
			foreach ( $tags as $tag ) {
				// Set up tag processor.
				$tags = new WP_HTML_Tag_Processor( $content );

				// Loop through tags.
				while ( $tags->next_tag( [ 'tag_name' => $tag ] ) ) {
					$tags->remove_attribute( 'id' );
					$tags->remove_attribute( 'class' );
					$tags->remove_attribute( 'style' );
				}

				$content = $tags->get_updated_html();
			}

			// Set up post args.
			$post_args = [
				'post_title'   => $title,
				'post_name'    => $slug,
				// 'post_content' => $content, // Added later, after we have a post ID for images.
				'post_status'  => $assoc_args['post_status'],
				'post_type'    => $assoc_args['post_type'],
			];

			// Insert the post.
			$post_id = wp_insert_post( $post_args );

			// Skip if no post ID.
			if ( is_wp_error( $post_id ) ) {
				WP_CLI::line( $post_id->get_error_message() );
				continue;
			}

			// Set up tag processor to import images.
			$tags = new WP_HTML_Tag_Processor( $content );

			// First, get the featured image.
			$first = true;

			// Loop through tags.
			while ( $tags->next_tag( [ 'tag_name' => 'img' ] ) ) {
				// Get src.
				$src = $tags->get_attribute( 'src' );

				// Skip if no src.
				if ( ! $src ) {
					continue;
				}

				// Parse the image src.
				$parse = wp_parse_url( $src );

				// If no host, add it.
				if ( ! wp_parse_url( $src, PHP_URL_HOST ) ) {
					$scheme = $parse['scheme'] ?: 'https';
					$src    = $scheme . '://' . trailingslashit( $host ) . ltrim( $src, '/' );
				}

				// Maybe upload the image.
				$image_id = $this->upload_image( $src, 'ref_url', $src, $post_id );

				// If we have an image ID.
				if ( $image_id ) {
					// Get image url.
					$image_url = wp_get_attachment_image_url( $image_id, 'large' );

					// Update the src.
					$tags->set_attribute( 'src', $image_url );

					// If first image, set as featured image.
					if ( $first ) {
						// Set the featured image.
						set_post_thumbnail( $post_id, $image_id );

						// Not first.
						$first = false;
					}

					WP_CLI::line( sprintf( 'Image imported: %s', $image_url ) );
				}
			}

			// Update the content.
			$content = $tags->get_updated_html();

			// Update the post content.
			wp_update_post(
				[
					'ID'           => $post_id,
					'post_content' => $content,
				]
			);

			// Log post inserted.
			WP_CLI::line( sprintf( '%s inserted: %s', ucwords( $assoc_args['post_type'] ), get_permalink( $post_id ) ) );
		}

		WP_CLI::success( 'Import complete.' );
	}

	/**
	 * Downloads a remote file and inserts it into the WP Media Library.
	 *
	 * @access private
	 *
	 * @see https://developer.wordpress.org/reference/functions/media_handle_sideload/
	 *
	 * @param string $ref_uri The reference URI of a remote file.
	 * @param string $ref_key The reference key of a remote file.
	 * @param string $url     HTTP URL address of a remote file.
	 * @param int    $post_id The post ID the media is associated with.
	 *
	 * @return int|WP_Error The ID of the attachment or a WP_Error on failure.
	 */
	function upload_image( $ref_uri, $ref_key, $image_url, $post_id ) {
		// Make sure we have the functions we need.
		if ( ! function_exists( 'download_url' ) || ! function_exists( 'media_handle_sideload' ) ) {
			require_once( ABSPATH . 'wp-admin/includes/media.php' );
			require_once( ABSPATH . 'wp-admin/includes/file.php' );
			require_once( ABSPATH . 'wp-admin/includes/image.php' );
		}

		// Check if there is an attachment with places_url meta key and value of $image_url.
		$existing_ids = get_posts(
			[
				'post_type'    => 'attachment',
				'post_status'  => 'any',
				'meta_key'     => $ref_key,
				'meta_value'   => $ref_uri,
				'meta_compare' => '=',
				'fields'       => 'ids',
			]
		);

		// Get existing ID.
		$existing_id = $existing_ids && isset( $existing_ids[0] ) ? $existing_ids[0] : 0;

		// Bail if the image already exists.
		if ( $existing_id ) {
			return $existing_id;
		}

		// Get contents of the image url.
		$image_hashed   = md5( $image_url ) . '.jpg';
		$image_contents = file_get_contents( $image_url );

		// If contents.
		if ( $image_contents ) {
			// Get the uploads directory.
			$upload_dir = wp_get_upload_dir();
			$upload_url = $upload_dir['baseurl'];

			// Specify the path to the destination directory within uploads.
			$destination_dir = $upload_dir['basedir'] . '/mai-sitemap-import/';

			// Create the destination directory if it doesn't exist.
			if ( ! file_exists( $destination_dir ) ) {
				mkdir( $destination_dir, 0755, true );
			}

			// Specify the path to the destination file.
			$destination_file = $destination_dir . $image_hashed;

			// Save the image to the destination file.
			file_put_contents( $destination_file, $image_contents );

			// Bail if the file doesn't exist.
			if ( ! file_exists( $destination_file ) ) {
				return 0;
			}

			$image_url = $image_hashed;
		}
		// Bail, no image contents.
		else {
			return 0;
		}

		// Build the image url.
		$image_url = untrailingslashit( $upload_url ) . '/mai-sitemap-import/' . $image_hashed;

		// Build a temp url.
		$tmp = download_url( $image_url );

		// Remove the temp file.
		wp_delete_file( $destination_file );

		// Bail if error.
		if ( is_wp_error( $tmp ) ) {
			// ray( $tmp->get_error_code() . ': upload_image() 1 ' . $image_url . ' ' . $tmp->get_error_message() );

			// Remove the original image and return the error.
			wp_delete_file( $tmp );

			return 0;
		}

		// Build the file array.
		$file_array = [
			'name'     => basename( $image_url ),
			'tmp_name' => $tmp,
		];

		// Add the image to the media library.
		$image_id = media_handle_sideload( $file_array, $post_id );

		// Bail if error.
		if ( is_wp_error( $image_id ) ) {
			// ray( $image_id->get_error_code() . ': upload_image() 2 ' . $image_url . ' ' . $image_id->get_error_message() );

			// Remove the original image and return the error.
			wp_delete_file( $file_array[ 'tmp_name' ] );
			return $image_id;
		}

		// Remove the original image.
		wp_delete_file( $file_array[ 'tmp_name' ] );

		// Set the reference url for possible reference later.
		update_post_meta( $image_id, $ref_key, $ref_uri );

		return $image_id;
	}
}