How to return only the Date from a SQL Server DateTime datatype

Question

SELECT GETDATE()

Returns: 2008-09-22 15:24:13.790

I want that date part without the time part: 2008-09-22 00:00:00.000

How can I get that?

Solution 1

SELECT DATEADD(dd, 0, DATEDIFF(dd, 0, @your_date))

for example

SELECT DATEADD(dd, 0, DATEDIFF(dd, 0, GETDATE()))

gives me

2008-09-22 00:00:00.000

Pros:

• No varchar<->datetime conversions required
• No need to think about locale

Solution 2

SQLServer 2008 now has a 'date' data type which contains only a date with no time component. Anyone using SQLServer 2008 and beyond can do the following:

SELECT CONVERT(date, GETDATE())

Solution 3

If using SQL 2008 and above:

select cast(getdate() as date)

Solution 4

DATEADD and DATEDIFF are better than CONVERTing to varchar. Both queries have the same execution plan, but execution plans are primarily about data access strategies and do not always reveal implicit costs involved in the CPU time taken to perform all the pieces. If both queries are run against a table with millions of rows, the CPU time using DateDiff can be close to 1/3rd of the Convert CPU time!

To see execution plans for queries:

set showplan_text on
GO 

Both DATEADD and DATEDIFF will execute a CONVERT_IMPLICIT.

Although the CONVERT solution is simpler and easier to read for some, it is slower. There is no need to cast back to DateTime (this is implicitly done by the server). There is also no real need in the DateDiff method for DateAdd afterward as the integer result will also be implicitly converted back to DateTime.

---

SELECT CONVERT(varchar, MyDate, 101) FROM DatesTable

  |--Compute Scalar(DEFINE:([Expr1004]=CONVERT(varchar(30),[TEST].[dbo].[DatesTable].[MyDate],101)))
       |--Table Scan(OBJECT:([TEST].[dbo].[DatesTable]))

---

SELECT DATEADD(dd, 0, DATEDIFF(dd, 0, MyDate)) FROM DatesTable

  |--Compute Scalar(DEFINE:([Expr1004]=dateadd(day,(0),CONVERT_IMPLICIT(datetime,datediff(day,'1900-01-01 00:00:00.000',CONVERT_IMPLICIT(datetime,[TEST].[dbo].[DatesTable].[MyDate],0)),0))))
       |--Table Scan(OBJECT:([TEST].[dbo].[DatesTable]))

Using FLOOR() as @digi suggested has performance closer to DateDiff, but is not recommended as casting the DateTime data type to float and back does not always yield the original value.

Remember guys: Don't believe anyone. Look at the performance statistics, and test it yourself!

Be careful when you're testing your results. Selecting many rows to the client will hide the performance difference because it takes longer to send the rows over the network than it does to perform the calculations. So make sure that the work for all the rows is done by the server but there is no row set sent to the client.

There seems to be confusion for some people about when cache optimization affects queries. Running two queries in the same batch or in separate batches has no effect on caching. So you can either expire the cache manually or simply run the queries back and forth multiple times. Any optimization for query #2 would also affect any subsequent queries, so throw out execution #1 if you like.

Here is full test script and performance results that prove DateDiff is substantially faster than converting to varchar.

Solution 5

Try this:

SELECT CONVERT(VARCHAR(10),GETDATE(),111)

The above statement converts your current format to YYYY/MM/DD, please refer to this link to choose your preferable format.

View more solutions

Comments

[Ben Hoffstein]

One thing to note is that SQL Server 2008 includes a separate DATE datatype for storing just dates without the time component. More info here: sql-server-performance.com/articles/dev/datetime_2008_p1.asp‌​x

[ErikE]

Don't miss this post showing performance testing results of the various time-removal methods.

[Adir D]

@Martin weird, thanks, let's try again.

[Rohit Vipin Mathews]

Don't be mislead by the votes and accepted answer, Take a look at stackoverflow.com/a/126984/1155650

[hktegner]

@Rohit not everyone has SQL Server 2008 or later.

[Rohit Vipin Mathews]

@hktegner - it must be misleading for those who have. Also there is one version below 2008 (ie:2005) where as after it we have 4 and counting.

[hktegner]

@Rohit You are incorrectly assuming that 2008 is the only version people care about. (There are more versions in the wild.) The votes speak for themselves.

[eaglei22]

For future readers: mssqltips.com/sqlservertip/1145/… is also useful for getting the correct format when using convert.

[Admin]

SELECT DATEADD(dd, 0, DATEDIFF(dd, 0, @your_date))

[Eddie Groves]

This returns '2008/09/22' for me

[Eddie Groves]

Is this way better or worse performance wise than using the convert methods other have suggested? Or is it negligible?

[Eddie Groves]

These methods are all great, but which single one do you suggest using?

[aku]

My method works faster. It doesn't require conversions to varchar and allows efficient date calculations

[Dane]

+1 Looks like this one is 35% faster than the double convert() method commonly used (which I also have used for years). Nice one.

[Cade Roux]

If this is 35% than the CONVERT method, you've got to wonder how much faster a built-in truncate would be - this has to be the most common datetime-related operation I ever do - I'm going to see about switching to this mechanism.

[aku]

Ricardo C, nice investigation! What version of SQL server do you use? On MSSQL2000 method with datediff performs slightly faster for me.

[aku]

Just to note, I performed test 1000.000 times. For real-world scenarios performance difference will not be noticeable, I guess

[Ricardo C]

Aku, I used SQL Server 2005 Express for this test. I work on 2000 at work, and I will test it with a table with over 24 million rows and see what comes out of it.

[Jim Birchall]

The only downside I can see to your solution is that unless you know what it is doing it is a bit obtuse. Using the double convert method makes your intentions more obvious to futire code maintainers. BTW I have not downvoted you. I think I'll start using your method too. Thankyou @aku

[Ricardo C]

Aku, same results. No difference in performance over ten million rows.

[ErikE]

The claims about equal performance are not true. Of course the execution plans will be the same!!! Measuring performance on these MUST be done by comparing CPU usage, not examining execution plans.

[ErikE]

+1 You may be interested to see Ricardo C's edited answer (since it is community wiki and factually incorrect, I corrected it). You also got a prop to your question.

[ErikE]

This method is not the fastest, and also implicitly teaches people that casting dates to float is accurate, which it is not. Please see this post for more detail.

[ErikE]

Also don't miss this post showing performance testing results.

[misteraidan]

There is also the 'time' data type in SQL2008 which answers the other half of the question of separating date and time.

[ErikE]

@pilavdzice Setting a datetime to midnight of that day does LEAVE OFF THE TIME. What result are you expecting? The datetime data type cannot have no time at all. I think you are confusing data storage with user presentation. If all you want is a way to show a user a string that has no time portion (not zeroes, just blanks) then you simply want Convert(varchar(30), @Date, 101) or something similar. See SQL Server Books Online • Cast and Convert for more info.

[Fredrick Gauss]

Msg 243, Level 16, State 1, Line 1 Type date is not a defined system type.

[abatishchev]

@FredrickGauss: What type, Date? What version of SQL Server do you use?

[Andrew Lazarus]

Except, say, if you want a query that retrieves all records matching a user-supplied date as the date-part of a certain time field. Good luck doing that only in the presentation layer. (You don't need convert, you can can use date arithmetic, but you get the idea…)

[N t]

Worth noting that this does not extend to aggregating monthly, or yearly data. You have to alter the terms in order for that to work. SELECT DATEADD(mm,DATEDIFF(mm, 0, @YourDate),0) is extensible for mm/yy iirc

[Praveen]

@aku Is there a way to get only Datepart as "2008-09-22" and not "2008-09-22 00:00:00.000" without converting it into VARCHAR. Now I'm using CONVERT(VARCHAR(10), @dateTime, 101) AS MyDate

[Magnus]

@user1671639 the datetime data type always contains both a date and a time, you can't sensibly store one without the other - unless you're using SQL Server 2008, in which case there are also separate 'date' and 'time' data types. If you use CONVERT() like that, you really want a string for later use, so you'll be stuck doing it like that - although it'd be better if you used date formatting functions instead of cutting the date off - or via CAST(... AS DATE) or CONVERT(DATE, ...), which has been mentioned quite often on this very page.

[Flea]

SELECT CONVERT(VARCHAR(10),GETDATE(),101) is mm/dd/yyyy format.

[aruno]

if you're sorting based on the raw text value (outside of the DB) then the 'japanese' format is better

[Adir D]

@Andrew why does that matter? You say WHERE col >= @Date AND col < DATEADD(DAY, 1, @Date); - there is absolutely no reason to strip time from the column.

[Andrew Lazarus]

@AaronBertrand That only works assuming the input @Date has a zero time part. In case that isn't true, you still need to know how to truncate times server-side. I agree with this answer that formatting should be left to the presentation layer, but I didn't agree with an implication that leaving that for the front end means you don't have to know a quick way to truncate.

[Adir D]

@Andrew all you have to do is make the input parameter DATE. My point is still that you should never have to apply any such truncation to the column, even though that is most people's first instinct.

[Andrew Lazarus]

@AaronBertrand and that assumes you have control over the datatype of the parameter. Fine in a stored procedure, not so possible in other situations. Why not cast to be sure the parameter is the type you want and need?

[Adir D]

@Andrew again, that's fine. My point - AGAIN - is that you should never have to cast the column. Please read.

[Bohemian]

This returns me date with zero time, not just date

[Mahesh ML]

can i know which version if sql server you are using?

[Marek]

@MaheshML it returns both date and time in MS SQL 2012.

[Martín Coll]

Works like a charm in SQL Azure

[UnhandledExcepSean]

FYI, I benchmarked different methods of trimming off time from dates and this was the fastest method. Granted the difference was small, but it was clearly faster over a large # of executions.

[Michael]

Note that the "correct" version of the top two is select dateadd(dd, datediff(dd, 0, getdate()), 0), because the dds can then be swapped out for any of the datepart keywords to clip the date at any segment you choose. (Also note that dd is just an abbreviation for day.)

[Michael]

I recommend changing the answer to SELECT DATEADD(dd, DATEDIFF(dd, 0, @your_date), 0) because then dd can be swapped out for any other datepart keyword to truncate your datetime at an arbitrary level.

[Andriy M]

This suggestion has been covered by other answers (more than once).

[Nick]

Beware! declare @date1 datetime = '2015-09-30 20:59:59.999'; select cast(@date1 as date) returns '2015-10-01'

[Nick]

@abatishchev sorry, that should have been declare @date1 datetime = '2015-09-30 23:59:59.999';select cast(@date1 as date)

[abatishchev]

@Nick: this is the issue with DateTime. use DateTime2 instead and it works fine. sqlfiddle.com/#!6/9eecb7/2833

[Dr. MAF]

wt about sqlserver 2005??

[Frédéric]

@Nick, to complement abatishchev response, your @date1 is indeed 2015-10-01, due to DateTime limitations. Try without any cast to Date, it yields 2015-10-01too! declare @date1 datetime = '2015-09-30 23:59:59.999';select @date1 => 2015-10-01

[The1nk]

The question states SQL Server. This seems like MySQL?

[NicVerAZ]

One of these easy to remember SQL tricks. As Mike says, only 2008 onward but, if you find a 2005 and previous DB somewhere, you may have a lot of issues :)

[abatishchev]

@NixVerAZ I believe there are exactly 0 reasons to run SQL Server 2005 in late 2016. This is pure idiocy, isn't it? A good sign of something terribly wrong there.

[Zack]

Your first example still has a time component. The point of the question was how to remove that.

[galaxis]

I like the use of the ANSI std "CAST()" - if portability is a concern (or even achievable) these days :). Also preferred if wanting to preserve precision: msdn.microsoft.com/en-us/library/ms187928.aspx

[Frosty840]

@Dr.MAF Completing the circle, the pre-2008 answer is here: stackoverflow.com/questions/113045/…

[SvenAelterman]

@MaheshML There is no such thing as SQL Server 2010.

[Luc VdV]

@abatischchev - 1 reason that applies to more instances than you'd hold for possible: management refusing to approve the budget for updating antiques as long as someone manages to keep them running. Even SQL Server 2000 running on Windows 2000 is still alive. SQL Server 2005 Express: about 50 instances in my company alone (no, not mine, the one I work for).

[Luc VdV]

The '2015-09-30 20:59:59.999' issue doesn't seem to apply to Sql Server 2016 anymore.

[David Faber]

One can go even further with the ANSI standard by using SELECT CAST(CURRENT_TIMESTAMP AS DATE) (CURRENT_TIMESTAMP is equivalent to GETDATE()).

[David Faber]

Not only is it for Oracle, not MS SQL - it's not even correct. To get the date part only from Oracle, one would use TRUNC(SYSDATE)

[karthik kasubha]

select cast(createddate as date) from table where createdate is your datetime column

[J. Chris Compton]

I expect "SQL Server 2010" means SQL Server 2008 R2 (it was released in 2010)

[ChrisM]

I think you have copied in 2 select statements but copied the second select inside the first select.

[brianary]

select {fn current_date()} as today works for me.

[lit]

@brianary - That's nice, but it is not ANSI SQL.

[brianary]

That's fair enough, and your answer is nicely portable, but I figured as long as we're working around T-SQL, this also works (and shows that implementing ANSI CURRENT_DATE would be trivial for MS).

[noobprogrammer]

DATE is not a function dude.

[ourmandave]

To skip the culture, the custom formats lets you set your own, e.g. FORMAT (@d, 'yyyyy-MM-dd') to get 2011-10-11.

[Nosajimiki]

To clarify... DATE() is a function in mySQL, but this question is about SQL Server. In mySQL you can use the DATE() function to extract the date from a date time element such that DATE("2021-06-15 09:34:21") returns "2021-06-15", but GETDATE() is not a mySQL function, so he is mix-matching SQL languages. The equivalent of this in mySQL would be DATE(NOW())... but mySQL makes this completely unnecessary since you can just call CURDATE() instead.

[Nosajimiki]

The last 3 are full dateTime stamps being truncated. 121 is not separated by a 'T' while the other two are. 127 is more precise when working with DATETIME2 formats, but with regular datetimes, they round to the same values; so, in a different context, these return 3 different datetime values. As for the second one, A = B can be used in slightly different ways than A AS B using other functions than this, but using these functions, they work the same. As for Convert vs Cast, Cast is a native ANSI function while Convert is SQL specific. Convert has more options but not always as performant.

[Auspex]

I prefer this over the CONVERT() syntax as it's more standards compliant (though, of course, CURRENT_TIMESTAMP)

[Aubrey Love]

Just my 2 cents worth, but the first SELECT statement: “select convert(getdate() as date)” does not work. The “date” and “getdate()” should have been inverted like in the code below. “SELECT CONVERT(DATE, GETDATE() )” You could also use the “CAST()” function as well. Either one will return the same results. “SELECT CAST(GETDATE() AS DATE)”

[variable]

In SQL 2019, what is preferred - this answer or CAST or CONVERT?

[Jeff Moden]

Using FORMAT in any case in SQL Server is at least 20 times slower than even some of the craziest conversions that you can think of. I'd stay away from it for everything.