Dates, we store them everywhere, DateOrdered, DateEntered, DateHired, DateShipped, DateUpdated, and on and on it goes. Up until and including SQL Server 2005, you really didn’t have much choice about how you stored your date values. But in SQL Server 2008 and higher you have alternatives to DateTime and they are all better than the original.
DATETIME stores a date and time, it takes 8 bytes to store, and has a precision of .001 seconds
In SQL Server 2008 you can use DATETIME2, it stores a date and time, takes 6-8 bytes to store and has a precision of 100 nanoseconds. So anyone who needs greater time precision will want DATETIME2. What if you don’t need the precision? Most of us don’t even need milliseconds. So you can specify DATETIME2(0) which will only take 6 bytes to store and has a precision of seconds. If you want to store the exact same value you had in DATETIME, just choose DATETIME2(3), you get the same precision but it only takes 7 bytes to store the value instead of 8. I know a lot of you are thinking, what’s one byte, memory is cheap. But it’s not a question of how much space you have on your disk. When you are performance tuning, you want to store as many rows on a page as you can for each table and index. that means less pages to read for a table or query, and more rows you can store in cache. Many of our tables have multiple date columns, and millions of rows. That one byte savings for every date value in your database is not going to make your users go ‘Wow everything is so much faster now’, but every little bit helps.
If you are building any brand new tables in SQL Server 2008, I recommend staying away from DATETIME and DATETIME2 altogether. Instead go for DATE and TIME. Yes, one of my happiest moments when I first started learning about SQL Server 2008 was discovering I could store the DATE without the time!! How many times have you used GETDATE() to populate a date column and then had problems trying to find all the records entered on ‘05-JUN-06’ and got no results back because of the time component. We end up truncating the time element before we store it, or when we query the date to ignore the time component. Now we can store a date in a column of datatype DATE. If you do want to store the time, store that in a separate column of datatype TIME. By storing the date and time in separate columns you can search by date or time, and you can index by date and or time as well! This will allow you to do much faster searches for time ranges.
Since we are talking about the date and time datatypes, I should also mention that there is another date datatype called DATETIMEOFFSET that is time zone aware. But that is a blog for another day if you are interested.
Here is a quick comparison of the different Date and Time Data types,
|Datatype||Range||Precision||Nbr Bytes||User Specified Precision|
|SMALL DATETIME||1900-01-01 to 2079-06-06||1 minute||4||No|
|DATETIME||1753-01-01 to 9999-12-31||.001 seconds||8||No|
|DATETIME2||0001-01-01 to 9999-12-31 23:59.59.9999999||100 ns||6-8||Yes|
|DATE||0001-01-01 to 9999-12-31||1 day||3||No|
|TIME||00:00:00.0000000 to 23:59.59.9999999||100 ns||3-5||Yes|
|DATETIMEOFFSET||0001-01-01 to 9999-12-31 23:59.59.9999999||100 ns||8-10||Yes|
Today’s My 5 is of course related to the Date and Time datatypes.
My 5 Important Date functions and their forward and backwards compatibility
- GETDATE() – Time to STOP using GETDATE(), it still works in SQL Server 2008, but it only returns a precision of milliseconds because it was developed for the DATETIME datatype.
- SYSDATETIME() – Time to START using SYSDATETIME(), it returns a precision of nanoseconds because it was developed for the DATETIME2 datatype and it also works for populating DATETIME columns.
- DATEDIFF() – This is a great little function that returns the number of minutes, hours, days, weeks, months, or years between two dates, it supports the new date datatypes.
- ISDATE() – This function is used to validate DATETIME values. It returns a 1 if you pass it a character string containing a valid date. However if you pass it a character string representing a datetime that has a precision greater than milliseconds it will consider this an invalid date and will return a 0.
- DATEPART() – This popular function returns a portion of a date, for example you can return the year, month, day, or hour. This date function supports all the new date datatypes.
Also one extra note, because I know there are some former Oracle developers who use this trick. If you have any select statements where you select OrderDate+1 to add one day to the date, that will not work with the new date and time datatypes. So you need to use the DATEADD() function.
This blog is also posted to the Canadian Solution Developer