From 1017005acab374f7f16cadfa34728425ca1730ed Mon Sep 17 00:00:00 2001
From: namtran1812 <158846154+namtran1812@users.noreply.github.com>
Date: Sun, 20 Sep 2026 14:24:55 -0400
Subject: [PATCH] [C++][Parquet] Avoid rebuilding column paths during schema
 initialization

---
 cpp/src/parquet/metadata_benchmark.cc |  5 +++++
 cpp/src/parquet/schema.cc             | 19 ++++++++++++++-----
 cpp/src/parquet/schema.h              |  2 +-
 cpp/src/parquet/schema_test.cc        | 16 ++++++++++++++++
 4 files changed, 36 insertions(+), 6 deletions(-)

diff --git a/cpp/src/parquet/metadata_benchmark.cc b/cpp/src/parquet/metadata_benchmark.cc
index 97a99be798cb..2864e4347ba8 100644
--- a/cpp/src/parquet/metadata_benchmark.cc
+++ b/cpp/src/parquet/metadata_benchmark.cc
@@ -125,6 +125,11 @@ void WriteMetadataSetArgs(benchmark::internal::Benchmark* bench) {
   for (int num_row_groups : {1, 100}) {
     bench->Args({/*num_columns=*/1000, num_row_groups});
   }
+
+  // Stress very wide schemas separately with a single row group.
+  for (int num_columns : {5000, 10000, 20000, 50000}) {
+    bench->Args({num_columns, /*num_row_groups=*/1});
+  }
 }
 
 void ReadMetadataSetArgs(benchmark::internal::Benchmark* bench) {
diff --git a/cpp/src/parquet/schema.cc b/cpp/src/parquet/schema.cc
index 3cb91f9a84eb..1af01c084e87 100644
--- a/cpp/src/parquet/schema.cc
+++ b/cpp/src/parquet/schema.cc
@@ -823,8 +823,9 @@ void SchemaDescriptor::Init(NodePtr schema) {
   group_node_ = static_cast<const GroupNode*>(schema_.get());
   leaves_.clear();
 
+  std::string path;
   for (int i = 0; i < group_node_->field_count(); ++i) {
-    BuildTree(group_node_->field(i), 0, 0, group_node_->field(i));
+    BuildTree(group_node_->field(i), 0, 0, group_node_->field(i), path);
   }
 }
 
@@ -853,7 +854,14 @@ bool SchemaDescriptor::Equals(const SchemaDescriptor& other,
 }
 
 void SchemaDescriptor::BuildTree(const NodePtr& node, int16_t max_def_level,
-                                 int16_t max_rep_level, const NodePtr& base) {
+                                 int16_t max_rep_level, const NodePtr& base,
+                                 std::string& path) {
+  const size_t path_size = path.size();
+  if (!path.empty()) {
+    path.push_back('.');
+  }
+  path.append(node->name());
+
   if (node->is_optional()) {
     ++max_def_level;
   } else if (node->is_repeated()) {
@@ -867,7 +875,7 @@ void SchemaDescriptor::BuildTree(const NodePtr& node, int16_t max_def_level,
   if (node->is_group()) {
     const GroupNode* group = static_cast<const GroupNode*>(node.get());
     for (int i = 0; i < group->field_count(); ++i) {
-      BuildTree(group->field(i), max_def_level, max_rep_level, base);
+      BuildTree(group->field(i), max_def_level, max_rep_level, base, path);
     }
   } else {
     node_to_leaf_index_[static_cast<const PrimitiveNode*>(node.get())] =
@@ -876,9 +884,10 @@ void SchemaDescriptor::BuildTree(const NodePtr& node, int16_t max_def_level,
     // Primitive node, append to leaves
     leaves_.push_back(ColumnDescriptor(node, max_def_level, max_rep_level, this));
     leaf_to_base_.emplace(static_cast<int>(leaves_.size()) - 1, base);
-    leaf_to_idx_.emplace(node->path()->ToDotString(),
-                         static_cast<int>(leaves_.size()) - 1);
+    leaf_to_idx_.emplace(path, static_cast<int>(leaves_.size()) - 1);
   }
+
+  path.resize(path_size);
 }
 
 int SchemaDescriptor::GetColumnIndex(const PrimitiveNode& node) const {
diff --git a/cpp/src/parquet/schema.h b/cpp/src/parquet/schema.h
index 65732603ea1d..26120d130407 100644
--- a/cpp/src/parquet/schema.h
+++ b/cpp/src/parquet/schema.h
@@ -485,7 +485,7 @@ class PARQUET_EXPORT SchemaDescriptor {
   const schema::GroupNode* group_node_;
 
   void BuildTree(const schema::NodePtr& node, int16_t max_def_level,
-                 int16_t max_rep_level, const schema::NodePtr& base);
+                 int16_t max_rep_level, const schema::NodePtr& base, std::string& path);
 
   // Result of leaf node / tree analysis
   std::vector<ColumnDescriptor> leaves_;
diff --git a/cpp/src/parquet/schema_test.cc b/cpp/src/parquet/schema_test.cc
index 6c8e6366adf8..268bda5f5ce0 100644
--- a/cpp/src/parquet/schema_test.cc
+++ b/cpp/src/parquet/schema_test.cc
@@ -870,6 +870,22 @@ TEST_F(TestSchemaDescriptor, BuildTree) {
   ASSERT_EQ(nleaves, descr_.num_columns());
 }
 
+TEST_F(TestSchemaDescriptor, ColumnIndexDuplicatePath) {
+  NodePtr first = Int32("duplicate", Repetition::REQUIRED);
+  NodePtr second = Int64("duplicate", Repetition::OPTIONAL);
+  NodePtr schema = GroupNode::Make("schema", Repetition::REQUIRED, {first, second});
+
+  descr_.Init(schema);
+
+  ASSERT_EQ(2, descr_.num_columns());
+  ASSERT_EQ("duplicate", descr_.Column(0)->path()->ToDotString());
+  ASSERT_EQ("duplicate", descr_.Column(1)->path()->ToDotString());
+
+  // Duplicate paths are disambiguated by node identity.
+  ASSERT_EQ(0, descr_.ColumnIndex(*first));
+  ASSERT_EQ(1, descr_.ColumnIndex(*second));
+}
+
 TEST_F(TestSchemaDescriptor, HasRepeatedFields) {
   NodeVector fields;
   NodePtr schema;