WordPress Migration: Convert HTML doms into the common Gutenberg blocks

functions.php

<?php
/**
 * This snippet is useful while doing migration for the WordPress.
 * Basically this script will convert the HTML dom into the common Gutenberg block comments, 
 * so the tags like Pagragraph will be converted into the "<!-- wp:paragraph -->SOME DATA<!-- /wp:paragraph -->".
 *
 * This script supports paragraph, headings, list tags and many more.
 */

/**
 * Parse the post content to convert to Gutenberg blocks.
 *
 * @param \DOMNode $dom_node         DOMNode object.
 * @param string   $parent_tag_start Parent tag content start string.
 * @param string   $parent_tag_end   Parent tag content end string.
 *
 * @return string
 */
function parse_dom_node( \DOMNode $dom_node, $parent_tag_start = '', $parent_tag_end = '' ) {
	$content      = '';
	$skip_content = false;

	// phpcs:disable WordPress.NamingConventions.ValidVariableName.UsedPropertyNotSnakeCase
	if ( ! empty( $dom_node->childNodes ) ) {
		foreach ( $dom_node->childNodes as $node ) {
			if ( in_array( $node->nodeName, [ 'html', 'body', 'head' ], true ) ) {
				$content .= $this->parse_dom_node( $node );
			} elseif ( '#text' === $node->nodeName && $node->nodeValue ) {
				$content .= $node->nodeValue;
			} elseif ( $node->hasChildNodes() ) {
				$block_start = '';
				$block_end   = '';
				$tag_attrs   = '';

				// Identify the tag names and convert to Gutenberg block markup.
				if ( 'p' === $node->nodeName ) {
					if ( 'blockquote' === $node->parentNode->nodeName && 'twitter-tweet' === $node->parentNode->getAttribute( 'class' ) ) {
						$content = $this->parse_dom_node( $node );
						continue;
					} elseif ( 'blockquote' !== $node->parentNode->nodeName ) {
						$block_start = '<!-- wp:paragraph -->';
						$block_end   = '<!-- /wp:paragraph -->';
					}
				} elseif ( in_array( $node->nodeName, [ 'ol', 'ul' ], true ) ) {
					$block_start = '<!-- wp:list -->';
					$block_end   = '<!-- /wp:list -->';
				} elseif ( in_array( $node->nodeName, [ 'h1', 'h2', 'h3', 'h4', 'h5', 'h6' ], true ) ) {
					$level = false;

					if ( 'h1' === $node->nodeName ) {
						$level = 1;
					} elseif ( 'h3' === $node->nodeName ) {
						$level = 3;
					} elseif ( 'h4' === $node->nodeName ) {
						$level = 4;
					} elseif ( 'h5' === $node->nodeName ) {
						$level = 5;
					} elseif ( 'h6' === $node->nodeName ) {
						$level = 6;
					}

					if ( $level ) {
						$block_start = '<!-- wp:heading {"level":' . $level . '} -->';
					} else {
						$block_start = '<!-- wp:heading -->';
					}
					$block_end = '<!-- /wp:heading -->';
				} elseif ( 'a' === $node->nodeName ) {
					$link = $node->getAttribute( 'href' );

					if ( 'p' === $node->parentNode->nodeName && 'img' === $node->firstChild->nodeName ) {
						$content      = $parent_tag_start . $content . $parent_tag_end;
						$skip_content = true;
					}

					if ( ! empty( $link ) ) {
						$content = sprintf( '%1$s <a href="%2$s">%3$s</a>', $content, esc_url( $link ), $node->nodeValue );
						continue;
					}

					if ( '_blank' === $node->getAttribute( 'target' ) ) {
						$tag_attrs .= ' rel="noreferrer noopener" aria-label=" (opens in a new tab)"';
					}
				} elseif ( 'blockquote' === $node->nodeName ) {
					if ( ! $node->nodeValue ) {
						$skip_content = true;
						continue;
					} elseif ( 'twitter-tweet' === $node->getAttribute( 'class' ) ) {
						$content .= $this->parse_dom_node( $node );
						continue;
					}

					$block_start = '<!-- wp:quote -->';
					$block_end   = '<!-- /wp:quote -->';
					$tag_attrs   = ' class="wp-block-quote"';
				} elseif ( 'pre' === $node->nodeName ) {
					$block_start = '<!-- wp:preformatted -->';
					$block_end   = '<!-- /wp:preformatted -->';
					$tag_attrs   = ' class="wp-block-preformatted"';
				} elseif ( 'object' === $node->nodeName ) {
					$content .= $this->parse_dom_node( $node );
					continue;
				} elseif ( 'strong' === $node->nodeName ) {
					if ( 'p' === $node->firstChild->nodeName ) {
						$content .= $this->parse_dom_node( $node );
						continue;
					}
				} elseif ( 'div' === $node->nodeName ) {
					$container_classes = $node->getAttribute( 'class' );
					$div_start         = '';
					$div_end           = '';

					if ( preg_match( '~(field-name-field-see-related)~', $container_classes ) ) {
						$div_start = '<!-- wp:paragraph -->';
						$div_end   = '<!-- /wp:paragraph -->';
					}

					$content .= $this->parse_dom_node( $node, $div_start, $div_end );
					continue;
				}

				// Final content string.
				$tag_start = "{$block_start}<{$node->nodeName}{$tag_attrs}>";
				if ( 'ol' === $node->nodeName ) {
					$tag_start = "{$block_start}<ul{$tag_attrs}>";
				}
				$tag_end = "</{$node->nodeName}>{$block_end}";
				if ( 'ol' === $node->nodeName ) {
					$tag_end = "</ul>{$block_end}";
				}
				$content .= $this->parse_dom_node( $node, $tag_start, $tag_end );
			} else {
				$block_start = '';
				$block_end   = '';

				if ( 'script' === $node->nodeName ) {
					$src = $node->getAttribute( 'src' );

					if ( false !== strpos( $src, 'playbuzz.com' ) ) {
						$tag_attrs = ' type="text/javascript" src="' . $src . '"';
						$content  .= '<!-- wp:html -->';
						$content  .= "<{$node->nodeName}{$tag_attrs} />";
					}
				} elseif ( 'param' === $node->nodeName ) {
					continue;
				} elseif ( 'img' === $node->nodeName ) {
					$parsed_url = wp_parse_url( $node->getAttribute( 'src' ) );
					$img_url    = isset( $parsed_url['path'] ) ? $parsed_url['path'] : '';

					$download_image = [];

					if ( false !== strpos( $node->getAttribute( 'src' ), '/media/' ) ) {
						$img_url        = str_replace( '/media/', $this->original_site_domain . '/media/', $img_url );
						$download_image = download_inline_image( $img_url );
					}

					if ( isset( $download_image['url'] ) && ! empty( $download_image['url'] ) ) {
						$tag_attrs    = ' src="' . $download_image['url'] . '" alt="' . $node->getAttribute( 'alt' ) . '"';
						$block_start  = '<!-- wp:image -->';
						$block_start .= '<figure class="wp-block-image">';

						if ( 'p' === $node->parentNode->nodeName ) {
							$content      = $parent_tag_start . $content . $parent_tag_end;
							$skip_content = true;
						} elseif ( 'a' === $node->parentNode->nodeName ) {
							$block_start .= $parent_tag_start;
							$block_end   .= $parent_tag_end;
							$skip_content = true;
						}

						$block_end .= '</figure>';
						$block_end .= '<!-- /wp:image -->';
						$content   .= "{$block_start}<{$node->nodeName}{$tag_attrs} />{$block_end}";
					}
				} elseif ( 'hr' === $node->nodeName ) {
					$block_start = '<!-- wp:separator -->';
					$block_end   = '<!-- /wp:separator -->';
					$tag_attrs   = ' class="wp-block-separator"';
					$content    .= "{$block_start}<{$node->nodeName}{$tag_attrs} />{$block_end}";
				}
			}
		}

		if ( ! $skip_content ) {
			$content = $parent_tag_start . $content . $parent_tag_end;
		}
	}
	// phpcs:enable WordPress.NamingConventions.ValidVariableName.UsedPropertyNotSnakeCase

	return $content;
}

/**
 * Download inline image and return new path.
 *
 * @param string $url The image URL.
 *
 * @return array
 */
function download_inline_image( $url ) {

	$url = $this->get_img_url( $url );

	$attach_id = $this->get_media_id_from_url( $url );

	$already_exists = false;

	if ( ! $attach_id ) {
		$name = basename( $url );

		// Download the image.
		$tmp_file = download_url( $url, 15 );

		if ( is_wp_error( $tmp_file ) ) {
			\WP_CLI::log( "\nERROR: Unable to download image: {$url}" );
		} else {
			$file = [
				'name'     => $name,
				'tmp_name' => $tmp_file,
				'error'    => 0,
				'size'     => filesize( $tmp_file ),
			];

			$attach_id = media_handle_sideload( $file );

			if ( is_wp_error( $attach_id ) ) {
				\WP_CLI::log( "\nERROR: Unable to open output file: {$url}" );
			}
		}
	} else {

		$already_exists = true;
	}

	if ( ! is_wp_error( $attach_id ) && ! empty( $attach_id ) ) {

		if ( ! $already_exists ) {
			set_media_meta( $attach_id, $url );
		}

		return [
			'url' => get_post_field( 'guid', $attach_id ),
			'id'  => $attach_id,
		];
	}

	return [];
}

/**
 * Update attachment meta with original url.
 *
 * @param int    $attach_id Attachment id.
 * @param string $url       Original URL.
 */
function set_media_meta( $attach_id, $url ) {

	update_post_meta( $attach_id, '_original_image_url', $url );
}