pitrou commented on a change in pull request #10204: URL: https://github.com/apache/arrow/pull/10204#discussion_r628278986
########## File path: cpp/src/arrow/engine/exec_plan.h ########## @@ -0,0 +1,245 @@ +// 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. + +#pragma once + +#include <memory> +#include <string> +#include <vector> + +#include "arrow/compute/type_fwd.h" +#include "arrow/type_fwd.h" +#include "arrow/util/macros.h" +#include "arrow/util/visibility.h" + +// NOTES: +// - ExecBatches only have arrays or scalars +// - data streams may be ordered, so add input number? +// - node to combine input needs to reorder + +namespace arrow { +namespace engine { + +class ExecNode; + +class ARROW_EXPORT ExecPlan : public std::enable_shared_from_this<ExecPlan> { + public: + using NodeVector = std::vector<ExecNode*>; + + virtual ~ExecPlan() = default; + + /// Make an empty exec plan + static Result<std::shared_ptr<ExecPlan>> Make(); + + void AddNode(std::unique_ptr<ExecNode> node); + + /// The initial inputs + const NodeVector& sources() const; + + /// The final outputs + const NodeVector& sinks() const; + + // XXX API question: + // There are clearly two phases in the ExecPlan lifecycle: + // - one construction phase where AddNode() and ExecNode::Bind() is called + // (with optional validation at the end) + // - one execution phase where the nodes are topo-sorted and then started + // + // => Should we separate out those APIs? e.g. have a ExecPlanBuilder + // for the first phase. + + Status Validate(); + + /// Start producing on all nodes + /// + /// Nodes are started in reverse topological order, such that any node + /// is started before all of its inputs. + Status StartProducing(); + + // XXX should we also have `void StopProducing()`? + + protected: + ExecPlan() = default; +}; + +class ARROW_EXPORT ExecNode { + public: + struct OutputNode { + ExecNode* output; + // Index of corresponding input in `output` node + int input_index; + }; + + using NodeVector = std::vector<ExecNode*>; + using OutputNodeVector = std::vector<OutputNode>; + using BatchDescr = std::vector<ValueDescr>; + + virtual ~ExecNode(); + + virtual const char* kind_name() = 0; + // The number of inputs and outputs expected by this node + // XXX should these simply return `input_descrs_.size()` + // (`output_descrs_.size()` respectively)? + virtual int num_inputs() const = 0; + virtual int num_outputs() const = 0; + + /// This node's predecessors in the exec plan + const NodeVector& inputs() const { return inputs_; } + + /// The datatypes for each input + // XXX Should it be std::vector<DataType>? + const std::vector<BatchDescr>& input_descrs() const { return input_descrs_; } + + /// This node's successors in the exec plan + const OutputNodeVector& outputs() const { return outputs_; } + + /// The datatypes for each output + // XXX Should it be std::vector<DataType>? + const std::vector<BatchDescr>& output_descrs() const { return output_descrs_; } + + /// This node's exec plan + ExecPlan* plan() { return plan_; } + std::shared_ptr<ExecPlan> plan_ref() { return plan_->shared_from_this(); } + + /// \brief An optional label, for display and debugging + /// + /// There is no guarantee that this value is non-empty or unique. + const std::string& label() const { return label_; } + + int AddInput(ExecNode* node) { Review comment: +1 -- 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
