SQL BQ Return 用户评价最好和最差的电影

Question

我有以下格式的数据：

user_id	user_name	movie_rating	movie_name
1	x	1	asd1
1	x	3	asd2
1	x	5	asd3
1	x	5	asd4
2	y	2	asd4
2	y	3	asd5
2	y	4	asd6

我想以 best_rated_movie 和 worst_rated_movie 为数组的格式提取每个用户评价最好和最差的电影（因为用户可以有多部最差的电影等）：

user_id	user_name	best_rated_movie	worst_rated_movie
1	x	(asd3,asd4)	asd1
2	y	asd6	asd4

我设法提取 table 输出我只有最好和最差评价的电影：

user	user_name	movie_rating	movie_name
1	x	1	asd1
1	x	5	asd3
1	x	5	asd5
2	y	2	asd4
2	y	4	asd6

这是我使用的代码：

with best_movie_rating as (
  select
    user_id,
    max(movie_rating) as max_rating
  from source_table
  group by user_id
),

worst_movie_rating as (
  select
    user_id,
    min(movie_rating) as min_rating
  from source_table
  group by user_id
),
columns_final as (
  select
    t1.user_id,
    t1.user_name,
    t1.company_name,
    t1.movie_rating,
    t1.movie_name 
  from source_table t1
  inner join best_movie_rating t2
    on t1.user_id = t2.user_id 
  inner join worst_movie_rating t3
    on t1.user_id = t3.user_id 
  where (t1.movie_rating= t2.max_rating and t1.user_id = t2.user_id)
    or (t1.movie_rating= t3.min_rating and t1.user_id = t3.user_id)
)
select * from columns_final

不幸的是我不知道如何前进，我尝试过聚合但没有成功（尤其是在使用数组的情况下）。至少我会很感激如何处理这个问题的建议。

Answer 1

您可以像下面这样使用 string_agg()：

with columns_final as (
  select
    t1.user_id,
    t1.user_name,    
    t1.movie_rating,
    t1.movie_name ,
    max (movie_rating)over (partition by user_id ) max_rating,
    min (movie_rating)over (partition by user_id ) min_rating
  from source_table t1  
)

select user_id,user_name,string_agg(case when movie_rating=max_rating then movie_name end,',')best_rated_movie,
string_agg(case when movie_rating=min_rating then movie_name end,',')worst_rated_movie
from columns_final   
where movie_rating =max_rating or movie_rating=min_rating
group by user_id,user_name

我还对您的查询进行了一些更改以使其紧凑。我使用 window 函数来计算最大值和最小值 movie_rating.

，而不是两个常见的 table 表达式

Answer 2

I want to extract the best and the worst rated movie per user in a format where best_rated_movie and worst_rated_movie is array (because user can have multiple worst movies etc)

考虑以下

select user_id, user_name,
  array_concat_agg(movies order by movie_rating desc limit 1) best_rated_movie,
  array_concat_agg(movies order by movie_rating limit 1) worst_rated_movie,
from (
  select user_id, user_name, movie_rating, array_agg(movie_name)  movies, 
  from `project.dataset.table` t
  group by user_id, user_name, movie_rating
)
group by user_id, user_name

如果应用于您问题中的示例数据 - 输出为

如果（正如其他答案所暗示的那样）您希望结果是一串以逗号分隔的电影 - 您可以在下面使用

select user_id, user_name,
  string_agg(movies order by movie_rating desc limit 1) best_rated_movie,
  string_agg(movies order by movie_rating limit 1) worst_rated_movie,
from (
  select user_id, user_name, movie_rating, string_agg(movie_name)  movies, 
  from `project.dataset.table` t
  group by user_id, user_name, movie_rating
)
group by user_id, user_name

有输出

SQL BQ Return 用户评价最好和最差的电影

SQL BQ Return user's best and worst rated movie

sql

google-bigquery

string-agg