T-SQL Between Dates Confusion

I am working with T-SQL in SQL Server 2000 and I have a table TRANSACTIONS which has a date column TRANDATE defined as DateTime, among many other columns which are irrelevant for this question..

The table is populated with transactions spanning many years. I ran into code, test, that has me confused. There is a simple SELECT, like this:

SELECT TRANDATE, RECEIPTNUMBER FROM TRANSACTIONS WHERE TRANDATE BETWEEN '12/01/2010' and '12/31/2010' ORDER BY TRANDATE

and its not returni开发者_开发百科ng two rows of data that I know are in that table.

With the statement above, the last row its returning, in order, has a TRANDATE of: 2010-12-31 00:00:00.000

When I modify the statement like below, I get the additional two rows for December 2010 that are in that table:

SELECT TRANDATE, RECEIPTNUMBER FROM TRANSACTIONS WHERE TRANDATE BETWEEN '12/01/2010 00:00:00' and '12/31/2010 23:59:59' ORDER BY TRANDATE

I have tried to find out why the BETWEEN operator doesnt include ALL rows for the 24 period in 12/31/2010 when using the first SELECT, above. And why does it need to have the explicit hours added to the SELECT statement as in the second, modified, statement to get it to pull the correct number of rows out?

Is it because of the way TRANDATE is defined as "DATETIME"?

Based on this finding, I think that am going to have to go through all of this old code because these BETWEEN operators are littered throughout this old system and it seems like its not pulling all of the data properly. I just wanted clarification from some folks first. Thanks!

A date is a point in time, not a time span.

'12/31/2010' is a point, too. Namely, it's the midnight of the 31st of December.
Everything that happened after this point is ignored.
That's exactly the behaviour you want (even if you haven't realised that yet).

Do not think that when you choose to omit the time part, it is magically assumed to be "any". It's going to be "all zeroes", that is, the midnight.

If you want to include the entire day in your query without having to specify 23:59:59 (which, by the way, excludes the last second of the day, between the moment 23:59:59 of the current day and the moment 00:00:00 of the next day), you can do that either by using strict inequalities (>, <) bounded by the first points of time you don't want:

WHERE TRANDATE >='12/01/2010 00:00:00' and TRANDATE < '01/01/2011'

or by comparing date values casted to DATE:

WHERE CAST(TRANDATE AS DATE) between '12/01/2010' and '12/31/2010'

(it is okay to put this type of cast in a WHERE clause, it is sargable).

As you have discovered, if you don't specify a time when entering a date, it defaults to midnight in the morning of the date. So 12/31/2010 stops at midnight when that day begins.

To get all dates for 12/31/2010, you can either specify the time, as you have done, or add one day to the ending date. Without a time, 1/1/2011 ends at the stroke of midnight on 12/31/2010. So, you could do BETWEEN 12/1/2010 AND 1/1/2011. You can use DATEADD to add the day in your SQL if that makes it easier.

There is some risk in that second approach of adding a day. You will get any records for 1/1/2011 that carry the time of 00:00:00.

Here's one way to perform the DATEADD:

DECLARE @FromDate datetime, @ToDate datetime
// These might be stored procedure input parameters
SET @FromDate = '12/1/2010'
SET @ToDate = '12/31/2010'

SET @ToDate = DATEADD(d, 1, @ToDate)

Then you use @ToDate in your WHERE clause in the BETWEEN phrase in the usual way.

'12/01/2010' means '12/01/2010 00:00:00' and '12/31/2010' means '12/31/2010 00:00:00'. This is why datetime values that fall later on the day on 12/31/2010 are excluded from your query results.

What would be your expected result if I would do this

Insert "12/31/2010" into your datetime column?

Exactly: 12-31-2010 00:00:00

So why would you expect it to be different as argument for a query?

You have kind of answered your own question already. What you have observed is the way SQL Server works.

If it is confirmation you need, this MSDN document has following to say about it

When the time part is unspecified, it defaults to 12:00 A.M. Note that a row that contains a time part that is after 12:00 A.M. on 1998-0105 would not be returned by this query because it falls outside the range.

Edit

As for your comment, a datetime essentially is a floating point value.

Following script shows what numbers SQL Server works with.
40541.9749 (12/31/2010 23:23:59) can't be included when your upper bound is 40541 (12/31/2010)

DECLARE @ADateTime1 DATETIME
DECLARE @ADateTime2 DATETIME
DECLARE @ADateTime1AsFloat FLOAT
DECLARE @ADateTime2AsFloat FLOAT

SET @ADateTime1 = '12/31/2010'
SET @ADateTime2 = '12/31/2010 23:23:59'

SET @ADateTime1AsFloat = CAST(@ADateTime1 AS FLOAT)
SET @ADateTime2AsFloat = CAST(@ADateTime2 AS FLOAT)

SELECT @ADateTime1AsFloat, @ADateTime2AsFloat