SQL：通过将子表连接到另一个表来计算百分比

Question

I've the above dataset, I need to report for each year the percentage of movies in that year with only female actors, and the total number of movies made that year. For example, one answer will be: 1990 31.81 13522 meaning that in 1990 there were 13,522 movies, and 31.81%

In order to get the moves with only female actors, wrote the following code:

SELECT a.year as Year, COUNT(a.title) AS Female_Movies, a.title 
FROM Movie a
WHERE a.title NOT IN (

  SELECT b.title from Movie b
  Inner Join M_cast c
  on TRIM(c.MID) = b.MID
  Inner Join Person d
  on TRIM(c.PID) = d.PID
  WHERE d.Gender='Male'
  GROUP BY b.title
  )
GROUP BY a.year,a.title
Order By a.year asc

The total movies in each year , can be found using the following:

SELECT a.year, count(a.title) AS Total_Movies
FROM Movie a
GROUP BY a.year
ORDER BY COUNT(a.title) DESC

Combinig the both I wrote, the following code:

SELECT z.year as Year, count(z.title) AS Total_Movies, count(x.title) as Female_movies, count(z.title)/ count(x.title) As percentage
FROM Movie z
Inner Join (
SELECT a.year as Year, COUNT(a.title) AS Female_Movies, a.title 
FROM Movie a
WHERE a.title NOT IN (

  SELECT b.title from Movie b
  Inner Join M_cast c
  on TRIM(c.MID) = b.MID
  Inner Join Person d
  on TRIM(c.PID) = d.PID
  WHERE d.Gender='Male'
  GROUP BY b.title
  )
GROUP BY a.year,a.title
Order By a.year asc
)x
on x.year = z.year 
GROUP BY z.year
ORDER BY COUNT(z.title) DESC

However, in th output I'm seeing the years with only female movies correctly, but the count of total movies is equal to female_movies so I'm getting 1%, I tried debugging the code, but not sure where this is going wrong. Any insights would be appreciated.

Answer 1

You assume that your 'z' contains all movies but since you do an inner join on the female movies, they'll also only contain female movies. You could fix that with a 'left join'.

Assuming your two queries are correct, you can join on them with a 'WITH' like this:

WITH allmovies (year, cnt) as
(SELECT a.year, count(a.title) AS Total_Movies
 FROM Movie a
 GROUP BY a.year
 ORDER BY COUNT(a.title) DESC)
,

femalemovies (year, cnt, title) as
(SELECT a.year as Year, COUNT(a.title) AS Female_Movies, a.title 
FROM Movie a
WHERE a.title NOT IN (

  SELECT b.title from Movie b
  Inner Join M_cast c
  on TRIM(c.MID) = b.MID
  Inner Join Person d
  on TRIM(c.PID) = d.PID
  WHERE d.Gender='Male'
  GROUP BY b.title
  )
GROUP BY a.year,a.title
Order By a.year asc)
select * from allmovies left join femalemovies on allmovies.year = femalemovies.year

Answer 2

You can use conditional aggregation. In a CASE expression check if no cast member that isn't female exists with a correlated subquery. If the check is successful, return something not NULL and count() that to get the number of movies with only female cast members (or none at all).

SELECT m.year,
       count(*) count_all,
       count(CASE
               WHEN NOT EXISTS (SELECT *
                                       FROM m_cast c
                                            INNER JOIN person p
                                                       ON p.pid = c.pid
                                       WHERE c.mid = m.mid
                                             AND p.gender <> 'Female') THEN
                 1
              END)
       /
       count(*)
       *
       100 percentage_only_female
       FROM movie m
       GROUP BY m.year;

Since in MySQL Boolean expressions in numerical context evaluate to 1 if true and to 0 otherwise, you could also use a sum() over the NOT EXISTS .

SELECT m.year,
       count(*) count_all,
       sum(NOT EXISTS (SELECT *
                              FROM m_cast c
                                   INNER JOIN person p
                                              ON p.pid = c.pid
                              WHERE c.mid = m.mid
                                    AND p.gender <> 'Female'))
       /
       count(*)
       *
       100 percentage_only_female
       FROM movie m
       GROUP BY m.year;

That however isn't compatible with most other DBMS in contrast to the first one.

Answer 3

I would use two levels of aggregation:

SELECT m.MID, m.title, m.year,
       COUNT(*) as num_actors,
       SUM(gender = 'Female') as num_female_actors
FROM Movie m JOIN
     M_cast c
     ON c.MID = b.MID JOIN
     Person p
     ON p.PID = c.PID
GROUP BY m.MID, m.title, m.year;

Then a simple outer aggregation:

SELECT year,
       COUNT(*) as num_movies,
       SUM( num_actors = num_female_actors ) as num_female_only,
       AVG( num_actors = num_female_actors ) as female_only_ratio
FROM (SELECT m.MID, m.title, m.year,
             COUNT(*) as num_actors,
             SUM(gender = 'Female') as num_female_actors
      FROM Movie m JOIN
           M_cast c
           ON c.MID = b.MID JOIN
           Person p
           ON p.PID = c.PID
      GROUP BY m.MID, m.title, m.year
     ) m
GROUP BY year;

Notes:

• Use meaningful table aliases, rather than arbitrary letters. You'll note that the table aliases are abbreviations for the table names.
• Do not use functions when filtering or JOIN ing unless necessary. I removed the TRIM() . If you need it use it. Or better yet, fix the data.

Answer 4

              SELECT m.Year,COUNT(m.Year),x.t, 
              (COUNT(m.Year)*1.0/x.t*1.0)*100
             FROM Movie m LEFT  JOIN
             (SELECT Year,COUNT(Year) AS t FROM Movie GROUP BY year) AS x

             ON  m.Year=x.Year
             WHERE m.MID IN
             (SELECT MID FROM M_Cast WHERE PID in
             (SELECT PID FROM Person WHERE Gender='Female') 
              AND m.MID NOT IN
             (SELECT MID FROM M_Cast WHERE PID in
             (SELECT PID FROM Person WHERE Gender='Male'))) GROUP BY m.year    

Answer 5

Check if this is what you're looking for.

select movie.year, count(movie.mid) as Year_Wise_Movie_Count,cast(x.Female_Cast_Only as real) / count(movie.mid) As Percentage_of_Female_Cast from movie
inner join
(
SELECT Movie.year as Year, COUNT(Movie.mid) AS Female_Cast_Only
FROM Movie
WHERE Movie.MID NOT IN (
  SELECT Movie.MID from Movie
  Inner Join M_cast
  on TRIM(M_cast.MID) = Movie.MID
  Inner Join Person
  on TRIM(M_cast.PID) = Person.PID
  WHERE Person.Gender!='Female'
  GROUP BY Movie.MID
  )
GROUP BY Movie.year
Order By Movie.year asc
) x
on x.year = movie.year 
GROUP BY movie.year
ORDER BY movie.year

Output:

year  Year_Wise_Movie_Count  Percentage_of_Female_Cast
----  ---------------------  -------------------------
1939  2                      0.5
1999  66                     0.0151515151515152
2000  64                     0.015625
2018  104                    0.00961538461538462

Note: This was executed in SQLIte3