How can I do the impossible?

Suppose you're tasked to report on sales figures for the company's sales force. The report has to include the individuals' percentile contribution to the team, region, and company sales in addition to the raw sales numbers. Team percentile contribution to region and total company sales and region percentile contribution to total sales are also needed. You could run multiple queries—one to get the company totals, another to calculate the region totals, a third for the team totals, and finally a detail query by salesperson. Then you would have to perform the matchups of salesperson to team and region and then team to region to calculate the percent contribution for each of those and against the company total sales all in application code. No sweat, right?

Another approach would be to insert correlated subqueries into the projection clause of the query and fetch the percentages that way, but that would be six correlated subqueries across hundreds of salespeople in dozens of teams and regions. This would be the quintessential query from hell!

OK, just to throw a monkey wrench into the mix, C-suite management wants a hard copy report for the Board of Directors, and the sales manager wants an intranet Web page showing the same data but sortable on various columns, and the numbers have to match the hard copy. Oh, and don't forget that the region figures have to go into the annual report for the shareholders, so you'll have to export the data to send to the PR team designing the annual report!

First you think, "Impossible!" Then you think, "I like my job! Or, at least I did before this request came in. So, I have to figure out a way to do this in a single place that can produce identical data for all of these delivery formats. There must be a way to do this in a single SQL statement!" Then you give it a try and rediscover that you cannot mix aggregation data with detail data in a single query. You are back to "Impossible!" Even you can't do the impossible...or can you?

Enter one of the most overused and misused features of SQL '92, derived tables in the FROM clause of a query used as a table entity. The ANSI SQL '92 standard provides the following definition:

A derived table is a table derived directly or indirectly from one or more other tables by the evaluation of a <query expression>. The values of a derived table are derived from the values of the underlying tables when the <query expression> is evaluated.

In effect, the result of any query is a derived table. This agrees with Codd's relational algebra definition that the result of combinations and selections from relations or tuples are themselves relations or tuples.

Enough tech talk, back to brass tacks. When most folk talk about derived tables, they mean using the results of a SELECT query in the FROM clause of a query as if it were a table. Simple example:

SELECT *
FROM ( SELECT col_1, col_2, col_3 FROM sometable) AS d
WHERE ….

Why not just "SELECT * FROM sometable WHERE ...", you ask? So have I. Many times. Too many times! Queries like this "dumb" example are rampant in the real world, but I digress. How can this help with the sales report outlined at the start? You would write queries for each of the aggregation levels and reference them in the FROM clause as derived table expressions. Here's how:

SELECT sm.fname, sm.lname, sm.team_id, sm.region_id, sp.sales,
       ((sp.sales/tm.sales)*100.00) as team_contribution,
       ((sp.sales/rs.sales)*100.00) as region_contribution,
       ((sp.sales/cw.sales)*100.00) as company_contribution,
       ((tm.sales/rs.sales)*100.00) as team_to_region,_contribution
       ((tm.sales/cw.sales)*100.00) as team_to_company_contribution,
       ((rs.sales/cw.sales)*100.00) as region_to_company_contribution,
       …
FROM sales_people as sm
JOIN sales_performance as sp
   ON sm.sales_id = sp.sales_id
JOIN (
         SELECT team_id, sum( sales )
         FROM sales_performance
         GROUP BY team_id
       ) AS tm( team_id, sales )
   ON sm.team_id = tm.team_id
JOIN (
         SELECT region_id, sum( sales )
         FROM sales_performance
         GROUP BY region_id
       ) AS rs( region_id, sales )
   ON sm.region_id = rs.region_id
JOIN (
         SELECT sum( sales )
         FROM sales_performance
       ) AS cw( sales )
ORDER BY sm.lname, sm.fname, sm.region_id, sm.team_id;

To further simplify the various reports and screens that must display these same results, you could create a VIEW based on this query and reference the VIEW in your applications, screens, and reports. VIEW is the common name for the entity type that the SQL standard refers to as a viewed table or a named derived table.

Anyway, this lesson is finished in more than one way. We've defined and described derived tables and used non-trivial ones in the FROM clause of a query in order to achieve "the impossible". All in all, a good day's work. Back to loving your job!

Also Read

SQL 101: DDL Recap—From Physical Files to SQL Tables, Starting the Journey

TechTip: Generate and Email a Spreadsheet with SQL

Practical SQL: Journal Forensics Using SQL