Postgres GROUP BY 数组列

Question

我有一个学生和 parent 的列表，我想使用学生 ID 将他们分组到家庭中。 Parents 共享相同学号的学生可以被视为一个家庭，而共享相同 parent id 的学生也可以被视为一个家庭。这是一个示例 table:

 p_id | parent_name | s_id | student_name |
------------------------------------------|
 1    | John Doe    | 100 |  Mike Doe     |
 3    | Jane Doe    | 100 |  Mike Doe     |
 3    | Jane Doe    | 105 |  Lisa Doe     |
 5    | Will Willy  | 108 |  William Son  |

我想以这样的方式结束：

     parents       |       students         |
-------------------|------------------------|
John Doe, Jane Doe |  Mike Doe, Lisa Doe    |
Will Willy         |  William Son           |

为此，我目前正在使用：

SELECT array_agg(parents) AS parents FROM (
    SELECT array_agg(p_id) AS par_ids, array_agg(parent_name) AS parents, student_name, s_id 
    FROM (
        /* sub query */
    )b
    GROUP BY s_id, student_name
    ORDER BY parents ASC
)c
GROUP BY unnest(par_ids)
ORDER BY parents ASC

但是我得到一个错误：ERROR: cannot accumulate arrays of different dimensionality。 SQL state: 2202E

我怎样才能达到预期的效果？上述语句的内部查询 returns:

| par_ids |       parents       |  student_name   | s_id |
--------------------------------|------------------------|
| {1,3}   | {John Doe, Jane Doe}|  Mike Doe       | 100  |
| {3}     | {Jane Doe}          |  Lisa Doe       | 105  |
| {5}     | {Will Willy}        |  William Son    | 108  |

现在将这些学生分组到 parents 是我遇到的问题。

Answer 1

我已经在这里做了类似的事情（但有点复杂）：

step-by-step demo:db<>fiddle

SELECT
    array_agg(parent_name) as parents,                -- 4
    array_agg(student_name) as students
FROM (
    SELECT DISTINCT ON (t.s_id)                       -- 3
        *
    FROM (
        SELECT
            s_id,
            array_agg(p_id) as parents                -- 1
        FROM mytable
        GROUP BY s_id
    ) s JOIN mytable t ON t.p_id = ANY(s.parents)     -- 2
    ORDER BY t.s_id, CARDINALITY(parents) DESC        -- 3
) s
GROUP BY parents

将 p_id 个值聚合到一个数组中：

s_id parents

108 {5}

105 {3}

100 {1,3}

在这个数组上自加入原来的table：

s_id	parents	p_id	parent_name	s_id	student_name
100	{1,3}	1	John Doe	100	Mike Doe
105	{3}	3	Jane Doe	100	Mike Doe
100	{1,3}	3	Jane Doe	100	Mike Doe
105	{3}	3	Jane Doe	105	Lisa Doe
100	{1,3}	3	Jane Doe	105	Lisa Doe
108	{5}	5	Will Willy	108	William Son

删除所有重复的学生记录。剩下的应该是p_id数组最全的记录。这可以通过数组长度的降序使用 DISTINCT ON(s_id) 来完成：

s_id parents p_id parent_name s_id student_name

100 {1,3} 1 John Doe 100 Mike Doe

100 {1,3} 3 Jane Doe 105 Lisa Doe

108 {5} 5 Will Willy 108 William Son
最后，您可以按 p_id 数组分组并聚合两个 name 列：

parents students

{"John Doe","Jane Doe"} {"Mike Doe","Lisa Doe"}

{"Will Willy"} {"William Son"}

如果你不想得到一个数组，而是一个字符串列表，你可以使用string_agg(name_colum, ',')代替array_agg(name_column)

Postgres GROUP BY 数组列

Postgres GROUP BY an array column

postgresql

group-by

array-agg

parents	students
{"John Doe","Jane Doe"}	{"Mike Doe","Lisa Doe"}
{"Will Willy"}	{"William Son"}