我确实在Oracle中声明了以下对象和数组：
create or replace
TYPE TestObj as
OBJECT (parent_key varchar2(20), child_key varchar2(20),
dt date，other1 varchar2（20），other2 varchar2（20））; 和
create or replace
TYPE TestArray as TABLE OF TestObj;
现在在我的一个SP中，我将以下类型作为输入：
IN_TestArray IN TestArray
现在在这个数组中，我得到的数据集包括一些重复项：
Id Parent_key Child_Key dt other1 other2
===========================================================
1 AT AA 2016-06-22 test1 test2
2 AT A0 2016-06-22 test3 test4
3 BT BA 2016-06-22 test5 test6
4 AT AA 2016-06-22 test1 test2
5 AT AA 2016-06-21 test21 test22
6 CT CA 2016-06-22 test23 test24
7 AT AA 2016-06-22 test31 test32
这里我希望通过 parent_key，child_key和dt的组合来消除数组中的重复记录。
因此更新的数组在进程后应该只有以下行：
(any one from 1, 4, 7)
2
3
5
6
任何帮助都会受到赞赏.....
答案 0 :(得分：1)
使用SET() collection function删除完全重复的内容。
Oracle安装程序：
CREATE TYPE TestObj AS OBJECT (
parent_key varchar2(20),
child_key varchar2(20),
dt date,
other1 varchar2(20),
other2 varchar2(20)
);
/
CREATE TYPE TestArray AS TABLE OF TestObj;
/
<强>查询：
SELECT *
FROM TABLE(
SET(
TestArray(
TestObj( 'AT', 'AA', DATE '2016-06-22', 'test1', 'test2' ),
TestObj( 'AT', 'AA', DATE '2016-06-22', 'test1', 'test2' )
)
)
);
<强>输出：
PARENT_KEY CHILD_KEY DT OTHER1 OTHER2
------------ ------------ ------------------- ------------ ------------
AT AA 2016-06-22 00:00:00 test1 test2
如果您只想在行的子集上获取重复项，则可以使用：
SELECT TestObj(
parent_key,
child_key,
dt,
MIN( other1 ) KEEP ( DENSE_RANK FIRST ORDER BY ROWNUM ),
MIN( other2 ) KEEP ( DENSE_RANK FIRST ORDER BY ROWNUM )
)
BULK COLLECT INTO array_unique
FROM TABLE( array_with_dups )
GROUP BY parent_key, child_key, dt;
像这样：
SET SERVEROUTPUT ON;
DECLARE
array_with_dups TestArray
:= TestArray(
TestObj( 'AT', 'AA', DATE '2016-06-22', 'test1', 'test2' ),
TestObj( 'AT', 'AA', DATE '2016-06-22', 'test1', 'test2' ),
TestObj( 'AT', 'AA', DATE '2016-06-22', 'test31', 'test32' )
);
array_unique TestArray;
BEGIN
SELECT TestObj(
parent_key,
child_key,
dt,
MIN( other1 ) KEEP ( DENSE_RANK FIRST ORDER BY ROWNUM ),
MIN( other2 ) KEEP ( DENSE_RANK FIRST ORDER BY ROWNUM )
)
BULK COLLECT INTO array_unique
FROM TABLE( array_with_dups )
GROUP BY parent_key, child_key, dt;
FOR i IN 1 .. array_unique.COUNT LOOP
DBMS_OUTPUT.PUT_LINE(
array_unique(i).parent_key
|| ', ' || array_unique(i).child_key
|| ', ' || array_unique(i).dt
|| ', ' || array_unique(i).other1
|| ', ' || array_unique(i).other2
);
END LOOP;
END;
/
<强>输出：
AT, AA, 2016-06-22 00:00:00, test1, test2
答案 1 :(得分：0)
您可以delete from a collection，但需要确定要使用的索引值。这有点乱，但应该有效：
for r in (
select i from (
select t.*, rownum as i,
row_number() over (partition by parent_key, child_key, dt
order by other1, other2) as key_rn
from table(in_testarray) t
)
where key_rn > 1
)
loop
in_testarray.delete(r.i);
end loop;
演示：
declare
in_testarray testarray;
begin
select *
bulk collect into in_testarray
from (
select testobj('AT', 'AA', date '2016-06-22', 'test1', 'test2') from dual
union all select testobj('AT', 'A0', date '2016-06-22', 'test3', 'test4') from dual
union all select testobj('BT', 'BA', date '2016-06-22', 'test5', 'test6') from dual
union all select testobj('AT', 'AA', date '2016-06-22', 'test1', 'test2') from dual
union all select testobj('AT', 'AA', date '2016-06-21', 'test21', 'test22') from dual
union all select testobj('CT', 'CA', date '2016-06-22', 'test23', 'test24') from dual
union all select testobj('AT', 'AA', date '2016-06-22', 'test31', 'test32') from dual
);
for r in (
select i from (
select t.*, rownum as i,
row_number() over (partition by parent_key, child_key, dt
order by other1, other2) as key_rn
from table(in_testarray) t
)
where key_rn > 1
)
loop
dbms_output.put_line('deleting ' || r.i);
in_testarray.delete(r.i);
end loop;
for i in in_testarray.first..in_testarray.last loop
if in_testarray.exists(i) then
dbms_output.put_line(i ||': '|| in_testarray(i).parent_key ||':'|| in_testarray(i).child_key
||':'|| in_testarray(i).dt ||':'|| in_testarray(i).other1 ||':'|| in_testarray(i).other2);
end if;
end loop;
end;
/
PL/SQL procedure successfully completed.
该系列最终有五名成员：
PL/SQL procedure successfully completed.
deleting 4
deleting 7
1: AT:AA:2016-06-22:test1:test2
2: AT:A0:2016-06-22:test3:test4
3: BT:BA:2016-06-22:test5:test6
5: AT:AA:2016-06-21:test21:test22
6: CT:CA:2016-06-22:test23:test24
但是直接修改传入的数组可能不是一个好主意（无论如何都不允许使用IN参数），所以你应该修改原始数组的副本。
或者你可以制作一个副本，但不包括重复的副本;如果你宣布说同一类型的my_testarray：
select testobj(parent_key, child_key, dt, other1, other2)
bulk collect into my_testarray
from (
select t.*,
row_number() over (partition by parent_key, child_key, dt
order by other1, other2) as key_rn
from table(in_testarray) t
)
where key_rn = 1;
然后使用该集合而不是原始集合。
快速演示：
declare
in_testarray testarray;
my_testarray testarray;
begin
select *
bulk collect into in_testarray
from (
select testobj('AT', 'AA', date '2016-06-22', 'test1', 'test2') from dual
union all select testobj('AT', 'A0', date '2016-06-22', 'test3', 'test4') from dual
union all select testobj('BT', 'BA', date '2016-06-22', 'test5', 'test6') from dual
union all select testobj('AT', 'AA', date '2016-06-22', 'test1', 'test2') from dual
union all select testobj('AT', 'AA', date '2016-06-21', 'test21', 'test22') from dual
union all select testobj('CT', 'CA', date '2016-06-22', 'test23', 'test24') from dual
union all select testobj('AT', 'AA', date '2016-06-22', 'test31', 'test32') from dual
);
select testobj(parent_key, child_key, dt, other1, other2)
bulk collect into my_testarray
from (
select t.*,
row_number() over (partition by parent_key, child_key, dt
order by other1, other2) as key_rn
from table(in_testarray) t
)
where key_rn = 1;
for i in my_testarray.first..my_testarray.last loop
dbms_output.put_line(i ||': '|| my_testarray(i).parent_key ||':'|| my_testarray(i).child_key
||':'|| my_testarray(i).dt ||':'|| my_testarray(i).other1 ||':'|| my_testarray(i).other2);
end loop;
end;
/
获得相同的输出，但该集合现在不稀疏：
PL/SQL procedure successfully completed.
1: AT:A0:2016-06-22:test3:test4
2: AT:AA:2016-06-21:test21:test22
3: AT:AA:2016-06-22:test1:test2
4: BT:BA:2016-06-22:test5:test6
5: CT:CA:2016-06-22:test23:test24
或者您可以在使用原始集合时执行类似的操作，例如：在表集合表达式中。