Spark：在数据集上应用地图功能 <T> 在java中

Question

I have below code which works fine with foreach function on Dataset. finalJoined is a DataFrame .

    KieServices ks = KieServices.Factory.get();
    KieContainer kContainer = ks.getKieClasspathContainer();
    ClassTag<KieBase> classTagTest =  scala.reflect.ClassTag$.MODULE$.apply(KieBase.class);
    Broadcast<KieBase> broadcastRules = context.broadcast(kContainer.getKieBase("rules"), classTagTest);


    Encoder<RuleParams> encoder = Encoders.bean(RuleParams.class);
        Dataset<RuleParams> ds = new Dataset<RuleParams>(sparkSession, finalJoined.logicalPlan(), encoder);
        System.out.println("Printing ruleParams DS");
        ds.show();
        ds.foreach(ruleParam -> droolprocess(broadcastRules.value(), ruleParam));

Here foreach method returns void .

I need Dataset<RuleParams> as return value . below is my droolprocess method which calls rule engine and updates RuleParams objects.

public static void droolprocess(KieBase base, RuleParams ruleParams) {
        StatelessKieSession session = base.newStatelessKieSession();
session.execute(CommandFactory.newInsert(ruleParams));
        System.out.println("After firing  rules");
        System.out.println(ruleParams.getPriceItemParam1());
        System.out.println(ruleParams.getCisDivision());
         }

I have seen some questions on stackoverflow and elsewhere but I am not sure how to write map function instead of foreach to return Dataset<RuleParams>

Can anyone help here?

Answer 1

You can use like below:

 Dataset<RuleParams> ds = new Dataset<RuleParams>(sparkSession, finalJoined.logicalPlan(), encoder);
    StructType schema = ds.schema();
    ds = ds.map(ruleParams -> {

RuleParams theRuleParams= ruleParams;

    ...//your processing
    return theRuleParams;
    }, RowEncoder.apply(schema));

Once mapping is done you need to return row by creating each of the row if you adding/deleting and modifying data in each row. Finally apply back the schema so that the dataset knows the schema that it will be returning after performing the map operation.