How to Search for Date and Time Values Using SQL Server 2000 - The Effect of Database Design on Querying
(Page 5 of 8 )
Database designers don’t always use date/time columns appropriately. At the time the database is designed, each date/time column should be identified as to whether it will store both dates and times, dates only, or times only. The designer, by using defaults, constraints, and triggers, can enforce these rules to prevent the accidental storage of data that are either unnecessary or not applicable.
For example, a column in an accounts payable system for the date an invoice is received is unlikely to need the time. In that case, the designer should plan to use the column solely for dates and never store the time component. A trigger could be assigned to prevent the non-integer portion of the date value from being stored when updating or inserting.
Generally, however, the programmer is forced to work with an existing database. In this case, you should examine the way in which the date/time values are being used before you assume the designer did his or her job correctly.
The simplest way to do that is to submit a query using a search condition similar to the following, where DT is the date/time column in question:
WHERE CAST(FLOOR(CAST(DT AS float))AS datetime) = 0 OR
DT - CAST(FLOOR(CAST(DT AS float))AS datetime) = 0
Note: The FLOOR function returns the largest integer that is less than or equal to the specified value. In this expression, FLOOR is applied to the floating-point representation of the DT column. This simply strips off the fractional portion of the number.
The first expression returns the date (integer) portion of the value, while the second returns the time portion. If this query returns no rows, it’s likely that the column has been used consistently to store both dates and times, since the date is never 0 and the time is never 0.
Keep in mind, of course, that if the above query returns rows, it doesn’t necessarily imply that the column has been used inconsistently. If the time happens to be exactly midnight or the date happens to be January 1, 1900, then it’ll show up in the result set. In that case, you can test for columns with time-only or date-only data by using either of these two queries:
WHERE TOnly <> Tonly - (CAST(FLOOR(CAST(TOnly AS float))AS datetime))
WHERE DOnly <> CAST(FLOOR(CAST(DOnly AS float))AS datetime)
Here, TOnly and DOnly are date/time columns that you expect contain only times or dates, respectively. If the query returns rows, then those rows don’t contain the type of data you expected.
Determining what kind of data are stored in the date/time columns of each table is important for intelligent querying. If the columns are used consistently, then your job is easier. However, even if the columns are used inconsistently, you’ll at least know which query pitfalls to watch out for as you code your queries.
Next: Performance Considerations in Querying >>
More SQL Server Articles
More By Bryan Syverson
/ 102 
