如何在 where 子句中使用 row_number

我正在尝试使用窗口函数来获取最近的 n 条记录，如下从这里 https://stackoverflow.com/questions/61570170/something-like-select-distinct-on-but-for-n-1/61570644#61570644:

I have:

select 
   id,
   blah,
   row_number () over (
     partition by blah, my_id
     order by datetime) rn,
   theme
from documents
where theme = 'cats';

我得到：

 id | blah | rn | theme 
----+-----+----+-------
  1 |   1 |  1 | cats
  2 |   1 |  2 | cats
  3 |   1 |  3 | cats
  4 |   1 |  4 | cats
  5 |   1 |  5 | cats
  9 |   2 |  1 | cats
  8 |   2 |  2 | cats
 11 |   3 |  1 | cats
 12 |   4 |  1 | cats
 13 |   5 |  1 | cats
 14 |   6 |  1 | cats
(11 rows)

这太棒了。但我想要不超过2行，rn <= 2， 例如。我把这想象成这样：

select 
   id,
   blah,
   row_number () over (
     partition by blah, my_id
     order by datetime) rn,
   theme
from documents
where theme = 'cats' and
rn <= 2;

但我得到：

ERROR:  column "rn" does not exist
LINE 15: rn <= 1;
         ^

我知道我可以将其设置为像链接问题一样的子查询，但是我必须缺少将 row_number 放入 where 子句中的语法，对吧？它是什么？

您需要一个派生表：

select id, blah, them
from (
  select id,
         blah,
         row_number () over (partition by blah, my_id order by datetime) rn,
         theme
  from documents
) x
where theme = 'cats' 
and rn <= 2;

这基本上是语法糖，不会导致性能开销。