sqlflow integrate with zeppelin (SQLFlowInterpreter.java)

[Klay1124]

fix #3

This file is to let sqlflow inherit the interpreter interface provided by Zeppelin, so as to connect the two and realize the abstract class of Zeppelin. In this file, you can read the configuration file parameters, construct the sqlflow client, start and close the interpreter, and so on.

Create a new class named org.Apache.Zeppelin.Sqlflow and let it inherit from org.Apache.Zeppelin.Interpreter.Interpreter. Since an abstract class is inherited, constructors and overridden methods are required, with 6 methods:

// Initialize interpreter
public void open();

// Close the interpreter and free resources
public void close();

// Run the interpreter and return the result. 'query' is the code entered in Zeppelin, and 'interpretercontext' is the additional running context information
public InterpreterResult interpret(String query, InterpreterContext interpreterContext);

// Canceling the run will interrupt the execution of the interpret method
public void cancel();

// For processing types of dynamic tables, please refer to: https://zeppelin.apache.org/docs/0.8.2/usage/dynamic_form/intro.html
public ForType getFormType();

// Get the running progress, the return value is 0-100
public int getProgress();

Among these methods, the most critical one is public InterpreterResult interpret(String query, InterpreterContext interpreterContext); It is a bridge between sqlflow and Zeppelin. The collection, creation and release of user information will be realized through this method.

Examples are as follows:

@Override
public InterpreterResult interpret(String query, InterpreterContext context)
		throws InterpreterException {

	if (StringUtils.isBlank(query)) {
		return new InterpreterResult(InterpreterResult.Code.SUCCESS);
	}

	// Create user
	SQLFlow client = EnvironmentSpecificSQLFlowClient.getClient(
			new MessageHandlerZeppelin(context), parameters);
        try {
	        // Run custom sqlflow code
	        client.run(query);
	        // Release user
		client.release();

                return new InterpreterResult(InterpreterResult.Code.SUCCESS);
	} catch (InterruptedException e) {
		LOGGER.error("client error", e);
                return new InterpreterResult(InterpreterResult.Code.ERROR);
	}
}

The specific code can refer to SQLFlowInterpreter.java

[weiguoz]

Do we only support mysql databases?

[Klay1124]

We don't just support mysql database. For other SQL engines, including Oracle, Hive, SparkSQL, Flink, etc., developers can expand and improve according to actual needs.

[weiguoz]

Shall we still use mysql.username to specific the hive username ?

[Klay1124]

This initial version has not yet opened the connection to the hive database.

[weiguoz]

This initial version has not yet opened the connection to the hive database.

Let's do this under the new issue.

[weiguoz]

useless blank.

[Klay1124]

useless blank removed.

[weiguoz]

blank.

[Klay1124]

useless blank removed.

[weiguoz]

blank

[Klay1124]

useless blank removed.