splmysql

(This is beta) MySQL wrapper library to split single UPDATE query into many tiny transaction queries.

How to use

Create Runner instance

Create runner instance with connection information.

sr, err := splmysql.NewByOptions("DB Name", "HostName", 3306, "USER", "PASSWORD", "CharacterSet")
if err != nil {
    return err
}
defer sr.Close()

You can use my.cnf.

sr, err := splmysql.NewByConf("DB Name", "path-to-my.cnf")
if err != nil {
    return err
}
defer sr.Close()

NewByOptions and NewByConf do not establish any connections to the database. They create connection information only.

Execute query

Create session object to get table information and test connection parameters. It's ok, execute query.

// Create session. Runner connect to DB and get table information.
// If connection information has invalid parameters, error returns. 
sessionData, err := sr.NewSession(sql)
if err != nil {
    return err
}

// Let's execute parallel
retrySessionData, err := sr.RunParallel(sessionData, numberOfParallel)

RunParallel() returns Session object retrySessionData to retry failed queries.

Fallback

If NewSession() returns NoUsableColumnError, you can run SimpleUpdate() as fallback. This method does not split query. Be careful to use Galera Cluster environment.

sess, err := sr.NewSession(sql)
if err != nil {
	e := reflect.ValueOf(err).Elem()
	switch {
	case e.Type() == reflect.TypeOf(splmysql.NoUsableColumnError{}):
        _, err := sr.SimpleUpdate(sql)
        if err != nil {
            return err
        }
    default:
        return err
    }
}

Logging

splmysql is using Logrus logger. You can overide it your own Logrus logger before call NewSession().

logger := logrus.New()
logger.Formatter = &logrus.TextFormatter{
    FullTimestamp: true,
}
logger.Out = os.Stdout

// Overide Logger
sr.Logger = logger