在同一表上的SQL嵌套查询

Question

I need to select every space a company has where its area is bigger than average per company. Here's my query:

SELECT *
FROM Space outer
WHERE area >
( SELECT AVG (area)
FROM Space inner
WHERE outer.address = inner.address );

MySQL returns a syntax error starting at area>. I know I shouldn't have used address as primary key, and I can't use JOIN. It's just an assignment. Please, help

Answer 1

First derive the average space across all records...

(Select avg(area) mavg from space)

This returns a single value that we can cross join to all space records and then just check for area > average.

Select * 
from Space
cross join (Select avg(area) mavg from space) Co
where area > Co.mavg

Since we know the result from the derived table/inline view will only 1 value every time, the cross join doesn't increase the number of rows being evaluated; and the rdbms only has to evaluate the average once.

However this assumes you want average across all records and not just across a company. If it is by company... then something like...

Select S.* 
from Space S
LEFT join (Select address, avg(area) mavg from space Group by address) Co
  on S.address= Co.address
where S.area > Co.mavg

This determines the average by company joins back to space on that company and then compares each space record for the company to the company's average.

Since we don't know how you define company in terms of data, I just assumed a "address" field.

a different approach...

Select S.* 
from space S
where S.area > (Select avg(area) from space)

However this assumes average across all companies

or

Select S.*
from space
where s.avg > (Select avg(area) from space S2 where S.Company = S2.company)

If this doesn't do it I would need to see the DDL of the table space (The structure columns, data types PK, FKs etc) and some sample data.

unless address is the same for every company area.... there's must be some other criteria to relate a company to all the other company records (a name perhaps or a consistent key?)

Personally I find the use of a correlated sub query in this case slow as it has to calculate the average for every record in Space. The cross join IMO would be more efficient.