Section 5.10. Querying Periods and Respecting Granularity

   

5.10 Querying Periods and Respecting Granularity

5.10.1 Problem

As in the previous recipe, you want to find all periods that include a particular date. However, you do not entirely trust the data within your database. Programmers are supposed to ensure that DATETIME values contain only dates (not times), but your experience has been that many DATETIME values do contain time-of-day components . You wish to write your query to avoid problems from time-of-day values.

With respect to our example, the date you are interested in is January 11, 2001, and you wish to find all contractors who are currently scheduled to work on that date.

5.10.2 Solution

Use the following query, which carefully avoids applying any functions to the two database columns , JobStart and JobEnd:

 SELECT     JobId, ContractorId, JobStart, JobEnd FROM ContractorsSchedules WHERE JobStart < (CAST('2001-1-11' AS DATETIME)+1)   AND JobEnd >= CAST('2001-1-11' AS DATETIME) JobId      ContractorId JobStart                    JobEnd                       ---------- ------------ --------------------------- ---------------------------  RF10001    Alex         2001-01-11 00:00:00.000     2001-01-20 00:00:00.000 RF10003    Bob          2001-01-05 00:00:00.000     2001-01-15 00:00:00.000 

5.10.3 Discussion

Why such a convoluted query? And why the need to avoid applying functions to database columns? Let's answer one question at a time. First, let's talk about the complexity. Your first thought when faced with the task of writing such a query that involves date ranges might be to use the following, simpler approach:

 SELECT     JobId, ContractorId, JobStart, JobEnd FROM ContractorsSchedules WHERE JobStart <= '2001-1-11'   AND JobEnd >= '2001-1-11' 

This approach is perfectly fine except for the implicit assumption that the time-of-day components of JobStart and JobEnd are always 00:00:00.000. But, if you had a JobStart date such as the one in the following record:

 RF10001    Alex         2001-01-11 15:30:00.000     2001-01-20 00:00:00.000 

the WHERE condition that SQL Server would evaluate would be:

 JobStart <= '2001-1-11' '2001-01-11 15:30:00.000' <= '2001-01-11 00:00:00.000' 

The first line represents the condition as you wrote it, while the second line represents the actual values, after the implicit conversion of your input date, that SQL Server evaluates . Clearly, this condition will fail, resulting in the exclusion of job RF10001 from the query results. Yet, the job does begin on January 11, 2001, and you really do want it to appear in the list of jobs underway as of that date. What then, do you do?

Now you might think to apply a function to the JobStart date in the database to truncate any time-of-day component. For example:

 CONVERT(CHAR(10),JobStart,120) <= CAST('2001-1-11' AS DATETIME) 

This approach will work, but it has the unfortunate side effect of applying a function (CONVERT in this case) to a database column. That potentially precludes the use of any index that happens to exist on the column in question. For large tables, this can be a significant performance issue. If you happen to index the JobStart column and you wish to allow SQL Server the possibility of using that index, you'll need to take an approach such as the following:

 JobStart < (CAST('2001-1-11' AS DATETIME)+1) 

Here, the JobStart column is left untouched. Instead, we manipulate the date value that we are supplying to the query. First, we CAST our date into a DATETIME value. Then, we add 1 to the result. Because we have been sure not to include any time-of-day component in our constant, the result will be the first instant of the following day:

 (CAST('2001-1-11' AS DATETIME)+1) = '2001-1-12 00:00:00.000' 

The maximum granularity of a DATETIME value is in milliseconds . There can be no earlier time recorded for our DATETIME value than 00:00:00.000. This means that any DATETIME value that is less than (but not less than or equal to) the value returned by our expression must, in fact, fall sometime during or before January 11, 2001. Thus, we use the less-than operator (<) in our WHERE condition.

For the JobEnd date, we still use greater-than-or-equal (>=) operator:

 AND JobEnd >= CAST('2001-1-11' AS DATETIME) 

The comparison that SQL Server makes in this case is:

 AND JobEnd >= '2001-1-11 00:00:00.000' 

Again, our expression evaluates to the earliest possible time of day on January 11, 2001. Any time-of-day component associated with a JobEnd date of January 11, 2001 must, by necessity, be equal to or greater than the result of our expression. Hence, we use the greater-than-or-equal (>=) operator in this case.

The result of our careful approach is a query that ignores any time-of-day values for JobStart and JobEnd, does not preclude the use of any indices on those columns, but still returns any jobs underway as of our specified date.

   


Transact-SQL Cookbook
Transact-SQL Cookbook
ISBN: 1565927567
EAN: 2147483647
Year: 2005
Pages: 152

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net