andygrove commented on a change in pull request #8961: URL: https://github.com/apache/arrow/pull/8961#discussion_r545948711
########## File path: rust/datafusion/src/optimizer/hash_build_probe_order.rs ########## @@ -0,0 +1,219 @@ +// 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 + +//! Optimizer rule to switch build and probe order of hash join +//! based on statistics of a `TableProvider`. If the number of +//! rows of both sources is known, the order can be switched +//! for a faster hash join. + +use crate::logical_plan::LogicalPlan; +use crate::optimizer::optimizer::OptimizerRule; +use crate::{error::Result, prelude::JoinType}; + +use super::utils; + +/// BuildProbeOrder reorders the build and probe phase of +/// hash joins. This uses the amount of rows that a datasource has. +/// The rule optimizes the order such that the left (build) side of the join +/// is the smallest. +/// If the information is not available, the order stays the same, +/// so that it could be optimized manually in a query. +pub struct HashBuildProbeOrder {} + +// Gets exact number of rows, if known by the statistics of the underlying +fn get_num_rows(logical_plan: &LogicalPlan) -> Option<usize> { Review comment: I have been thinking about what we can do to estimate the number of rows coming out of joins so that we can extend this optimization to nested joins. We can't do anything accurate with the current statistics in this case but I feel that we should try and do something rather than just pick the left side as the build side. One idea is to assume that all joins produce a cartesian product (left row count * right row count). This would at least help in the case where two small tables are joined, and then joined with a huge table, or the other way around. ---------------------------------------------------------------- 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: [email protected]
