SELECT Query between dates, only selecting items between start and end fields - sql

Question

I have two tables that I will be using for tracking purposes, a Date Table and a Item Table. The Date Table is used to track the start and end dates of a tracked id. The Item Table is the amount of items that are pulled on a specific date for an id. The id is the foreign key between these two tables.
What I want to do, is a sum of the items with a GROUP BY of the id of the items, but only by summing the items based on if the date of the pulled item falls between the start_date and end_date of the tracked id.
The Date Table
id start_date end_date
1 2014-01-01 NULL
2 2014-01-01 2014-01-02
3 2014-01-25 NULL
The Item Table
id items date
1 3 2014-01-01
1 5 2014-01-02
1 5 2014-01-26
2 2 2014-01-01
2 3 2014-01-05
2 2 2014-01-26
3 2 2014-01-01
3 3 2014-01-05
3 2 2014-01-26
SQL I have so far, but I'm lost as to what to add to it from here.
SELECT
a.id,
SUM(items)
FROM
ww_test.dbo.items a
INNER JOIN ww_test.dbo.dates b ON
a.id = b.id
WHERE
a.date >= '2014-01-01' AND a.date <= '2014-01-30'
GROUP BY
a.id
ORDER BY
a.id
The output should be:
id items
1 13
2 2
3 2
Instead of:
id items
1 13
2 7
3 7

Answer 1

First of all, I strongly recommend that you stop using NULL in your date ranges to represent "no end date" and instead use a sentinel value such as 9999-12-31. The reason for this is primarily performance and secondarily query simplicity--a benefit to yourself now in writing the queries and to you or others later who have to maintain them. In front-end or middle-tier code, there is little difference to comparing a date range to Null or to 9999-12-31, and in fact you get some of the same benefits of simplified code there as you do in your SQL. I base this recommendation on over 10 years of full-time professional SQL query writing experience.
To fix your query as is, I think this would work:
SELECT
a.id,
ItemsSum = SUM(items)
FROM
ww_test.dbo.items a
INNER JOIN ww_test.dbo.dates b
ON a.id = b.id
AND a.date >= Coalesce(b.start_date, 0)
AND a.date <= Coalesce(b.end_date, '99991231')
WHERE
a.date >= '20140101'
AND a.date <= '20140130'
GROUP BY
a.id
ORDER BY
a.id
;
Note that if you followed my recommendation, your query JOIN conditions could look like this:
INNER JOIN ww_test.dbo.dates b
ON a.id = b.id
AND a.date >= b.start_date
AND a.date <= b.end_date
You will find that if your data sets become large, having to put a Coalesce or IsNull in there will hurt performance in a significant way. It doesn't help to use OR clauses, either:
INNER JOIN ww_test.dbo.dates b
ON a.id = b.id
AND (a.date >= b.start_date OR b.start_date IS NULL)
AND (a.date <= b.end_date OR b.end_date IS NULL)
That's going to have the same problems (for example converting what could have been a seek when there's a suitable index, into a scan, which would be very sad).
Last, I also recommend that you change your end dates to be exclusive instead of inclusive. This means that for the end date, instead of entering the date of the beginning of the final day the information is true, you put the date of the first day it is no longer true. There are several reasons for this recommendation:
If your date resolution ever changes to hours, or minutes, or seconds, every piece of code you have ever written dealing with this data will have to change (and it won't if you use exclusive end dates).
If you ever have to compare date ranges to each other (to collapse date ranges together or locate contiguous ranges or even locate non-contiguous ranges), you now have to do all the comparisons on a.end_date + 1 = b.start_date instead of a simple equijoin of a.end_date = b.start_date. This is painful, and easy to make mistakes.
Always thinking of dates as suggesting time of day will be extremely salutary to your coding ability in any language. Many mistakes are made, over and over, by people forgetting that dates, even ones in formats that can't denote a time portion (such as the date data type in SQL 2008 and up) still have an implicit time portion, and can be converted directly to date data types that do have a time portion, and that time portion will always be 0 or 12 a.m..
The only drawback is that in some cases, you have to do some twiddling about what date you show users (to convert to the inclusive date) and then convert dates they enter into the exclusive date for storing into the database. But this is confined to UI-handling code and is not throughout your database, so it's not that big a drawback.
The only change to your query would be:
INNER JOIN ww_test.dbo.dates b
ON a.id = b.id
AND a.date >= b.start_date
AND a.date < b.end_date -- no equal sign now
One last thing: be aware that the date format 'yyyy-mm-dd' is not culture-safe.
SET LANGUAGE FRENCH;
SELECT Convert(datetime, '2014-01-30'); -- fails with an error
The only invariantly culture-safe formats for datetime in SQL Server are:
yyyymmdd
yyyy-mm-ddThh:mm:ss

Answer 2

I think what you want to do is to compare the dates to be between the start_date and end_date of your Data table.
Change your query to the following and try
SELECT
a.id,
SUM(items)
FROM
ww_test.dbo.items a
INNER JOIN ww_test.dbo.dates b ON a.id = b.id
WHERE
a.date >= ISNULL(b.start_date, GETDATE())
AND a.date <= ISNULL(b.end_date, GETDATE())
GROUP BY a.id
ORDER BY a.id

Answer 3

The problem with the query is the condition part.
Also, since you need to retrieve data based on the condition defined in Dates table, you do not have to explicitly hard code the condition.
Assuming that your End Date can either be null or have values, you can use the following
query:
SELECT
a.id,
SUM(items)
FROM
ww_test.dbo.items a
INNER JOIN ww_test.dbo.dates b ON
a.id = b.id
where (b.end_date is not null and a.date between b.start_date and b.end_date)
or (b.end_date is null and a.date >= b.start_date)
GROUP BY
a.id
ORDER BY
a.id