SQL

Question

我有以下数据：

ID  --- GRP_ID  --- REC_VAL
1   --- 1       --- A
2   --- 2       --- A
3   --- 2       --- B
4   --- 3       --- A
5   --- 3       --- B
6   --- 3       --- C
7   --- 4       --- A
8   --- 4       --- B
9   --- 4       --- C
10  --- 5       --- A
11  --- 5       --- B
12  --- 5       --- E

有没有办法找到具有相同值的记录组的 ID？（在这种情况下只有 grp_id 3 和 4 具有相同的值）

第二个问题：

当我有一组值时，是否有有效的方法来找到准确的 grp_id？我的解决方案不是很快，因为 table 与组有超过 600 万。记录：

-- Large table - up to 6m records
create table tmp_grp (id number, grp_id number, rec_val varchar2(10));
--
insert into tmp_grp(id, grp_id, rec_val) values (1, 1, 'A');
insert into tmp_grp(id, grp_id, rec_val) values (2, 2, 'A');
insert into tmp_grp(id, grp_id, rec_val) values (3, 2, 'B');
insert into tmp_grp(id, grp_id, rec_val) values (4, 3, 'A');
insert into tmp_grp(id, grp_id, rec_val) values (5, 3, 'B');
insert into tmp_grp(id, grp_id, rec_val) values (6, 3, 'C');
insert into tmp_grp(id, grp_id, rec_val) values (7, 4, 'A');
insert into tmp_grp(id, grp_id, rec_val) values (8, 4, 'B');
insert into tmp_grp(id, grp_id, rec_val) values (9, 4, 'C');
insert into tmp_grp(id, grp_id, rec_val) values (10, 5, 'A');
insert into tmp_grp(id, grp_id, rec_val) values (11, 5, 'B');
insert into tmp_grp(id, grp_id, rec_val) values (12, 5, 'E');
commit;
--
-- CTE representing record group for asking  
WITH datrec AS
 (SELECT 'A' rec FROM dual UNION ALL 
  SELECT 'B' rec FROM dual)
--
SELECT x.grp_id
  FROM (
  -- Count of joined records 
  SELECT COUNT(1) cnt, t.grp_id
          FROM tmp_grp t
          JOIN datrec d
            ON d.rec = t.rec_val
         GROUP BY t.grp_id
  -- 
  ) x
 WHERE 
 -- Count of all data records
 x.cnt = (SELECT COUNT(1) FROM datrec)
 -- Count of all group records
 AND x.cnt = (SELECT COUNT(1) FROM tmp_grp g WHERE x.grp_id = g.grp_id);
 --

这个问题类似于，但是这个主题只涵盖精确的一组值（值的数量和 datrec 的 rec 列中的值将由另一个查询提供）和查询 return包含此集合的组。我只需要 return 完全匹配。

更新 - 在 table 中添加数据以便更好地说明

也与 How to compare groups of tuples in sql

相关

Answer 1

这是一种避免将基数 table 与其自身连接的方法。如果每个 grp_id 有几个（很多？）可能的 rec_val 值，它会更有效。如果不同的 grp_id 已经存在于数据中的某处，它仍然可以变得更快；我即时创建它们。

with gid ( grp_id ) as (
       select distinct grp_id from tmp_grp
     ),
     prep ( grp_id_1, grp_id_2, rec_val ) as (
       select t.grp_id, g.grp_id, t.rec_val
         from tmp_grp t join gid g on t.grp_id < g.grp_id
       union all
       select g.grp_id, t.grp_id, t.rec_val
         from gid g join tmp_grp t on g.grp_id < t.grp_id
     ),
     counts ( grp_id_1, grp_id_2, cnt ) as (
       select   grp_id_1, grp_id_2, count(*)
       from     prep
       group by grp_id_1, grp_id_2, rec_val
     )
select   grp_id_1, grp_id_2
from     counts
group by grp_id_1, grp_id_2
having min(cnt) = 2
;

SQL - 在大型 table 中找到确切的记录组

SQL - find exact group of records in large table

oracle

grouping

database-performance

relational-division