From c1de3a64160eab9d8eacc039ba4c256fc587ec50 Mon Sep 17 00:00:00 2001
From: Mustafa Akur <mustafa.akur@synnada.ai>
Date: Tue, 20 Feb 2024 11:04:22 +0300
Subject: [PATCH] Add comment to explain ratioanale of using IndexMap, and
 IndexSet

---
 datafusion/physical-expr/src/equivalence/properties.rs | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/datafusion/physical-expr/src/equivalence/properties.rs b/datafusion/physical-expr/src/equivalence/properties.rs
index de50ab5bd002..5a9a4f64876d 100644
--- a/datafusion/physical-expr/src/equivalence/properties.rs
+++ b/datafusion/physical-expr/src/equivalence/properties.rs
@@ -1118,6 +1118,11 @@ impl DependencyNode {
     }
 }
 
+// Using `IndexMap` and `IndexSet` makes sure to generate consistent results across different executions for the same query.
+// We could have used `HashSet`, `HashMap` in place of them without any loss of functionality.
+// As an example, if existing orderings are `[a ASC, b ASC]`, `[c ASC]` for output ordering
+// both `[a ASC, b ASC, c ASC]` and `[c ASC, a ASC, b ASC]` are valid (e.g. concatenated version of the alternative orderings).
+// When using `HashSet`, `HashMap` it is not guaranteed to generate consistent result, among the possible 2 results in the example above.
 type DependencyMap = IndexMap<PhysicalSortExpr, DependencyNode>;
 type Dependencies = IndexSet<PhysicalSortExpr>;