Best Practices for using DATETIME column in sql server 2005 and 2008
In production environment we have seen that there are certain times when you make a good query by applying best practices but you don’t get good results at the time of sorting.
This might happen because of DATETIME columns which are mostly used for sorting purpose and indexes for these columns do not give you appropriate results.
Mentioned below is tip for using DATETIME column which can enhance the speed of your query while maintaining same number of indexes.
This technique requires mentioned below changes in your table: -
- One Persisted Computed Column (BIGINT) which will store DATETIME in integer format by removing special symbols from DATETIME
- This computed column will be added in the covering index
- Query will use this column instead of DATETIME column
Example :
/* Create Sample Table */
CREATE TABLE [dbo].[MyTable](
[C1] [int] IDENTITY(1,1) NOT NULL,
[C2] [DATETIME] NULL,
CONSTRAINT [PK_MyTable] PRIMARY KEY CLUSTERED
(
[C1] ASC
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, IGNORE_DUP_KEY = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [PRIMARY]
) ON [PRIMARY]
GO
/* Data Insert */
SET IDENTITY_INSERT [dbo].[MyTable] ON
INSERT [dbo].[MyTable] ([C1], [C2]) VALUES (1, CAST(0x00009E5E00000000 AS DATETIME))
INSERT [dbo].[MyTable] ([C1], [C2]) VALUES (2, CAST(0x00009E5F00000000 AS DATETIME))
INSERT [dbo].[MyTable] ([C1], [C2]) VALUES (3, CAST(0x00009E6000000000 AS DATETIME))
INSERT [dbo].[MyTable] ([C1], [C2]) VALUES (4, CAST(0x00009E6100000000 AS DATETIME))
INSERT [dbo].[MyTable] ([C1], [C2]) VALUES (5, CAST(0x00009E7D00000000 AS DATETIME))
INSERT [dbo].[MyTable] ([C1], [C2]) VALUES (6, CAST(0x00009E7E00000000 AS DATETIME))
SET IDENTITY_INSERT [dbo].[MyTable] OFF
/* Query */
DECLARE @D1 DATETIME
DECLARE @D2 DATETIME
SET @D1 = GETDATE()-10
SET @D2 = GETDATE()
SELECT C1
,C2
FROM MyTable
WHERE C2 BETWEEN @D1 AND @D2
/* Change in Table Add New Column that will store DATETIME in Integer Format */
ALTER TABLE MyTable
ADD [C3] AS (CONVERT([bigint],CONVERT([varchar],[C2],(112))+replace(CONVERT([varchar],[C2],(114)),':',''),0)) PERSISTED
/* Existing Index for query */
CREATE NONCLUSTERED INDEX [Index_For_C2] ON MyTable
(
[C2] ASC,
[C1] ASC
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, SORT_IN_TEMPDB = OFF,
IGNORE_DUP_KEY = OFF, DROP_EXISTING = OFF, ONLINE = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [PRIMARY]
GO
/* Add New Column to Existing Index for query */
CREATE NONCLUSTERED INDEX [Index_For_C2] ON MyTable
(
[C3] ASC,
[C2] ASC,
[C1] ASC
)WITH (STATISTICS_NORECOMPUTE = OFF, SORT_IN_TEMPDB = OFF, IGNORE_DUP_KEY = OFF,
DROP_EXISTING = ON, ONLINE = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [PRIMARY]
GO
/*Updated Query using Integer Column instead of DATETIME having same result set*/
DECLARE @D1 BIGINT
DECLARE @D2 BIGINT
SET @D1 = (CONVERT([bigint],CONVERT([varchar],GETDATE()-10,(112))+replace(CONVERT([varchar],GETDATE()-10,(114)),':',''),0))
SET @D2 = (CONVERT([bigint],CONVERT([varchar],GETDATE(),(112))+replace(CONVERT([varchar],GETDATE(),(114)),':',''),0))
SELECT C1
,C2
FROM MyTable
WHERE C3 BETWEEN @D1 AND @D2
The above query will now use the new updated index and will be much faster as it will get records by using an integer column instead of DATETIME column.
