Merge pull request ClickHouse#79158 from petern48/fix_arraySymmetricDifference

Bug Fix arraySymmetricDifference() to work properly for tables with batch inserts

Author: jsc0218

src/Functions/array/arrayIntersect.cpp

@@ -640,6 +640,13 @@ ColumnPtr FunctionArrayIntersect<Mode>::execute(const UnpackedArrays & arrays, M
         // We have NULL in output only once if it should be there
         bool null_added = false;
         bool use_null_map;
+        const auto & arg = arrays.args[0];
+        size_t off;
+        // const array has only one row
+        if (arg.is_const)
+            off = (*arg.offsets)[0];
+        else
+            off = (*arg.offsets)[row];
 
         if constexpr (std::is_same_v<Mode, ArrayModeUnion>)
         {
@@ -657,12 +664,6 @@ ColumnPtr FunctionArrayIntersect<Mode>::execute(const UnpackedArrays & arrays, M
                 null_map.push_back(1);
                 null_added = true;
             }
-            const auto & arg = arrays.args[0];
-            // const array has only one row
-            if (arg.is_const)
-                prev_off[0] = 0;
-            else
-                prev_off[0] = (*arg.offsets)[row];
         }
         else if constexpr (std::is_same_v<Mode, ArrayModeSymmetricDifference>)
         {
@@ -684,13 +685,6 @@ ColumnPtr FunctionArrayIntersect<Mode>::execute(const UnpackedArrays & arrays, M
         else if constexpr (std::is_same_v<Mode, ArrayModeIntersect>)
         {
             use_null_map = all_nullable;
-            const auto & arg = arrays.args[0];
-            size_t off;
-            // const array has only one row
-            if (arg.is_const)
-                off = (*arg.offsets)[0];
-            else
-                off = (*arg.offsets)[row];
 
             for (auto i : collections::range(prev_off[0], off))
             {
@@ -719,9 +713,6 @@ ColumnPtr FunctionArrayIntersect<Mode>::execute(const UnpackedArrays & arrays, M
                     const char * data = nullptr;
                     pair = map.find(columns[0]->serializeValueIntoArena(i, arena, data));
                 }
-                prev_off[0] = off;
-                if (arg.is_const)
-                    prev_off[0] = 0;
 
                 if (!current_has_nullable)
                     all_has_nullable = false;
@@ -732,6 +723,10 @@ ColumnPtr FunctionArrayIntersect<Mode>::execute(const UnpackedArrays & arrays, M
                     insertElement<Map, ColumnType, is_numeric_column>(pair, result_offset, result_data, null_map, use_null_map);
             }
         }
+        // Now we update the offsets for the first array
+        prev_off[0] = off;
+        if (arg.is_const)
+            prev_off[0] = 0;
 
         result_offsets.getElement(row) = result_offset;
     }

tests/queries/0_stateless/03357_arraySymmetricDifference.reference

@@ -14,3 +14,13 @@ Array(Tuple(Nullable(UInt8), Array(Nullable(String))))
 Non-const arguments
 [(NULL,['c']),(2,['c',NULL])]
 [(NULL,['c']),(2,['c',NULL])]
+1	['1','2']
+2	[]
+3	['2']
+4	['1','4']
+5	['5']
+6	['5','4']
+7	['0','7']
+8	['10','8','9']
+9	['-1','9']
+10	['5','10']

tests/queries/0_stateless/03357_arraySymmetricDifference.sql

@@ -27,4 +27,33 @@ SELECT arraySymmetricDifference(f, s);
 WITH
     materialize([(1, ['a', 'b']::Array(LowCardinality(String))), (NULL, ['c']::Array(LowCardinality(String)))]) AS f,
     materialize([(2, ['c', NULL]::Array(LowCardinality(Nullable(String)))), (1, ['a', 'b']::Array(LowCardinality(String)))]) AS s
-SELECT arraySymmetricDifference(f, s)
+SELECT arraySymmetricDifference(f, s);
+
+-- Table with batch inserts
+DROP TABLE IF EXISTS test_arraySymmetricDifference;
+CREATE TABLE test_arraySymmetricDifference
+(
+    `id` Int8,
+    `arr1` Array(String),
+    `arr2` Array(String)
+)
+ENGINE = MergeTree
+ORDER BY id;
+
+INSERT INTO test_arraySymmetricDifference
+VALUES
+(1, ['1'], ['2']),
+(2, ['2'], ['2']),
+(3, ['3'], ['3', '2']),
+(4, ['4'], ['1']),
+(5, ['5'], []),
+(6, ['6', '4'], ['5', '6']),
+(7, ['7', '0'], []),
+(8, ['8', '9', '10'], []),
+(9, ['9'], ['-1']),
+(10, ['10'], ['5']);
+
+SELECT
+	ta.id AS id,
+	arraySymmetricDifference(ta.arr1, ta.arr2) AS symmetricDifference
+FROM test_arraySymmetricDifference ta;