Extract attachment URLs from WordPress WXR export files

wp-wxr-extract-images.php

<?php

// Get a list of all WRX export files
$files = glob( 'wrx-export/*.xml' );

// Remove invalid XML attributes and values which prevent SimpleXML from parsing the file
// This shouldn't be necessary, if the exported creates a valid XML file.
$replace = array(
	'&uuml;' => '',
	'&hellip;' => '',
	'<";' => '-";',
);

$invalid = '';

foreach ( $files as $file_no => $xml_file ) {
	$attachments = '';
	$xml = file_get_contents( $xml_file );
	$filaname_save = sprintf( 'wrx-files/%d.txt', $file_no );

	if ( file_exists( $filaname_save ) ) {
		continue;
	}

	// The Photon ?quality=100&strip=into thing makes this invalid XML
	$xml = preg_replace( '#&(?=[a-z_0-9]+=)#', '&amp;', $xml );

	$xml = str_replace( array_keys( $replace ), array_values( $replace ), $xml );

	$xml = simplexml_load_string( $xml, 'SimpleXMLElement' );

	if ( ! $xml ) {
		die( sprintf( 'Error at %s.', $xml_file ) );
	}

	$ns = $xml->getNamespaces( true );

	foreach ( $xml->channel->item as $post ) {
		$wp_data = $post->children( $ns['wp'] );

		if ( isset( $wp_data->attachment_url ) && ! empty( $wp_data->attachment_url ) ) {
			$image_url = strtok( $wp_data->attachment_url, '?' );
			
			// Mark URLs from other domains as "invalid" for further inspection
			if ( false !== strstr( $image_url, 'example.files.wordpress.com' ) ) {
				$attachments .= $image_url . "\n";
			} else {
				$invalid .= $image_url . "\n";
			}
		}
	}

	file_put_contents( $filaname_save, $attachments );
}

file_put_contents( 'wrx-files-invalid.txt', $invalid );

echo "Done!";