Craig S. Mullins |
|
April 2000 |
|
|
Creating SQL Results Sets with Aggregate and Detail Data Every SQL manipulation statement operates on a table and results in another table. All operations native to SQL, therefore, are performed at a set level. One retrieval statement can return multiple rows; one modification statement can modify multiple rows. This feature of relational databases is called relational closure. Relational closure is the major reason that relational databases such as SQL Server generally are easier to maintain and query than other types of databases. Additionally, SQL specifies what data to retrieve or manipulate, but does not specify how to accomplish these tasks. This keeps SQL intrinsically simplistic. If you can remember the set-at-a-time orientation of SQL Server, and that your queries specify "what" to retrieve and not "how" to retrieve it, then you grasp the essence and nature of SQL. The capability to act on a set of data coupled with the lack of need for establishing how to retrieve and manipulate data defines SQL as a non-procedural language. SQL Server also provides a series of functions that can be used to aggregate data. Certain column functions can be used to compute, from a group of rows, a single value for a designated column or expression. For example, the sum function can be used to add values in multiple rows, returning the sum of the values instead of each individual value. Details and Aggregates Now, taking the previous discussion into consideration, what if you need to return detail data and aggregated data in the same results set using SQL. For example, consider the sample pubs database that comes with Microsoft SQL Server. The sales table contains information about each title sold (refer to Figure 1 for a sample of the data contained in the sales table). Figure 1. Sample sales table data. To return a simple list of the detail information contained in this table you could issue the following query: select stor_id, ord_num, ord_date, qty, But what if, instead of just returning qty on each row, you also wished to return the average quantity, maximum quantity, and minimum quantity. This is not quite so easy. You can use the avg, max, and min functions to return the average, maximum, and minimum qty of a sale by store as follows: select stor_id, avg(qty), max(qty), min(qty) This query returns a single row for each store containing the average, maximum, and minimum sale. It does not contain any detail information, only aggregated information. However, there are times when it is desirable for a single query to return both detail and aggregate information. For example, what if you wished to compare each sale qty to the average sale qty? There are several ways to do this, but it is desirable to be able return both values in a single query to simplify your application. One solution is to create a view. Consider the following view named sales_qty created on the sales table, as shown here:
create view sales_qty as select stor_id, avg(qty), max(qty), min(qty) from sales group by stor_id After the view is created, you can issue the following select statement joining the view to the base table, thereby providing both detail and aggregate information on each report row: select s.stor_id, s.title_id, s.qty, from sales s, sales_qty q where s.stor_id = q.stor_id order by 1 This query returns one row for each sale, listing both the store identifier and title identifier, along with the actual quantity, and the maximum, minimum, and average quantity for that store. The data will be returned in stor_id order. This can be changed simply by changing the order by clause to specify which column the results should be sorted by. Additionally, it is important to remember the meaning of the aggregates returned by the view. The aggregated data is by store, and not by title. Neither does the aggregated data represent the average, minimum, and maximum of all sales (independent of store). This is because that is the way the view was created. Do not forget the definition of the view. View definitions can be retrieved from the SQL Server system catalog tables using the sp_helptext stored procedure as follows:
This will return the actual SQL DDL that was used to create the view. If additional information, such as the publication title is required, you can add the title table to the join as follows: select s.stor_id, s.title_id, t.title, s.qty, from sales s, sales_qty q, titles t where s.stor_id = q.stor_id and s.title_id = t.title_id order by 1 Indeed, once the view is created, you can use it in any way allowable by SQL Server. Using views in this manner you can create any number of detail and aggregate combination reports that your organization may desire. Other Approaches Of course, this is only one example of how to return detail and aggregate data in a single row. You could also use a temporary table. A temporary table is created by prefacing the table name (in the create statement) with a number sign (#). Temporary tables are stored in the temporary database, tempdb. The tempdb database is used to hold all temporary tables and any other temporary storage needs. It is a global resource; the temporary tables and stored procedures for all users connected to the system are stored there. SQL Server re-creates tempdb every time it is started. This clears out data so the system starts with a clean copy of tempdb. Since tempdb is be definition, temporary, this is not a problem. So, you could create a table in tempdb as follows:
( stor_id char(4), max_qty smallint, min_qty smallint, avg_qty smallint )
Then you would have to populate the table with data using a bulk insert, for example: insert into #sales_qty group by stor_id) Then you can run the query to retrieve the detail information from sales and the aggregate information from the temporary table as follows: select s.stor_id, s.title_id, s.qty, Views, however, are easier to implement than this method. The advantage of a view over a temporary table is that the view does not require storage. Once created, it is always in sync with the base table(s) that it is defined on. The temporary table will consume storage, but only temporarily. Temporary tables are automatically dropped on disconnect. So even though you will need to create the temporary table only once, you will need to populate it with the aggregated data every time you run a query that needs the information. Furthermore, when tempdb is used you will need to concern yourself with the possibility that the data you need to store in temporary tables exceeds the amount currently allocated to tempdb. For these reasons, views are preferred over temporary tables. Another method, in this case preferable to the view implementation, is to use in-line views. SQL Server supports the capability to code a select statement in the from clause of another select statement. This is often referred to as an in-line view because the SQL statement is placed in-line in the from clause of another SQL select statement. So, you can use an in-line view to embed the aggregate query into the detail query as follows: select s.stor_id, s.title_id, s.qty, from sales group by stor_id) as q (stor_id, avg_qty, max_qty, min_qty) where s.stor_id = q.stor_id order by 1 This returns the same results as the other methods. Refer to Figure 2 for an example of the results of this query. Figure 2. Using an in-line view. This approach is the most complicated because it combines everything into a single query. However, it is also the easiest to implement once you understand the process. Synopsis This article has discussed three different ways to combine detail and aggregate information on a single line in a result set from a SQL Server query. Of the three, the view method is probably the easiest to understand. However, the last one, the in-line view method, is the one with the least administrative overhead because not additional SQL Server objects need to be created. Only the middle method, using temporary tables, is not recommended in practice because of the programming and administrative overhead required to implement. From SQL Server Update (Xephon) April 2000. © 2000 Craig S. Mullins, All rights reserved.
|