背景

假设有一张学生成绩表

sql">create table student_scores (
  id BIGSERIAL PRIMARY KEY,
  name varchar,
  subject varchar,
  exam varchar,
  score int
);

DO $DO$ 
DECLARE 
	name_var student_scores.name%type;
	subject_var student_scores.subject%type;
	exam_var student_scores.exam%type;
	score_var student_scores.score%type;
BEGIN
	FOR name_iter IN 1..10 LOOP 
		FOR subject_iter IN 1..10 LOOP 
			FOR exam_iter IN 1..10 LOOP 
				name_var = 'student_' || name_iter;
				subject_var = 'subject_' || subject_iter;
				exam_var = 'exam_' || exam_iter;
				score_var = ceil(random()*40+60);
				insert into student_scores(name, subject, exam, score) values (name_var, subject_var, exam_var, score_var);
			END LOOP;
		END LOOP;
	END LOOP;
END$DO$;

获取学生在所有考试中subject_1和subject_2的平均分

使用FILTER

sql">select name, 
AVG(score) FILTER (WHERE subject = 'subject_1') as avg_subject_1,
AVG(score) FILTER (WHERE subject = 'subject_2') as avg_subject_2 from student_scores group by name;

使用多一个group by

sql">select name, subject, avg(score) from student_scores where subject in ('subject_1', 'subject_2') group by name, subject;

FILTER优势

在聚合语句中，默认的数据集合是group by筛选出来的所有符合条件的数据，FILTER可以在此基础上再进一步过滤数据。这在很多时候是非常方便的。再看一个例子：获取学生在所有考试中subject_1和subject_2平均分，只考虑subject1高于65，subject2 高于70分的情况

sql">select name, 
AVG(score) FILTER (WHERE subject = 'subject_1' and score > 65) as avg_subject_1,
AVG(score) FILTER (WHERE subject = 'subject_2' and score > 70) as avg_subject_2 from student_scores group by name;

是不是很方便？当然了，不使用filter，也可以通过在group by里面多加一行并且添加where条件的方式达到目的，不过没有FILTER更简洁

使用窗口函数实现topn

现在要求在exam_2这次考试中，各科排名前3的同学

sql">with ranked_students_scores_at_exam2 as (
	select *, rank() over (partition by subject order by score desc) as rank, sum(1) over (partition by subject) as participants, avg(score) over (partition by subject) as avg_score  from student_scores where exam = 'exam_2'
) select * from ranked_students_scores_at_exam2  where rank <= 3;

development=# with ranked_students_scores_at_exam2 as ( select *, rank() over (partition by subject order by score desc) as rank, sum(1) over (partition by subject) as particip
ants, avg(score) over (partition by subject) as avg_score  from student_scores where exam = 'exam_2') select * from ranked_students_scores_at_exam2  where rank <= 3;
 id  |    name    |  subject   |  exam  | score | rank | participants |      avg_score
-----+------------+------------+--------+-------+------+--------------+---------------------
   2 | student_1  | subject_1  | exam_2 |    98 |    1 |           10 | 82.1000000000000000
 602 | student_7  | subject_1  | exam_2 |    90 |    2 |           10 | 82.1000000000000000
 502 | student_6  | subject_1  | exam_2 |    89 |    3 |           10 | 82.1000000000000000
 302 | student_4  | subject_1  | exam_2 |    89 |    3 |           10 | 82.1000000000000000
 292 | student_3  | subject_10 | exam_2 |    99 |    1 |           10 | 76.1000000000000000
 692 | student_7  | subject_10 | exam_2 |    91 |    2 |           10 | 76.1000000000000000
 592 | student_6  | subject_10 | exam_2 |    90 |    3 |           10 | 76.1000000000000000
 112 | student_2  | subject_2  | exam_2 |    99 |    1 |           10 | 79.8000000000000000
 412 | student_5  | subject_2  | exam_2 |    93 |    2 |           10 | 79.8000000000000000
 212 | student_3  | subject_2  | exam_2 |    86 |    3 |           10 | 79.8000000000000000
 522 | student_6  | subject_3  | exam_2 |    90 |    1 |           10 | 70.6000000000000000
 322 | student_4  | subject_3  | exam_2 |    83 |    2 |           10 | 70.6000000000000000
 122 | student_2  | subject_3  | exam_2 |    75 |    3 |           10 | 70.6000000000000000
 532 | student_6  | subject_4  | exam_2 |   100 |    1 |           10 | 84.5000000000000000
 132 | student_2  | subject_4  | exam_2 |   100 |    1 |           10 | 84.5000000000000000
 232 | student_3  | subject_4  | exam_2 |    99 |    3 |           10 | 84.5000000000000000
 842 | student_9  | subject_5  | exam_2 |    91 |    1 |           10 | 80.1000000000000000
 342 | student_4  | subject_5  | exam_2 |    89 |    2 |           10 | 80.1000000000000000
 642 | student_7  | subject_5  | exam_2 |    87 |    3 |           10 | 80.1000000000000000
 552 | student_6  | subject_6  | exam_2 |    98 |    1 |           10 | 76.8000000000000000
 152 | student_2  | subject_6  | exam_2 |    95 |    2 |           10 | 76.8000000000000000
  52 | student_1  | subject_6  | exam_2 |    88 |    3 |           10 | 76.8000000000000000
 762 | student_8  | subject_7  | exam_2 |    97 |    1 |           10 | 82.9000000000000000
 662 | student_7  | subject_7  | exam_2 |    95 |    2 |           10 | 82.9000000000000000
 362 | student_4  | subject_7  | exam_2 |    95 |    2 |           10 | 82.9000000000000000
 872 | student_9  | subject_8  | exam_2 |    98 |    1 |           10 | 81.1000000000000000
 672 | student_7  | subject_8  | exam_2 |    96 |    2 |           10 | 81.1000000000000000
 772 | student_8  | subject_8  | exam_2 |    89 |    3 |           10 | 81.1000000000000000
 382 | student_4  | subject_9  | exam_2 |    92 |    1 |           10 | 77.6000000000000000
 982 | student_10 | subject_9  | exam_2 |    91 |    2 |           10 | 77.6000000000000000
 782 | student_8  | subject_9  | exam_2 |    82 |    3 |           10 | 77.6000000000000000
 (31 行记录)

这里，仅仅一次查询，就返回了第二次考试中各科目排名前三的同学，以及该科目的平均分数和参与人数。非常的方便。