我在80多个不同的表中有一个ReferenceID varchar(6)列。在需要分配ID的政府机构实施更改后,我需要将此扩展到整个数据库中的varchar(8)。
我希望声明一个游标来获取表名,如下所示:
DECLARE @TableName AS VARCHAR(200)
DECLARE TableCursor CURSOR LOCAL READ_ONLY FOR
SELECT t.name AS TableName
FROM sys.columns c
JOIN sys.tables t ON c.object_id = t.object_id
WHERE c.name = 'ReferenceID'
OPEN TableCursor
FETCH NEXT FROM TableCursor
INTO @TableName
然后按如下方式编辑类型:
ALTER TABLE @TableName ALTER COLUMN ReferenceID VARCHAR(8)
这会失败,因为该列是某些表中主键的一部分(PK中包含的列因表而异)。
我真的不想为每个表手动删除并重新创建每个PK。
在游标中,有没有办法在更改数据类型之前禁用PK然后重新启用它,或者在更改数据类型的任何一侧删除并重新创建PK,同时记住PK将依赖于我们目前正在看哪个表?
答案 0 :(得分:6)
您需要在NOT NULL
中明确指定ALTER TABLE ... ALTER COLUMN
,否则默认允许NULL
。 PK列中不允许这样做。
以下工作正常。
CREATE TABLE p
(
ReferenceID VARCHAR(6) NOT NULL PRIMARY KEY
)
INSERT INTO p VALUES ('AAAAAA')
ALTER TABLE p ALTER COLUMN ReferenceID VARCHAR(8) NOT NULL
当省略NOT NULL
时,会出现以下错误
Msg 5074, Level 16, State 1, Line 1
The object 'PK__p__E1A99A792180FB33' is dependent on column 'ReferenceID'.
Msg 4922, Level 16, State 9, Line 1
ALTER TABLE ALTER COLUMN ReferenceID failed because one or more objects access this column.
您的程序化方法需要考虑的几件事情是 需要暂时删除引用ReferenceID
列的任何外键,并确保不包含{ {1}}表示当前 可以为空的(非PK)NOT NULL
列。
答案 1 :(得分:3)
编辑如果你有一个混乱的varchar(6)和char(6)列的混乱数据库需要这个解决方案,这个数据库的开发时间超过10年(政府政策的变化很大)导致任何“良好的数据库设计”的尝试最终崩溃。) 结束编辑
对于那些说我必须放弃并重建PK的人,你是对的。索引和外键也需要删除和重新创建。
幸运的是,有一些可管理的索引和FK,因此我将这些作为“例外”处理,并在脚本开头一次一个地删除它们,然后一次一个地重新添加它们,在脚本的末尾(参见下面/ * * /中的两个部分)。
然后,SQL脚本的主体会将有关FK的完整详细信息提示到临时表中,然后遍历每个表名,删除FK,更改数据类型,重新添加FK。
汇编的SQL字符串在下面的脚本中是PRINTed。如果您打算重复使用此项(不提供任何保证,等等,请等等),请将这些评论从执行时间降低50%。
SET NOCOUNT ON
/* Handle exceptional tables here
* Remove indexes and foreign keys
* --Lots of "IF EXISTS ... ALTER TABLE <name> DROP CONSTRAINT <constraint name>, etc.
*/
--Declare variables
DECLARE @SQL VARCHAR(8000)
DECLARE @TableName VARCHAR(512)
DECLARE @ConstraintName VARCHAR(512)
DECLARE @tColumn VARCHAR(512)
DECLARE @Columns VARCHAR(8000)
DECLARE @IsDescending BIT
--Set up temporary table
SELECT
tbl.[schema_id],
tbl.name AS TableName,
i.NAME AS IndexName,
i.type_desc,
c.[column],
c.key_ordinal,
c.is_desc,
i.[object_id],
s.no_recompute,
i.[ignore_dup_key],
i.[allow_row_locks],
i.[allow_page_locks],
i.[fill_factor],
dsi.type,
dsi.name AS DataSpaceName
INTO #PKBackup
FROM
sys.tables AS tbl
INNER JOIN sys.indexes AS i
ON (
i.index_id > 0
AND i.is_hypothetical = 0
)
AND ( i.[object_id] = tbl.[object_id] )
INNER JOIN (
SELECT
ic.[object_id] ,
c.[name] [column] ,
ic.is_descending_key [is_desc],
ic.key_ordinal
FROM
sys.index_columns ic
INNER JOIN
sys.indexes i
ON
i.[object_id] = ic.[object_id]
AND
i.index_id = 1
AND
ic.index_id = 1
INNER JOIN
sys.tables t
ON
t.[object_id] = ic.[object_id]
INNER JOIN
sys.columns c
ON
c.[object_id] = t.[object_id]
AND
c.column_id = ic.column_id
) AS c
ON c.[object_id] = i.[object_id]
LEFT OUTER JOIN
sys.key_constraints AS k
ON
k.parent_object_id = i.[object_id]
AND
k.unique_index_id = i.index_id
LEFT OUTER JOIN
sys.data_spaces AS dsi
ON
dsi.data_space_id = i.data_space_id
LEFT OUTER JOIN
sys.xml_indexes AS xi
ON
xi.[object_id] = i.[object_id]
AND
xi.index_id = i.index_id
LEFT OUTER JOIN
sys.stats AS s
ON
s.stats_id = i.index_id
AND
s.[object_id] = i.[object_id]
WHERE
k.TYPE = 'PK'
DECLARE TableCursor CURSOR LOCAL READ_ONLY FOR
SELECT t.name AS TableName
FROM sys.columns c
JOIN sys.tables t ON c.object_id = t.object_id
WHERE
c.name = 'ReferenceID'
OPEN TableCursor
FETCH NEXT FROM TableCursor
INTO @TableName
WHILE @@FETCH_STATUS = 0
BEGIN
PRINT('--Updating ' + @TableName + '...')
SELECT @ConstraintName = PK.CONSTRAINT_NAME
FROM
INFORMATION_SCHEMA.TABLE_CONSTRAINTS PK
WHERE
PK.TABLE_NAME = @TableName
AND
PK.CONSTRAINT_TYPE = 'PRIMARY KEY'
--drop the constraint
--Some tables don't have a PK defined, only do the next bit if they do
IF (SELECT COUNT(*) FROM #PKBackup PK WHERE PK.TableName = @TableName) > 0
BEGIN
SET @SQL = 'ALTER TABLE @TableName DROP CONSTRAINT @ConstraintName'
SET @SQL = REPLACE(@SQL, '@TableName', @TableName)
SET @SQL = REPLACE(@SQL, '@ConstraintName', @ConstraintName)
PRINT @SQL
EXEC (@SQL)
END
--This is where we actually change the datatype of the column
SET @SQL = 'ALTER TABLE @TableName ALTER COLUMN ReferenceID VARCHAR(8)' + (SELECT CASE WHEN C.Is_Nullable = 'NO' THEN ' NOT NULL' ELSE '' END
FROM INFORMATION_SCHEMA.COLUMNS C
WHERE C.TABLE_NAME = @TableName AND C.COLUMN_NAME = 'ReferenceID')
SET @SQL = REPLACE(@SQL, '@TableName', @TableName)
PRINT(@SQL)
EXEC(@SQL)
--Recreate the constraint
--Some tables don't have a PK defined, only do the next bit if they do
IF (SELECT COUNT(*) FROM #PKBackup PK WHERE PK.TableName = @TableName) > 0
BEGIN
--First set up @SQL template
SELECT @SQL = 'ALTER TABLE [' + SCHEMA_NAME(PK.schema_id) + '].[' + PK.TableName
+ '] ADD CONSTRAINT [' + PK.IndexName
+ '] PRIMARY KEY ' + Type_desc + ' ( @Columns ) WITH '
+ '( PAD_INDEX = ' + CASE WHEN CAST(INDEXPROPERTY(pk.[object_id], PK.IndexName, N'IsPadIndex') AS BIT) = 0 THEN 'OFF'
ELSE 'ON'
END + ', '
+ 'STATISTICS_NORECOMPUTE = ' + CASE WHEN pk.no_recompute = 0 THEN 'OFF'
ELSE 'ON'
END
+ ', SORT_IN_TEMPDB = OFF, '
+ 'IGNORE_DUP_KEY = ' + CASE WHEN pk.[ignore_dup_key] = 0 THEN 'OFF'
ELSE 'ON'
END + ', '
+ 'ONLINE = OFF, '
+ 'ALLOW_ROW_LOCKS = ' + CASE WHEN pk.allow_row_locks = 0 THEN 'OFF'
ELSE 'ON'
END + ', '
+ 'ALLOW_PAGE_LOCKS = ' + CASE WHEN pk.allow_page_locks = 0 THEN 'OFF'
ELSE 'ON'
END + ', '
+ 'FILLFACTOR = ' + CASE WHEN pk.[fill_factor] = 0 THEN '100'
ELSE CONVERT(NVARCHAR, pk.[fill_factor])
END + ' '
+ ') ON [' + CASE WHEN 'FG' = pk.[type] THEN pk.DataSpaceName
ELSE N''
END + ']'
FROM
#PKBackup PK WHERE PK.TableName = @TableName
SET @SQL = REPLACE(@SQL, '@TableName', @TableName)
SET @SQL = REPLACE(@SQL, '@ConstraintName', @ConstraintName)
--Second, build up @Columns
SET @Columns = ' '
DECLARE ColumnCursor CURSOR LOCAL READ_ONLY FOR
SELECT pk.[column], PK.is_desc
FROM #PKBackup PK
WHERE PK.TableName = @TableName
ORDER BY PK.key_ordinal ASC
OPEN ColumnCursor
FETCH NEXT FROM ColumnCursor
INTO @tColumn, @IsDescending
WHILE @@FETCH_STATUS = 0
BEGIN
SET @Columns = @Columns + @tColumn + CASE WHEN @IsDescending = 1 THEN ' DESC, ' ELSE ' ASC, ' END
--Get the next TableName
FETCH NEXT FROM ColumnCursor
INTO @tColumn, @IsDescending
END
--Tidy up
CLOSE ColumnCursor
DEALLOCATE ColumnCursor
--Delete the last comma
SET @Columns = LEFT(@Columns, LEN(@Columns) - 1)
END
--Recreate the constraint
SET @SQL = REPLACE(@SQL, '@Columns', @Columns)
PRINT @SQL
EXEC (@SQL)
PRINT('--Done
')
SET @SQL = ''
--Get the next TableName
FETCH NEXT FROM TableCursor
INTO @TableName
END
--Tidy up
CLOSE TableCursor
DEALLOCATE TableCursor
DROP TABLE #PKBackup
/* Handle exceptional tables here
* Replace indexes and foreign keys that were removed at the start
*/
SET NOCOUNT OFF
答案 2 :(得分:0)
您需要将ALTER
语句作为动态SQL执行:将语句构建为SQL字符串并将其传递给sp_executesql
。
答案 3 :(得分:0)
根据我30多年的数据库经验,一个常数是随着数据需求的变化,需要不断更改您正在使用的任何数据库的结构。此外,有很多实例,其中自动增量主键不是最合适的,特别是当您想要确保在使用数据库的程序时,通过DBMS(例如SQL Server)直接可以有效地访问数据的情况。不再可用。数据库管理的一个重大缺陷是程序死亡时数据库的通常不可穿透性 - 这与长期数据管理原理完全相反。
因此,无法轻易更改主键字段的大小并不是因为数据库设计不佳,这是因为处理数据库的DBMS和SQL工具严重不足,显然是由程序员用理论设计而不是对现实世界的真正理解。此类编程缺陷的其他示例是从0开始而不是1开始的数组索引(由于必须添加或减去1以从索引切换到计数而产生的错误数量很多),数值变量无法原生处理空值,等等我期待有一天数据库结构修改被视为主流必需品,而不是所谓的糟糕的数据库设计。