HappenLee commented on code in PR #65837: URL: https://github.com/apache/doris/pull/65837#discussion_r3777937168
########## fe/fe-core/src/main/java/org/apache/doris/nereids/glue/translator/RuntimeFilterBucketPruneClassifier.java: ########## @@ -0,0 +1,121 @@ +// 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.doris.nereids.glue.translator; + +import org.apache.doris.analysis.Expr; +import org.apache.doris.analysis.SlotRef; +import org.apache.doris.catalog.Column; +import org.apache.doris.catalog.DistributionInfo; +import org.apache.doris.catalog.HashDistributionInfo; +import org.apache.doris.catalog.OlapTable; +import org.apache.doris.catalog.Partition; +import org.apache.doris.planner.OlapScanNode; +import org.apache.doris.planner.PlanNode; +import org.apache.doris.thrift.TRuntimeFilterType; + +/** Classifies direct single-column HASH targets for BE-side runtime-filter bucket pruning. */ +final class RuntimeFilterBucketPruneClassifier { + private RuntimeFilterBucketPruneClassifier() { + } + + static Classification classify(TRuntimeFilterType filterType, Expr targetExpr, PlanNode scanNode) { + if (filterType != TRuntimeFilterType.IN && filterType != TRuntimeFilterType.IN_OR_BLOOM) { + return Classification.unsupported("runtime filter is not IN or IN_OR_BLOOM"); + } + if (!(scanNode instanceof OlapScanNode)) { + return Classification.unsupported("target scan is not an OlapScanNode"); + } + if (!(targetExpr instanceof SlotRef)) { + return Classification.unsupported("target expression is not a direct SlotRef"); + } + + Column targetColumn = ((SlotRef) targetExpr).getColumn(); + if (targetColumn == null) { + return Classification.unsupported("target SlotRef has no column"); + } + + OlapScanNode olapScanNode = (OlapScanNode) scanNode; + if (olapScanNode.isPointQuery()) { + return Classification.unsupported("target scan is a point query"); + } + OlapTable table = olapScanNode.getOlapTable(); + if (table == null || olapScanNode.getSelectedPartitionIds().isEmpty()) { + return Classification.unsupported("target scan has no selected partitions"); + } + + Column distributionColumn = null; + for (Long partitionId : olapScanNode.getSelectedPartitionIds()) { + Partition partition = table.getPartition(partitionId); + if (partition == null) { + return Classification.unsupported("selected partition does not exist"); + } + DistributionInfo distributionInfo = partition.getDistributionInfo(); + if (!(distributionInfo instanceof HashDistributionInfo)) { + return Classification.unsupported("distribution type is not HASH"); + } + HashDistributionInfo hashDistributionInfo = (HashDistributionInfo) distributionInfo; + if (hashDistributionInfo.getDistributionColumns().size() != 1) { + return Classification.unsupported("HASH distribution is not single-column"); + } + Column currentDistributionColumn = hashDistributionInfo.getDistributionColumns().get(0); + if (!sameColumn(targetColumn, currentDistributionColumn)) { + return Classification.unsupported("target SlotRef is not the HASH distribution column"); + } + if (distributionColumn != null && !sameColumn(distributionColumn, currentDistributionColumn)) { + return Classification.unsupported("selected partitions use different distribution columns"); + } + distributionColumn = currentDistributionColumn; + } + return Classification.supported(); + } + + private static boolean sameColumn(Column targetColumn, Column distributionColumn) { + if (targetColumn == distributionColumn) { + return true; + } + return targetColumn.tryGetBaseColumnName() Review Comment: Fixed in 6b7a97a4f24. The classifier now resolves MV columns only through a direct `SlotRef` to a non-MV base column; complex and nested MV definitions are rejected before name/type comparison. Added a historical alias/type collision case where `mv.dist_col := abs(base.dist_col)` and verified bucket pruning is disabled. ########## be/src/exec/scan/parallel_scanner_builder.h: ########## @@ -54,7 +55,17 @@ class ParallelScannerBuilder { _is_preaggregation(is_preaggregation), _tablets(tablets.cbegin(), tablets.cend()), _key_ranges(key_ranges.cbegin(), key_ranges.cend()), - _read_sources(read_sources) {} + _read_sources(read_sources) { + DORIS_CHECK_EQ(_tablets.size(), scan_ranges.size()); + for (size_t i = 0; i < _tablets.size(); ++i) { Review Comment: Fixed in 6b7a97a4f24. Removed the per-tablet `unordered_map`; `ParallelScannerBuilder` now reuses the already-owned, index-aligned `TPaloScanRange` vector and forwards each range directly to every split scanner. This adds no identity allocation for ineligible or old-FE ranges. Added 20,000-range ineligible coverage and retained the split-scanner bucket identity test. ########## be/src/exec/scan/scanner_scheduler.cpp: ########## @@ -176,14 +176,21 @@ void ScannerScheduler::_scanner_scan(std::shared_ptr<ScannerContext> ctx, Status status = Status::OK(); bool eos = false; + auto append_late_arrival_runtime_filter = [&] { + Status rf_status = scanner->try_append_late_arrival_runtime_filter(); + if (!rf_status.ok()) { + LOG(WARNING) << "Failed to append late arrival runtime filter: " + << rf_status.to_string(); + } + }; ASSIGN_STATUS_IF_CATCH_EXCEPTION( RuntimeState* state = ctx->state(); DCHECK(nullptr != state); // scanner->open may alloc plenty amount of memory(read blocks of data), // so better to also check low memory and clear free blocks here. if (ctx->low_memory_mode()) { ctx->clear_free_blocks(); } - if (scanner->check_partition_pruned()) { eos = true; } + if (scanner->is_pruned_by_runtime_filter()) { eos = true; } Review Comment: Fixed in 6b7a97a4f24. The cleanup hook is now `release_unopened_resources()` and is valid both before and after `prepare()`. The scheduler invokes it on the initial pre-prepare pruning path and on the post-prepare/pre-open path. `OlapScanner` clears the complete `ReaderParams` (including tablet, rowset splits, and delete metadata) plus prepared-only resources. Added bounded-concurrency coverage proving a pending scanner is pruned before its first `prepare()`, and lifecycle coverage proving constructor-owned inputs are released. -- 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]
