pitrou commented on a change in pull request #9742: URL: https://github.com/apache/arrow/pull/9742#discussion_r598576236
########## File path: cpp/src/arrow/engine/exec_plan.cc ########## @@ -0,0 +1,131 @@ +// 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. + +#include "arrow/engine/exec_plan.h" + +#include "arrow/compute/exec.h" +#include "arrow/engine/query_plan.h" +#include "arrow/result.h" +#include "arrow/status.h" +#include "arrow/util/checked_cast.h" +#include "arrow/util/future.h" +#include "arrow/util/logging.h" +#include "arrow/util/thread_pool.h" + +namespace arrow { + +using internal::checked_cast; + +namespace engine { + +ExecNode::InputBatch::InputBatch(std::vector<util::optional<compute::ExecBatch>> batches) + : batches(std::move(batches)), ready_fut(Future<>::Make()) {} + +ExecNode::ExecNode(ExecPlan* plan, const QueryNode* query_node) + : plan_(plan), + query_node_(query_node), + num_inputs_(query_node->num_inputs()), + finished_fut_(Future<>::Make()) {} + +ExecNode::~ExecNode() = default; + +void ExecNode::ReserveBatches(int32_t num_batches) { + // Should be called with mutex locked + if (static_cast<size_t>(num_batches) > input_batches_.size()) { + input_batches_.resize(num_batches); + } +} + +ExecNode::InputBatch* ExecNode::EnsureBatch(int32_t batch_index) { + // Should be called with mutex locked + if (input_batches_[batch_index] == nullptr) { + input_batches_[batch_index].reset( + new InputBatch{std::vector<util::optional<compute::ExecBatch>>(num_inputs_)}); + } + return input_batches_[batch_index].get(); +} + +Status ExecNode::InputReceived(int32_t input_index, int32_t batch_index, + compute::ExecBatch batch) { + auto lock = mutex_.Lock(); + + if (input_index >= num_inputs_) { + return Status::Invalid("Invalid input index"); + } + if (finish_at_ >= 0 && batch_index >= finish_at_) { + return Status::Invalid("Input batch index out of bounds"); + } + + ReserveBatches(batch_index + 1); + auto* input_batch = EnsureBatch(batch_index); + + if (input_batch->batches[input_index].has_value()) { + return Status::Invalid("Batch #", batch_index, " for input #", input_index, + " already received"); + } + input_batch->batches[input_index] = std::move(batch); + if (++input_batch->num_ready == num_inputs_) { + input_batch->ready_fut.MarkFinished(); + } + return Status::OK(); +} + +Status ExecNode::InputFinished(int32_t num_batches) { + auto lock = mutex_.Lock(); + + if (finish_at_ >= 0) { + return Status::Invalid("InputFinished already called"); + } + finish_at_ = num_batches; + ReserveBatches(num_batches); + + std::vector<Future<>> batch_futures; + for (int32_t i = 0; i < num_batches; ++i) { + auto* input_batch = EnsureBatch(i); + batch_futures[i] = input_batch->ready_fut; + } + + // TODO lifetime + AllComplete(std::move(batch_futures)) + .AddCallback([this](const Result<detail::Empty>& res) { + finished_fut_.MarkFinished(res.status()); + }); + return Status::OK(); +} + +Future<> ExecNode::RunAsync(int32_t batch_index, internal::Executor* executor) { + auto lock = mutex_.Lock(); + + ReserveBatches(batch_index + 1); + auto* input_batch = EnsureBatch(batch_index); + + // TODO lifetime (take strong ref to ExecPlan?) + return executor->Transfer(input_batch->ready_fut) Review comment: I'm not sure what "root nodes" are in this context. You mean the sources? Indeed, we may simply want to start executing as soon as all inputs are ready. Though this will depend on the node: some can execute as soon as one batch is received, some need the whole input to be received. -- 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. For queries about this service, please contact Infrastructure at: us...@infra.apache.org
