Question

I have three columns that look something like this:

Column1       Date1         Date2
Test2         2012-06-10    
Test8         2012-05-05    2012-06-10

I’ll start off by describing my desired output given this data set. It would look like this:

Year    Month    Sum
2012    05       1
2012    06       2
2012    07       1

If Column1 contains a number (which can be found by something like: WHERE Column1 LIKE ‘%2%’), this number should be added to the month value in Date1. This is the case with the first row, i.e. a 1 is added for month 06 and 07. However, for the case of row 2, if there is a date in Date2, it should only add 1 to Date1 and every month up to and including the month in Date2. This is why a 1 is added to 05 and 06.

I’m guessing this query would use the INTERVAL function but I’m not sure as to how I should add values to months in the future.

Update:

@CraigRinger – I’ll try to explain it again. I’d like to use the number in Column1 to determine how many months into the future the date in Date1 will run. For the first row, it will run from 2012-06 to 2012-07 (as there’s a 2 in Column1).

Date2 can be seen as a cancellation date. So for row two, Date1 would continue for 8 months, but as it is being cancelled (by Date2) it runs only from 05 to 06.

In other words, the query should add 1 to the beginning month (Date1) and to every month up to and including the end (or Date2 if this exists).

I’d like to add 1 to every one of these months, so that I know the sum of months for all my rows. I’m guessing this would involve adding a date interval (equal to the number in Column1) to Date1, extracting the months from the dates between these two numbers and adding a one to them. Unfortunately I have no idea how best to implement this.

Hope I’ve explained it better this time!

Answer 1

Based on the data sample above then the following should suffice:

create table ddata(test text, d1 date, d2 date);
insert into ddata values ('Test2','2012-06-10'::date,null),('Test8','2012-05-05','2012-6-10')

To get the number of months from the string value

SELECT substring(test,E'\\d+$') FROM ddata;

To get the difference in months where the second date is populated

SELECT ((extract(year FROM d2) - extract(year FROM d1)) *12) + extract(month FROM d2) - extract(month FROM d1)
  FROM ddata;

Putting it together using a case statement to decide which month calculation to use and generate_series to produce a list of intermediate dates to count:

SELECT extract(year FROM ccal.cdate),
       extract(month FROM ccal.cdate),
       count(*) AS test_count
FROM
(SELECT generate_series(mcal.d1, mcal.d1 + CAST((mcal.num_mths || ' months') AS INTERVAL), '1 month') AS cdate
  FROM
(SELECT d1,
       CASE (d2 IS NULL)
            WHEN TRUE THEN
                substring(test,E'\\d+$')::integer - 1
            ELSE 
                ((extract(year FROM d2) - extract(year FROM d1)) *12) + extract(month FROM d2) - extract(month FROM d1)
            END as num_mths
 from ddata) as mcal

 )AS ccal
GROUP BY 1,2
ORDER BY 1,2;