我正在研究Flink,我想建立一个扩展的操作函数ProcessWindowFunction,并重载一个新的构造函数,其参数为该类的字段值,但当这个类被实例化时,没有这个字段,我感到困惑。
import com.aliyun.datahub.client.model.Field;
import com.aliyun.datahub.client.model.FieldType;
import com.aliyun.datahub.client.model.PutRecordsResult;
import io.github.streamingwithflink.chapter8.PoJoElecMeterSource;
import org.apache.flink.streaming.api.TimeCharacteristic;
import org.apache.flink.streaming.api.datastream.DataStream;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
import org.apache.flink.streaming.api.windowing.assigners.TumblingProcessingTimeWindows;
import org.apache.flink.streaming.api.windowing.time.Time;
public class DataHubSinkDemo {
public static void main(String[] args) throws Exception {
StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
env.setStreamTimeCharacteristic(TimeCharacteristic.ProcessingTime);
env.enableCheckpointing(10_000L);
env.setParallelism(2);
RecordSchemaSer schema = new RecordSchemaSer();
schema.addField(new Field("id", FieldType.STRING));
DataStream<PutRecordsResult> out = env
.addSource(new PoJoElecMeterSource())
.keyBy( r -> r.getId())
.window(TumblingProcessingTimeWindows.of(Time.seconds(3)))
.process(new PutDatahubFunction<>(schema)); // PutDatahubFunction is my build a new Operator function class
env.execute();
}
}
变量schema是我想发送给构造函数的一个参数,它是RecordSchemaSer类的一个实例。
import com.aliyun.datahub.client.model.RecordSchema;
import java.io.Serializable;
public class RecordSchemaSer
extends RecordSchema
implements Serializable {
}
PutDatahubFunction是ProcessWindowFunction的扩展类,代码如下。
import com.aliyun.datahub.client.model.*;
import io.github.streamingwithflink.chapter8.PUDAPoJo;
import org.apache.flink.configuration.Configuration;
import org.apache.flink.streaming.api.functions.windowing.ProcessWindowFunction;
import org.apache.flink.streaming.api.windowing.windows.TimeWindow;
import org.apache.flink.util.Collector;
import java.lang.reflect.Field;
import java.util.ArrayList;
import java.util.List;
public class PutDatahubFunction<IN extends PUDAPoJo, KEY>
extends ProcessWindowFunction<IN, PutRecordsResult, KEY, TimeWindow> {
private DataHubBase dataHubHandler;
private List<RecordEntry> recordEntries;
private RecordSchema schema;
public PutDatahubFunction(RecordSchema schema) {
this.schema = schema;
System.out.println("field 'id' not exist ? " + this.schema.containsField("id")); // it's true
}
@Override
public void open(Configuration parameters) throws Exception {
.........
}
@Override
public void process(KEY KEY,
Context context,
Iterable<IN> elements,
Collector<PutRecordsResult> out)
throws Exception {
RecordEntry entry = new RecordEntry();
for (IN e : elements) {
System.out.println("field 'id' not exist ? " + this.schema.containsField("id")); // it's false
......
}
}
}
第一个system.out在构造函数中,this.schema.containsField("id")是true,但第二个system.out在处理方法中,this.schema.containsField("id")是false!为什么?我有system.out的两个类名的实例都是PutDatahubFunction。
使用ValueState不工作,因为构造函数没有调用getRuntimeContext(),否则在线程 "main "中出现了java.lang.IllegalStateException。运行时上下文没有被初始化,代码如下:
private ValueState<RecordSchema> schema;
public PutTupleDatahubFunction(RecordSchema schema) throws IOException {
ValueStateDescriptor schemaDes =
new ValueStateDescriptor("datahub schema", TypeInformation.of(RecordSchema.class));
/*
* error Exception in thread "main" java.lang.IllegalStateException:
* The runtime context has not been initialized.
*/
this.schema = getRuntimeContext().getState(schemaDes);
this.schema.update(schema);
}
我很困惑,谁能告诉我原因,有什么方法可以把参数传给这个操作函数类的构造函数吗?
我终于知道为什么了,原因是Serialize和Deserialize。我在编写RecordSchemaSer的时候原因是Serialize内容,由于null的原因,我没有编写RecordSchemaSer。
public class RecordSchemaSer
extends RecordSchema
implements Serializable
{
}