NTH_VALUE

Returns the value at the specified position (N) within the window frame.

See also:

• FIRST_VALUE
• LAST_VALUE

Syntax

NTH_VALUE(
    expression,
    n
)
[ { RESPECT | IGNORE } NULLS ]
OVER (
    [ PARTITION BY partition_expression ]
    ORDER BY order_expression
    [ window_frame ]
)

Arguments:

• expression: Required. The column or expression to evaluate.
• n: Required. Position number (1-based) of the value to return.
• IGNORE NULLS: Optional. Skips null values when counting positions.
• RESPECT NULLS: Optional. Keeps null values when counting positions (default).
• window_frame: Optional. Defines the window frame. The default is RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW.

Notes:

• n must be a positive integer; n = 1 is equivalent to FIRST_VALUE.
• Returns NULL if the specified position does not exist in the frame.
• Combine with ROWS BETWEEN ... to control whether the position is evaluated over the whole partition or the rows seen so far.
• For window frame syntax, see Window Frame Syntax.

Examples

-- Sample order data
CREATE OR REPLACE TABLE orders_window_demo (
    customer VARCHAR,
    order_id INT,
    order_time TIMESTAMP,
    amount INT,
    sales_rep VARCHAR
);

INSERT INTO orders_window_demo VALUES
    ('Alice', 1001, to_timestamp('2024-05-01 09:00:00'), 120, 'Erin'),
    ('Alice', 1002, to_timestamp('2024-05-01 11:00:00'), 135, NULL),
    ('Alice', 1003, to_timestamp('2024-05-02 14:30:00'), 125, 'Glen'),
    ('Bob',   1004, to_timestamp('2024-05-01 08:30:00'),  90, NULL),
    ('Bob',   1005, to_timestamp('2024-05-01 20:15:00'), 105, 'Kai'),
    ('Bob',   1006, to_timestamp('2024-05-03 10:00:00'),  95, NULL),
    ('Carol', 1007, to_timestamp('2024-05-04 09:45:00'),  80, 'Lily');

Find the second order and illustrate null-handling for the second sales rep:

SELECT customer,
       order_id,
       order_time,
       NTH_VALUE(order_id, 2) OVER (
           PARTITION BY customer
           ORDER BY order_time
           ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW
       ) AS second_order_so_far,
       NTH_VALUE(sales_rep, 2) RESPECT NULLS OVER (
           PARTITION BY customer
           ORDER BY order_time
           ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW
       ) AS second_rep_respect,
       NTH_VALUE(sales_rep, 2) IGNORE NULLS OVER (
           PARTITION BY customer
           ORDER BY order_time
           ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW
       ) AS second_rep_ignore
FROM orders_window_demo
ORDER BY customer, order_time;

Result:

customer | order_id | order_time           | second_order_so_far | second_rep_respect | second_rep_ignore
---------+----------+----------------------+---------------------+--------------------+-------------------
Alice    |     1001 | 2024-05-01 09:00:00  |                NULL | NULL               | NULL
Alice    |     1002 | 2024-05-01 11:00:00  |                1002 | NULL               | NULL
Alice    |     1003 | 2024-05-02 14:30:00  |                1002 | NULL               | Glen
Bob      |     1004 | 2024-05-01 08:30:00  |                NULL | NULL               | NULL
Bob      |     1005 | 2024-05-01 20:15:00  |                1005 | Kai                | Kai
Bob      |     1006 | 2024-05-03 10:00:00  |                1005 | Kai                | Kai
Carol    |     1007 | 2024-05-04 09:45:00  |                NULL | NULL               | NULL