From 98cb6dddeea43ce6ddbe26153087f8167f3f245b Mon Sep 17 00:00:00 2001
From: drrtuy <drrtuy@gmail.com>
Date: Mon, 16 Jun 2025 13:59:29 +0000
Subject: [PATCH] feat(optimizer): replace simple walk with iterative DFS with
 convergence

---
 dbcon/execplan/calpontselectexecutionplan.h |  3 +++
 dbcon/mysql/rulebased_optimizer.cpp         | 21 ++++++++++++++++++---
 dbcon/mysql/rulebased_optimizer.h           |  1 +
 3 files changed, 22 insertions(+), 3 deletions(-)

diff --git a/dbcon/execplan/calpontselectexecutionplan.h b/dbcon/execplan/calpontselectexecutionplan.h
index 0914acb86..e9425ed45 100644
--- a/dbcon/execplan/calpontselectexecutionplan.h
+++ b/dbcon/execplan/calpontselectexecutionplan.h
@@ -159,6 +159,9 @@ class CalpontSelectExecutionPlan : public CalpontExecutionPlan
    */
   ~CalpontSelectExecutionPlan() override;
 
+  /**
+   * Clones this CSEP without recursive selects for optimizer purposes
+   */
   execplan::SCSEP cloneWORecursiveSelects();
   /**
    * Access and mutator methods
diff --git a/dbcon/mysql/rulebased_optimizer.cpp b/dbcon/mysql/rulebased_optimizer.cpp
index d3de06055..4e9c22b96 100644
--- a/dbcon/mysql/rulebased_optimizer.cpp
+++ b/dbcon/mysql/rulebased_optimizer.cpp
@@ -21,6 +21,7 @@
 
 namespace optimizer {
 
+// Apply a list of rules to a CSEP
 bool optimizeCSEPWithRules(execplan::CalpontSelectExecutionPlan& root, const std::vector<Rule>& rules) {
 
   bool changed = false;
@@ -31,6 +32,7 @@ bool optimizeCSEPWithRules(execplan::CalpontSelectExecutionPlan& root, const std
   return changed;
 }
 
+// high level API call for optimizer
 bool optimizeCSEP(execplan::CalpontSelectExecutionPlan& root)
 {
   optimizer::Rule parallelCES{"parallelCES", optimizer::matchParallelCES, optimizer::applyParallelCES};
@@ -40,8 +42,22 @@ bool optimizeCSEP(execplan::CalpontSelectExecutionPlan& root)
   return optimizeCSEPWithRules(root, rules);
 }
 
-// DFS walk
-bool Rule::apply(execplan::CalpontSelectExecutionPlan& csep) const
+// Apply iteratively until CSEP is converged by rule
+bool Rule::apply(execplan::CalpontSelectExecutionPlan& root) const
+{
+  bool changedThisRound = false;
+  bool hasBeenApplied = false;
+  do 
+  {
+    changedThisRound = walk(root);
+    hasBeenApplied = changedThisRound;
+  } while (changedThisRound);
+
+  return hasBeenApplied;
+}
+
+// DFS walk to match CSEP and apply rules if match
+bool Rule::walk(execplan::CalpontSelectExecutionPlan& csep) const
 {
   bool rewrite = false;
 
@@ -78,7 +94,6 @@ bool Rule::apply(execplan::CalpontSelectExecutionPlan& csep) const
   return rewrite;
 }
 
-
 bool tableIsInUnion(const execplan::CalpontSystemCatalog::TableAliasName& table, execplan::CalpontSelectExecutionPlan& csep)
 {
   return std::any_of(csep.unionVec().begin(), csep.unionVec().end(), 
diff --git a/dbcon/mysql/rulebased_optimizer.h b/dbcon/mysql/rulebased_optimizer.h
index 65fa3f359..9296f2002 100644
--- a/dbcon/mysql/rulebased_optimizer.h
+++ b/dbcon/mysql/rulebased_optimizer.h
@@ -32,6 +32,7 @@ struct Rule
   bool (*matchRule)(execplan::CalpontSelectExecutionPlan&);
   void (*applyRule)(execplan::CalpontSelectExecutionPlan&);
   bool apply(execplan::CalpontSelectExecutionPlan& csep) const;
+  bool walk(execplan::CalpontSelectExecutionPlan& csep) const;
 };
 
 bool matchParallelCES(execplan::CalpontSelectExecutionPlan& csep);