Merge pull request ClickHouse#70695 from CurtizJ/fix-filling-missed-a…

…rrays

Fixed crash in filling of missed arrays

src/Storages/MergeTree/IMergeTreeReader.cpp

@@ -156,13 +156,21 @@ void IMergeTreeReader::evaluateMissingDefaults(Block additional_columns, Columns
         auto it = original_requested_columns.begin();
         for (size_t pos = 0; pos < num_columns; ++pos, ++it)
         {
-            auto name_in_storage = it->getNameInStorage();
-
-            if (full_requested_columns_set.emplace(name_in_storage).second)
-                full_requested_columns.emplace_back(name_in_storage, it->getTypeInStorage());
-
             if (res_columns[pos])
+            {
+                /// If column is already read, request it as is.
+                if (full_requested_columns_set.emplace(it->name).second)
+                    full_requested_columns.emplace_back(it->name, it->type);
+
                 additional_columns.insert({res_columns[pos], it->type, it->name});
+            }
+            else
+            {
+                /// If column or subcolumn is missed, request full column for correct evaluation of defaults of subcolumns.
+                auto name_in_storage = it->getNameInStorage();
+                if (full_requested_columns_set.emplace(name_in_storage).second)
+                    full_requested_columns.emplace_back(name_in_storage, it->getTypeInStorage());
+            }
         }
 
         auto dag = DB::evaluateMissingDefaults(
@@ -183,6 +191,12 @@ void IMergeTreeReader::evaluateMissingDefaults(Block additional_columns, Columns
         it = original_requested_columns.begin();
         for (size_t pos = 0; pos < num_columns; ++pos, ++it)
         {
+            if (additional_columns.has(it->name))
+            {
+                res_columns[pos] = additional_columns.getByName(it->name).column;
+                continue;
+            }
+
             auto name_in_storage = it->getNameInStorage();
             res_columns[pos] = additional_columns.getByName(name_in_storage).column;
 

tests/queries/0_stateless/03203_fill_missed_subcolumns.reference

@@ -27,5 +27,5 @@
 2	('aaa','bbb')	[1,NULL,3]
 3	('ccc','ddd')	[4,5,6]
 1	foo	bar	3	[0,1,0]
-2	foo	bar	3	[0,1,0]
+2	aaa	bbb	3	[0,1,0]
 3	ccc	ddd	3	[0,0,0]

tests/queries/0_stateless/03252_fill_missed_arrays.reference

@@ -0,0 +1 @@
+20000

tests/queries/0_stateless/03252_fill_missed_arrays.sql

@@ -0,0 +1,22 @@
+DROP TABLE IF EXISTS t_fill_arrays;
+
+CREATE TABLE t_fill_arrays
+(
+    `id` String,
+    `mapCol` Map(String, Array(String)),
+)
+ENGINE = MergeTree
+ORDER BY id
+SETTINGS vertical_merge_algorithm_min_rows_to_activate = 1, vertical_merge_algorithm_min_columns_to_activate = 1, min_bytes_for_full_part_storage = 0;
+
+INSERT INTO t_fill_arrays (id) SELECT hex(number) FROM numbers(10000);
+
+ALTER TABLE t_fill_arrays ADD COLUMN arrCol Array(String) DEFAULT [];
+
+INSERT INTO t_fill_arrays (id) SELECT hex(number) FROM numbers(10000);
+
+SELECT count() FROM t_fill_arrays WHERE NOT ignore(arrCol, mapCol.values);
+
+OPTIMIZE TABLE t_fill_arrays FINAL;
+
+DROP TABLE t_fill_arrays;