为什么PostgreSQL以错误的方式组合系列？

Question

我得到了一些组合generate_series的奇怪行为.在我试图用网格填充的2个不同的多边形中,一个网格非常罕见:

查询是这样的:

SELECT
    osm_id ,
        generate_series(floor(st_xmin(way))::int, ceiling(st_xmax(way))::int, 150) x,
        generate_series(floor(st_ymin(way))::int, ceiling(st_ymax(way))::int, 150) y 
from osm_polygon
order by osm_id, x, y;

我试着追踪问题,只输入了最小/最大坐标.从最小值/最大值生成序列创建正确的值数:分别为9行和12行.

  => select generate_series(9237195, 9238873, 150) x;
      x    
  ---------
   9237195
   9237345
   9237495
   9237645
   9237795
   9237945
   9238095
   9238245
   9238395
   9238545
   9238695
   9238845
  (12 rows)

  => select generate_series(7371701, 7372922, 150) y order by y;
      y    
  ---------
   7371701
   7371851
   7372001
   7372151
   7372301
   7372451
   7372601
   7372751
   7372901
  (9 rows)

结合起来,他们应该制作108排吧？不,只有36行:

=> select generate_series(9237195, 9238873, 150) x, generate_series(7371701, 7372922, 150) y order by x, y;
      x    |    y    
  ---------+---------
   9237195 | 7371701
   9237195 | 7372151
   9237195 | 7372601
   9237345 | 7371851
   9237345 | 7372301
   9237345 | 7372751
   9237495 | 7372001
   9237495 | 7372451
   9237495 | 7372901
   9237645 | 7371701
   9237645 | 7372151
   9237645 | 7372601
   9237795 | 7371851
   9237795 | 7372301
   9237795 | 7372751
   9237945 | 7372001
   9237945 | 7372451
   9237945 | 7372901
   9238095 | 7371701
   9238095 | 7372151
   9238095 | 7372601
   9238245 | 7371851
   9238245 | 7372301
   9238245 | 7372751
   9238395 | 7372001
   9238395 | 7372451
   9238395 | 7372901
   9238545 | 7371701
   9238545 | 7372151
   9238545 | 7372601
   9238695 | 7371851
   9238695 | 7372301
   9238695 | 7372751
   9238845 | 7372001
   9238845 | 7372451
   9238845 | 7372901
(36 rows)

蛮力实际上在这里工作:

with a as (select generate_series(9237195, 9238873, 150) x),
    b as (select generate_series(7371701, 7372922, 150) y)
select x, y from a, b;

制作108排.很好,除了查询变得更复杂.

为什么这样？

Postgres版本是9.1

Answer 1

在SELECT-list中调用时,如何执行多个set-returns函数是一个怪癖.你希望结果是两者的交叉乘积,但这不是它的工作原理.它实际上是两者行数的最低公倍数.

看到:

• 为什么这些根据大小加入不同？
• 如何重写旧的PostgreSQL版本的SELECT ... CROSS JOIN LATERAL ...语句？

相比:

test=>     SELECT generate_series(1,3) aval, generate_series(1,4) bval;
 aval | bval 
------+------
    1 |    1
    2 |    2
    3 |    3
    1 |    4
    2 |    1
    3 |    2
    1 |    3
    2 |    4
    3 |    1
    1 |    2
    2 |    3
    3 |    4
(12 rows)

test=>     SELECT generate_series(1,3) aval, generate_series(1,3) bval;
 aval | bval 
------+------
    1 |    1
    2 |    2
    3 |    3
(3 rows)

因此,在PostgreSQL 9.2及更高版本中,您应该使用LATERAL在FROM子句中调用set-returns函数的查询:

test=>     SELECT aval, bval FROM generate_series(1,3) aval CROSS JOIN LATERAL generate_series(1,3) bval;
 aval | bval 
------+------
    1 |    1
    1 |    2
    1 |    3
    2 |    1
    2 |    2
    2 |    3
    3 |    1
    3 |    2
    3 |    3
(9 rows)

test=>     SELECT aval, bval FROM generate_series(1,3) aval CROSS JOIN LATERAL generate_series(1,4) bval;
 aval | bval 
------+------
    1 |    1
    1 |    2
    1 |    3
    1 |    4
    2 |    1
    2 |    2
    2 |    3
    2 |    4
    3 |    1
    3 |    2
    3 |    3
    3 |    4
(12 rows)

在旧版本中,您可以使用子查询在FROM中避免在SELECT术语中使用多个SRF:

test=> SELECT generate_series(1,3) aval, bval FROM (SELECT generate_series(1,4)) AS x(bval);
 aval | bval 
------+------
    1 |    1
    2 |    1
    3 |    1
    1 |    2
    2 |    2
    3 |    2
    1 |    3
    2 |    3
    3 |    3
    1 |    4
    2 |    4
    3 |    4
(12 rows)