Merge pull request #1067 from Altinity/backports/24.8.14/87863

24.8.14 Backport of ClickHouse#87863: ArrowStream processing crash if non unique dictionary

Author: zvonand

src/Processors/Formats/Impl/ArrowColumnToCHColumn.cpp

@@ -1024,6 +1024,16 @@ static ColumnWithTypeAndName readNonNullableColumnFromArrowColumn(
                 auto tmp_lc_column = lc_type->createColumn();
                 auto tmp_dict_column = IColumn::mutate(assert_cast<ColumnLowCardinality *>(tmp_lc_column.get())->getDictionaryPtr());
                 dynamic_cast<IColumnUnique *>(tmp_dict_column.get())->uniqueInsertRangeFrom(*dict_column.column, 0, dict_column.column->size());
+                size_t expected_dictionary_size = dict_column.column->size() + (dict_info.default_value_index == -1) + is_lc_nullable;
+                if (tmp_dict_column->size() != expected_dictionary_size)
+                {
+                    throw Exception(
+                        ErrorCodes::INCORRECT_DATA,
+                        "Expected Dictionary size {}, real Dictionary size is {}. The discrepancy probably caused by duplicated values",
+                        expected_dictionary_size,
+                        tmp_dict_column->size());
+                }
+
                 dict_column.column = std::move(tmp_dict_column);
                 dict_info.values = std::make_shared<ColumnWithTypeAndName>(std::move(dict_column));
                 dict_info.dictionary_size = arrow_dict_column->length();

tests/queries/0_stateless/02904_arrow_dictionary_indexes.sh

@@ -18,3 +18,4 @@ $CLICKHOUSE_LOCAL -q "select uniqExact(a) from file('$CLICKHOUSE_TMP/$CLICKHOUSE
 
 $CLICKHOUSE_LOCAL -q "select * from file('$CUR_DIR/data_arrow/different_dicts.arrowstream') order by x"
 
+$CLICKHOUSE_LOCAL -q "select * from file('$CUR_DIR/data_arrow/non_unique_dict.arrowstream') -- { serverError INCORRECT_DATA }"