Sql server - 递归删除

Question

我正在尝试删除位于不同表中的用户数据及其所有相关数据。 所有表都有外键但没有级联删除。

我调查了一些选项：

1. 在所有 FK 上启用级联删除，删除并移除级联删除。
2. 从底部向上删除，循环向上删除所有叶子并重复此操作直到 Root。

还有更聪明的选择或其他技巧吗？

我使用的是 Microsoft SQL Server 2012 (SP1)

Answer 1

那些是最好和最有效的。对于生产查询，我会使用2。

我能想到的唯一其他方法（IMO）仅适用于在测试环境中快速而肮脏地删除数据（避免需要分析正确的顺序）

1. 禁用所有 FK 删除所需数据，然后重新启用 FK。这是低效的，因为它们需要重新启用 WITH CHECK 以避免使 FK 处于不受信任的状态，这意味着需要重新验证所有保留的数据。
2. 以任意顺序列出受影响表上的所有 DELETE 语句，并根据需要多次运行批处理，直到成功且没有 FK 错误。

Answer 2

Oracle 解决方案：

How to generate DELETE statements in PL/SQL, based on the tables FK relations?

SQL Server 解决方案：

Generate Delete Statement From Foreign Key Relationships in SQL 2008?

希望对你有帮助

Answer 3

级联删除简单、性能好且可靠。适用于多个级别。查询计划很有趣，而且看起来优化得很好。

如果您想要手动删除，请确保为每个表只发出一个查询以提高效率。您可以在删除语句中使用连接来连接到父级别以过滤要删除的行。

Answer 4

我找到了一个很好的解决方案， 下：How to generate DELETE statements in PL/SQL, based on the tables FK relations?

生成函数：

IF OBJECT_ID('dbo.udfGetFullQualName') IS NOT NULL
    DROP FUNCTION dbo.udfGetFullQualName;

GO
CREATE FUNCTION dbo.udfGetFullQualName
(@ObjectId INT)
RETURNS VARCHAR (300)
AS
BEGIN
    DECLARE @schema_id AS BIGINT;
    SELECT @schema_id = schema_id
    FROM   sys.tables
    WHERE  object_id = @ObjectId;
    RETURN '[' + SCHEMA_NAME(@schema_id) + '].[' + OBJECT_NAME(@ObjectId) + ']';
END

GO
--============ Supporting Function dbo.udfGetOnJoinClause
IF OBJECT_ID('dbo.udfGetOnJoinClause') IS NOT NULL
    DROP FUNCTION dbo.udfGetOnJoinClause;

GO
CREATE FUNCTION dbo.udfGetOnJoinClause
(@fkNameId INT)
RETURNS VARCHAR (1000)
AS
BEGIN
    DECLARE @OnClauseTemplate AS VARCHAR (1000);
    SET @OnClauseTemplate = '[<@pTable>].[<@pCol>] = [<@cTable>].[<@cCol>] AND ';
    DECLARE @str AS VARCHAR (1000);
    SET @str = '';
    SELECT @str = @str + REPLACE(REPLACE(REPLACE(REPLACE(@OnClauseTemplate, '<@pTable>', OBJECT_NAME(rkeyid)), '<@pCol>', COL_NAME(rkeyid, rkey)), '<@cTable>', OBJECT_NAME(fkeyid)), '<@cCol>', COL_NAME(fkeyid, fkey))
    FROM   dbo.sysforeignkeys AS fk
    WHERE  fk.constid = @fkNameId; --OBJECT_ID('FK_ProductArrearsMe_ProductArrears')
    RETURN LEFT(@str, LEN(@str) - LEN(' AND '));
END

GO
--=========== CASECADE DELETE STORED PROCEDURE dbo.uspCascadeDelete
IF OBJECT_ID('dbo.uspCascadeDelete') IS NOT NULL
    DROP PROCEDURE dbo.uspCascadeDelete;

GO
CREATE PROCEDURE dbo.uspCascadeDelete
@ParentTableId VARCHAR (300), @WhereClause VARCHAR (2000), @ExecuteDelete CHAR (1)='N', --'N' IF YOU NEED DELETE SCRIPT
@FromClause VARCHAR (8000)='', @Level INT=0 -- TABLE NAME OR OBJECT (TABLE) ID (Production.Location) WHERE CLAUSE (Location.LocationID = 7) 'Y' IF WANT TO DELETE DIRECTLY FROM SP,  IF LEVEL 0, THEN KEEP DEFAULT
AS -- writen by Daniel Crowther 16 Dec 2004 - handles composite primary keys
SET NOCOUNT ON;
/* Set up debug */
DECLARE @DebugMsg AS VARCHAR (4000), 
@DebugIndent AS VARCHAR (50);
SET @DebugIndent = REPLICATE('---', @@NESTLEVEL) + '> ';
IF ISNUMERIC(@ParentTableId) = 0
    BEGIN -- assume owner is dbo and calculate id
        IF CHARINDEX('.', @ParentTableId) = 0
            SET @ParentTableId = OBJECT_ID('[dbo].[' + @ParentTableId + ']');
        ELSE
            SET @ParentTableId = OBJECT_ID(@ParentTableId);
    END
IF @Level = 0
    BEGIN
        PRINT @DebugIndent + ' **************************************************************************';
        PRINT @DebugIndent + ' *** Cascade delete ALL data from ' + dbo.udfGetFullQualName(@ParentTableId);
        IF @ExecuteDelete = 'Y'
            PRINT @DebugIndent + ' *** @ExecuteDelete = Y *** deleting data...';
        ELSE
            PRINT @DebugIndent + ' *** Cut and paste output into another window and execute ***';
    END
DECLARE @CRLF AS CHAR (2);
SET @CRLF = CHAR(13) + CHAR(10);
DECLARE @strSQL AS VARCHAR (4000);
IF @Level = 0
    SET @strSQL = 'SET NOCOUNT ON' + @CRLF;
ELSE
    SET @strSQL = '';
SET @strSQL = @strSQL + 'PRINT ''' + @DebugIndent + dbo.udfGetFullQualName(@ParentTableId) + ' Level=' + CAST (@@NESTLEVEL AS VARCHAR) + '''';
IF @ExecuteDelete = 'Y'
    EXECUTE (@strSQL);
ELSE
    PRINT @strSQL;
DECLARE curs_children CURSOR LOCAL FORWARD_ONLY
    FOR SELECT DISTINCT constid AS fkNameId, -- constraint name
                        fkeyid AS cTableId
        FROM   dbo.sysforeignkeys AS fk
        WHERE  fk.rkeyid <> fk.fkeyid -- WE DO NOT HANDLE self referencing tables!!!
               AND fk.rkeyid = @ParentTableId;
OPEN curs_children;
DECLARE @fkNameId AS INT, 
@cTableId AS INT, 
@cColId AS INT, 
@pTableId AS INT, 
@pColId AS INT;
FETCH NEXT FROM curs_children INTO @fkNameId, @cTableId; --, @cColId, @pTableId, @pColId
DECLARE @strFromClause AS VARCHAR (1000);
DECLARE @nLevel AS INT;
IF @Level = 0
    BEGIN
        SET @FromClause = 'FROM ' + dbo.udfGetFullQualName(@ParentTableId);
    END
WHILE @@FETCH_STATUS = 0
    BEGIN
        SELECT @strFromClause = @FromClause + @CRLF + '      INNER JOIN ' + dbo.udfGetFullQualName(@cTableId) + @CRLF + '       ON ' + dbo.udfGetOnJoinClause(@fkNameId);
        SET @nLevel = @Level + 1;
        EXECUTE dbo.uspCascadeDelete @ParentTableId = @cTableId, @WhereClause = @WhereClause, @ExecuteDelete = @ExecuteDelete, @FromClause = @strFromClause, @Level = @nLevel;
        SET @strSQL = 'DELETE FROM ' + dbo.udfGetFullQualName(@cTableId) + @CRLF + @strFromClause + @CRLF + 'WHERE   ' + @WhereClause + @CRLF;
        SET @strSQL = @strSQL + 'PRINT ''---' + @DebugIndent + 'DELETE FROM ' + dbo.udfGetFullQualName(@cTableId) + '     Rows Deleted: '' + CAST(@@ROWCOUNT AS VARCHAR)' + @CRLF + @CRLF;
        IF @ExecuteDelete = 'Y'
            EXECUTE (@strSQL);
        ELSE
            PRINT @strSQL;
        FETCH NEXT FROM curs_children INTO @fkNameId, @cTableId;
    --, @cColId, @pTableId, @pColId
    END
IF @Level = 0
    BEGIN
        SET @strSQL = @CRLF + 'PRINT ''' + @DebugIndent + dbo.udfGetFullQualName(@ParentTableId) + ' Level=' + CAST (@@NESTLEVEL AS VARCHAR) + ' TOP LEVEL PARENT TABLE''' + @CRLF;
        SET @strSQL = @strSQL + 'DELETE FROM ' + dbo.udfGetFullQualName(@ParentTableId) + ' WHERE ' + @WhereClause + @CRLF;
        SET @strSQL = @strSQL + 'PRINT ''' + @DebugIndent + 'DELETE FROM ' + dbo.udfGetFullQualName(@ParentTableId) + ' Rows Deleted: '' + CAST(@@ROWCOUNT AS VARCHAR)' + @CRLF;
        IF @ExecuteDelete = 'Y'
            EXECUTE (@strSQL);
        ELSE
            PRINT @strSQL;
    END
CLOSE curs_children;
DEALLOCATE curs_children;

用法：

EXEC uspCascadeDelete
@ParentTableId = '[VAULT].[Package]',
@WhereClause = 'CreatedBy  =''VZBMNSTEST'''

Answer 5

对于任何来这里寻找涉及同一表中的分层父/子关系的递归删除的人，这里有一个如何使用 CTE 执行此操作的示例：

declare @headid int

;with RowsToDelete as (
    SELECT head.id
    FROM ..hierarchicaltable head
    WHERE id = @headid

    UNION ALL

    SELECT child.id
    FROM ..hierarchicaltable child
    INNER JOIN RowsToDelete parent
        ON parent.id = child.parentid
)
delete ht
FROM ..hierarchicaltable ht
INNER JOIN RowsToDelete d
    ON ht.id = d.id

Answer 6

我相信这是可能的

DELETE FROM tbl1
INNER JOIN tbl2
ON tbl1.ID = tbl2.tbl1ID
WHERE
tbl1.somthing = x