Zakelly commented on code in PR #25861: URL: https://github.com/apache/flink/pull/25861#discussion_r1912867539
########## flink-table/flink-table-runtime/src/main/java/org/apache/flink/table/runtime/operators/aggregate/asyncwindow/buffers/AsyncStateRecordsWindowBuffer.java: ########## @@ -0,0 +1,226 @@ +/* + * 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.aggregate.asyncwindow.buffers; + +import org.apache.flink.api.common.functions.RuntimeContext; +import org.apache.flink.api.common.state.v2.StateFuture; +import org.apache.flink.runtime.memory.MemoryManager; +import org.apache.flink.table.data.RowData; +import org.apache.flink.table.runtime.generated.GeneratedRecordEqualiser; +import org.apache.flink.table.runtime.generated.RecordEqualiser; +import org.apache.flink.table.runtime.operators.window.async.tvf.combines.AsyncStateRecordsCombiner; +import org.apache.flink.table.runtime.operators.window.async.tvf.state.AsyncStateKeyContext; +import org.apache.flink.table.runtime.operators.window.async.tvf.state.WindowAsyncState; +import org.apache.flink.table.runtime.operators.window.tvf.common.WindowTimerService; +import org.apache.flink.table.runtime.typeutils.AbstractRowDataSerializer; +import org.apache.flink.table.runtime.typeutils.PagedTypeSerializer; +import org.apache.flink.table.runtime.typeutils.WindowKeySerializer; +import org.apache.flink.table.runtime.util.KeyValueIterator; +import org.apache.flink.table.runtime.util.WindowKey; +import org.apache.flink.table.runtime.util.collections.binary.BytesMap; +import org.apache.flink.table.runtime.util.collections.binary.WindowBytesMultiMap; + +import javax.annotation.Nullable; + +import java.io.EOFException; +import java.time.ZoneId; +import java.util.ArrayList; +import java.util.Iterator; +import java.util.List; + +import static org.apache.flink.table.runtime.util.AsyncStateUtils.REUSABLE_VOID_STATE_FUTURE; +import static org.apache.flink.table.runtime.util.TimeWindowUtil.isWindowFired; + +/** + * An implementation of {@link AsyncStateWindowBuffer} that buffers input elements in a {@link + * WindowBytesMultiMap} and combines buffered elements into async state when flushing. + */ +public final class AsyncStateRecordsWindowBuffer implements AsyncStateWindowBuffer { + + private final AsyncStateRecordsCombiner combineFunction; + private final WindowBytesMultiMap recordsBuffer; + private final WindowKey reuseWindowKey; + private final AbstractRowDataSerializer<RowData> recordSerializer; + private final ZoneId shiftTimeZone; + private final RecordEqualiser keyEqualiser; + private final AsyncStateKeyContext keyContext; + + private long minSliceEnd = Long.MAX_VALUE; + + public AsyncStateRecordsWindowBuffer( + Object operatorOwner, + MemoryManager memoryManager, + long memorySize, + AsyncStateRecordsCombiner combineFunction, + PagedTypeSerializer<RowData> keySer, + AbstractRowDataSerializer<RowData> inputSer, + RecordEqualiser keyEqualiser, + AsyncStateKeyContext keyContext, + ZoneId shiftTimeZone) { + this.combineFunction = combineFunction; + this.recordsBuffer = + new WindowBytesMultiMap( + operatorOwner, memoryManager, memorySize, keySer, inputSer.getArity()); + this.recordSerializer = inputSer; + this.keyEqualiser = keyEqualiser; + this.keyContext = keyContext; + this.reuseWindowKey = new WindowKeySerializer(keySer).createInstance(); + this.shiftTimeZone = shiftTimeZone; + } + + @Override + public StateFuture<Void> addElement(RowData dataKey, long sliceEnd, RowData element) + throws Exception { + StateFuture<Void> resultFuture = REUSABLE_VOID_STATE_FUTURE; + + // track the lowest trigger time, if watermark exceeds the trigger time, + // it means there are some elements in the buffer belong to a window going to be fired, + // and we need to flush the buffer into state for firing. + minSliceEnd = Math.min(sliceEnd, minSliceEnd); + + reuseWindowKey.replace(sliceEnd, dataKey); + BytesMap.LookupInfo<WindowKey, Iterator<RowData>> lookup = + recordsBuffer.lookup(reuseWindowKey); + try { + recordsBuffer.append(lookup, recordSerializer.toBinaryRow(element)); + } catch (EOFException e) { + // buffer is full, flush it to state + resultFuture = flush(dataKey); + // remember to add the input element again + addElement(dataKey, sliceEnd, element); + } + return resultFuture; + } + + @Override + public StateFuture<Void> advanceProgress(@Nullable RowData currentKey, long progress) + throws Exception { + if (isWindowFired(minSliceEnd, progress, shiftTimeZone)) { + // there should be some window to be fired, flush buffer to state first + return flush(currentKey); + } + return REUSABLE_VOID_STATE_FUTURE; + } + + @Override + public StateFuture<Void> flush(@Nullable RowData currentKey) throws Exception { + StateFuture<Void> flushFuture = REUSABLE_VOID_STATE_FUTURE; + if (recordsBuffer.getNumKeys() > 0) { + // due to the delayed processing of async requests, all objects cannot be reused, so + // they must be copied. + KeyValueIterator<WindowKey, Iterator<RowData>> entryIterator = + recordsBuffer.getEntryIterator(true); + while (entryIterator.advanceNext()) { + WindowKey windowKey = entryIterator.getKey(); + long window = windowKey.getWindow(); + List<RowData> allData = itertorToList(entryIterator.getValue()); Review Comment: As offline discussed, I'd suggest we use two `recordsBuffer` and switch between them when flushing, to avoid heap memory usage as well as the copying overhead. Maybe we can leave this as a further optimization. -- 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
