How to insert a table with only one Row Type column?

[boson234]

steps:
1.
CREATE TABLE hive.default.row_table (
col1 ROW(col2 integer, col3 varchar)
) ;
2. insert into hive.default.row_table values (row(2, 'item'));
3. then the error as fellow:
Insert query has mismatched column types: Table: [row(col2 integer, col3 varchar)], Query: [integer, varchar(10)]

as rules in SqlBase.g4 ,rowConstructor considers that Row(1,'item1') equal to (1,'item1')
| '(' expression (',' expression)+ ')' #rowConstructor
| ROW '(' expression (',' expression)* ')' #rowConstructor

if I want to Insert data into the table row_type, I have to write sql as fellow:
insert into hive.default.row_table values (row(row(2, 'item')));

that's not cool, so is there any better way to do this?

[ebyhr]

You can use INSERT INTO ... SELECT ... instead.

[tschaub]

@ebyhr (or anyone else) - I'm trying to INSERT large batches of data coming from Parquet files with group/struct columns.

Using ROW types, it sounds like I would have to serialize this as INSERT INTO ... SELECT ... UNION ALL SELECT ... UNION ALL SELECT ... (and so on). Are there performance tradeoffs to consider with this syntax vs INSERT INTO ... VALUES ...?

I'm trying to decide whether to represent this data using ROW type columns or whether to flatten each nested field into its own column.