Related
Closed. This question needs to be more focused. It is not currently accepting answers.
Want to improve this question? Update the question so it focuses on one problem only by editing this post.
Closed last year.
Improve this question
Can anyone recommend how I can speed up this code and primarily the cursor? The code is an SQL Server db query that creates a trigger on INSERT, UPDATE, or DELETE. It writes a record to a changlog table identifying the type of change (I, U, or D) and then saves the old value and new value of each affected column for each row in a details table.
I want this to be generic so I can easily reuse it for any table I throw at it that has a unique column I can filter on. Writing the whole row of changes to a cloned structure audit table is not an option unfortunately.
Any help is greatly appreciated, I am not the greatest at query optimization and welcome any feedback or rewrites.. Thanks!
ALTER TRIGGER [dbo].[tbl_Address_ChangeTracking] ON [dbo].[tbl_Address]
AFTER INSERT, DELETE, UPDATE
AS
BEGIN
-- SET NOCOUNT ON added to prevent extra result sets from
-- interfering with SELECT statements.
SET NOCOUNT ON
--SET XACT_ABORT ON
-- Get the table name of the current process
DECLARE #TableName VARCHAR(25)
SET #TableName = COALESCE(
(
SELECT SCHEMA_NAME(schema_id) + '.' + OBJECT_NAME(parent_object_id)
FROM sys.objects
WHERE sys.objects.name = OBJECT_NAME(##PROCID) AND
SCHEMA_NAME(sys.objects.schema_id) = OBJECT_SCHEMA_NAME(##PROCID)
), 'Unknown')
--Declare our cursor to navigate the records in inserted and deleted
DECLARE #cursorSQL AS NVARCHAR(MAX) = ''
DECLARE #PrimaryID AS VARCHAR(MAX) = ''
DROP TABLE IF EXISTS #inserted1TableTemp
DROP TABLE IF EXISTS #inserted2TableTemp
DROP TABLE IF EXISTS #deletedTableTemp
DECLARE #ourLogCursor CURSOR
--If we have a record in inserted and deleted this is an update record and we should pull from the inserted table and assume
--this is one update or many update statements
IF EXISTS
(
SELECT 1
FROM inserted
) AND
EXISTS
(
SELECT 1
FROM deleted
)
BEGIN
SELECT *
INTO #inserted1TableTemp
FROM inserted
SET #cursorSQL = 'SET #ourLogCursor = CURSOR FOR SELECT AddressID FROM #inserted1TableTemp; OPEN #ourLogCursor;'
END
--If we have an inserted record and no deleted record this is an insert and we pull from the inserted table
IF EXISTS
(
SELECT 1
FROM inserted
) AND
NOT EXISTS
(
SELECT 1
FROM deleted
)
BEGIN
DROP TABLE IF EXISTS #inserted2TableTemp
DROP TABLE IF EXISTS #inserted1TableTemp
DROP TABLE IF EXISTS #deletedTableTemp
SELECT *
INTO #inserted2TableTemp
FROM inserted
SET #cursorSQL = 'SET #ourLogCursor = CURSOR FOR SELECT AddressID FROM #inserted2TableTemp; OPEN #ourLogCursor;'
END
--If we have a deleted record and no insert record this is a deletion and we pull from the deleted table
IF NOT EXISTS
(
SELECT 1
FROM inserted
) AND
EXISTS
(
SELECT 1
FROM deleted
)
BEGIN
DROP TABLE IF EXISTS #inserted1TableTemp
DROP TABLE IF EXISTS #inserted2TableTemp
DROP TABLE IF EXISTS #deletedTableTemp
SELECT *
INTO #deletedTableTemp
FROM deleted
SET #cursorSQL = 'SET #ourLogCursor = CURSOR FOR SELECT AddressID FROM #deletedTableTemp; OPEN #ourLogCursor;'
END
--If we have a deleted record and no insert record this is a deletion and we pull from the deleted table
IF NOT EXISTS
(
SELECT 1
FROM inserted
) AND
NOT EXISTS
(
SELECT 1
FROM deleted
)
BEGIN
RETURN;
END
--Execute our dynamic SQL that sets the correct FOR SELECT statment for the cursor. Pass #ourCursorLog as an input param, and then grab the output
--so the results are available outside the scope of the executesql call
EXEC sys.sp_executesql #cursorSQL, N'#ourLogCursor CURSOR OUTPUT', #ourLogCursor OUTPUT;
FETCH NEXT FROM #ourLogCursor INTO #PrimaryID
DECLARE #xmlOld XML
DECLARE #xmlNew XML
DECLARE #SummaryID INT
SET #TableName = COALESCE(
(
SELECT SCHEMA_NAME(schema_id) + '.' + OBJECT_NAME(parent_object_id)
FROM sys.objects
WHERE sys.objects.name = OBJECT_NAME(##PROCID) AND
SCHEMA_NAME(sys.objects.schema_id) = OBJECT_SCHEMA_NAME(##PROCID)
), 'Unknown')
--Navigate all our rows
WHILE ##FETCH_STATUS = 0
BEGIN
DROP TABLE IF EXISTS #tmp_AddressesChangelogTrigger
DROP TABLE IF EXISTS #tmp_AddressesChangelogTriggerXML1
DROP TABLE IF EXISTS #tmp_AddressesChangelogTriggerXML2
DROP TABLE IF EXISTS #tmp_AddressesChangelogTriggerXMLsWithDifferences
--Get the deleted and inserted records as xml for comparison against each other
SET #xmlNew =
(
SELECT *
FROM deleted AS [TABLE]
WHERE AddressID = #PrimaryID
ORDER BY AddressID FOR XML AUTO, ELEMENTS
)
SET #xmlOld =
(
SELECT *
FROM inserted AS [TABLE]
WHERE AddressID = #PrimaryID
ORDER BY AddressID FOR XML AUTO, ELEMENTS
)
CREATE TABLE #tmp_AddressesChangelogTriggerXML1
(
NodeName VARCHAR(MAX), Value VARCHAR(MAX)
)
CREATE TABLE #tmp_AddressesChangelogTriggerXML2
(
NodeName VARCHAR(MAX), Value VARCHAR(MAX)
)
--Extract the values and column names
INSERT INTO #tmp_AddressesChangelogTriggerXML2( NodeName, Value )
--Throw the XML into temp tables with the column name and value
SELECT N.value( 'local-name(.)', 'nvarchar(MAX)' ) AS NodeName, N.value( 'text()[1]', 'nvarchar(MAX)' ) AS VALUE
FROM #xmlNew.nodes( '/TABLE/*' ) AS T(N)
INSERT INTO #tmp_AddressesChangelogTriggerXML1( NodeName, Value )
SELECT N.value( 'local-name(.)', 'nvarchar(MAX)' ) AS NodeName, N.value( 'text()[1]', 'nvarchar(MAX)' ) AS VALUE
FROM #xmlOld.nodes( '/TABLE/*' ) AS T(N)
--Get the differences into a temp table
SELECT *
INTO #tmp_AddressesChangelogTriggerXMLsWithDifferences
FROM
(
SELECT COALESCE(A.NodeName, B.NodeName) AS NodeName, B.Value AS OldValue, A.Value AS NewValue
FROM #tmp_AddressesChangelogTriggerXML1 AS A
FULL OUTER JOIN #tmp_AddressesChangelogTriggerXML2 AS B ON A.NodeName = B.NodeName
WHERE A.Value <> B.Value
) AS tmp
--If anything changed thhen start our write statments
IF
(
SELECT COUNT(*)
FROM #tmp_AddressesChangelogTriggerXMLsWithDifferences
) > 0
BEGIN
BEGIN TRY
-- Now create the Summary record
--BEGIN TRANSACTION WRITECHANGELOGRECORDS
INSERT INTO TableChangeLogSummary( ID, ModifiedDate, ChangeType, TableName )
--Get either insert, or if no insert value, get the delete value
--Set the update type, I, D, U
--Compare values with a full outer join
--Filter on the ID we are on in the CURSOR
SELECT COALESCE(I.AddressID, D.AddressID), GETDATE(),
CASE
WHEN D.AddressID IS NULL THEN 'I'
WHEN I.AddressID IS NULL THEN 'D'
ELSE 'U'
END, #TableName
FROM inserted AS I
FULL OUTER JOIN deleted AS D ON I.AddressID = D.AddressID
WHERE( I.AddressID = #PrimaryID OR
I.AcesAddressID IS NULL
) AND
( D.AddressID = #PrimaryID OR
D.AcesAddressID IS NULL
)
--Get the last summary id that was inserted so we can use it in the detail record
SET #SummaryID = (SELECT IDENT_CURRENT('TableChangeLogSummary'))
--Insert our
INSERT INTO TableChangeLogDetail( SummaryID, ColumnName, OldValue, NewValue )
SELECT #SummaryID, T.NodeName, T.OldValue, T.NewValue
FROM #tmp_AddressesChangelogTriggerXMLsWithDifferences AS T
--COMMIT TRANSACTION WRITECHANGELOGRECORDS
--PRINT 'RECORD WRITTEN'
END TRY
BEGIN CATCH
DECLARE #errorXML XML
SET #errorXML = (SELECT ERROR_NUMBER() AS ErrorNumber, ERROR_STATE() AS ErrorState, ERROR_SEVERITY() AS ErrorSeverity, ERROR_PROCEDURE() AS ErrorProcedure, ERROR_LINE() AS ErrorLine, ERROR_MESSAGE() AS ErrorMessage FOR XML RAW)
DECLARE #errorXMLText NVARCHAR(MAX) = ''
SET #errorXMLText = (SELECT CAST(#errorXML AS NVARCHAR(MAX)))
RAISERROR(#errorXMLText, 16, 1) WITH NOWAIT
END CATCH
END
--Go to the next record and process
FETCH NEXT FROM #ourLogCursor INTO #PrimaryID
END
CLOSE #ourLogCursor
DEALLOCATE #ourLogCursor
END
Acknowledging the recommendation for using change data tracking and caution against putting too much logic into triggers, the following is a refactoring (and some outright rewriting) of your change capture logic.
The updated logic makes a single pass through the data, handing all affected records at once. Given the requirements, I think it is pretty close to optimal, but there may still be room for improvements. The conversion to and from XML likely adds a significant bit of overhead. The alternative would be to dynamically generate and apply custom triggers for each table that explicitly reference all of the data columns individually to get the details and UNION them together.
I also refined the value comparison to better handle nulls, case sensitivity, and potential trailing space changes.
The code below is not in the form of a trigger, but in a form suitable for stand-alone testing. I figured you (and any others who may be interested) would want to test Once checked out, you should be able to retrofit it back into your trigger.
Note that this is not a 100% generalized solution. Some column types may not be supported. The logic currently assumes a single column primary key of type integer. Changes would be required to handle deviations from these (and possibly some currently unidentified) constraints.
-- Simulated change log tables
DECLARE #TableChangeLogSummary TABLE (ID INT IDENTITY(1,1), KeyValue INT NOT NULL, ModifiedDate DATETIME NOT NULL, ChangeType CHAR(1) NOT NULL, TableName NVARCHAR(1000) NOT NULL )
DECLARE #TableChangeLogDetails TABLE (ID INT IDENTITY(1,1), SummaryID int NOT NULl, ColumnName NVARCHAR(1000) NOT NULL, OldValue NVARCHAR(MAX), NewValue NVARCHAR(MAX))
-- Simulated system defined inserted/deleted tables
DECLARE #inserted TABLE (ID INTEGER, Value1 NVARCHAR(100), Value2 BIT, Value3 FLOAT)
DECLARE #deleted TABLE (ID INTEGER, Value1 NVARCHAR(100), Value2 BIT, Value3 FLOAT)
-- Test data
INSERT #inserted
VALUES
(1, 'AAA', 0, 3.14159), -- Insert
(2, 'BBB', 1, null), -- Mixed updates including null to non-null and non-null to null
(3, 'CCC', 0, 0), -- Trailing space change
(4, 'DDD', null, 1.68), -- No changes
(5, '', 0, null), -- No changes with blanks and nulls
(6, null, null, null), -- No changes all nulls
(7, null, null, null) -- Insert all nulls (summary with key, but no details will be logged)
INSERT #deleted
VALUES
(2, 'bbb', null, 2.73),
(3, 'CCC ', 0, 0),
(4, 'DDD', null, 1.68),
(5, '', 0, null),
(6, null, null, null),
(8, null, null, null), -- Delete all null values (summary with key, but no details will be logged)
(9, 'ZZZ', 999, 999.9) -- Delete non-nulls
--- Now the real work begins...
-- Set table and information. Assumes table has exactly one PK column. Later logic assumes an INT.
DECLARE #TableName NVARCHAR(1000) = 'MyTable' -- To be extracted from the parent object of the trigger
DECLARE #KeyColumnName SYSNAME = 'ID' -- This can be fixed if known or derived on the fly from the primary key definition
-- Extract inserted and/or deleted data
DECLARE #InsertedXml XML = (
SELECT *
FROM #inserted
FOR XML PATH('inserted'), TYPE
)
DECLARE #DeletedXml XML = (
SELECT *
FROM #deleted
FOR XML PATH('deleted'), TYPE
)
-- Parse and reassange the captured key and data values
DECLARE #TempDetails TABLE(
KeyValue INT NOT NULL,
ChangeType CHAR(1) NOT NULL,
ColumnName VARCHAR(1000) NOT NULL,
IsKeyColumn BIT NOT NULL,
NewValue NVARCHAR(MAX),
OldValue NVARCHAR(MAX))
INSERT #TempDetails
SELECT
KeyValue = COALESCE(I.KeyValue, D.KeyValue),
ChangeType = CASE WHEN D.KeyValue IS NULL THEN 'I' WHEN I.KeyValue IS NULL THEN 'D' ELSE 'U' END,
ColumnName = COALESCE(I.ColumnName, D.ColumnName),
IsKeyColumn = K.IsKeyColumn,
NewValue = I.Value,
OldValue = D.Value
FROM (
SELECT K.KeyValue, C.ColumnName, C.Value
FROM #InsertedXml.nodes( '/inserted' ) R(Row)
CROSS APPLY (
SELECT KeyValue = C.Col.value('text()[1]', 'int')
FROM R.Row.nodes( './*' ) C(Col)
WHERE C.Col.value( 'local-name(.)', 'nvarchar(MAX)' ) = #KeyColumnName
) K
CROSS APPLY (
SELECT ColumnName = C.Col.value('local-name(.)', 'nvarchar(MAX)'), Value = C.Col.value('text()[1]', 'nvarchar(MAX)')
FROM R.Row.nodes( './*' ) C(Col)
) C
) I
FULL OUTER JOIN (
SELECT K.KeyValue, C.ColumnName, C.Value
FROM #DeletedXml.nodes( '/deleted' ) R(Row)
CROSS APPLY (
SELECT KeyValue = C.Col.value('text()[1]', 'int')
FROM R.Row.nodes( './*' ) C(Col)
WHERE C.Col.value( 'local-name(.)', 'nvarchar(MAX)' ) = #KeyColumnName
) K
CROSS APPLY (
SELECT ColumnName = C.Col.value('local-name(.)', 'nvarchar(MAX)'), Value = C.Col.value('text()[1]', 'nvarchar(MAX)')
FROM R.Row.nodes( './*' ) C(Col)
) C
) D
ON D.KeyValue = I.KeyValue
AND D.ColumnName = I.ColumnName
CROSS APPLY (
SELECT IsKeyColumn = CASE WHEN COALESCE(I.ColumnName, D.ColumnName) = #KeyColumnName THEN 1 ELSE 0 END
) K
WHERE ( -- We need to be careful about edge cases here
(I.Value IS NULL AND D.Value IS NOT NULL)
OR (I.Value IS NOT NULL AND D.Value IS NULL)
OR I.Value <> D.Value COLLATE Latin1_General_Bin -- Precise compare (case and accent sensitive)
OR DATALENGTH(I.Value) <> DATALENGTH(D.Value) -- Catch trailing space cases
OR K.IsKeyColumn = 1
)
-- Get rid of updates with no changes, but keep key-only inserts or deletes
DELETE T
FROM #TempDetails T
WHERE T.IsKeyColumn = 1
AND T.ChangeType = 'U'
AND NOT EXISTS (
SELECT *
FROM #TempDetails T2
WHERE T2.KeyValue = T.KeyValue
AND T2.IsKeyColumn = 0
)
-- Local table to capture and link SummaryID between the summary and details tables
DECLARE #CaptureSummaryID TABLE (SummaryID int, KeyValue INT NOT NULL)
-- Insert change summary and capture the assigned Summary ID via the OUTPUT clause
INSERT INTO #TableChangeLogSummary (KeyValue, ModifiedDate, ChangeType, TableName)
OUTPUT INSERTED.id, INSERTED.KeyValue INTO #CaptureSummaryID
SELECT T.KeyValue, ModifiedDate = GETDATE(), T.ChangeType, TableName = #TableName
FROM #TempDetails T
WHERE T.IsKeyColumn = 1
ORDER BY T.KeyValue -- Optional, but adds consistancy
-- Insert change details
INSERT INTO #TableChangeLogDetails (SummaryID, ColumnName, OldValue, NewValue)
SELECT S.SummaryID, T.ColumnName, T.OldValue, T.NewValue
FROM #CaptureSummaryID S
JOIN #TempDetails T ON T.KeyValue = S.KeyValue
WHERE T.IsKeyColumn = 0
ORDER BY T.ColumnName -- Optional, but adds consistancy
-- View test results
SELECT 'Change Log:', *
FROM #TableChangeLogSummary S
LEFT JOIN #TableChangeLogDetails D ON D.SummaryID = S.ID
ORDER BY S.ID, D.ID
I am using Sql-Server 2016 in a C# application.
Let's say I have two tables:
CREATE TABLE Table_A
(
UserID NVARCHAR2(15),
FullName NVARCHAR2(25),
Available NUMBER(1),
MachineID NVARCHAR2(20),
myDate date
);
and
CREATE TABLE Table_B
(
UserID NVARCHAR2(15),
FullName NVARCHAR2(25),
Team NVARCHAR2(15),
MachineID NVARCHAR2(20),
Stuff NUMBER(2)
);
I want to perform a global select so that I will get as result data from both tables, somehow concatenated and of course, when a column does not exist in one of the tables, that column to be automatically populated with NULL, and if a column exists on both tables the results must be merged in a single column.
The first solution that pops-up is a UNION with NULL aliases for the missing columns, sure. The problem is that at runtime I will not be able to know in advance which tables are interrogated so that I could anticipate the column names. I need a more general solution.
The expected result from the two tables must look like this:
user_Table_A; fullName_Table_A; 1; machineID_Table_A; 12-JUN-18; NULL; 10;
user_Table_B; fullName_Table_B; NULL; machineID_Table_B; NULL; team_Table_B; 20;
The data for the two tables is inserted with the following commands:
INSERT INTO Table_A VALUES ('user_Table_A', 'fullName_Table_A', 1, 'machineID_Table_A', TO_DATE('12-06-2018', 'DD-MM-YYYY'));
INSERT INTO Table_B VALUES ('user_Table_B', 'fullName_Table_B', 'team_Table_B', 'machineID_Table_B', 20);
You can do something like this. I havent have time to completely tweak it, so there can be something the order of the columns. But perhaps it can get you started:
You also write that you use Oracle - Im not sure what you wanted, but this is in pure sql-server version.
SQL:
IF OBJECT_ID('tempdb..#temp') IS NOT NULL
/*Then it exists*/
DROP TABLE #temp;
GO
DECLARE #SQLList nvarchar(max)
DECLARE #SQLList2 nvarchar(max)
DECLARE #SQL nvarchar(max)
with table_a as (
select column_name as Table_aColumnName,ORDINAL_POSITION from INFORMATION_SCHEMA.COLUMNS
where TABLE_NAME = 'table_a'
)
,
table_b as (
select column_name as Table_bColumnName,ORDINAL_POSITION from INFORMATION_SCHEMA.COLUMNS
where TABLE_NAME = 'table_b'
)
,preresult as (
select case when Table_aColumnName IS null then 'NULL as ' + Table_bColumnName else Table_aColumnName end as Table_a_ColumnName,case when Table_bColumnName IS null then 'NULL as ' +Table_aColumnName else Table_bColumnName end as Table_b_ColumnName
,a.ORDINAL_POSITION,b.ORDINAL_POSITION as Table_b_Ordinal from table_a a full join Table_B b on a.Table_aColumnName = b.Table_bColumnName
)
select * into #temp from preresult
SET #SQLList = (
select distinct display = STUFF((select ','+table_a_columnName from #temp b order by table_b_ordinal FOR XML PATH('')),1,1,'') from #temp a
)
SET #SQLList2 = (
select distinct display = STUFF((select ','+table_b_columnName from #temp b order by Table_b_Ordinal FOR XML PATH('')),1,1,'') from #temp a
)
SET #SQL = 'select ' +#SQLList +' from dbo.Table_a union all select ' + #SQLList2 + ' from dbo.table_b'
exec(#SQL)
Result:
I am trying to import data from a source that I have no control of the data quality. I know how to remove/replace most anything but I have encountered a field that contains a single ` (backtick), I need to replace this with a NULL. I am needing to remove the backtick because I convert the data from the varchar field that it is to a numeric. I have tried versions of this:
Update dbo.Sale
Set Stamps = NULL
where Stamps = ''''
You can use CHAR(39) for clarity's sake:
UPDATE dbo.Sale
SET Stamps = NULL
WHERE Stamps = CHAR(39);
this may help you
DECLARE #t TABLE (MyColumn VARCHAR(255))
INSERT INTO #t
( MyColumn )
VALUES ( '''' -- MyColumn - varchar(255)
)
SELECT *
FROM #t
UPDATE #t
SET MyColumn = NULL
WHERE MyColumn = CHAR(39)
SELECT *
FROM #t
OR
UPDATE #t
SET MyColumn = NULL
WHERE MyColumn = ''''
I am working on a tool (using c# winforms) which generates SQL scripts of data which are independent of identities but yet not violating constraints. While walking through all possibilities I got into below trouble. I need to insert data considering the fact a column (which allows null) may exists or not.
For example, a column as VAL2 exists in dev box, but may or may not exists in prod box.
CREATE TABLE TEMP ( COL1 VARCHAR(50) , VAL1 VARCHAR(50)) -- OPTIONAL/MAY EXISTS VAL2 VARCHAR(50)
IF EXISTS (SELECT TOP(1) COLUMN_NAME FROM INFORMATION_SCHEMA.COLUMNS WHERE TABLE_NAME = 'TEMP' AND COLUMN_NAME = 'VAL2')
BEGIN
INSERT INTO TEMP (COL1,VAL1,VAL2) VALUES ('1','YES','NO')
INSERT INTO TEMP (COL1,VAL1,VAL2) VALUES ('2','NO','NO')
END
ELSE
BEGIN
INSERT INTO TEMP (COL1,VAL1) VALUES ('1','YES')
INSERT INTO TEMP (COL1,VAL1) VALUES ('2','NO')
END
GO
But this method fails miserably if that column is missing even though its should be flowing through else by syntax.
Note: I had for now made a work around but adding that column and dropping it at end if only created through current script. But the problem is the execution is through limited access grant which rules possibility of CREATE/DROP statement.This may be wrong, but that's the way it is for me.
If there is any alternate way preferably which doesn't mandates adding this column , please do guide me.
You'll have to use dynamic sql to accomplish that:
CREATE TABLE TEMP
(
COL1 VARCHAR(50) ,
VAL1 VARCHAR(50)
)
-- OPTIONAL/MAY EXISTS VAL2 VARCHAR(50)
IF EXISTS ( SELECT TOP ( 1 )
COLUMN_NAME
FROM INFORMATION_SCHEMA.COLUMNS
WHERE TABLE_NAME = 'TEMP'
AND COLUMN_NAME = 'VAL2' )
BEGIN
DECLARE #sql NVARCHAR(MAX)
SELECT #sql = 'INSERT INTO TEMP (COL1,VAL1,VAL2) VALUES (''1'',''YES'',''NO'')
INSERT INTO TEMP (COL1,VAL1,VAL2) VALUES (''2'',''NO'',''NO'')'
EXEC sp_executesql #sql
END
ELSE
BEGIN
INSERT INTO TEMP
( COL1, VAL1 )
VALUES ( '1', 'YES' )
INSERT INTO TEMP
( COL1, VAL1 )
VALUES ( '2', 'NO' )
END
I have a table that has a column with a default value:
create table t (
value varchar(50) default ('something')
)
I'm using a stored procedure to insert values into this table:
create procedure t_insert (
#value varchar(50) = null
)
as
insert into t (value) values (#value)
The question is, how do I get it to use the default when #value is null? I tried:
insert into t (value) values ( isnull(#value, default) )
That obviously didn't work. Also tried a case statement, but that didn't fair well either. Any other suggestions? Am I going about this the wrong way?
Update: I'm trying to accomplish this without having to:
maintain the default value in multiple places, and
use multiple insert statements.
If this isn't possible, well I guess I'll just have to live with it. It just seems that something this should be attainable.
Note: my actual table has more than one column. I was just quickly writing an example.
Christophe,
The default value on a column is only applied if you don't specify the column in the INSERT statement.
Since you're explicitiy listing the column in your insert statement, and explicity setting it to NULL, that's overriding the default value for that column
What you need to do is "if a null is passed into your sproc then don't attempt to insert for that column".
This is a quick and nasty example of how to do that with some dynamic sql.
Create a table with some columns with default values...
CREATE TABLE myTable (
always VARCHAR(50),
value1 VARCHAR(50) DEFAULT ('defaultcol1'),
value2 VARCHAR(50) DEFAULT ('defaultcol2'),
value3 VARCHAR(50) DEFAULT ('defaultcol3')
)
Create a SPROC that dynamically builds and executes your insert statement based on input params
ALTER PROCEDURE t_insert (
#always VARCHAR(50),
#value1 VARCHAR(50) = NULL,
#value2 VARCHAR(50) = NULL,
#value3 VARCAHR(50) = NULL
)
AS
BEGIN
DECLARE #insertpart VARCHAR(500)
DECLARE #valuepart VARCHAR(500)
SET #insertpart = 'INSERT INTO myTable ('
SET #valuepart = 'VALUES ('
IF #value1 IS NOT NULL
BEGIN
SET #insertpart = #insertpart + 'value1,'
SET #valuepart = #valuepart + '''' + #value1 + ''', '
END
IF #value2 IS NOT NULL
BEGIN
SET #insertpart = #insertpart + 'value2,'
SET #valuepart = #valuepart + '''' + #value2 + ''', '
END
IF #value3 IS NOT NULL
BEGIN
SET #insertpart = #insertpart + 'value3,'
SET #valuepart = #valuepart + '''' + #value3 + ''', '
END
SET #insertpart = #insertpart + 'always) '
SET #valuepart = #valuepart + + '''' + #always + ''')'
--print #insertpart + #valuepart
EXEC (#insertpart + #valuepart)
END
The following 2 commands should give you an example of what you want as your outputs...
EXEC t_insert 'alwaysvalue'
SELECT * FROM myTable
EXEC t_insert 'alwaysvalue', 'val1'
SELECT * FROM myTable
EXEC t_insert 'alwaysvalue', 'val1', 'val2', 'val3'
SELECT * FROM myTable
I know this is a very convoluted way of doing what you need to do.
You could probably equally select the default value from the InformationSchema for the relevant columns but to be honest, I might consider just adding the default value to param at the top of the procedure
Try an if statement ...
if #value is null
insert into t (value) values (default)
else
insert into t (value) values (#value)
As far as I know, the default value is only inserted when you don't specify a value in the insert statement. So, for example, you'd need to do something like the following in a table with three fields (value2 being defaulted)
INSERT INTO t (value1, value3) VALUES ('value1', 'value3')
And then value2 would be defaulted. Maybe someone will chime in on how to accomplish this for a table with a single field.
Probably not the most performance friendly way, but you could create a scalar function that pulls from the information schema with the table and column name, and then call that using the isnull logic you tried earlier:
CREATE FUNCTION GetDefaultValue
(
#TableName VARCHAR(200),
#ColumnName VARCHAR(200)
)
RETURNS VARCHAR(200)
AS
BEGIN
-- you'd probably want to have different functions for different data types if
-- you go this route
RETURN (SELECT TOP 1 REPLACE(REPLACE(REPLACE(COLUMN_DEFAULT, '(', ''), ')', ''), '''', '')
FROM information_schema.columns
WHERE table_name = #TableName AND column_name = #ColumnName)
END
GO
And then call it like this:
INSERT INTO t (value) VALUES ( ISNULL(#value, SELECT dbo.GetDefaultValue('t', 'value') )
This is the best I can come up with. It prevents sql injection uses only one insert statement and can ge extended with more case statements.
CREATE PROCEDURE t_insert ( #value varchar(50) = null )
as
DECLARE #sQuery NVARCHAR (MAX);
SET #sQuery = N'
insert into __t (value) values ( '+
CASE WHEN #value IS NULL THEN ' default ' ELSE ' #value ' END +' );';
EXEC sp_executesql
#stmt = #sQuery,
#params = N'#value varchar(50)',
#value = #value;
GO
chrisofspades,
As far as I know that behavior is not compatible with the way the db engine works,
but there is a simple (i don't know if elegant, but performant) solution to achive your two objectives of DO NOT
maintain the default value in multiple places, and
use multiple insert statements.
The solution is to use two fields, one nullable for insert, and other one calculated to selections:
CREATE TABLE t (
insValue VARCHAR(50) NULL
, selValue AS ISNULL(insValue, 'something')
)
DECLARE #d VARCHAR(10)
INSERT INTO t (insValue) VALUES (#d) -- null
SELECT selValue FROM t
This method even let You centralize the management of business defaults in a parameter table, placing an ad hoc function to do this, vg changing:
selValue AS ISNULL(insValue, 'something')
for
selValue AS ISNULL(insValue, **getDef(t,1)**)
I hope this helps.
The best option by far is to create an INSTEAD OF INSERT trigger for your table, removing the default values from your table, and moving them into the trigger.
This will look like the following:
create trigger dbo.OnInsertIntoT
ON TablenameT
INSTEAD OF INSERT
AS
insert into TablenameT
select
IsNull(column1 ,<default_value>)
,IsNull(column2 ,<default_value>)
...
from inserted
This makes it work NO MATTER what code tries to insert NULLs into your table, avoids stored procedures, is completely transparent, and you only need to maintain your default values in one place, namely this trigger.
You can use default values for the parameters of stored procedures:
CREATE PROCEDURE MyTestProcedure ( #MyParam1 INT,
#MyParam2 VARCHAR(20) = ‘ABC’,
#MyParam3 INT = NULL)
AS
BEGIN
-- Procedure body here
END
If #MyParam2 is not supplied, it will have the 'ABC' value...
You can use the COALESCE function in MS SQL.
INSERT INTO t ( value ) VALUES( COALESCE(#value, 'something') )
Personally, I'm not crazy about this solution as it is a maintenance nightmare if you want to change the default value.
My preference would be Mitchel Sellers proposal, but that doesn't work in MS SQL. Can't speak to other SQL dbms.
Don't specify the column or value when inserting and the DEFAULT constaint's value will be substituted for the missing value.
I don't know how this would work in a single column table. I mean: it would, but it wouldn't be very useful.
Hope To help to -newbie as i am- Ones who uses Upsert statements in MSSQL.. (This code i used in my project on MSSQL 2008 R2 and works simply perfect..May be It's not Best Practise.. Execution time statistics shows execution time as 15 milliSeconds with insert statement)
Just set your column's "Default value or binding" field as what you decide to use as default value for your column and Also set the column as Not accept null values from design menu and create this stored Proc..
`USE [YourTable]
GO
SET ANSI_NULLS ON
GO
SET QUOTED_IDENTIFIER ON
GO
CREATE PROC [dbo].[YourTableName]
#Value smallint,
#Value1 bigint,
#Value2 varchar(50),
#Value3 varchar(20),
#Value4 varchar(20),
#Value5 date,
#Value6 varchar(50),
#Value7 tinyint,
#Value8 tinyint,
#Value9 varchar(20),
#Value10 varchar(20),
#Value11 varchar(250),
#Value12 tinyint,
#Value13 varbinary(max)
-- in my project #Value13 is a photo column which storing as byte array..
--And i planned to use a default photo when there is no photo passed
--to sp to store in db..
AS
--SET NOCOUNT ON
IF #Value = 0 BEGIN
INSERT INTO YourTableName (
[TableColumn1],
[TableColumn2],
[TableColumn3],
[TableColumn4],
[TableColumn5],
[TableColumn6],
[TableColumn7],
[TableColumn8],
[TableColumn9],
[TableColumn10],
[TableColumn11],
[TableColumn12],
[TableColumn13]
)
VALUES (
#Value1,
#Value2,
#Value3,
#Value4,
#Value5,
#Value6,
#Value7,
#Value8,
#Value9,
#Value10,
#Value11,
#Value12,
default
)
SELECT SCOPE_IDENTITY() As InsertedID
END
ELSE BEGIN
UPDATE YourTableName SET
[TableColumn1] = #Value1,
[TableColumn2] = #Value2,
[TableColumn3] = #Value3,
[TableColumn4] = #Value4,
[TableColumn5] = #Value5,
[TableColumn6] = #Value6,
[TableColumn7] = #Value7,
[TableColumn8] = #Value8,
[TableColumn9] = #Value9,
[TableColumn10] = #Value10,
[TableColumn11] = #Value11,
[TableColumn12] = #Value12,
[TableColumn13] = #Value13
WHERE [TableColumn] = #Value
END
GO`
With enough defaults on a table, you can simply say:
INSERT t DEFAULT VALUES
Note that this is quite an unlikely case, however.
I've only had to use it once in a production environment. We had two closely related tables, and needed to guarantee that neither table had the same UniqueID, so we had a separate table which just had an identity column, and the best way to insert into it was with the syntax above.
The most succinct solution I could come up with is to follow the insert with an update for the column with the default:
IF OBJECT_ID('tempdb..#mytest') IS NOT NULL DROP TABLE #mytest
CREATE TABLE #mytest(f1 INT DEFAULT(1), f2 INT)
INSERT INTO #mytest(f1,f2) VALUES (NULL,2)
INSERT INTO #mytest(f1,f2) VALUES (3,3)
UPDATE #mytest SET f1 = DEFAULT WHERE f1 IS NULL
SELECT * FROM #mytest
The pattern I generally use is to create the row without the columns that have default constraints, then update the columns to replace the default values with supplied values (if not null).
Assuming col1 is the primary key and col4 and col5 have a default contraint
-- create initial row with default values
insert table1 (col1, col2, col3)
values (#col1, #col2, #col3)
-- update default values, if supplied
update table1
set col4 = isnull(#col4, col4),
col5 = isnull(#col5, col5)
where col1 = #col1
If you want the actual values defaulted into the table ...
-- create initial row with default values
insert table1 (col1, col2, col3)
values (#col1, #col2, #col3)
-- create a container to hold the values actually inserted into the table
declare #inserted table (col4 datetime, col5 varchar(50))
-- update default values, if supplied
update table1
set col4 = isnull(#col4, col4),
col5 = isnull(#col5, col5)
output inserted.col4, inserted.col5 into #inserted (col4, col5)
where col1 = #col1
-- get the values defaulted into the table (optional)
select #col4 = col4, #col5 = col5 from #inserted
Cheers...
The easiest way to do this is to modify the table declaration to be
CREATE TABLE Demo
(
MyColumn VARCHAR(10) NOT NULL DEFAULT 'Me'
)
Now, in your stored procedure you can do something like.
CREATE PROCEDURE InsertDemo
#MyColumn VARCHAR(10) = null
AS
INSERT INTO Demo (MyColumn) VALUES(#MyColumn)
However, this method ONLY works if you can't have a null, otherwise, your stored procedure would have to use a different form of insert to trigger a default.
The questioner needs to learn the difference between an empty value provided and null.
Others have posted the right basic answer: A provided value, including a null, is something and therefore it's used. Default ONLY provides a value when none is provided. But the real problem here is lack of understanding of the value of null.
.