SQL for last six "full" months - sql

Question

I have table containing one datetime column. I need to return rows for only last 6 months. This can be done by
WHERE CloseTime >= DATEADD(Month, DATEDIFF(Month, 0, DATEADD(m, - 6, CURRENT_TIMESTAMP)), 0)
This gets me data for the month I am starting this script + 6 last months. So e.g. if I run this script today, Ill get the data for this month + all previous months till April (04).
Now I need to modify the condition so if I run the script today, the data will be obtained only for months 03-09 only, exluding days in this month (10).
Any advice, please?

Answer 1

If you want to have the previous 6 months regardless of whether today is the 1st, 3rd, 9th, 29th, whatever, then just subtract 7 months. Here is one way to do that: get the first of the month into a variable, then use an open-ended range in the query.
DECLARE #ThisMonth DATETIME;
SET #ThisMonth = DATEADD(MONTH, DATEDIFF(MONTH, '19000101', GETDATE()), '19000101');
SELECT...
WHERE CloseTime >= DATEADD(MONTH, -7, #ThisMonth)
AND CloseTime < #ThisMonth;
You could also use 0 in place of '19000101' but I prefer an explicit date than implicit shorthand (it was a very tough habit to break).
If you really don't like variables, then you can make the query a lot more complex by repeating the expression to calculate the first of this month (and in the start of the range, subtract 7 from the number of months):
SELECT...
WHERE CloseTime >= DATEADD(MONTH, DATEDIFF(MONTH, '19000101', GETDATE())-7, '19000101')
AND CloseTime < DATEADD(MONTH, DATEDIFF(MONTH, '19000101', GETDATE()), '19000101');
Yuck. Variables make this much tidier.

Answer 2

When creating queries you do not want to use a function on the search column since it will result in a full table scan.
The solution works and should pick up any index on CloseTime.
-- Get me data in months 3 (mar) to 9 (sep) of this year
select
*
from
my_table
where
CloseTime between
DATEADD(d, -1, '03-01-2013') and DATEADD(d, +1, '09-20-2013')
If the table is small and a full table scan is not a issue, a simple solution is to use the MONTH function.
-- Get me data in months 3 (mar) to 9 (sep) of this year
select
*
from
my_table
where
MONTH(CloseTime) IN (3,4,5,6,7,8,9) and YEAR(CloseTime) = 2013

Answer 3

I looked at Aaron's article. Its a very good read.
I wondered if there was some new function that was not tested since that article.
If you are using 2012, why not use the format function? Logically, you want date variables with a 01 day. The query plan still gets a clustered index scan.
Let's see how this solution stacks up using Aaron's test database.
-- Use the sample
use [DateTesting]
go
-- Johns - log by 6 months
CREATE PROCEDURE dbo.Johns_LogBy6Months
#date SMALLDATETIME
AS
BEGIN
SET NOCOUNT ON;
DECLARE #cmp_date SMALLDATETIME = format(#date, 'yyyyMM01');
DECLARE #c INT;
SELECT #c = COUNT(*)
FROM dbo.SomeLogTable
WHERE DateColumn >= dateadd(m, -7, #cmp_date)
AND DateColumn < #cmp_date
END
GO
-- Aarons - log by 6 months
CREATE PROCEDURE dbo.Aarons_LogBy6Months
#date SMALLDATETIME
AS
BEGIN
SET NOCOUNT ON;
DECLARE #c INT;
DECLARE #cmp_date SMALLDATETIME = DATEADD(MONTH, DATEDIFF(MONTH, '19000101', #date), '19000101');
SELECT #c = COUNT(*)
FROM dbo.SomeLogTable
WHERE
DateColumn >= dateadd(m, -7, #cmp_date)
AND DateColumn < #cmp_date
END
GO
Lets make 1000 calls to the functions.
-- Sample calls x 1000
PRINT CONVERT(char(23), GETDATE(), 121);
GO
EXEC dbo.Johns_LogCountByDay #date = '20091005';
GO 1000
PRINT CONVERT(char(23), GETDATE(), 121);
GO
EXEC dbo.Aarons_LogBy6Months #date = '20091005';
GO 1000
PRINT CONVERT(char(23), GETDATE(), 121);
GO
Here are the executions times.
2013-10-10 11:58:49.547
Beginning execution loop
Batch execution completed 1000 times.
2013-10-10 11:58:52.837
Beginning execution loop
Batch execution completed 1000 times.
2013-10-10 11:58:55.883
In summary a call to the new format() function and a implicit cast to a small date time takes a little more time than a dateadd() and datediff() with two string (date) literals.
The format() solution seems more intuitive or self documenting to me. The time difference was 3.3 versus 3.0 sec.
I have to give the speed test a win to Aaron's solution. Stick with in-equality comparisons of date variables. They are faster.
In short, I will have to fix my bad habits.

Answer 4

I had a similar request. I needed last six months and next six months of appointments. But just like you I needed the full months. So a simple getdate +- 180 wouldn't do.
I took a more simple approach. I get the year and month and turn it into a number like 201912. Then do a between clause. Its dynamic and I get full months.
I'm sure there's more sophisticated approaches, but this is what I came up with.
WHERE Year([ApptDate])*100 + Month([ApptDate]) between Year(getdate()-180)*100 + Month(getdate()-180) and Year(getdate()+180)*100 + Month(getdate()+180)