Nielsen c16.tex V4 - 07/21/2009 12:53pm Page 382 Part II Manipulating Data with Select VALUES (DEFAULT, ‘From Default’); GUID from function INSERT dbo.ProductCategory (ProductCategoryID, ProductCategoryName) VALUES (NewID(), ‘From Function’); GUID in variable DECLARE @NewGUID UniqueIdentifier; SET @NewGUID = NewID(); INSERT dbo.ProductCategory (ProductCategoryID, ProductCategoryName) VALUES (@NewGUID, ‘From Variable’); To view the results of the previous three methods of inserting a GUID, the following SELECT statement is filtered to those rows that are like `From %´: SELECT ProductCategoryID, ProductCategoryName FROM dbo.ProductCategory WHERE ProductCategoryName like ‘From %’; Result: ProductCategoryID ProductCategoryName 25894DA7-B5BB-435D-9540-6B9207C6CF8F From Default 393414DC-8611-4460-8FD3-4657E4B49373 From Function FF868338-DF9A-4B8D-89B6-9C28293CA25F From Variable This INSERT statement uses the NEWID() function to insert multiple GUIDs: INSERT dbo.ProductCategory (ProductCategoryID, ProductCategoryName) Select NewID(), LastName From CHA2.dbo.Guide; The following SELECT statement retrieves the new GUIDs: SELECT ProductCategoryID, ProductCategoryName FROM dbo.ProductCategory; Result: ProductCategoryID ProductCategoryName 1B2BBE15-B415-43ED-BCA2-293050B7EFE4 Kite 23FC5D45-8B60-4800-A505-D2F556F863C9 Accessory 3889671A-F2CD-4B79-8DCF-19F4F4703693 Video 382 www.getcoolebook.com Nielsen c16.tex V4 - 07/21/2009 12:53pm Page 383 Modification Obstacles 16 5471F896-A414-432B-A579-0880757ED097 Fletcher 428F29B3-111B-4ECE-B6EB-E0913A9D34DC Atlas E4B7D325-8122-48D7-A61B-A83E258D8729 Bistier The final GUID insert example features my recommended method, NEWSEQUENTIALID(). The function is defined as the default for the primary key column. Three rows are inserted and then selected: USE tempdb; CREATE TABLE GUIDtable ( GUIDtableID UNIQUEIDENTIFIER NOT NULL DEFAULT (NewSequentialID()) PRIMARY KEY, Col1 CHAR(1), ); INSERT GUIDtable (Col1) VALUES (’a’), (’b’), (’c’); SELECT GUIDtableID, Col1 FROM GUIDtable; Result: GUIDtableID Col1 748040E1-210D-DE11-8196-002170BF2EB9 a 758040E1-210D-DE11-8196-002170BF2EB9 b 768040E1-210D-DE11-8196-002170BF2EB9 c SQL Server provides the flexibility of two excellent candidates for surrogate key generation. Whether the database relies on identity columns or GUIDs may be based on other factors. Either way, there are mul- tiple methods for inserting new rows. You, as the SQL developer or DBA, are in control. If a column has a unique index (even if it’s not a key), then attempting to insert a new value, or an update to a new value that’s already in use, will fail. Typically, the entire transaction, including all the inserted or updated rows, will fail. However, there’s an index option, IGNORE DUP KEY, that enables the transaction to succeed with only a warning, and just skips any duplicate rows. This should only be used for staging tables in a scenario where an external application may accidentally include the same data more than once. Deleting Duplicate Rows A common question on the newsgroups is ‘‘How can I delete a single row when I don’t have a primary key?’’ 383 www.getcoolebook.com Nielsen c16.tex V4 - 07/21/2009 12:53pm Page 384 Part II Manipulating Data with Select <rant>The first order of business is to question why the table doesn’t have a primary key. Without a primary key, it’s not a legitimate table. My beef with SQL is that it even allows a table without a primary key. Horrors! One reason I like policy-based management is that it can be configured to require a primary key on every table. As soon as the duplicates are deleted, be sure to apply a primary key to the table.</rant> Fortunately, there are several methods to clean up duplicate data. The following sections include the windowing, surrogate key, and SELECT DISTINCT INTO methods. To experiment with duplicate data, the following script sets up a poorly designed (no primary key) table and inserts some duplicate data: USE tempdb ; go CREATE TABLE DupsNoPK ( Col1 INT NULL, Col2 CHAR(5) NULL ); go Insert multiple dup rows (can be executed mulitple times) INSERT DupsNoPK (Col1, Col2) VALUES (1, ‘abc’), (2, ‘abc’), (2, ‘abc’), (2, ‘abc’), (7, ‘xyz’), (7, ‘xyz’) ; To verify that the table does in fact have duplicate data, the following query uses a GROUP BY and HAVING clause to return only the duplicated rows, with a count of the number of duplicates: SELECT Col1, Col2, COUNT(*) AS DupCount FROM DupsNoPK GROUP BY Col1, Col2 HAVING COUNT(*) > 1; Result: Col1 Col2 DupCount 2 abc 3 7 xyz 2 Deleting duplicate rows using windowing Of the three methods to remove duplicate rows, this method is the most straightforward because it doesn’t need to alter the table or generate a second table. 384 www.getcoolebook.com Nielsen c16.tex V4 - 07/21/2009 12:53pm Page 385 Modification Obstacles 16 The key to this method is using the windowing’s OVER() clause with a ROW_NUMBER() function and a partition. The partition will begin renumbering with every new partition. Set the OVER() clause to PARTITION BY every column to be checked for duplicate data. In this case, every column is being checked. Running the windowing query first shows how it applies the row number: SELECT Col1, Col2, ROW_NUMBER() OVER (PARTITION BY Col1, Col2 ORDER BY Col1) AS rn FROM DupsNoPK Result: Col1 Col2 rn 1 abc 1 2 abc 1 2 abc 2 2 abc 3 7 xyz 1 7 xyz 2 Every duplicate row has an rn value of greater than 1, so it’s now easy to delete the duplicates: WITH DupsNumbered AS ( SELECT Col1, Col2, ROW_NUMBER() OVER (PARTITION BY Col1, Col2 ORDER BY Col1) AS rn FROM DupsNoPK ) DELETE DupsNumbered WHERE rn > 1; The next SELECT tests the effect of the windowing remove duplicates query: SELECT Col1, Col2 FROM DupsNoPK; Result: Col1 Col2 1 abc 2 abc 7 xyz Deleting duplicate rows using a surrogate key A traditional method of removing duplicate rows uses a surrogate key to uniquely identify each row. This means the table itself must be altered to add the surrogate key column. 385 www.getcoolebook.com Nielsen c16.tex V4 - 07/21/2009 12:53pm Page 386 Part II Manipulating Data with Select Assuming the DupsNoPK table is reset with the original rows, the following script applies an IDENTITY surrogate key and looks at the altered table: ALTER TABLE dbo.DupsNoPK ADD PK INT IDENTITY NOT NULL CONSTRAINT PK_DupsNoPK PRIMARY KEY; SELECT * FROM DupsNoPK; Result: Col1 Col2 PK 1 abc 74 2 abc 75 2 abc 76 2 abc 77 7 xyz 78 7 xyz 79 To search and destroy the duplicate data, the next query finds and deletes all the rows with matching Col1 and Col2 data but higher primary key values: DELETE DupsNoPK WHERE EXISTS ( SELECT * FROM DupsNoPK AS D1 WHERE D1.Col1 = DupsNoPK.Col1 AND D1.Col2 = DupsNoPK.Col2 AND D1.PK > DupsNoPK.PK ); SELECT * FROM DupsNoPK; Result: Col1 Col2 PK 1 abc 74 2 abc 77 7 xyz 79 Deleting duplicate rows using select distant into The third method of removing duplicate data may seem crude, but if the goal is to remove duplicates while creating a new table — perhaps as part of an ETL process — it may be the best choice. 386 www.getcoolebook.com Nielsen c16.tex V4 - 07/21/2009 12:53pm Page 387 Modification Obstacles 16 A SELECT DISTINCT will automatically pass the data through a filter that eliminates duplicate rows. The INTO option causes the results of the select to be placed into a new table instead of going to the client application. Mix the two options together and you have an instant duplicate row remover. Again, assuming the DupsNoPK table is reset with its original duplicate data, the following query gener- ates a new table without duplicates and then examines the contents of the new table: SELECT distinct Col1, Col2 INTO NoDups FROM DupsNoPK; SELECT Col1, Col2 FROM NoDups; Result: Col1 Col2 1 abc 2 abc 7 xyz Foreign Key Constraints Foreign keys may affect INSERT, UPDATE,andDELETE commands by blocking those operations. Inserting a new secondary table row with a foreign key value that doesn’t match an existing primary key will cause the secondary row insert to fail. In the following insert example, the ProductCategoryID supplied does not exist in the ProductCategory table. This causes the foreign key constraint to block the INSERT operation, as the error message indicates: Foreign Key: Insert Obstacle INSERT Product (ProductID, Code, ProductCategoryID, ProductName) VALUES (’9562C1A5-4499-4626-BB33-E5E140ACD2AC’, ‘999’ ‘DB8D8D60-76F4-46C3-90E6-A8648F63C0F0’, ‘Basic Box Kite 21"’); Result: Server: Msg 547, Level 16, State 1, Line 1 INSERT statement conflicted with COLUMN FOREIGN KEY constraint ‘FK__Product__Product__7B905C75’. The conflict occurred in database ‘OBXKites’, table ‘ProductCategory’, column ‘ProductCategoryID’. The statement has been terminated. Note that because every GUID is unique, the GUIDs you use on your system will be different. 387 www.getcoolebook.com Nielsen c16.tex V4 - 07/21/2009 12:53pm Page 388 Part II Manipulating Data with Select Foreign key constraints can also block updates to either the primary or the secondary table. If the pri- mary key is updated and a foreign key is pointed to that primary key, then the update will fail. In the following sample code, the update is blocked because the secondary table update is trying to set the foreign key, ProductCategoryID, to a value that does not exist in the ProductCategory table: Foreign Key: Secondary table Update Obstacle UPDATE Product SET ProductCategoryID = ‘DB8D8D60-76F4-46C3-90E6-A8648F63C0F0’ WHERE ProductID = ‘67804443-7E7C-4769-A41C-3DD3CD3621D9’; Result: Server: Msg 547, Level 16, State 1, Line 1 UPDATE statement conflicted with COLUMN FOREIGN KEY Constraint ‘FK__Product__Product__7B905C75’. The conflict occurred in database ‘OBXKites’, table ‘ProductCategory’, column ‘ProductCategoryID’. The statement has been terminated. If foreign keys are pointing to a primary key, updating the primary key to a new value has the same effect as deleting a primary table row with an existing secondary table row referring to it. In both cases the error is caused not by the primary key but by the foreign key referencing the primary key. In the following code, the error is generated not by the ProductCategory table, even though it is the table being updated, but by the Product table. This is because the Product table has both the row and the foreign key reference constraint that will be violated if the primary key value no longer exists: Foreign Key: Primary table Update Obstacle UPDATE ProductCategory SET ProductCategoryID = ‘DB8D8D60-76F4-46C3-90E6-A8648F63C0F0’ WHERE ProductCategoryID = ‘1B2BBE15-B415-43ED-BCA2-293050B7EFE4’; Result: Server: Msg 547, Level 16, State 1, Line 1 UPDATE statement conflicted with COLUMN REFERENCE constraint ’FK__Product__Product__7B905C75’. The conflict occurred in database ‘OBXKites’, table ‘Product’, column ‘ProductCategoryID’. The statement has been terminated. For more information about creating unique index constraints, refer to Chapter 20, ‘‘Creat- ing the Physical Database Schema.’’ 388 www.getcoolebook.com Nielsen c16.tex V4 - 07/21/2009 12:53pm Page 389 Modification Obstacles 16 Null and Default Constraints Column nullability and defaults may affect INSERT and UPDATE operations. An INSERT or UPDATE operation can send one of four possible values to a table column: data values, null, default,ornoth- ing at all. The table column can be configured with a default value and nullability. Table 16-2 indicates the result of the operation according to the column configuration, and the new value to be inserted or updated. For example, if the column properties are set so that the column has a default and does not accept nulls (see the far-right column) and the SQL insert or update sends a null, then the result is an error. TABLE 16-2 Data Modifications, Defaults, and Nulls Column Properties No default, allow null No default, not allow null Has default, allow null Has default, not allow null SQL Sent: Result: data data data data data null null error null error default null error default default nothing sent null most common error default default By far, the most common error in the table is submitting nothing when no default exists and nulls are not permitted. For more information about dealing with nulls when retrieving data, see Chapter 9, ‘‘Data Types, Expressions, and Scalar Functions.’’ Check Constraints Check constraints may affect INSERT and UPDATE operations. Each table column may have multiple check constraints. These are fast Boolean operations that deter- mine whether the update will pass or fail. 389 www.getcoolebook.com Nielsen c16.tex V4 - 07/21/2009 12:53pm Page 390 Part II Manipulating Data with Select The following check constraint permits Dr. Johnson’s insert but blocks Greg’s insert (note that the check constraint is already applied to the database by the Create_CHA2.sql script): USE CHA2; go ALTER TABLE dbo.Guide ADD CONSTRAINT CK_Guide_Age21 CHECK (DateDiff(yy,DateOfBirth, DateHire) >= 21); The following query inserts Dr. Johnson’s data. Because she was 26 years old at the time of hire, her row is accepted by the check constraint: INSERT Guide(LastName, FirstName, Qualifications, DateOfBirth, DateHire) VALUES (’Johnson’, ‘Mary’, ‘E.R. Physician’, ‘19710114’, ‘19970601’); Greg, conversely, was only 18 at the time he applied, so his insert is rejected by the check constraint: INSERT Guide (LastName, FirstName, Qualifications, DateOfBirth, DateHire) VALUES (’Franklin’, ‘Greg’, ‘Guide’, ‘19831212’, ‘20020101’); Result: Server: Msg 547, Level 16, State 1, Line 1 INSERT statement conflicted with TABLE CHECK constraint ’CK_Guide_Age21’. The conflict occurred in database ‘CHA2’, table ‘Guide’. The statement has been terminated. Instead of Triggers INSTEAD OF triggers may affect INSERT, UPDATE,andDELETE operations. Triggers are special stored procedures that are attached to a table and fire when certain data- modification operations hit that table. Two types of triggers exist: INSTEAD OF and AFTER.Theydiffer both in their timing and in how they handle the data-modification operation. An INSTEAD OF trigger always causes the INSERT, UPDATE,orDELETE operation to be canceled. The SQL command submitted to SQL Server is discarded by the INSTEAD OF trigger; the code within the INSTEAD OF trigger is executed instead of the submitted SQL command, hence the name. The INSTEAD OF trigger might be programmed to repeat the requested operation so that it looks like it went through, or it could do something else altogether. The problem with the INSTEAD OF trigger is that it reports back ‘‘n row(s) affected’’ when in fact noth- ing is written to the database. There is no error warning because the INSTEAD OF trigger works prop- erly; however, the operation doesn’t go through. 390 www.getcoolebook.com Nielsen c16.tex V4 - 07/21/2009 12:53pm Page 391 Modification Obstacles 16 In the following code sample, the InsteadOfDemo trigger causes the INSERT operation to disappear into thin air: USE CHA2; go CREATE TRIGGER InsteadOfDemo ON Guide INSTEAD OF INSERT AS Print ‘Instead of trigger demo’; With the INSTEAD OF trigger in place, the following query inserts a test row: INSERT Guide(lastName, FirstName, Qualifications, DateOfBirth, DateHire) VALUES (’Jamison’, ‘Tom’, ‘Biologist, Adventurer’, ‘19560114’, ‘19990109’); Result: Instead of trigger demo (1 row(s) affected) The INSERT operation appears to have worked, but is the row in the table? SELECT GuideID FROM Guide WHERE LastName = ‘Jamison’; Result: GuideID (0 row(s) affected) Building triggers is explained in detail in Chapter 26, ‘‘Creating DML Triggers.’’ The flow of data-modification transactions and the timing of triggers are also discussed in Chapter 66, ‘‘Managing Transactions, Locking, and Blocking.’’ Note that the sample code for this chapter drops the InsteadOfDemo trigger before moving on. After Triggers AFTER triggers may affect INSERT, UPDATE,andDELETE operations. AFTER triggers are often used for complex data validation. These triggers can roll back, or undo, the insert, update, or delete if the code inside the trigger doesn’t like the operation in question. The code can then do something else, or it can just fail the transaction. However, if the trigger doesn’t 391 www.getcoolebook.com . OVER() clause with a ROW_NUMBER() function and a partition. The partition will begin renumbering with every new partition. Set the OVER() clause to PARTITION BY every column to be checked for duplicate. to be canceled. The SQL command submitted to SQL Server is discarded by the INSTEAD OF trigger; the code within the INSTEAD OF trigger is executed instead of the submitted SQL command, hence the. ProductCategoryID = ‘DB8D8D60-76F4-46C3-90E6-A8648F63C0F0’ WHERE ProductID = ‘67804 443- 7E7C-4769-A41C-3DD3CD3621D9’; Result: Server: Msg 547, Level 16, State 1, Line 1 UPDATE statement conflicted with