SQL - 查询 - max（count（））

Question

I'm studying for my Database Systems Exam (tomorrow) and I'm having trouble in exercises where I'm asked to write queries. Here's an example:

I am being asked to write a query to answer the following question: Among the authors with lowest Age, who has most books written?

The problem is my teacher forbids me to use sub-queries inside FROM clause, and to use TOP .

I have written an answer, but one I know is incorrect:

SELECT W.AName, COUNT(W.ID_B) AS NBooks
FROM Write W, Author A1
WHERE (A1.AName = W.AName) AND
      (A1.AAge = (SELECT MIN(A2.Age)
                  FROM Author A2))
GROUP BY W.AName
ORDER BY NBooks DESC

This one gives all the authors with the lower age, and their respective number of books written (I hope..). The correct answer should be only the first line of this one.

Let me be clear:

Table Author
AName    | AAge
---------------
John     | 25
Rick     | 30
Sean     | 26
Lena     | 25

Table Writes
AName    | ID_B
---------------
John     | 2
Lena     | 1
John     | 3
Lena     | 4
Rick     | 5
Rick     | 6
Lena     | 6
Rick     | 7
Rick     | 8

(notice that Sean did't write any book, book nº6 has 2 authors, and Rick is the author with most books (4) )

Now, the code I wrote above gives this result (I guess):

AName    | NBooks
-----------------
Lena     | 3
John     | 2

(The lowest age is 25 and both Lena and John are 25)

Whats asked is:

AName    | NBooks
-----------------
Lena     | 3

(Lena is the Author, among all the authors with the lowest age (25), with most books written)

Thanks in advance

Answer 1

Because you are a student, I'll answer part of the question. Here is an answer, ignoring the youngest part:

select a.AName, COUNT(*) as NumBooks
from Author a join
     Write w
     on a.AName = w.AName
group by a.AName
having count(*) >= all(select COUNT(*) as NumBooks
                       from write w
                       group by w.AName
                      )

I think you can figure out how to modify it.

By the way, the restriction on limit and top is, I hope, only for this example. Otherwise, you should get another teacher, since these are very important constructs.

Also, you need to learn conventional join syntax, rather than , in the from clause. Once again, if you teacher is not teaching the modern sytnax (since about 1988), get a new teacher. And, I assume the restriction on subqueries applies to CTEs as well.

I also want to point out the "correct" version of the query:

select top 1 a.aname, count(*) as NumBooks
from Author a join
     Write w
     on a.AName = w.AName
group by author.name, author.Aage
order by author.Age asc, count(*) desc

This query is better than the query above along almost any dimension. It does one join , one group by and one sort. The complete version of my query does two join s explicitly, two join s implicitly (the age clause), and two group by s. The former is going to have better performance than the latter.

From a readability perspective, this version is shorter and cleaner. I also think that it is much easier to teach what this is doing, rather than the "unusual" constructs in the first version. Most students will understand what top and order by are doing and can emulate this. Mimicking what happens in that having clause requires some mental gymnastics.

If you want to get all the authors with the maximum count, the first thing is to realize that the previous query is equivalent to:

select aname, NumBooks
from (select a.aname, count(*) as NumBooks,
             row_number() over (partition by author.Name order by a.aAge, count(*) desc) as seqnum
      from Author a join
           Write w
           on a.AName = w.AName
      group by author.name, author.Aage
     ) aw
where seqnum = 1

Switching this to get all the authors is easy:

select aname, NumBooks
from (select a.aname, count(*) as NumBooks,
             dense_rank() over (partition by author.Name order by a.aAge, count(*) desc) as seqnum
      from Author a join
           Write w
           on a.AName = w.AName
      group by author.name, author.Aage
     ) aw
where seqnum = 1

This is also more efficient than the query that answers the question. Not being able to use top or subqueries in the from clause is like running a three-legged race. Yes, you can probably get there, but you'll get there much faster running on your own two legs.

Answer 2

This is some restriction, but it makes one use it's creativity.

So you want one of the youngest authors, that has written a number of books that is higher than (or equal to) any other number of books written by another of the youngest authors...

SELECT
  [a1].[AName],
  [a1].[AAge],
  COUNT(*) AS [NBooks]
FROM [Author] [a1], [Writes] [w1]
WHERE 
  [a1].[AName] = [w1].[AName]
  AND [a1].[AAge] = (SELECT MIN([a2].[AAge]) FROM [Author] [a2])
GROUP BY 
  [a1].[AName],
  [a1].[AAge]
HAVING COUNT(*) >= ALL
  (SELECT
    COUNT(*) AS [NBooks]
  FROM [Author] [a3], [Writes] [w2]
  WHERE 
    [a3].[AName] = [w2].[AName]
    AND [a3].[AAge] = (SELECT MIN([a4].[AAge]) FROM [Author] [a4])
    AND [a3].[AName] <> [a1].[AName]
  GROUP BY 
    [a3].[AName],
    [a3].[AAge])

PS: Got to admit, I learned about ALL from Gordon Linoff .

Answer 3

If you only want one result select a top one and the ordering should do the rest. I personally would do a ranking function to explicitly get rankings using the Aggregate() Over() windowed function. But since you are learning maybe they don't want to bring that up yet and show you how the 'top' works.

declare @Person Table ( personID int identity, person varchar(8), age int);

insert into @Person values ('Brett', 34),('John', 34),('Peter', 52);

declare @Books Table ( BookID int identity, personID int);

insert into @Books values (1),(1),(1),(2),(2),(3)

Select top 1 -- TOP WILL LIMIT TO CHOICE YOU WANT BASED ON ORDER BY CLAUSE
    p.person
,   p.age
,   count(b.BookID) as cnts
from @Person p, @Books b
where p.personID = b.personID
group by p.person, p.age
order by age, cnts desc

Answer 4

I understand that you just want 1 row as a result;

You could limit the Author first and then by using inner join you could retrieve his name and book count from Write table.

SELECT W.AName, COUNT(W.ID_B) AS NBooks
FROM Write W INNER JOIN Author A1 ON A1.AName = W.AName
WHERE 
A1.AName = (SELECT AName FROM Write GROUP BY AName ORDER BY COUNT(ID_B) DESC)
AND A1.AAge = (SELECT MIN(A2.Age) FROM Author A2)
GROUP BY W.AName
ORDER BY NBooks DESC

Answer 5

If you're allowed to use CTE's and RANK its trival.

WITH cte 
     AS (SELECT a.aname, 
                A.aage, 
                Count(id_b)                             Book_Count, 
                RANK() 
                  OVER( 
                    ORDER BY a.aage, Count(id_b) DESC ) rn 
         FROM   author a 
                INNER JOIN writes w 
                        ON a.aname = w.aname 
         GROUP  BY a.aname, 
                   a.aage) 
SELECT aname, 
       Book_Count
FROM   cte 
WHERE  rn = 1 

SQL Fiddle

Demo Where John writes another book