Filling in missing balance and dates in table to track balance - sql

Question

I hope you can help me with this problem. I just started out on SQL using Bigquery so my problem can seem a bit tedious.
So I have a table that basically records the date and balance whenever the balance changes. It looks somewhat like this:
+------------+-----------+------+---------+
| Date | seller_ID | Name | Balance |
+------------+-----------+------+---------+
| 2020-09-10 | 1 | John | 10 |
| 2020-09-13 | 1 | John | 8 |
| 2020-09-15 | 1 | John | 6 |
+------------+-----------+------+---------+
However, I need to create a new table with the daily balances that looks like this
+------------+-----------+------+---------+
| Date | seller_ID | Name | Balance |
+------------+-----------+------+---------+
| 2020-09-10 | 1 | John | 10 |
| 2020-09-11 | 1 | John | 10 |
| 2020-09-12 | 1 | John | 10 |
| 2020-09-13 | 1 | John | 8 |
| 2020-09-14 | 1 | John | 8 |
| 2020-09-15 | 1 | John | 6 |
+------------+-----------+------+---------+
I tried creating a separate table of all the dates between the first and final date, and then LEFT JOIN the original table with it but the resulting table isn't very helpful to draw from.
Does anyone have an idea of what to do in this case?

Answer 1

To fill null value with previous non-null value in BigQuery you can use LAST_VALUE with IGNORE NULLS:
WITH test_table AS (
SELECT DATE '2020-09-10' AS Date, 1 AS seller_Id, 'John' AS Name, 10 AS Balance UNION ALL
SELECT '2020-09-13', 1, 'John' AS Name, 8 UNION ALL
SELECT '2020-09-15', 1, 'John' AS Name, 6
)
SELECT Date,
LAST_VALUE(seller_Id IGNORE NULLS) OVER (ORDER BY Date) AS seller_Id,
LAST_VALUE(Name IGNORE NULLS) OVER (ORDER BY Date) AS Name,
LAST_VALUE(Balance IGNORE NULLS) OVER (ORDER BY Date) AS purchase_date
FROM UNNEST(GENERATE_DATE_ARRAY('2020-09-10', '2020-09-15')) AS Date
LEFT JOIN test_table USING (Date)
ORDER BY Date

Answer 2

You can do this without window functions for the balance. The key is the window function only for the date:
WITH t AS (
SELECT DATE '2020-09-10' AS Date, 1 AS seller_Id, 'John' AS Name, 10 AS Balance UNION ALL
SELECT '2020-09-13', 1, 'John' AS Name, 8 UNION ALL
SELECT '2020-09-15', 1, 'John' AS Name, 6
),
tt as (
SELECT t.*, LEAD(date) OVER (PARTITION BY name ORDER BY date) as next_date
FROM t
)
SELECT dte, tt.name, tt.balance
FROM tt LEFT JOIN
UNNEST(GENERATE_DATE_ARRAY(tt.date, COALESCE(DATE_ADD(tt.next_date, INTERVAL - 1 DAY), DATE '2020-09-15'))) dte
ON true;
(Note: The ON clause is optional in this case. However, I am not a fan of having joins without ON -- unless it is a CROSS JOIN.)
This has two important advantages over Sergey's solution. The most important is that it will work for multiple names with different time periods.
The second advantage is that it is more efficient, because it is not using window functions to fetch values from previous rows.