[Python] Generate Python database connector from data source string

[tonyyang-svail]

To avoid all the code generator having the same database connection logic, we may have a Python function like

def connect_with_data_source(data_source):
    driver, source = data_source.split("://")
    if driver != "mysql":
        raise ValueError("connect_with_data_source doesn't support driver type {}".format(driver))

    if driver == "mysql":
        user, passwd, _, host, port = re.findall("^(\w*):(\w*)@(\w*)\((.*):(.*)\)/?.*$", source)[0]
        from MySQLdb import connect
        conn = connect(user=user,
                       passwd=passwd,
                       host=host,
                       port=int(port))
    conn.driver = driver
    return conn

@weiguoz @Yancey1989 What do you think?

[Yancey0623]

Hi @tonyyang-svail, I'm not sure we would implement parsing datasource logic in Python, each DBdriver has ParseDSN function to parse the datasource string, such as gohive.ParseDSN would parse datasource string to fields username, password, session_cfg etc... . re-implement this logic may easy to cause a bug.

[tonyyang-svail]

Hi @tonyyang-svail, I'm not sure we would implement parsing datasource logic in Python, each DBdriver has ParseDSN function to parse the datasource string, such as gohive.ParseDSN would parse datasource string to fields username, password, session_cfg etc... . re-implement this logic may easy to cause a bug.

@Yancey1989 Sure

[weiguoz]

I agree that making a connection with Datasource in python. And this my opinions:

• the format of Datasource needs to be standard, whether in go or in python
  I think sqlalchemy can give some reference.

- SQLFlow splits the training dataset to train&val, reads schema, does verify and runs the standard SQL in go. Can SQLFlow do these just in python, replace the driver database/sql in go?