HeartSaVioR commented on code in PR #43961: URL: https://github.com/apache/spark/pull/43961#discussion_r1445740654
########## sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/ValueStateImpl.scala: ########## @@ -0,0 +1,105 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +package org.apache.spark.sql.execution.streaming + +import java.io.Serializable + +import org.apache.commons.lang3.SerializationUtils + +import org.apache.spark.internal.Logging +import org.apache.spark.sql.catalyst.InternalRow +import org.apache.spark.sql.catalyst.expressions._ +import org.apache.spark.sql.execution.streaming.state.StateStore +import org.apache.spark.sql.streaming.ValueState +import org.apache.spark.sql.types._ + +/** + * Class that provides a concrete implementation for a single value state associated with state + * variables used in the streaming transformWithState operator. + * @param store - reference to the StateStore instance to be used for storing state + * @param stateName - name of logical state partition + * @tparam S - data type of object that will be stored + */ +class ValueStateImpl[S]( + store: StateStore, + stateName: String) extends ValueState[S] with Logging{ + + private def encodeKey(): UnsafeRow = { + val keyOption = ImplicitKeyTracker.getImplicitKeyOption + if (!keyOption.isDefined) { + throw new UnsupportedOperationException("Implicit key not found for operation on" + + s"stateName=$stateName") + } + + val schemaForKeyRow: StructType = new StructType().add("key", BinaryType) + val keyByteArr = SerializationUtils.serialize(keyOption.get.asInstanceOf[Serializable]) Review Comment: https://www.databricks.com/blog/2016/01/04/introducing-apache-spark-datasets.html Check `Lightning-fast Serialization with Encoders`. You can see how much performance benefit we get on serde if we make the custom types be bound to Spark SQL type system. Combining two UnsafeRows into one could be also done via either flattening schema and data (concatenate) or nested model. Users may have to provide the encoder for types on state (value for all types of state, key for map state) which may be burdensome, but that is what Flink requires users to do as well - it requires type information (at least class type), which is to perform the magic behind the scene. They build their own type system and serializer, not just simply going through Java's Serializable. ``` @Override public void open(OpenContext openContext) throws Exception { state = getRuntimeContext().getState(new ValueStateDescriptor<>("myState", CountWithTimestamp.class)); } ``` https://nightlies.apache.org/flink/flink-docs-release-1.18/docs/dev/datastream/fault-tolerance/serialization/types_serialization/#data-types--serialization -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
