How to Search for Date and Time Values Using SQL Server 2000 - How Dates and Times are Stored in SQL Server
(Page 2 of 8 )
Before you can effectively query date/time (or temporal) data, you have to know something about how date/time values are stored. SQL Server supports two date/time data types: datetime and smalldatetime. The difference between the two is the amount of storage used. Datetime uses 8 bytes of storage, while smalldatetime uses only 4 bytes. For this reason, datetime can represent date/time values within a wider range and with more precision than smalldatetime. These differences are summarized in the table below.
| Type | Minimum | Maximum | Precision |
| datetime | Jan 1, 1753
midnight | Dec 31, 9999
23:59:59.997
(0.003 seconds until midnight) | To the nearest
3.33 milliseconds |
| smalldatetime | Jan 1, 1900
midnight | Jun 6, 2079
23:59
(1 minute until midnight) | To the nearest
minute |
Both datetime and smalldatetime represent the date and time as a value that’s equal to the number of days in relationship to a base date. In SQL Server, that base date is midnight on January 1, 1900. As you can see in the table, the smalldatetime type can only represent dates from this base date on. In contrast, the datetime type can also represent dates that are before January 1, 1900. To do that, it stores those values as negative numbers.
To visualize how date/time values are stored, you can think of them as consisting of two parts. The integer portion represents the number of whole days since January 1, 1900. The fractional portion represents the fraction of a day that’s passed since midnight. For example, the date/time value representing noon on January 4, 1900 is stored as 3.5. In this case, 3 represents three full days since the base date and 0.5 represents one half of a day between midnight and noon. To see this, submit the following query:
SELECT CAST(CAST('1900-01-04 12:00' AS datetime) AS float)
Note: The CAST function explicitly changes the data type of a value as specified. So in this statement, the inner CAST changes the string literal '1900-01-04 12:00' to a value of data type datetime. Then, the outer CAST changes that datetime value to a value of data type float. The final result is a floating-point representation of the datetime value that represents noon on January 4, 1900.
So far, so good. But the problems that crop up in querying date/time data are caused by confusion over two fundamental facts that aren’t so obvious. First, date/time data types are approximate numerics, not exact numerics. Second, date/time data types can’t store a date without a time or a time without a date.
Next: Date/Time Values are Approximate Numerics >>
More SQL Server Articles
More By Bryan Syverson
/ 108 
