bvarghese1 commented on code in PR #25753: URL: https://github.com/apache/flink/pull/25753#discussion_r1928136710
########## flink-table/flink-table-runtime/src/main/java/org/apache/flink/table/runtime/operators/over/NonTimeUnboundedPrecedingFunction.java: ########## @@ -0,0 +1,591 @@ +/* + * 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.flink.table.runtime.operators.over; + +import org.apache.flink.api.common.functions.OpenContext; +import org.apache.flink.api.common.state.MapState; +import org.apache.flink.api.common.state.MapStateDescriptor; +import org.apache.flink.api.common.state.ValueState; +import org.apache.flink.api.common.state.ValueStateDescriptor; +import org.apache.flink.api.common.typeinfo.Types; +import org.apache.flink.api.java.typeutils.ListTypeInfo; +import org.apache.flink.streaming.api.functions.KeyedProcessFunction; +import org.apache.flink.table.api.DataTypes; +import org.apache.flink.table.data.GenericRowData; +import org.apache.flink.table.data.RowData; +import org.apache.flink.table.data.RowData.FieldGetter; +import org.apache.flink.table.data.utils.JoinedRowData; +import org.apache.flink.table.runtime.dataview.PerKeyStateDataViewStore; +import org.apache.flink.table.runtime.functions.KeyedProcessFunctionWithCleanupState; +import org.apache.flink.table.runtime.generated.AggsHandleFunction; +import org.apache.flink.table.runtime.generated.GeneratedAggsHandleFunction; +import org.apache.flink.table.runtime.generated.GeneratedRecordComparator; +import org.apache.flink.table.runtime.generated.GeneratedRecordEqualiser; +import org.apache.flink.table.runtime.generated.RecordEqualiser; +import org.apache.flink.table.runtime.typeutils.InternalTypeInfo; +import org.apache.flink.table.types.logical.LogicalType; +import org.apache.flink.table.types.logical.RowType; +import org.apache.flink.types.RowKind; +import org.apache.flink.util.Collector; + +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +import java.util.ArrayList; +import java.util.Comparator; +import java.util.List; +import java.util.ListIterator; + +/** A basic implementation to support non-time unbounded over aggregate with retract mode. */ +public class NonTimeUnboundedPrecedingFunction<K> + extends KeyedProcessFunctionWithCleanupState<K, RowData, RowData> { + private static final long serialVersionUID = 1L; + + private static final Logger LOG = + LoggerFactory.getLogger(NonTimeUnboundedPrecedingFunction.class); + + private final int keyIdx; + private final int sortKeyIdx; + private final FieldGetter sortKeyFieldGetter; + + private final GeneratedAggsHandleFunction genAggsHandler; + private final GeneratedRecordEqualiser generatedRecordEqualiser; + private final GeneratedRecordComparator generatedRecordComparator; + private final GeneratedRecordComparator keyGeneratedRecordComparator; + + // The util to compare two rows based on the sort attribute. + private transient Comparator<RowData> sortKeyComparator; + private transient Comparator<RowData> newSortKeyComparator; + // The record equaliser used to equal RowData. + private transient RecordEqualiser equaliser; + + private final LogicalType[] accTypes; + private final LogicalType[] inputFieldTypes; + protected transient JoinedRowData output; + + // state to hold the accumulators of the aggregations + private transient ValueState<RowData> accState; + // state to hold rows until state ttl expires + private transient ValueState<List<RowData>> inputState; + + // state to hold the Long ID counter + private transient ValueState<Long> idState; + + // state to hold a list of sorted keys with an artificial id + // The artificial id acts as the key in the valueMapState + private transient ValueState<List<RowData>> sortedKeyState; + // state to hold rows until state ttl expires + private transient MapState<Long, RowData> valueMapState; + // state to hold row ID and its associated accumulator + private transient MapState<Long, RowData> accMapState; + + protected transient AggsHandleFunction currFunction; + protected transient AggsHandleFunction prevFunction; + + public NonTimeUnboundedPrecedingFunction( + long minRetentionTime, + long maxRetentionTime, + GeneratedAggsHandleFunction genAggsHandler, + GeneratedRecordEqualiser genRecordEqualiser, + GeneratedRecordComparator genRecordComparator, + GeneratedRecordComparator keyGenRecordComparator, + LogicalType[] accTypes, + LogicalType[] inputFieldTypes, + FieldGetter sortKeyFieldGetter, + int sortKeyIdx) { + super(minRetentionTime, maxRetentionTime); + this.genAggsHandler = genAggsHandler; + this.generatedRecordEqualiser = genRecordEqualiser; + this.generatedRecordComparator = genRecordComparator; + this.keyGeneratedRecordComparator = keyGenRecordComparator; + this.accTypes = accTypes; + this.inputFieldTypes = inputFieldTypes; + this.sortKeyFieldGetter = sortKeyFieldGetter; + this.sortKeyIdx = sortKeyIdx; + this.keyIdx = 0; + } + + @Override + public void open(OpenContext openContext) throws Exception { + // Initialize agg functions + currFunction = genAggsHandler.newInstance(getRuntimeContext().getUserCodeClassLoader()); + currFunction.open(new PerKeyStateDataViewStore(getRuntimeContext())); + + prevFunction = genAggsHandler.newInstance(getRuntimeContext().getUserCodeClassLoader()); + prevFunction.open(new PerKeyStateDataViewStore(getRuntimeContext())); + + // Initialize output record + output = new JoinedRowData(); + + // Initialize record equaliser + equaliser = + generatedRecordEqualiser.newInstance(getRuntimeContext().getUserCodeClassLoader()); + + // Initialize sort comparator + sortKeyComparator = + generatedRecordComparator.newInstance(getRuntimeContext().getUserCodeClassLoader()); + + newSortKeyComparator = + keyGeneratedRecordComparator.newInstance( + getRuntimeContext().getUserCodeClassLoader()); + + // Initialize accumulator state + InternalTypeInfo<RowData> accTypeInfo = InternalTypeInfo.ofFields(accTypes); + ValueStateDescriptor<RowData> accStateDesc = + new ValueStateDescriptor<RowData>("accState", accTypeInfo); + accState = getRuntimeContext().getState(accStateDesc); + + // Input elements are all binary rows as they came from network + InternalTypeInfo<RowData> inputType = InternalTypeInfo.ofFields(inputFieldTypes); + ListTypeInfo<RowData> rowListTypeInfo = new ListTypeInfo<RowData>(inputType); + + ValueStateDescriptor<List<RowData>> inputStateDescriptor = + new ValueStateDescriptor<List<RowData>>("inputState", rowListTypeInfo); + + // Initialize state which maintains records in sorted(ASC) order + inputState = getRuntimeContext().getState(inputStateDescriptor); + + ValueStateDescriptor<Long> idStateDescriptor = + new ValueStateDescriptor<Long>("idState", Long.class); + idState = getRuntimeContext().getState(idStateDescriptor); + + LogicalType idType = DataTypes.BIGINT().getLogicalType(); + LogicalType sortKeyType = inputType.toRowType().getTypeAt(sortKeyIdx); + RowType sortedRow = RowType.of(idType, sortKeyType); + LogicalType[] sortedRowTypes = sortedRow.getChildren().toArray(new LogicalType[0]); + InternalTypeInfo<RowData> sortedKeyRowType = InternalTypeInfo.ofFields(sortedRowTypes); + ListTypeInfo<RowData> sortedKeyTypeInfo = new ListTypeInfo<>(sortedKeyRowType); + // Initialize state which maintains a sorted list of pair(ID, sortKey) + ValueStateDescriptor<List<RowData>> sortedKeyStateDescriptor = + new ValueStateDescriptor<List<RowData>>("sortedKeyState", sortedKeyTypeInfo); + sortedKeyState = getRuntimeContext().getState(sortedKeyStateDescriptor); + + MapStateDescriptor<Long, RowData> valueStateDescriptor = + new MapStateDescriptor<Long, RowData>("valueState", Types.LONG, inputType); + valueMapState = getRuntimeContext().getMapState(valueStateDescriptor); + + MapStateDescriptor<Long, RowData> accStateDescriptor = + new MapStateDescriptor<Long, RowData>("accMapState", Types.LONG, accTypeInfo); + accMapState = getRuntimeContext().getMapState(accStateDescriptor); + + initCleanupTimeState("NonTimeUnboundedPrecedingFunctionCleanupTime"); + } + + /** + * Puts an element from the input stream into state. Emits the aggregated value for the newly + * inserted element. For append stream emits updates(UB, UA) for all elements which are present + * after the newly inserted element. For retract stream emits an UB for the element and + * thereafter emits updates(UB, UA) for all elements which are present after the retracted + * element. + * + * @param input The input value. + * @param ctx A {@link Context} that allows querying the timestamp of the element and getting + * TimerService for registering timers and querying the time. The context is only valid + * during the invocation of this method, do not store it. + * @param out The collector for returning result values. + * @throws Exception + */ + @Override + public void processElement( + RowData input, + KeyedProcessFunction<K, RowData, RowData>.Context ctx, + Collector<RowData> out) + throws Exception { + // register state-cleanup timer + registerProcessingCleanupTimer(ctx, ctx.timerService().currentProcessingTime()); + + // get last accumulator + RowData lastAccumulatorCurr = accState.value(); + if (lastAccumulatorCurr == null) { + // initialize accumulator + lastAccumulatorCurr = currFunction.createAccumulators(); + } + // set accumulator in function context first + currFunction.setAccumulators(lastAccumulatorCurr); + + // get last accumulator + RowData lastAccumulatorPrev = accState.value(); + if (lastAccumulatorPrev == null) { + // initialize accumulator + lastAccumulatorPrev = prevFunction.createAccumulators(); + } + // set accumulator in function context first + prevFunction.setAccumulators(lastAccumulatorPrev); + + RowKind rowKind = input.getRowKind(); + + if (rowKind == RowKind.INSERT || rowKind == RowKind.UPDATE_AFTER) { + // insertIntoSortedList(input, out); + // insertIntoSortedListWithValueMap(input, out); + insertIntoSortedListWithValueMapAndAccumulatorMap(input, out); + } else if (rowKind == RowKind.DELETE || rowKind == RowKind.UPDATE_BEFORE) { + // removeFromSortedList(input, out); + // removeFromSortedListWithValueMap(input, out); + removeFromSortedListWithValueMapAndAccumulatorMap(input, out); + } + + // Reset acc state since we can have out of order inserts into the ordered list + currFunction.resetAccumulators(); + prevFunction.resetAccumulators(); + currFunction.cleanup(); + prevFunction.cleanup(); + } + + private void insertIntoSortedList(RowData rowData, Collector<RowData> out) throws Exception { + List<RowData> rowList = inputState.value(); + if (rowList == null) { + rowList = new ArrayList<>(); + } + boolean isInserted = false; + RowKind origRowKind = rowData.getRowKind(); + rowData.setRowKind(RowKind.INSERT); + ListIterator<RowData> iterator = rowList.listIterator(); + + while (iterator.hasNext()) { + RowData curRow = iterator.next(); + if (sortKeyComparator.compare(curRow, rowData) > 0) { + iterator.previous(); + iterator.add(rowData); + isInserted = true; + break; + } + currFunction.accumulate(curRow); + prevFunction.accumulate(curRow); + } + + // Add to the end of the list + if (!isInserted) { + iterator.add(rowData); + } + + // Only accumulate rowData with currFunction + currFunction.accumulate(rowData); + + // Update state with the newly inserted row + inputState.update(rowList); + + // prepare output row + output.setRowKind(origRowKind); + output.replace(rowData, currFunction.getValue()); + out.collect(output); + + // Emit updated agg value for all records after newly inserted row + while (iterator.hasNext()) { + RowData curRow = iterator.next(); + currFunction.accumulate(curRow); + prevFunction.accumulate(curRow); + // Generate UPDATE_BEFORE + output.setRowKind(RowKind.UPDATE_BEFORE); + output.replace(curRow, prevFunction.getValue()); + out.collect(output); + // Generate UPDATE_AFTER + output.setRowKind(RowKind.UPDATE_AFTER); + output.replace(curRow, currFunction.getValue()); + out.collect(output); + } + } + + private void insertIntoSortedListWithValueMap(RowData input, Collector<RowData> out) + throws Exception { + List<RowData> sortedKeyList = sortedKeyState.value(); + if (sortedKeyList == null) { + sortedKeyList = new ArrayList<>(); + } + Long id = idState.value(); + if (id == null) { + id = 0L; + } + boolean isInserted = false; + RowKind origRowKind = input.getRowKind(); + input.setRowKind(RowKind.INSERT); + ListIterator<RowData> iterator = sortedKeyList.listIterator(); + + while (iterator.hasNext()) { + RowData curKey = iterator.next(); // (ID, sortKey) + RowData inputKey = GenericRowData.of(-1L, sortKeyFieldGetter.getFieldOrNull(input)); + + if (newSortKeyComparator.compare(curKey, inputKey) > 0) { + iterator.previous(); + iterator.add(GenericRowData.of(id, sortKeyFieldGetter.getFieldOrNull(input))); + valueMapState.put(id, input); + isInserted = true; + id++; + break; + } + // Can also add the accKey to the sortedKeyList to avoid reading from the valueMapState + RowData curRow = valueMapState.get(curKey.getLong(keyIdx)); + currFunction.accumulate(curRow); + prevFunction.accumulate(curRow); + } + + // Add to the end of the list + if (!isInserted) { + iterator.add(GenericRowData.of(id, sortKeyFieldGetter.getFieldOrNull(input))); + valueMapState.put(id, input); + id++; + } + + // Only accumulate rowData with currFunction + currFunction.accumulate(input); + + // Update sorted key state with the newly inserted row's key + sortedKeyState.update(sortedKeyList); + idState.update(id); + + // prepare output row + output.setRowKind(origRowKind); + output.replace(input, currFunction.getValue()); + out.collect(output); + + // Emit updated agg value for all records after newly inserted row + while (iterator.hasNext()) { + RowData curKey = iterator.next(); + RowData curValue = valueMapState.get(curKey.getLong(keyIdx)); + currFunction.accumulate(curValue); + prevFunction.accumulate(curValue); + // Generate UPDATE_BEFORE + output.setRowKind(RowKind.UPDATE_BEFORE); + output.replace(curValue, prevFunction.getValue()); + out.collect(output); + // Generate UPDATE_AFTER + output.setRowKind(RowKind.UPDATE_AFTER); + output.replace(curValue, currFunction.getValue()); + out.collect(output); + } + } + + private void insertIntoSortedListWithValueMapAndAccumulatorMap( + RowData input, Collector<RowData> out) throws Exception { + List<RowData> sortedKeyList = sortedKeyState.value(); + if (sortedKeyList == null) { + sortedKeyList = new ArrayList<>(); + } + Long id = idState.value(); + if (id == null) { + id = 0L; + } + boolean isInserted = false; + RowKind origRowKind = input.getRowKind(); + input.setRowKind(RowKind.INSERT); + ListIterator<RowData> iterator = sortedKeyList.listIterator(); + + RowData inputKey = GenericRowData.of(-1L, sortKeyFieldGetter.getFieldOrNull(input)); + + while (iterator.hasNext()) { + RowData curKey = iterator.next(); // (ID, sortKey) + if (newSortKeyComparator.compare(curKey, inputKey) > 0) { + iterator.previous(); + RowData prev = sortedKeyList.get(iterator.previousIndex()); + iterator.add(GenericRowData.of(id, sortKeyFieldGetter.getFieldOrNull(input))); + valueMapState.put(id, input); + Long prevKey = prev.getLong(keyIdx); + RowData prevAcc = accMapState.get(prevKey); + currFunction.setAccumulators(prevAcc); + currFunction.accumulate(input); + accMapState.put(id, currFunction.getValue()); + isInserted = true; + id++; + break; + } + } + + // Add to the end of the list + if (!isInserted) { + if (iterator.hasPrevious()) { + RowData prev = iterator.previous(); + Long prevKey = prev.getLong(0); + RowData prevAcc = accMapState.get(prevKey); + currFunction.setAccumulators(prevAcc); + iterator.next(); + } + // Only accumulate rowData with currFunction + currFunction.accumulate(input); + iterator.add(GenericRowData.of(id, sortKeyFieldGetter.getFieldOrNull(input))); + valueMapState.put(id, input); + accMapState.put(id, currFunction.getAccumulators()); + id++; + } + + // Update sorted key state with the newly inserted row's key + sortedKeyState.update(sortedKeyList); + idState.update(id); + + // prepare output row + output.setRowKind(origRowKind); + output.replace(input, currFunction.getValue()); + out.collect(output); + + // Emit updated agg value for all records after newly inserted row + while (iterator.hasNext()) { + RowData curKey = iterator.next(); + RowData curValue = valueMapState.get(curKey.getLong(keyIdx)); + currFunction.accumulate(curValue); + RowData prevAcc = accMapState.get(curKey.getLong(keyIdx)); + if (prevAcc.equals(currFunction.getValue())) { + // If previously accumulated value is same as the newly accumulated value + // then no need to send updates downstream since nothing has changed + break; + } + // Generate UPDATE_BEFORE + output.setRowKind(RowKind.UPDATE_BEFORE); + output.replace(curValue, prevAcc); + out.collect(output); + // Generate UPDATE_AFTER + output.setRowKind(RowKind.UPDATE_AFTER); + output.replace(curValue, currFunction.getValue()); + out.collect(output); + + accMapState.put(curKey.getLong(keyIdx), currFunction.getValue()); + } + } + + private void removeFromSortedList(RowData rowData, Collector<RowData> out) throws Exception { + boolean isRetracted = false; + rowData.setRowKind(RowKind.INSERT); + List<RowData> rowList = inputState.value(); + ListIterator<RowData> iterator = rowList.listIterator(); + + while (iterator.hasNext()) { + RowData curRow = iterator.next(); + currFunction.accumulate(curRow); + prevFunction.accumulate(curRow); + if (isRetracted) { + // Emit updated agg value for all records after retraction + output.setRowKind(RowKind.UPDATE_BEFORE); + output.replace(curRow, prevFunction.getValue()); + out.collect(output); + + output.setRowKind(RowKind.UPDATE_AFTER); + output.replace(curRow, currFunction.getValue()); + out.collect(output); + } else if (equaliser.equals(curRow, rowData)) { + // Retract record + output.setRowKind(RowKind.UPDATE_BEFORE); + output.replace(rowData, currFunction.getValue()); + out.collect(output); + iterator.remove(); + currFunction.retract(curRow); + isRetracted = true; + } + } + + // Update state without the retracted row + inputState.update(rowList); + } + + private void removeFromSortedListWithValueMap(RowData input, Collector<RowData> out) + throws Exception { + boolean isRetracted = false; + input.setRowKind(RowKind.INSERT); + List<RowData> sortedKeyList = sortedKeyState.value(); + ListIterator<RowData> iterator = sortedKeyList.listIterator(); + + while (iterator.hasNext()) { + RowData curKey = iterator.next(); + RowData curValue = valueMapState.get(curKey.getLong(keyIdx)); + currFunction.accumulate(curValue); + prevFunction.accumulate(curValue); + if (isRetracted) { + // Emit updated agg value for all records after retraction + output.setRowKind(RowKind.UPDATE_BEFORE); + output.replace(curValue, prevFunction.getValue()); + out.collect(output); + + output.setRowKind(RowKind.UPDATE_AFTER); + output.replace(curValue, currFunction.getValue()); + out.collect(output); + } else if (equaliser.equals(curValue, input)) { + // Retract record + output.setRowKind(RowKind.UPDATE_BEFORE); + output.replace(input, currFunction.getValue()); + out.collect(output); + iterator.remove(); + valueMapState.remove(curKey.getLong(keyIdx)); + currFunction.retract(curValue); + isRetracted = true; + } + } + + // Update sorted key state without the retracted row + sortedKeyState.update(sortedKeyList); + } + + private void removeFromSortedListWithValueMapAndAccumulatorMap( + RowData input, Collector<RowData> out) throws Exception { + boolean isRetracted = false; + input.setRowKind(RowKind.INSERT); + List<RowData> sortedKeyList = sortedKeyState.value(); + ListIterator<RowData> iterator = sortedKeyList.listIterator(); + + while (iterator.hasNext()) { + RowData curKey = iterator.next(); + RowData curValue = valueMapState.get(curKey.getLong(keyIdx)); + if (isRetracted) { + RowData prevAcc = accMapState.get(curKey.getLong(keyIdx)); + if (prevAcc.equals(currFunction.getValue())) { + // If previously accumulated value is same as the newly accumulated value + // then no need to send updates downstream since nothing has changed + break; + } + // Emit updated agg value for all records after retraction + output.setRowKind(RowKind.UPDATE_BEFORE); + output.replace(curValue, prevAcc); + out.collect(output); + + currFunction.accumulate(curValue); + accMapState.put(curKey.getLong(keyIdx), currFunction.getValue()); + output.setRowKind(RowKind.UPDATE_AFTER); + output.replace(curValue, currFunction.getValue()); + out.collect(output); + } else if (equaliser.equals(curValue, input)) { + // Retract record + RowData curAcc = accMapState.get(curKey.getLong(keyIdx)); + currFunction.setAccumulators(curAcc); + output.setRowKind(RowKind.UPDATE_BEFORE); + output.replace(input, currFunction.getValue()); + out.collect(output); + iterator.remove(); + + valueMapState.remove(curKey.getLong(keyIdx)); + accMapState.remove(curKey.getLong(keyIdx)); + if (iterator.hasPrevious()) { + RowData prev = sortedKeyList.get(iterator.previousIndex()); + RowData prevAcc = accMapState.get(prev.getLong(keyIdx)); + currFunction.setAccumulators(prevAcc); Review Comment: Stale hence resolving -- 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: issues-unsubscr...@flink.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
