Question

我想在SQL Server数据库中获取一部分数据，因此我可以为应用程序运行一些冒烟测试。我的典型工作流程如下，但删除查询需要一个多小时，因为X，Y和Z中有数亿行。如何才能更快地完成此操作？为清楚起见，请在答案中使用T-SQL代码。

通过为每个表运行此命令来备份每个表：

SELECT * INTO BACKUP_SRC_PATIENT FROM [SRC_PATIENT] -- has PK patient_id
SELECT * INTO BACKUP_SRC_X FROM [SRC_X] -- has FK patient_id
SELECT * INTO BACKUP_SRC_Y FROM [SRC_Y] -- has FK patient_id
SELECT * INTO BACKUP_SRC_Z FROM [SRC_Z] -- has FK patient_id

删除所有100名患者

DELETE FROM [SRC_PATIENT] 
WHERE [PATIENT_ID_NONNUM] NOT IN 
      (SELECT top 100 [PATIENT_ID] FROM [SRC_PATIENT]
       WHERE BLA = 'BLA')

从SRC_X，SRC_Y和SRC_Z删除数据，其中patient_id不在SRC_PATIENT中

DELETE FROM [SRC_X] 
WHERE [PATIENT_ID] NOT IN 
      (SELECT [PATIENT_ID] FROM [SRC_PATIENT])

DELETE FROM [SRC_Y] 
WHERE [PATIENT_ID] NOT IN 
      (SELECT [PATIENT_ID] FROM [SRC_PATIENT])

DELETE FROM [SRC_Z] 
WHERE [PATIENT_ID] NOT IN 
      (SELECT [PATIENT_ID] FROM [SRC_PATIENT])

Answer 1

而不是删除除100之外的所有行 - 截断表并从备份中复制超过100行。

TRUNCATE剪辑表格中的所有数据，而不必逐行删除（或逐页逐页删除）：

TRUNCATE TABLE [SRC_PATIENT]
TRUNCATE TABLE [SRC_X]
TRUNCATE TABLE [SRC_Y]
TRUNCATE TABLE [SRC_Z]

INSERT INTO [SRC_PATIENT]
SELECT TOP 100 * FROM [BACKUP_SRC_PATIENT]
WHERE BLA = 'BLA' 

INSERT INTO [SRC_X]
SELECT * FROM [BACKUP_SRC_X]
WHERE [PATIENT_ID] IN (SELECT [PATIENT_ID] FROM [SRC_PATIENT])

INSERT INTO [SRC_Y]
SELECT * FROM [BACKUP_SRC_Y]
WHERE [PATIENT_ID] IN (SELECT [PATIENT_ID] FROM [SRC_PATIENT])

INSERT INTO [SRC_Z]
SELECT * FROM [BACKUP_SRC_Z]
WHERE [PATIENT_ID] IN (SELECT [PATIENT_ID] FROM [SRC_PATIENT])

Answer 2

使用架构开关，然后将所需的行拉入非备份表

您需要创建一个具有相同元数据的表，但切换几乎立即发生。

CREATE TABLE dbo.Backup-Table1
(Col1 int
Col2 int)

ALTER TABLE dbo.Table1
SWITCH TO dbo.Backup-Table1

INSERT INTO dbo.Table1
SELECT TOP 100 *
FROM dbo.Backup-Table1
WHERE Condition = Met

将数据库减少到连接在一起的表中的数据子集

2 个答案: