1.需求說明
TEST表情況說明:
- 按月進行分區的分區表
- 未定義主鍵或唯一索引
- 包含COL1,COL2,COL3,INSERTTIME四列
現需要刪除2019年3月31日當天存在的重複數據
2.解決方法
2.1 確認無重複數據的記錄數
SELECT COUNT(1) FROM (
SELECT COL1,COL2,COL3,INSERTTIME FROM TEST PARTITION(P201903) A
WHERE INSERTTIME >= DATE'2019-03-31' AND INSERTTIME < DATE'2019-04-01'
GROUP BY COL1,COL2,COL3
);
2.2 梳理需要篩選的數據
由於原表A數據量特別大,此處新建一張表將需要處理的數據單獨存放
CREATE TABLE TEST_TMP NOLOGGING AS
SELECT /*PARALLEL +8 */ A.*,A.ROWID ROWID_OLD FROM TEST PARTITION(P201903) A
WHERE INSERTTIME >= DATE'2019-03-31' AND INSERTTIME < DATE'2019-04-01';
2.2 確認需要刪除的數據
理論上而言需要刪除和需要保留的數據記錄數應相等
--需要刪除的數據記錄數
SELECT COUNT(1) FROM TEST PARTITION(P201903) A WHERE ROWID IN (
SELECT MIN(ROWID_OLD) ROWID_OLD FROM TEST_TMP
WHERE INSERTTIME >= DATE'2019-03-31' AND INSERTTIME < DATE'2019-04-01'
GROUP BY COL1,COL2,COL3,INSERTTIME
HAVING COUNT(1) > 1)
AND INSERTTIME >= DATE'2019-03-31' AND INSERTTIME < DATE'2019-04-01'
--需要保留的數據記錄數
SELECT COUNT(1) FROM TEST PARTITION(P201903) A WHERE ROWID NOT IN (
SELECT MIN(ROWID_OLD) ROWID_OLD FROM TEST_TMP
WHERE INSERTTIME >= DATE'2019-03-31' AND INSERTTIME < DATE'2019-04-01'
GROUP BY COL1,COL2,COL3,INSERTTIME
HAVING COUNT(1) > 1)
AND INSERTTIME >= DATE'2019-03-31' AND INSERTTIME < DATE'2019-04-01'
2.3 利用分批提交刪除重複數據
DECLARE
TYPE ROWID_LIST IS TABLE OF UROWID INDEX BY BINARY_INTEGER;
ROWID_INFOS ROWID_LIST;
I NUMBER;
CURSOR C_ROWIDS IS (SELECT MIN(ROWID_OLD) ROWID_OLD
FROM TEST_TMP
WHERE INSERTTIME >= DATE'2019-03-31' AND INSERTTIME < DATE'2019-04-01'
GROUP BY COL1,COL2,COL3,INSERTTIME
HAVING COUNT(1) > 1);
BEGIN
OPEN C_ROWIDS;
LOOP
--此處LIMIT後的值爲分批提交的記錄數,可以根據實際情況調整
FETCH C_ROWIDS BULK COLLECT INTO ROWID_INFOS LIMIT 10000;
FORALL I IN 1..ROWID_INFOS.COUNT
--如下的DELETE語句爲分批提交實際需要執行的部分
DELETE FROM TEST WHERE ROWID=ROWID_INFOS(I);
COMMIT;
EXIT WHEN ROWID_INFOS.COUNT<10000;
END LOOP;
CLOSE C_ROWIDS;
END;
2.4 確認無重複數據
SELECT * FROM (
SELECT COL1,COL2,COL3,INSERTTIME FROM TEST PARTITION(P201903) A
WHERE INSERTTIME >= DATE'2019-03-31' AND INSERTTIME < DATE'2019-04-01'
GROUP BY COL1,COL2,COL3,INSERTTIME
HAVING COUNT(1)>1 )
);