wjones127 commented on code in PR #14200: URL: https://github.com/apache/arrow/pull/14200#discussion_r979084272
########## docs/source/cpp/gandiva.rst: ########## @@ -0,0 +1,151 @@ +.. 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. + +.. default-domain:: cpp +.. highlight:: cpp +.. cpp:namespace:: arrow::compute + +=============================== +The Gandiva Expression Compiler +=============================== + +Gandiva is a runtime expression compiler that uses `LLVM`_ to generate +efficient native code for projections and filters on Arrow record batches. +Gandiva only handles projections and filters. For other transformations, see +:ref:`Compute Functions <compute-cpp>`. + +Gandiva was designed to take advantage of the Arrow memory format and modern +hardware. Compiling expressions using LLVM allows the execution to be optimized +to the local runtime environment and hardware, including available SIMD +instructions. To reduce optimization overhead, many Gandiva functions are +pre-compiled into LLVM IR (intermediate representation). + +.. _LLVM: https://llvm.org/ + + +Building Expressions +==================== + +Gandiva provides a general expression representation where expressions are +represented by a tree of nodes. The expression trees are built using +:class:`gandiva::TreeExprBuilder`. The leaves of the expression tree are typically +field references, created by :func:`gandiva::TreeExprBuilder::MakeField`, and +literal values, created by :func:`gandiva::TreeExprBuilder::MakeLiteral`. Nodes +can be combined into more complex expression trees using: + +* :func:`gandiva::TreeExprBuilder::MakeFunction` to create a function + node. +* :func:`gandiva::TreeExprBuilder::MakeIf` to create if-else logic. +* :func:`gandiva::TreeExprBuilder::MakeAnd` and :func:`gandiva::TreeExprBuilder::MakeOr` + to create boolean expressions. (For "not", use the ``not(bool)`` function in ``MakeFunction``.) +* :func:`gandiva::TreeExprBuilder::MakeInExpressionInt32` and the other "in expression" + functions to create set membership tests. + +Once an expression tree is built, they are wrapped in either :class:`gandiva::Expression` +or :class:`gandiva::Condition`, depending on how they will be used. +``Expression`` is used in projections while ``Condition`` is used filters. + +As an example, here is how to create an Expression representing ``x + 3`` and a +Condition representing ``x < 3``: + +.. code-block:: cpp + + auto field_x_raw = arrow::field("x", arrow::int32()); + auto field_x = TreeExprBuilder::MakeField(field_x_raw); + auto literal_3 = TreeExprBuilder::MakeLiteral(3); + auto field_result = arrow::field("result", arrow::int32()); + + auto add_node = TreeExprBuilder::MakeFunction("add", {field_x, literal_3}, arrow::int32()); + auto expression = TreeExprBuilder::MakeExpression(add_node, field_result); + + auto less_than_node = TreeExprBuilder::MakeFunction("less_than", {field_x, literal_3}, + boolean()); + auto condition = TreeExprBuilder::MakeCondition(less_than_node); + +For simpler expressions, there are also convenience functions that allow you to +use functions directly in ``MakeExpression`` and ``MakeCondition``: + +.. code-block:: cpp + + auto expression = TreeExprBuilder::MakeExpression("add", {field_x, literal_3}, field_result); + + auto condition = TreeExprBuilder::MakeCondition("less_than", {field_x, literal_3}); + + +Projectors and Filters +====================== + +Gandiva's two execution kernels are :class:`gandiva::Projector` and +:class:`gandiva::Filter`. ``Projector`` consumes a record batch and projects +into a new record batch. ``Filter`` consumes a record batch and produces a +:class:`gandiva::SelectionVector` containing the indices that matched the condition. + +For both ``Projector`` and ``Filter``, optimization of the expression IR happens +when creating instances. They are compiled against a static schema, so the +schema of the record batches must be known at this point. + +Continuing with the ``expression`` and ``condition`` created in the previous +section, here is an example of creating a Projector and a Filter: + +.. code-block:: cpp + + auto schema = arrow::schema({field_x}); + std::shared_ptr<Projector> projector; + auto status = Projector::Make(schema, {expression}, &projector); + ARROW_CHECK_OK(status); + + std::shared_ptr<Filter> filter; + status = Filter::Make(schema, condition, &filter); + ARROW_CHECK_OK(status); + + +Once a Projector or Filter is created, it can be evaluated on Arrow record batches. +These execution kernels are single-threaded on their own, but are designed to be +reused to process distinct record batches in parallel. + +Execution is performed with :func:`gandiva::Projector::Evaluate` and +:func:`gandiva::Filter::Evaluate`. Filters produce :class:`gandiva::SelectionVector`, +a vector of row indices that matched the filter condition. When filtering and +projecting record batches, you can pass the selection vector into the projector +so that the projection is only evaluated on matching rows. Review Comment: > So you can pass the SelectionVector into Projector::Evaluate()? Yes, there is an overload for that. It's documented in the API reference. I'm not sure I understand your suggestion. Are you saying something like this? ``` For filtering and projecting record batches simultaneously, you can pass the selection vector into the projector (using one of the overloads) so that the projection is only evaluated on matching rows. ``` -- 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]
