Backport ClickHouse#88879 to 25.8: Fix handling of non materialized Nested(Tuple(...)) (added via ALTER)

Author: robot-clickhouse

src/DataTypes/ObjectUtils.cpp

@@ -70,12 +70,19 @@ DataTypePtr getBaseTypeOfArray(const DataTypePtr & type)
 DataTypePtr getBaseTypeOfArray(DataTypePtr type, const Names & tuple_elements)
 {
     auto it = tuple_elements.begin();
+
+    /// Get underlying type for array, but w/o processing tuple elements that are not part of the nested, so it is done in 3 steps:
+    /// 1. Find Nested type (since it can be part of Tuple/Array)
+    /// 2. Process all Nested types (this is Array(Tuple()), it is responsibility of the caller to re-create proper Array nesting)
+    /// 3. Strip all nested arrays (it is responsibility of the caller to re-create proper Array nesting)
+
+    /// 1. Find Nested type (since it can be part of Tuple/Array)
     while (true)
     {
-        if (const auto * type_array = typeid_cast<const DataTypeArray *>(type.get()))
-        {
+        if (type->hasCustomName())
+            break;
+        else if (const auto * type_array = typeid_cast<const DataTypeArray *>(type.get()))
             type = type_array->getNestedType();
-        }
         else if (const auto * type_tuple = typeid_cast<const DataTypeTuple *>(type.get()))
         {
             if (it == tuple_elements.end())
@@ -84,16 +91,38 @@ DataTypePtr getBaseTypeOfArray(DataTypePtr type, const Names & tuple_elements)
             auto pos = type_tuple->tryGetPositionByName(*it);
             if (!pos)
                 break;
-
             ++it;
+
             type = type_tuple->getElement(*pos);
         }
         else
-        {
             break;
+    }
+
+    /// 2. Process all Nested types (this is Array(Tuple()), it is responsibility of the caller to re-create proper Array nesting)
+    while (type->hasCustomName())
+    {
+        if (const auto * type_nested = typeid_cast<const DataTypeNestedCustomName *>(type->getCustomName()))
+        {
+            if (it == tuple_elements.end())
+                break;
+
+            const auto & names = type_nested->getNames();
+            auto pos = std::find(names.begin(), names.end(), *it);
+            if (pos == names.end())
+                break;
+            ++it;
+
+            type = type_nested->getElements().at(std::distance(names.begin(), pos));
         }
+        else
+            break;
     }
 
+    /// 3. Strip all nested arrays (it is responsibility of the caller to re-create proper Array nesting)
+    while (const auto * type_array = typeid_cast<const DataTypeArray *>(type.get()))
+        type = type_array->getNestedType();
+
     return type;
 }
 

tests/queries/0_stateless/03672_nested_array_nested_tuple.reference

@@ -0,0 +1,2 @@
+[1,2,3]	[(0),(0),(0)]
+[1,2,3]	[((0)),((0)),((0))]

tests/queries/0_stateless/03672_nested_array_nested_tuple.sql

@@ -0,0 +1,13 @@
+DROP TABLE IF EXISTS nest;
+CREATE TABLE nest (nested_field Nested(e1 Int32)) ENGINE = MergeTree() ORDER BY nested_field.e1;
+INSERT INTO nest (nested_field.e1) VALUES ([1, 2, 3]);
+ALTER TABLE nest ADD COLUMN nested_field.e2 Array(Tuple(some_value Int32));
+OPTIMIZE TABLE nest FINAL;
+SELECT * FROM nest;
+
+DROP TABLE IF EXISTS nest_2;
+CREATE TABLE nest_2 (nested_field Nested(e1 Int32)) ENGINE = MergeTree() ORDER BY nested_field.e1;
+INSERT INTO nest_2 (nested_field.e1) VALUES ([1, 2, 3]);
+ALTER TABLE nest_2 ADD COLUMN nested_field.e2 Array(Tuple(some_value Tuple(another_value Int32)));
+OPTIMIZE TABLE nest_2 FINAL;
+SELECT * FROM nest_2;