基于列之间的比较从两个表中聚合和连接

Question

I have two tables like this:

Table1

A    T
a1   t1
a2   t2
a3   t3
a4   t4
a5   t5
...
...

Table2

E    T
e1   t1
e2   t2
e3   t3
e4   t4
e5   t5
...
...

what I wanted to achieve is this:

Table 3

E    A'
e1   a1,a2,a3
e2   a4,a5,a6
...
...

The aggregation A' is done like this: In table 2 for each e there is a value in column T : t and with that t you look for the last 3 values in Table 1 that are less than the t in question. So a1, a2, a3 are values of A whose t values are less than t1 in Table 2 whose E is e1 .

I know that I could write two queries for this like this:

ResultSet (rt) -> select t from e

and then iterate ResultSet and do something like this:

select A from Table1 where t < rt[i] limit 3 - not sure how to concatenate here:)

but I m pretty sure this is utterly inefficient. There should be a better way to do this.

I m working with Postgresql.

If it had been a dataframe from a file I would use python's pandas. Also I know that python has read_sql but the tables are very huge I don't want to load the whole table in memory which I think it won't but not sure either - anyway its a separate story.

How do we solve this in SQL? Any ideas please.

Answer 1

In table 2 for each e there is a value in column T: t and with that t you look for the last 3 values in Table 1 that are less than the t in question.

I don't understand the results follow this logic. But based on your description, you can use a lateral join:

select t2.*, t1.the_as
from t2 left join lateral
     (select array_agg(t1.a) as the_as
      from (select t1.*
            from t1
            where t1.T <= t2.T
            order by t1.T desc
            limit 3
           ) t1
     ) t1
     on 1=1;

Note that this uses arrays rather than strings because I think arrays are a better data structure for storing multiple values. That said, you can just use string_agg() instead, if you really want a string. The syntax would be string_agg(t1.a, ',') .