fix: add identifier-field-ids JSON serialization/deserialization

src/iceberg/json_internal.cc

@@ -309,7 +309,9 @@ nlohmann::json ToJson(const Type& type) {
 nlohmann::json ToJson(const Schema& schema) {
   nlohmann::json json = ToJson(static_cast<const Type&>(schema));
   SetOptionalField(json, kSchemaId, schema.schema_id());
-  // TODO(gangwu): add identifier-field-ids.
+  if (!schema.IdentifierFieldIds().empty()) {
+    json[kIdentifierFieldIds] = schema.IdentifierFieldIds();
+  }
   return json;
 }
 
@@ -473,8 +475,19 @@ Result<std::unique_ptr<Schema>> SchemaFromJson(const nlohmann::json& json) {
     return JsonParseError("Schema must be a struct type, but got {}", SafeDumpJson(json));
   }
 
+  // Parse identifier-field-ids if present
+  ICEBERG_ASSIGN_OR_RAISE(
+      auto identifier_field_ids,
+      GetJsonValueOrDefault<std::vector<int32_t>>(json, kIdentifierFieldIds));
+
   auto& struct_type = static_cast<StructType&>(*type);
-  return FromStructType(std::move(struct_type), schema_id);
+  std::vector<SchemaField> fields;
+  fields.reserve(struct_type.fields().size());
+  for (auto& field : struct_type.fields()) {
+    fields.emplace_back(std::move(field));
+  }
+  return std::make_unique<Schema>(std::move(fields), schema_id,
+                                  std::move(identifier_field_ids));
 }
 
 nlohmann::json ToJson(const PartitionField& partition_field) {

src/iceberg/test/metadata_serde_test.cc

@@ -144,7 +144,8 @@ TEST(MetadataSerdeTest, DeserializeV2Valid) {
       std::vector<SchemaField>{SchemaField::MakeRequired(1, "x", int64()),
                                SchemaField::MakeRequired(2, "y", int64()),
                                SchemaField::MakeRequired(3, "z", int64())},
-      /*schema_id=*/1);
+      /*schema_id=*/1,
+      /*identifier_field_ids=*/std::vector<int32_t>{1, 2});
 
   auto expected_spec_result = PartitionSpec::Make(
       /*spec_id=*/0,

src/iceberg/test/schema_json_test.cc

@@ -133,4 +133,51 @@ TEST(SchemaJsonTest, RoundTrip) {
   ASSERT_EQ(dumped_json, json);
 }
 
+TEST(SchemaJsonTest, IdentifierFieldIds) {
+  // Test schema with identifier-field-ids
+  constexpr std::string_view json_with_identifiers =
+      R"({"fields":[{"id":1,"name":"id","required":true,"type":"long"},{"id":2,"name":"data","required":false,"type":"string"}],"identifier-field-ids":[1],"schema-id":1,"type":"struct"})";
+
+  // Deserialize
+  auto from_json_result = SchemaFromJson(nlohmann::json::parse(json_with_identifiers));
+  ASSERT_TRUE(from_json_result.has_value());
+  auto schema = std::move(from_json_result.value());
+  ASSERT_EQ(schema->fields().size(), 2);
+  ASSERT_EQ(schema->schema_id(), 1);
+  ASSERT_EQ(schema->IdentifierFieldIds().size(), 1);
+  ASSERT_EQ(schema->IdentifierFieldIds()[0], 1);
+
+  // Serialize back
+  auto serialized_json = ToJson(*schema).dump();
+  ASSERT_EQ(serialized_json, json_with_identifiers);
+
+  // Test schema without identifier-field-ids
+  constexpr std::string_view json_without_identifiers =
+      R"({"fields":[{"id":1,"name":"id","required":true,"type":"int"},{"id":2,"name":"name","required":false,"type":"string"}],"schema-id":1,"type":"struct"})";
+
+  auto from_json_result2 =
+      SchemaFromJson(nlohmann::json::parse(json_without_identifiers));
+  ASSERT_TRUE(from_json_result2.has_value());
+  auto schema2 = std::move(from_json_result2.value());
+  ASSERT_TRUE(schema2->IdentifierFieldIds().empty());
+
+  // Serialize - should not include identifier-field-ids field
+  auto serialized_json2 = ToJson(*schema2).dump();
+  ASSERT_EQ(serialized_json2, json_without_identifiers);
+
+  // Test schema with multiple identifier fields
+  constexpr std::string_view json_multi_identifiers =
+      R"({"fields":[{"id":1,"name":"user_id","required":true,"type":"long"},{"id":2,"name":"org_id","required":true,"type":"long"},{"id":3,"name":"data","required":false,"type":"string"}],"identifier-field-ids":[1,2],"schema-id":2,"type":"struct"})";
+
+  auto from_json_result3 = SchemaFromJson(nlohmann::json::parse(json_multi_identifiers));
+  ASSERT_TRUE(from_json_result3.has_value());
+  auto schema3 = std::move(from_json_result3.value());
+  ASSERT_EQ(schema3->IdentifierFieldIds().size(), 2);
+  ASSERT_EQ(schema3->IdentifierFieldIds()[0], 1);
+  ASSERT_EQ(schema3->IdentifierFieldIds()[1], 2);
+
+  auto serialized_json3 = ToJson(*schema3).dump();
+  ASSERT_EQ(serialized_json3, json_multi_identifiers);
+}
+
 }  // namespace iceberg