csv-url-processor.php

<?php
/*
Plugin Name: CSV URL Processor
Plugin URI: https://gist.github.com/mgratch/ec801047eee7ae3bc06d717200c4196a
Description: A WP-CLI command to process URLs from a CSV file.
Version: 1.0
Author: Marc Gratch
Author URI: https://marcgratch.com
*/

if ( defined( 'WP_CLI' ) && WP_CLI ) {

	/**
	 * Class for WP-CLI command to process URLs from CSV.
	 */
	class CSV_URL_Processor_Command extends WP_CLI_Command {

		/**
		 * Processes the CSV file and outputs the path to the failed URLs file.
		 *
		 * ## OPTIONS
		 *
		 * <file>
		 * : The path to the CSV file.
		 *
		 * ## EXAMPLES
		 *
		 *     wp csv_process file_path.csv
		 *
		 * @when after_wp_load
		 *
		 * @param array $args The positional arguments.
		 * @param array $assoc_args The associative arguments.
		 */
		public function __invoke( $args, $assoc_args ) {
			global $wpdb;
			list( $file_path ) = $args;

			// Truncate $wpdb->prefix . 'redirection_items' table.
			$wpdb->query( "TRUNCATE TABLE {$wpdb->prefix}redirection_items" );

			// flush rewrite rules via cli.
			WP_CLI::runcommand( 'rewrite flush' );

			// clear cache and transients with cli.
			WP_CLI::runcommand( 'cache flush' );
			WP_CLI::runcommand( 'transient delete --all' );

			$failed_urls = $this->process_csv_file( $file_path );

			// Generate a timestamped filename for failed URLs.
			$timestamp        = gmdate( 'Ymd_His' );
			$failed_file_path = 'failed_urls_' . $timestamp . '.csv';
			$failed_handle    = fopen( $failed_file_path, 'w' );

			foreach ( $failed_urls as $failed_row ) {
				fputcsv( $failed_handle, $failed_row );
			}
			fclose( $failed_handle );

			WP_CLI::success( 'Processed CSV. Failed URLs saved to: ' . $failed_file_path );
		}

		/**
		 * Reads the CSV file, processes URLs, and returns the failed URLs.
		 *
		 * @param string $file_path The path to the CSV file.
		 *
		 * @return array
		 */
		private function process_csv_file( $file_path ) {
			$csv_data = array();
			$handle   = fopen( $file_path, 'r' );
			if ( $handle ) {
				fgetcsv( $handle ); // Skip header row.
				while ( ( $data = fgetcsv( $handle ) ) !== false ) {
					$csv_data[] = $data;
				}
				fclose( $handle );
			}

			$total_rows = count( $csv_data );
			$progress   = \WP_CLI\Utils\make_progress_bar( 'Processing URLs', $total_rows );

			$redirects             = array();
			$failed_urls           = array();
			$source_to_destination = array_column( $csv_data, 1, 0 );
			foreach ( $csv_data as $index => $row ) {
				$source_url      = $row[0];
				$destination_url = $row[1];

				WP_CLI::log( "Processing URL $index: $source_url -> $destination_url" );

				// Resolve the final destination URL for the current source URL.
				$final_destination_url = $this->resolve_final_destination( $destination_url, $source_to_destination );

				// Store the source URL with its final destination URL.
				$redirects[ $source_url ] = $final_destination_url;

				$progress->tick();
			}

			$progress->finish();
			WP_CLI::log( 'Completed processing of CSV file. Now validating and creating redirects.' );

			// Initialize a new progress bar for redirects validation.
			$progress = \WP_CLI\Utils\make_progress_bar( 'Validating and Creating Redirects', count( $redirects ) );

			// Validate and create redirects.
			foreach ( $redirects as $source => $destination ) {
				WP_CLI::log( "Validating URL: $source -> $destination" );

				// If it is a relative domain, prepend the site URL.
				if ( ! str_contains( $destination, 'http' ) ) {
					$site_url = get_site_url();
					// remove trailing slash from site URL.
					$site_url = rtrim( $site_url, '/' );
					$abs_url  = $site_url . $destination;
				} else {
					$abs_url = $destination;
				}

				$response = wp_remote_get(
					$abs_url,
					array(
						'timeout'   => 10,
						'sslverify' => false,
					)
				);
				$code     = wp_remote_retrieve_response_code( $response );

				if ( 200 === $code ) {
					$item = array(
						'url'         => trim( $source ),
						'action_data' => array( 'url' => trim( $destination ) ),
						'regex'       => false,
						'group_id'    => 1,
						'match_type'  => 'url',
						'action_type' => 'url',
						'action_code' => 301,
					);

					$created = Red_Item::create( $item );
					if ( is_wp_error( $created ) ) {
						WP_CLI::warning( "Failed to create redirect for: $source -> $destination. Error: " . $created->get_error_message() );
						$failed_urls[] = array( $source, $destination, '500', $created->get_error_message() );
					} else {
						WP_CLI::success( "Redirect created for: $source -> $destination" );
					}
				} else {
					$message = wp_remote_retrieve_response_message( $response );
					WP_CLI::warning( "URL validation failed for: $source -> $abs_url. HTTP Status: $code $message" );
					$failed_urls[] = array( $source, $destination, $code, $message );
				}

				$progress->tick();
			}

			$progress->finish();

			return $failed_urls;
		}

		/**
		 * Resolves the final destination URL for a given source URL by following the redirect chain.
		 *
		 * @param string $destination_url The initial destination URL to resolve.
		 * @param array  $source_to_destination An associative array mapping source URLs to their immediate destinations.
		 * @return string The resolved final destination URL.
		 */
		private function resolve_final_destination( string $destination_url, array $source_to_destination ): string {
			$domain = '4cornerresources.com';

			// Keep following the destination URL until it's not found as a source URL in the array.
			while ( $this->url_exists_in_column_a( $destination_url, $source_to_destination, $domain ) ) {
				// make sure the destination URL is normalized before checking for existence in the array.
				$destination_url = $this->normalize_url( $destination_url, $domain );
				$destination_url = $source_to_destination[ $destination_url ];
			}

			return $destination_url;
		}

		/**
		 * Checks if the normalized URL exists in the source column, accounting for trailing slashes.
		 *
		 * @param string $url The URL to check for existence as a source URL.
		 * @param array  $source_to_destination An associative array mapping source URLs to their immediate destinations.
		 * @param string $domain The domain within which to normalize URLs.
		 * @return bool True if the URL exists as a source, false otherwise.
		 */
		private function url_exists_in_column_a( string $url, array $source_to_destination, string $domain ): bool {
			// Check the URL with and without a trailing slash.
			$normalized_url            = $this->normalize_url( $url, $domain );
			$normalized_url_with_slash = rtrim( $normalized_url, '/' ) . '/';

			return array_key_exists( $normalized_url, $source_to_destination ) || array_key_exists( $normalized_url_with_slash, $source_to_destination );
		}

		/**
		 * Normalizes a URL by stripping away the domain and protocol if it belongs to the specified domain.
		 * Returns the URL path if it is a relative URL or belongs to the specified domain.
		 * Returns the original URL if it's absolute and not of the specified domain.
		 *
		 * @param string $url The URL to normalize.
		 * @param string $domain The domain within which to normalize URLs.
		 * @return string The normalized URL path or the original URL.
		 */
		private function normalize_url( string $url, string $domain ): string {
			// Parse the URL to get components.
			$parsed_url = wp_parse_url( $url );
			$host       = isset( $parsed_url['host'] ) ? $parsed_url['host'] : '';

			// If the URL is relative or the host matches the domain, normalize it.
			if ( empty( $host ) || str_contains( $host, $domain ) ) {
				// Ensure the path is set and prepend a slash if it's not there.
				$path = isset( $parsed_url['path'] ) ? $parsed_url['path'] : '';
				return '/' . ltrim( $path, '/' );
			}

			// Return the original URL if it's absolute and not of the specified domain.
			return $url;
		}
	}

	WP_CLI::add_command( 'csv_process', 'CSV_URL_Processor_Command' );
}