拆分相等组的数据（PL / SQL，SQL）

Question

我有一些id：

Select count(*) Into count_id From table_I;--4

我知道我有总total_user = 109（记录数）。所以我想将它拆分为相同的组：

Select round(count(*)/count_user,0) Into mapUser From table_U;

所以我有4 group。在first three will be 27和last should be 28用户。

现在我想for each group分配唯一的ID。

set serveroutput on 
declare 
      CURSOR cur IS Select * From table_U FOR UPDATE OF USER_ID;
      mapUser NUMBER;
      l_rec table_U%rowtype;
      x_mapUser Number := 0;--number between 0-27
      c_obj_id NUMBER := 1;
      count_id NUMBER := 0;

      type T1 is table of number(10) index by binary_integer;
      V1 T1;    

begin
     Select count(*) Into count_id From table_I;--count_id = 4
     Select round(count(*)/count_id,0) Into mapUser From table_U; --mapUser = 27

     SELECT id BULK COLLECT INTO V1 FROM table_I;--it's 4 id (id_1, id_2, id_3, id_4)

    OPEN cur;
        LOOP FETCH cur INTO l_rec;
         EXIT WHEN cur%notfound;

           IF x_mapUser > mapUser Then --0-27 > 27
                x_mapUser := 1;                   
                c_obj_id := c_obj_id +1;--next value from V1
           End if;

          UPDATE table_U SET USER_ID = V1(c_obj_id) WHERE CURRENT OF cur;

         x_mapUser := x_mapUser +1;

        END LOOP;
       CLOSE cur;        
end;

但我不知道如何改变我的IF和cur分配的最后价值以及id_4。我在这里做错了什么：/

Answer 1

这对我有用：

merge into table_u a
using (select rd, i.id
         from (select u.rowid rd, cnt - mod(rownum-1, cnt) rn 
                 from table_u u, (select count(1) cnt from table_i) ) u
         join (select row_number() over( order by id) rn, id from table_i) i using (rn)) b
on (a.rowid = b.rd)         
when matched then update set a.user_id = b.id

我的测试表：

create table table_i as (
  select level*10 id from dual connect by level <= 4);

create table table_u as (
  select cast(null as number(3)) user_id, level id from dual connect by level <= 109);

第二个表的最高值分配了28次，其他27次。这是因为我用过

cnt - mod(rownum-1, cnt) rn

计算加入列。我不知道它对你来说是否重要。 :)这个解决方案的基础是mod()，它允许我们在1和cnt之间循环（在这种情况下为4）。

您可以在PLSQL中执行此操作，但SQL解决方案通常更快，更可取。

Answer 2

我会用ntile()：

select u.*, ntile(4) over (order by user_id) as grp
from table_u u;

我不知道你想要什么订购，如果有的话。如果您愿意，可以使用随机数。

如果要枚举每个组中的值，请使用子查询：

select u.*, row_number() over (partition by grp order by grp) as seqnum
from (select u.*, ntile(4) over (order by user_id) as grp
      from table_u u
     ) u;

Answer 3

我发现了错误并使用了我的PL / SQL代码：

...
x_mapUser Number := 0
...
OPEN cur;
 LOOP FETCH cur INTO l_rec;
  EXIT WHEN cur%notfound;
     UPDATE table_U SET USER_ID = V1(c_obj_id) WHERE CURRENT OF cur;

     IF x_mapUser > mapUser-1 AND c_obj_id < count_id Then
            x_mapUser := 0;
            c_obj_id := c_obj_id +1;
     End if;

 x_mapUser := x_mapUser +1;

 END LOOP;
CLOSE cur;