UTF-8 Multibyte (utf8mb4) support

Added alterTableCharacterSet method to JDatabaseDriver to allow Joomla! to transparently convert tables to utf8/utf8mb4

libraries/joomla/database/driver.php

@@ -461,6 +461,98 @@ public function alterDbCharacterSet($dbName)
 		return $this->execute();
 	}
 
+	/**
+	 * Alter a table's character set, obtaining an array of queries to do so from a protected method. The conversion is
+	 * wrapped in a transaction, if supported by the database driver. Otherwise the table will be locked before the
+	 * conversion. This prevents data corruption.
+	 *
+	 * @param   string   $tableName  The name of the table to alter
+	 * @param   boolean  $rethrow    True to rethrow database exceptions. Default: false (exceptions are suppressed)
+	 *
+	 * @return  boolean  True if successful
+	 *
+	 * @since   CMS 3.5.0
+	 * @throws  RuntimeException  If the table name is empty
+	 * @throws  Exception  Relayed from the database layer if a database error occurs and $rethrow == true
+	 */
+	public function alterTableCharacterSet($tableName, $rethrow = false)
+	{
+		if (is_null($tableName))
+		{
+			throw new RuntimeException('Table name must not be null.');
+		}
+
+		$queries = $this->getAlterTableCharacterSet($tableName);
+
+		if (empty($queries))
+		{
+			return false;
+		}
+
+		$hasTransaction = true;
+
+		try
+		{
+			$this->transactionStart();
+		}
+		catch (Exception $e)
+		{
+			$hasTransaction = false;
+			$this->lockTable($tableName);
+		}
+
+		foreach ($queries as $query)
+		{
+			try
+			{
+				$this->setQuery($query)->execute();
+			}
+			catch (Exception $e)
+			{
+				if ($hasTransaction)
+				{
+					$this->transactionRollback();
+				}
+				else
+				{
+					$this->unlockTables();
+				}
+
+				if ($rethrow)
+				{
+					throw $e;
+				}
+
+				return false;
+			}
+		}
+
+		if ($hasTransaction)
+		{
+			try
+			{
+				$this->transactionCommit();
+			}
+			catch (Exception $e)
+			{
+				$this->transactionRollback();
+
+				if ($rethrow)
+				{
+					throw $e;
+				}
+
+				return false;
+			}
+		}
+		else
+		{
+			$this->unlockTables();
+		}
+
+		return true;
+	}
+
 	/**
 	 * Connects to the database if needed.
 	 *
@@ -631,6 +723,93 @@ protected function getAlterDbCharacterSet($dbName)
 		return 'ALTER DATABASE ' . $this->quoteName($dbName) . ' CHARACTER SET `' . $charset .'`';
 	}
 
+	/**
+	 * Get the query strings to alter the character set and collation of a table.
+	 *
+	 * @param   string  $tableName  The name of the table
+	 *
+	 * @return  string[]  The queries required to alter the table's character set and collation
+	 *
+	 * @since   CMS 3.5.0
+	 */
+	public function getAlterTableCharacterSet($tableName)
+	{
+		$charset = $this->utf8mb4 ? 'utf8mb4' : 'utf8';
+		$collation = $charset . '_general_ci';
+
+		$quotedTableName = $this->quoteName($tableName);
+
+		$queries = array();
+		$queries[] = "ALTER TABLE $quotedTableName CONVERT TO CHARACTER SET $charset COLLATE $collation";
+
+		/**
+		 * We also need to convert each text column, modifying their character set and collation. This allows us to
+		 * change, for example, a utf8_bin collated column to a utf8mb4_bin collated column.
+		 */
+		$sql = "SHOW FULL COLUMNS FROM $quotedTableName";
+		$this->setQuery($sql);
+		$columns = $this->loadAssocList();
+		$columnMods = array();
+
+		if (is_array($columns))
+		{
+			foreach ($columns as $column)
+			{
+				// Make sure we are redefining only columns which do support a collation
+				$col = (object)$column;
+
+				if (empty($col->Collation))
+				{
+					continue;
+				}
+
+				// Default new collation: utf8_general_ci or utf8mb4_general_ci
+				$newCollation = $charset . '_general_ci';
+				$collationParts = explode('_', $col->Collation);
+
+				/**
+				 * If the collation is in the form charset_collationType_ci or charset_collationType we have to change
+				 * the charset but leave the collationType intact (e.g. utf8_bin must become utf8mb4_bin, NOT
+				 * utf8mb4_general_ci).
+				 */
+				if (count($collationParts) >= 2)
+				{
+					$ci = array_pop($collationParts);
+					$collationType = array_pop($collationParts);
+					$newCollation = $charset . '_' . $collationType . '_' . $ci;
+
+					/**
+					 * When the last part of the old collation is not _ci we have a charset_collationType format,
+					 * something like utf8_bin. Therefore the new collation only has *two* parts.
+					 */
+					if ($ci != 'ci')
+					{
+						$newCollation = $charset . '_' . $ci;
+					}
+				}
+
+				// If the old and new collation is the same we don't have to change the collation type
+				if (strtolower($newCollation) == strtolower($col->Collation))
+				{
+					continue;
+				}
+
+				$null = $col->Null == 'YES' ? 'NULL' : 'NOT NULL';
+				$default = is_null($col->Default) ? '' : "DEFAULT '" . $this->q($col->Default) . "'";
+				$columnMods[] = "MODIFY COLUMN `{$col->Field}` {$col->Type} CHARACTER SET $charset COLLATE $newCollation $null $default";
+			}
+		}
+
+		if (count($columnMods))
+		{
+			$queries[] = "ALTER TABLE $quotedTableName " .
+				implode(',', $columnMods) .
+				" CHARACTER SET $charset COLLATE $collation";
+		}
+
+		return $queries;
+	}
+
 	/**
 	 * Return the query string to create new Database.
 	 * Each database driver, other than MySQL, need to override this member to return correct string.