Question

基于this我有一个表，其值将是PostgreSQL查询结果的列名。

id    col1     col2
----------------------
0      name    ax
0      name2   bx
0      name3   cx
1      name    dx
1      name2   ex
1      name2   fx
...     ...     ...

现在我希望查询结果看起来像这样

id   name    name2     name3   ...
0    ax      bx         cx     ...
1    dx      ex         fx     ...

每次添加新字段时，col1上的字段数都会更改。因此，我需要生成一个以动态方式返回结果的函数。

这样做了：

SELECT
  id,
  /* if col1 matches the name string of this CASE, return col2, otherwise return NULL */
  /* Then, the outer MAX() aggregate will eliminate all NULLs and collapse it down to one row per id */
  MAX(CASE WHEN (col1 = 'name') THEN col2 ELSE NULL END) AS name,
  MAX(CASE WHEN (col1 = 'name2') THEN col2 ELSE NULL END) AS name2,
  MAX(CASE WHEN (col1 = 'name3') THEN col2 ELSE NULL END) AS name3
FROM mytable
GROUP BY id

但我需要动态，因为col1名称可以是一个大的列表，所以每次在col1中添加新名称时我都无法更新查询。

我使用数据透视表检查了如何执行此操作，我尝试按照此example进行操作，但是字段众所周知，请有人帮助我吗？

Answer 1

对于PostgreSQL 9.4 +

-- Test data
create table t(id int, col1 text, col2 text);
insert into t values
  (0, 'name', 'ax'),
  (0, 'name2', 'bx'),
  (0, 'name3', 'cx'),
  (1, 'name', 'dx'),
  (1, 'name2', 'ex'),
  (1, 'name3', 'fx');

create or replace function fn_pivot(
  p_sql text,
  p_row_field text,
  p_col_field text,
  p_data_field text,
  p_cursor refcursor) returns refcursor language plpgsql as $$
declare
  cols text[];
  a text[];
  q text;
  --f text;
begin
  -- Get dynamic columns
  q := format('select array_agg(distinct %s::text) from (%s) t', p_col_field, p_sql);
  execute q into cols;
  -- Generate SELECT part
  select array_agg(format('%s filter (where %s::text = %L) as %I', p_data_field, p_col_field, x, x)) into a from unnest(cols) as t(x);
  q := format('%s, %s', p_row_field, array_to_string(a, ', '));
  -- Complete the whole statement
  q := format('select %s from (%s) t group by %s order by %s', q, p_sql, p_row_field, p_row_field);
  raise info '%', q;
  open p_cursor for execute q;
  return p_cursor;
end $$;

用法（带一些调试输出）：

nd@postgres=# start transaction;
START TRANSACTION
*nd@postgres=# select * from fn_pivot('select * from t', 'id', 'col1', 'max(col2)', 'cur');
INFO:  select id, max(col2) filter (where col1::text = 'name') as name, max(col2) filter (where col1::text = 'name2') as name2, max(col2) filter (where col1::text = 'name3') as name3 from (select * from t) t group by id order by id
╔══════════╗
║ fn_pivot ║
╠══════════╣
║ cur      ║
╚══════════╝
(1 row)

*nd@postgres=# fetch all in cur;
╔════╤══════╤═══════╤═══════╗
║ id │ name │ name2 │ name3 ║
╠════╪══════╪═══════╪═══════╣
║  0 │ ax   │ bx    │ cx    ║
║  1 │ dx   │ ex    │ fx    ║
╚════╧══════╧═══════╧═══════╝
(2 rows)

*nd@postgres=# rollback;

列值作为psql查询中的列名

1 个答案: