Merge pull request #70868 from ClickHouse/cherrypick/24.7/688c9717db9…

…0cf1c8f242c3fe583d6add0a86718 Cherry pick #70103 to 24.7: Avoid reusing columns among different named tuples
ClickHouse · Oct 21, 2024 · f3a6430 · f3a6430
2 parents 0cbe01a + 4cfc569
commit f3a6430
Show file tree

Hide file tree

Showing 3 changed files with 56 additions and 0 deletions.
diff --git a/src/Planner/PlannerActionsVisitor.cpp b/src/Planner/PlannerActionsVisitor.cpp
@@ -36,6 +36,10 @@
 
 namespace DB
 {
+namespace Setting
+{
+    extern const SettingsBool enable_named_columns_in_function_tuple;
+}
 
 namespace ErrorCodes
 {
@@ -178,6 +182,33 @@ class ActionNodeNameHelper
                     break;
                 }
 
+                if (planner_context.getQueryContext()->getSettingsRef()[Setting::enable_named_columns_in_function_tuple])
+                {
+                    /// Function "tuple" which generates named tuple should use argument aliases to construct its name.
+                    if (function_node.getFunctionName() == "tuple")
+                    {
+                        if (const DataTypeTuple * type_tuple = typeid_cast<const DataTypeTuple *>(function_node.getResultType().get()))
+                        {
+                            if (type_tuple->haveExplicitNames())
+                            {
+                                const auto & names = type_tuple->getElementNames();
+                                size_t size = names.size();
+                                WriteBufferFromOwnString s;
+                                s << "tuple(";
+                                for (size_t i = 0; i < size; ++i)
+                                {
+                                    if (i != 0)
+                                        s << ", ";
+                                    s << backQuoteIfNeed(names[i]);
+                                }
+                                s << ")";
+                                result = s.str();
+                                break;
+                            }
+                        }
+                    }
+                }
+
                 String in_function_second_argument_node_name;
 
                 if (isNameOfInFunction(function_node.getFunctionName()))

diff --git a/tests/queries/0_stateless/03240_insert_select_named_tuple.reference b/tests/queries/0_stateless/03240_insert_select_named_tuple.reference
@@ -0,0 +1,3 @@
+1	('dete','ok')	('dete','ok')
+{"id":1,"a":{"col_a":"dete","type":"ok"},"b":{"col_b":"dete","type":"ok"}}
+{"id":1,"a":{"col_a":"dete","type":"ok"},"b":{"col_b":"dete","type":"ok"}}
diff --git a/tests/queries/0_stateless/03240_insert_select_named_tuple.sql b/tests/queries/0_stateless/03240_insert_select_named_tuple.sql
@@ -0,0 +1,22 @@
+SET enable_analyzer = 1;
+SET enable_named_columns_in_function_tuple = 1;
+
+DROP TABLE IF EXISTS src;
+DROP TABLE IF EXISTS dst;
+
+CREATE TABLE src (id UInt32, type String, data String) ENGINE=MergeTree ORDER BY tuple();
+CREATE TABLE dst (id UInt32, a Tuple (col_a Nullable(String), type String), b Tuple (col_b Nullable(String), type String)) ENGINE = MergeTree ORDER BY id;
+
+INSERT INTO src VALUES (1, 'ok', 'data');
+INSERT INTO dst (id, a, b) SELECT id, tuple(replaceAll(data, 'a', 'e') AS col_a, type) AS a, tuple(replaceAll(data, 'a', 'e') AS col_b, type) AS b FROM src;
+SELECT * FROM dst;
+
+DROP TABLE src;
+DROP TABLE dst;
+
+DROP TABLE IF EXISTS src;
+CREATE TABLE src (id UInt32, type String, data String) ENGINE=MergeTree ORDER BY tuple();
+INSERT INTO src VALUES (1, 'ok', 'data');
+SELECT id, tuple(replaceAll(data, 'a', 'e') AS col_a, type) AS a, tuple(replaceAll(data, 'a', 'e') AS col_b, type) AS b FROM cluster(test_cluster_two_shards, currentDatabase(), src) SETTINGS prefer_localhost_replica=0 FORMAT JSONEachRow;
+
+DROP TABLE src;