mingmwang commented on code in PR #3855:
URL: https://github.com/apache/arrow-datafusion/pull/3855#discussion_r998927646


##########
datafusion/core/src/physical_optimizer/enforcement.rs:
##########
@@ -0,0 +1,858 @@
+// 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.
+
+//! Enforcement optimizer rules are used to make sure the plan's Distribution 
and Ordering
+//! requirements are met by inserting necessary [[RepartitionExec]] and 
[[SortExec]].
+//!
+use crate::error::Result;
+use crate::physical_optimizer::utils::transform_up;
+use crate::physical_optimizer::PhysicalOptimizerRule;
+use crate::physical_plan::coalesce_partitions::CoalescePartitionsExec;
+use crate::physical_plan::repartition::RepartitionExec;
+use crate::physical_plan::sorts::sort::SortExec;
+use crate::physical_plan::{with_new_children_if_necessary, Distribution, 
ExecutionPlan};
+use crate::prelude::SessionConfig;
+use datafusion_physical_expr::expressions::Column;
+use datafusion_physical_expr::{
+    normalize_sort_expr_with_equivalence_properties, PhysicalSortExpr,
+};
+use std::sync::Arc;
+
+/// BasicEnforcement rule, it ensures the Distribution and Ordering 
requirements are met
+/// in the strictest way. It might add additional [[RepartitionExec]] to the 
plan tree
+/// and give a non-optimal plan, but it can avoid the possible data skew in 
joins
+///
+/// For example for a HashJoin with keys(a, b, c), the required 
Distribution(a, b, c) can be satisfied by
+/// several alternative partitioning ways: [(a, b, c), (a, b), (a, c), (b, c), 
(a), (b), (c), ( )].
+///
+/// This rule only chooses the exactly match and satisfies the Distribution(a, 
b, c) by a HashPartition(a, b, c).
+#[derive(Default)]
+pub struct BasicEnforcement {}

Review Comment:
   In some Cascade style optimizers, they have the enforcement rules which do 
the similar things. So I just name the rule to `BasicEnforcement`, and this is 
for Phase1.
   
   In the following PRs, I will also try to implement a  more dynamic 
enforcement rule which will consider different alternative partitioning options 
and choose a best plan which will have the least number of RepartitionExec as 
the goal. Following are some papers I read recently, I will try to refer to 
them and  implement the `DynamicEnforcement`  rule.
   
   Incorporating Partitioning and Parallel Plans into the SCOPE Optimizer
   http://www.cs.albany.edu/~jhh/courses/readings/zhou10.pdf
   
   Another recent paper, see the **EXCHANGE PLACEMENT** sections.
   https://vldb.org/pvldb/vol15/p936-rajan.pdf
   
   And we can have an optimizer conf option like  
'perfer_join_keys_eactly_match', if this option is set to `true`, DataFusion 
physical optimizer will run `BasicEnforcement`, otherwise run 
`DynamicEnforcement`.
    



-- 
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]

Reply via email to