Add detection of posts table character set

Author: schlessera

features/db-import.feature

@@ -157,16 +157,25 @@ Feature: Import a WordPress database
     Given a WP install
 
     When I run `wp post create --post_title="🍣"`
-    And I run `wp db export wp_cli_test.sql`
-    Then the wp_cli_test.sql file should exist
-
-    When I run `wp post list`
+    And I run `wp post list`
     Then the return code should be 0
     And STDOUT should contain:
       """
       🍣
       """
 
+    When I try `wp db export wp_cli_test.sql --debug`
+    Then the return code should be 0
+    And the wp_cli_test.sql file should exist
+    And STDERR should contain:
+      """
+      Detected character set of the posts table: utf8mb4
+      """
+    And STDERR should contain:
+      """
+      Setting missing default character set to utf8mb4
+      """
+
     When I run `wp db import --dbuser=wp_cli_test --dbpass=password1`
     Then STDOUT should be:
       """

src/DB_Command.php

@@ -27,6 +27,20 @@
  */
 class DB_Command extends WP_CLI_Command {
 
+	/**
+	 * Legacy UTF-8 encoding for MySQL.
+	 *
+	 * @var string
+	 */
+	const ENCODING_UTF8 = 'utf8';
+
+	/**
+	 * Standards-compliant UTF-8 encoding for MySQL.
+	 *
+	 * @var string
+	 */
+	const ENCODING_UTF8MB4 = 'utf8mb4';
+
 	/**
 	 * A list of incompatible SQL modes.
 	 *
@@ -556,11 +570,19 @@ public function export( $args, $assoc_args ) {
 
 		/*
 		 * In case that `--default-character-set` is not given and `DB_CHARSET` is `utf8`,
-		 * use `utf8mb4` as a `default-character-set` to ensure emojis are encoded correctly.
+		 * we try to deduce what the actual character set for the posts table of the
+		 * current database is and use `utf8mb4` as a `default-character-set` if that
+		 * seems like the safer default, to ensure emojis are encoded correctly.
 		 */
-		if ( ! isset( $assoc_args['default-character-set'] ) &&
-			defined( 'DB_CHARSET' ) && 'utf8' === constant( 'DB_CHARSET' ) ) {
-			$assoc_args['default-character-set'] = 'utf8mb4';
+		if (
+			! isset( $assoc_args['default-character-set'] )
+			&&
+			( defined( 'DB_CHARSET' ) && self::ENCODING_UTF8 === constant( 'DB_CHARSET' ) )
+			&&
+			self::ENCODING_UTF8MB4 === $this->get_posts_table_charset( $assoc_args )
+		) {
+			WP_CLI::debug( 'Setting missing default character set to ' . self::ENCODING_UTF8MB4, 'db' );
+			$assoc_args['default-character-set'] = self::ENCODING_UTF8MB4;
 		}
 
 		$initial_command = sprintf( "{$mysqldump_binary}%s ", $this->get_defaults_flag_string( $assoc_args ) );
@@ -616,6 +638,45 @@ public function export( $args, $assoc_args ) {
 		}
 	}
 
+	/**
+	 * Get the current character set of the posts table.
+	 *
+	 * @param array Associative array of associative arguments.
+	 * @return string Posts table character set.
+	 */
+	private function get_posts_table_charset( $assoc_args ) {
+		$query = 'SELECT CCSA.character_set_name '
+				. 'FROM information_schema.`TABLES` T, '
+				. 'information_schema.`COLLATION_CHARACTER_SET_APPLICABILITY` CCSA '
+				. 'WHERE CCSA.collation_name = T.table_collation '
+				. 'AND T.table_schema = "' . DB_NAME . '" '
+				. 'AND T.table_name LIKE "%\_posts";';
+
+		list( $stdout, $stderr, $exit_code ) = self::run(
+			sprintf(
+				'/usr/bin/env mysql%s --no-auto-rehash --batch --skip-column-names',
+				$this->get_defaults_flag_string( $assoc_args )
+			),
+			[ 'execute' => $query ],
+			false
+		);
+
+		if ( $exit_code ) {
+			WP_CLI::warning(
+				'Failed to get current character set of the posts table.'
+				. ( ! empty( $stderr ) ? " Reason: {$stderr}" : '' )
+			);
+
+			return self::ENCODING_UTF8MB4;
+		}
+
+		$stdout = trim( $stdout );
+
+		WP_CLI::debug( "Detected character set of the posts table: {$stdout}.", 'db' );
+
+		return $stdout;
+	}
+
 	/**
 	 * Imports a database from a file or from STDIN.
 	 *
@@ -1244,7 +1305,7 @@ public function search( $args, $assoc_args ) {
 		}
 
 		$encoding = null;
-		if ( 0 === strpos( $wpdb->charset, 'utf8' ) ) {
+		if ( 0 === strpos( $wpdb->charset, self::ENCODING_UTF8 ) ) {
 			$encoding = 'UTF-8';
 		}
 
@@ -1616,7 +1677,7 @@ private static function is_text_col( $type ) {
 	 * @return string|array An escaped string if given a string, or an array of escaped strings if given an array of strings.
 	 */
 	private static function esc_sql_ident( $idents ) {
-		$backtick = function ( $v ) {
+		$backtick = static function ( $v ) {
 			// Escape any backticks in the identifier by doubling.
 			return '`' . str_replace( '`', '``', $v ) . '`';
 		};