Merge pull request #70494 from ClickHouse/backport/24.8/70374

Backport #70374 to 24.8: Fix crash during GROUP BY JSON sub-object subcolumn

src/Columns/ColumnObject.cpp

@@ -323,6 +323,19 @@ void ColumnObject::setDynamicPaths(const std::vector<String> & paths)
     }
 }
 
+void ColumnObject::setDynamicPaths(const std::vector<std::pair<String, ColumnPtr>> & paths)
+{
+    if (paths.size() > max_dynamic_paths)
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "Cannot set dynamic paths to Object column, the number of paths ({}) exceeds the limit ({})", paths.size(), max_dynamic_paths);
+
+    for (const auto & [path, column] : paths)
+    {
+        auto it = dynamic_paths.emplace(path, column).first;
+        dynamic_paths_ptrs[path] = assert_cast<ColumnDynamic *>(it->second.get());
+        sorted_dynamic_paths.insert(it->first);
+    }
+}
+
 void ColumnObject::insert(const Field & x)
 {
     const auto & object = x.safeGet<Object>();

src/Columns/ColumnObject.h

@@ -222,6 +222,7 @@ public:
     void addNewDynamicPath(std::string_view path);
 
     void setDynamicPaths(const std::vector<String> & paths);
+    void setDynamicPaths(const std::vector<std::pair<String, ColumnPtr>> & paths);
     void setMaxDynamicPaths(size_t max_dynamic_paths_);
     void setStatistics(const StatisticsPtr & statistics_) { statistics = statistics_; }
 

src/DataTypes/DataTypeObject.cpp

@@ -348,17 +348,13 @@ std::unique_ptr<ISerialization::SubstreamData> DataTypeObject::getDynamicSubcolu
                     result_typed_columns[getSubPath(path, prefix)] = column;
             }
 
-            auto & result_dynamic_columns = result_object_column.getDynamicPaths();
+            std::vector<std::pair<String, ColumnPtr>> result_dynamic_paths;
-            auto & result_dynamic_columns_ptrs = result_object_column.getDynamicPathsPtrs();
             for (const auto & [path, column] :  object_column.getDynamicPaths())
             {
                 if (path.starts_with(prefix) && path.size() != prefix.size())
-                {
+                    result_dynamic_paths.emplace_back(getSubPath(path, prefix), column);
-                    auto sub_path = getSubPath(path, prefix);
-                    result_dynamic_columns[sub_path] = column;
-                    result_dynamic_columns_ptrs[sub_path] = assert_cast<ColumnDynamic *>(result_dynamic_columns[sub_path].get());
-                }
             }
+            result_object_column.setDynamicPaths(result_dynamic_paths);
 
             const auto & shared_data_offsets = object_column.getSharedDataOffsets();
             const auto [shared_data_paths, shared_data_values] = object_column.getSharedDataPathsAndValues();

tests/queries/0_stateless/03250_json_group_by_sub_object_subcolumn.reference

@@ -0,0 +1,2 @@
+{"b":{"c":0,"d":"0","e":"str_0"}}
+{"b":{"c":0,"d":"1","e":"str_1"}}

tests/queries/0_stateless/03250_json_group_by_sub_object_subcolumn.sql

@@ -0,0 +1,10 @@
+set allow_experimental_json_type = 1;
+set allow_experimental_variant_type = 1;
+set use_variant_as_common_type = 1;
+
+drop table if exists test;
+create table test (json JSON(max_dynamic_paths = 20, `a.b.c` UInt32)) engine = Memory;
+insert into test select toJSONString(map('a.b.d', number::UInt32, 'a.b.e', 'str_' || toString(number))) from numbers(2);
+select json.^a from test group by json.^a order by toString(json.^a);
+drop table test;
+