I want to create one row per id per month till the month of end_date.

e.g. first customer id started in Oct and ended in Nov. So I want to get two rows for each month that the customer was active. Besides that, I want to create a column that flags if it was active in that month.

| id | start_date |  end_date  |
|----|------------|------------|
| a  | 2021-10-02 | 2021-11-15 |
| b  | 2021-11-13 | 2021-11-30 |
| c  | 2021-11-16 |            |

When there is no end_date, meaning it is still active, it has to be till the current month.

Example data:

WITH t1 AS (
SELECT 'a' AS id, '2021-10-02'::date AS start_date, '2021-11-15'::date AS end_date UNION ALL
SELECT 'b' AS id, '2021-11-13'::date AS start_date, '2021-11-30'::date AS end_date UNION ALL
SELECT 'c' AS id, '2021-11-16'::date AS start_date, NULL::date AS end_date
    )

The expected result:

| id | start_date |  end_date  |   months   | is_active |
|----|------------|------------|------------|-----------|
| a  | 2021-10-02 | 2021-11-15 | 2021-10-01 | TRUE      |
| a  | 2021-10-02 | 2021-11-15 | 2021-11-01 | FALSE     |
| b  | 2021-11-13 | 2021-11-30 | 2021-11-01 | FALSE     |
| c  | 2021-11-16 |            | 2021-11-01 | TRUE      |
| c  | 2021-11-16 |            | 2021-12-01 | TRUE      |
| c  | 2021-11-16 |            | 2022-01-01 | TRUE      |

How can I achieve that in Snowflake?

CodePudding user response：

so given you have a range, you will need something that spans time to join against, this is where a generator can be used, which I will put into a CTE. I will also use ROW_NUMBER() to generate the sequence of month steps, to make sure there is no gaps. the 200 needs to be hard coded, so put a value that spans enough data for you needs, or pop this into a table.

WITH months AS (
    SELECT 
        ROW_NUMBER() OVER (ORDER BY NULL) - 1 as rn
    FROM TABLE(generator(rowcount => 200))
)

next we want to truncate the start_date and find the number of months later that end_date is, and join that to our range

), range_prep AS (
    SELECT id,
        start_date,
        end_date,
        date_trunc(month, start_date) as start_month,
        datediff(month, start_month, coalesce(end_date, CURRENT_DATE())) as month_count
    FROM data
)

joining those together, and doing :

SELECT id,
    r.start_date,
    r.end_date,
    dateadd(month, m.rn, r.start_month) as months,
    (r.end_date is null) OR (date_trunc(month, r.end_date) > months) AS is_active
FROM range_prep as r
JOIN months as m
    ON m.rn <= r.month_count
ORDER BY 1,2;

putting all together with a CTE for data we have:

WITH data AS (
    SELECT id, 
        to_date(start_date) as start_date, 
        to_date(end_date) as end_date  
    FROM VALUES 
        ('a','2021-10-02','2021-11-15'),
        ('b','2021-11-13','2021-11-30'),
        ('c','2021-11-16',null)
    v( id, start_date, end_date)
), months AS (
    SELECT 
        ROW_NUMBER() OVER (ORDER BY NULL) - 1 as rn
    FROM TABLE(generator(rowcount => 200))
), range_prep AS (
    SELECT id,
        start_date,
        end_date,
        date_trunc(month, start_date) as start_month,
        datediff(month, start_month, coalesce(end_date, CURRENT_DATE())) as month_count
    FROM data
)
SELECT id,
    r.start_date,
    r.end_date,
    dateadd(month, m.rn, r.start_month) as months,
    (r.end_date is null) OR (date_trunc(month, r.end_date) > months) AS is_active
FROM range_prep as r
JOIN months as m
    ON m.rn <= r.month_count
ORDER BY 1,2;

gives: