获取最早日期的SQL方法

Question

What would be the best method to fetch the Sale date for each client when the records are stored in columns? Each record is flattened and store the sales data for each month for the last 15 years. Hence it contains Client_ID and the sale month (Dec_08, Jan_09, Feb_09...) as the column header and the column values contain the Sale value (number of Boxes) as Integer. I would need to do a calculation of datediff to get the number of years since the first sale, hence the Month Year needs to be some sort of date answer here. I am doing this in SQL in Snowflake enter image description here

Example of the Sale is in the picture below

Any help would be appreciated.

Answer 1

You could use the UNPIVOT construct to rotate the table by transforming columns into rows.

Then calculate min date (you need to cast you month_year value from string to date using to_date( yourcolumn,'Mon_YY')) , group by ClientID, where box# is not null.

You can also apply datediff function in order to retrieve time between min date and current date ( https://docs.snowflake.com/en/sql-reference/functions/datediff.html ).

Below is an example from Snowflake documentation of UNPIVOT construct ( https://docs.snowflake.com/en/sql-reference/constructs/unpivot.html ).

Given a table, monthly_sales, with the following structure and data, unpivot the individual month columns to return a single sales value by month for each employee:

-- example setup
create or replace table monthly_sales(empid int, dept text, jan int, feb int, mar int, april int);

insert into monthly_sales values
    (1, 'electronics', 100, 200, 300, 100),
    (2, 'clothes', 100, 300, 150, 200),
    (3, 'cars', 200, 400, 100, 50);

-- UNPIVOT example
select * from monthly_sales
    unpivot(sales for month in (jan, feb, mar, april))
    order by empid;


+-------+-------------+-------+-------+
| EMPID | DEPT        | MONTH | SALES |
|-------+-------------+-------+-------|
|     1 | electronics | JAN   |   100 |
|     1 | electronics | FEB   |   200 |
|     1 | electronics | MAR   |   300 |
|     1 | electronics | APRIL |   100 |
|     2 | clothes     | JAN   |   100 |
|     2 | clothes     | FEB   |   300 |
|     2 | clothes     | MAR   |   150 |
|     2 | clothes     | APRIL |   200 |
|     3 | cars        | JAN   |   200 |
|     3 | cars        | FEB   |   400 |
|     3 | cars        | MAR   |   100 |
|     3 | cars        | APRIL |    50 |
+-------+-------------+-------+-------+

Answer 2

You can't work with that data structure directly, it's too unwieldy. You have to first get it into a different form. The way to do that is to introduce a new table and write a query to populate it like this:

insert into YourNewTable
(YearNum, MonthNum, Client_ID, Sale_Value)
select
    2008 as Year_Num,
    12 as Month_Num,
    Client_ID,
    Dec_08 as Sale_Value
from
    Sale
where
    Dec_08 <> 0

union all

select
    2009 as Year_Num,
    1 as Month_Num,
    Client_ID,
    Jan_09 as Sale_Value
from
    Sale
where
    Jan_09 <> 0

union all

[... etc., all the way to the most recent month]

You can't write the query above manually, you have to code some sort of automated routine to write the query, in the language of your choice. But once it's done, it's done.

After you've run this new query to populate YourNewTable, you can query YourNewTable in a sensible fashion.