Question

Below is my new sql so far as i do not manage to use Dale M advice,

SELECT
all_months.a_month_id AS month,
year($P{date}) as year,
count(case when clixsteraccount.rem_joindate between DATE_FORMAT($P{date}-INTERVAL 2   MONTH, '%Y-%m-01') AND $P{date} THEN clixsteraccount.rem_registerbycn end) AS
total_activation,
'ACTIVATION(No)' AS fake_column

FROM clixsteraccount right join all_months on all_months.a_month_id = date_format(clixsteraccount.rem_joindate,'%m') and
(clixsteraccount.rem_registrationtype = 'Normal')and(clixsteraccount.rem_kapowstatus='pending' or clixsteraccount.rem_kapowstatus='success')

GROUP BY year,month

HAVING month BETWEEN month(date_sub($P{date},interval 2 month)) and month($P{date})

So, what i do is create a table with two fields, a_month_id(1,2,3…,12) and a_month(name of months). Sql above does give me what i want which is to display previous 3 months even the months before is not exist.
exp: data start on July. So, i want to display May,June and July data like 0,0,100.

The problem occur when it comes to next months or next year. When i try to generate sql based on parameter on Jan, it doesn’t work like i thought. I do realize the problem are with ‘Having’ condition. Do anyone have idea how to improvised this sql to make it continue generate in the next,next year.

Thank you in advanced.

Answer 1

OK, I will make a few suggestions and give you an answer that will work on SQL Server – you will need to make any translations yourself.

I note that your query will aggregate all years together, i.e. Dec 2012 + Dec 2013 + Dec 2014 etc. Based on your question I don’t think that is your intention so I will keep each distinct. You can change the query if that was your intention. I have also not included your selection criteria (other than by the month).

I suggest that you utilize an index table. This is a table stored in your database (or the master database if possible) with an clustered indexed integer column running from 0 to n where n is a sufficiently large number – 10,000 will be more than enough for this application (there are 12 months in a year so 10,000 represents 833 years). This table is so useful everyone should have one.

SELECT  DATEADD(month, it.id, 0 ) AS month
    ,ISNULL(COUNT(clixsteraccount.rem_registerbycn), 0) AS registration
    ,'REGISTRATION(No)' AS fake_column
FROM    cn
    INNER JOIN ON ca.rem_registerbycn = cn.cn_id
    clixsteraccount ca 
    RIGHT JOIN
    IndexTable it ON it.id = DATEDIFF(month, 0, clixsteraccount.rem_joindate)
WHERE   it.id BETWEEN DATEDIFF(month, 0, @StartDate) - 3 AND DATEDIFF(month, 0, GETDATE())
GROUP BY it.id

The way it works is by converting the clixsteraccount.rem_joindate to an integer that represents the number of months since date 0 (01-01-1900 in SQL Server). This is then matched to the id column of the IndexTable and limited by the dates you select. Because every number exists in the index table and we are using an outer join it doesn’t matter if there are months missing from your data.