💎🏭SQL Reloaded: Splitting a String (Before and After)

Starting with SQL Server 2016 Microsoft introduced the STRING_SPLIT table-valued function, which splits a string into rows of substrings, based on a specified separator character (e.g. “:”).

-- splitting a string (SQL Server 2016+)
SELECT *
FROM STRING_SPLIT('100001::M:black:', ':') 

The table-valued function object allowed also earlier to implement the same functionality, either by looping through the string or by using a common table expression. Here’s the implementation based on a loop (it was modified from a previous version to include an index):

-- split table-valued function with an index
CREATE FUNCTION [dbo].[SplitListWithIndex]( 
  @ListValues varchar(500) 
, @Delimiter char(1)) 
RETURNS @Temp TABLE(
  Ind int
, Value varchar(50)) 
AS 
BEGIN 
 DECLARE @Index int 
 DECLARE @Length int 
 DECLARE @Ind int 

 SET @Index = CharIndex(@Delimiter, @ListValues) 
 SET @Length = Len(@ListValues) - @Index 
 SET @Ind = 1
   
 WHILE @Index > 0 --if the fatch was successful
 BEGIN 
  INSERT @Temp 
  VALUES(@Ind, Substring(@ListValues, 0, @Index)) 

  SET @ListValues = Substring(@ListValues, @Index+1, @Length) 
  SET @Index = CharIndex(@Delimiter, @ListValues) 
  SET @Length = @Length - @Index 
  SET @Ind = @Ind + 1
 END  
   
 INSERT @Temp 
 VALUES(@Ind, @ListValues) 
RETURN 
END  
GO

The function could be called in code same as the STRING_SPLIT:

-- splitting a string (SQL Server 2000+)
SELECT *
FROM dbo.SplitListWithIndex('100001::M:black:', ':') 

The two functions are helpful when a column stores delimited values. It’s the case of Dynamics 365 which stores a SKU (Stock Keeping Unit) – the Product Numbers together with its Dimensions (ItemId, Configid, InventSizeId, InventColorId, StyleId) delimited by “:”, like in the above examples. Therefore, to parse the records one could write such code:

-- parsing delimited values (SQL Server 2000+)
SELECT DAT.ProductNumber
, Max(CASE WHEN LV.Ind = 1 THEN LV.Value END) ItemId
, Max(CASE WHEN LV.Ind = 2 THEN LV.Value END) Configid
, Max(CASE WHEN LV.Ind = 3 THEN LV.Value END) InventSizeId
, Max(CASE WHEN LV.Ind = 4 THEN LV.Value END) InventColorId
, Max(CASE WHEN LV.Ind = 5 THEN LV.Value END) StyleId
FROM ( VALUES ('100001::S:black:')
 , ('100001::M:black:')
 , ('100001::L:black:')
 , ('100001::XL:black:')
 , ('100001::S:white:')
 , ('100001::M:white:')
 , ('100001::L:white:')
 , ('100001::XL:white:')
) DAT (ProductNumber)
CROSS APPLY dbo.SplitListWithIndex(DAT.ProductNumber, ':') LV
GROUP BY DAT.ProductNumber
ORDER BY DAT.ProductNumber

 

Similar output can be obtained via the STRING_SPLIT with the help of row_number() ranking window function introduced with SQL Server 2005:

 

-- parsing delimited values (SQL Server 2016+)
SELECT LV.ProductNumber
, Max(CASE WHEN LV.Ind = 1 THEN LV.Value END) ItemId
, Max(CASE WHEN LV.Ind = 2 THEN LV.Value END) Configid
, Max(CASE WHEN LV.Ind = 3 THEN LV.Value END) InventSizeId
, Max(CASE WHEN LV.Ind = 4 THEN LV.Value END) InventColorId
, Max(CASE WHEN LV.Ind = 5 THEN LV.Value END) StyleId
FROM (
 SELECT DAT.ProductNumber
 , XT.VALUE
 , ROW_NUMBER() OVER (PARTITION BY DAT.ProductNumber ORDER BY DAT.ProductNumber) Ind
 FROM ( VALUES ('100001::S:black:')
  , ('100001::M:black:')
  , ('100001::L:black:')
  , ('100001::XL:black:')
  , ('100001::S:white:')
  , ('100001::M:white:')
  , ('100001::L:white:')
  , ('100001::XL:white:')
 ) DAT (ProductNumber)
 CROSS APPLY STRING_SPLIT(DAT.ProductNumber, ':') XT
) LV
GROUP BY LV.ProductNumber
ORDER BY LV.ProductNumber

As can be seen the introduction of an index into the dbo.SplitListWithIndex function simplified the code, making the use of a ranking window function unnecessary. It would be useful for the STRING_SPLIT to provide the same, as this time of processing is pretty common.  
Here’s another example based on the PIVOT clause introduced also in SQL 2005:

-- parsing delimited values (SQL Server 2016+)
SELECT P.ProductNumber
, Cast(Trim([1]) as nvarchar(20)) ItemId
, Cast(Trim([2]) as nvarchar(20)) ConfigId
, Cast(Trim([3]) as nvarchar(20)) InventSizeid
, Cast(Trim([4]) as nvarchar(20)) InventColorId
, Cast(Trim([5]) as nvarchar(20)) StyleId
FROM (
 SELECT DAT.ProductNumber
 , XT.VALUE
 , ROW_NUMBER() OVER (PARTITION BY DAT.ProductNumber ORDER BY DAT.ProductNumber) Ranking
 FROM ( VALUES ('100001::S:black:')
  , ('100001::M:black:')
  , ('100001::L:black:')
  , ('100001::XL:black:')
  , ('100001::S:white:')
  , ('100001::M:white:')
  , ('100001::L:white:')
  , ('100001::XL:white:')
 ) DAT (ProductNumber)
 CROSS APPLY STRING_SPLIT(DAT.ProductNumber, ':') XT
) DAT
PIVOT (MAX(DAT.[VALUE]) FOR DAT.Ranking IN ([1],[2],[3],[4],[5])) P

Notes:
The queries work also in SQL databases in Microsoft Fabric (see file in GitHub repository). You might want to use another schema (e.g. Test), not to interfere with the existing code. 

Happy coding!

💎🏭SQL Reloaded: Drop If Exists (Before and After)

    One of the activities of a database developer/administrator is to create and drop objects on the fly. If in objects' creation there are always some aspects to take into account that are implied by object's definition, the verbose syntax for their destruction seemed to be an unnecessary thing. For example for dropping a table, view, stored procedure, function or index, the most used objects, one would need to write such statements:

-- dropping a table 
IF  EXISTS (SELECT * FROM sys.objects WHERE object_id = OBJECT_ID(N'[dbo].[TestTable]') AND type in (N'U'))
DROP TABLE [dbo].[TestTable]

-- dropping a view 
IF  EXISTS (SELECT * FROM sys.views WHERE object_id = OBJECT_ID(N'[dbo].[TestView]'))
DROP VIEW [dbo].[TestView]

-- dropping a stored procedure 
IF  EXISTS (SELECT * FROM sys.objects WHERE object_id = OBJECT_ID(N'[dbo].[TestProcedure]') AND type in (N'P', N'PC'))
DROP PROCEDURE [dbo].[TestProcedure]

-- dropping a fucntion
IF  EXISTS (SELECT * FROM sys.objects WHERE object_id = OBJECT_ID(N'[dbo].[TestFunction]') AND type in (N'FN', N'IF', N'TF', N'FS', N'FT'))
DROP FUNCTION [dbo].[TestFunction]

-- dropping an index
IF EXISTS (SELECT Name FROM sysindexes WHERE Name = 'IX_TestTable') 
DROP INDEX dbo.TestTable.IX_TestTable

   Even if Copy-Paste does its magic and SQL Server allows generating scripts for existing objects, there’s still some work do be done in order to drop an object. Fortunately, with SQL Server 2016 Microsoft introduced a simplified syntax for dropping an object, namely DROP IF EXISTS.

   The general syntax:

DROP <object_type> [ IF EXISTS ] <object_name>

  The above statements can be written as follows:

-- dropping the table 
DROP TABLE IF EXISTS dbo.TestTable

-- dropping the view 
DROP VIEW IF EXISTS dbo.TestView 

-- dropping the procedure 
DROP PROCEDURE IF EXISTS dbo.TestProcedure

-- dropping the function 
DROP FUNCTION IF EXISTS dbo.TestFunction

-- dropping the index 
DROP INDEX IF EXISTS dbo.TestTable.IX_TestTable

  Similarly can be dropped aggregates, assemblies, roles, triggers, rules, databases, schemas, users, sequences, synonyms, etc. The scripts will run also when the objects don’t exist.

  An object can't be dropped if explicit dependencies exist on them, e.g. when the table is referenced by a  FOREIGN KEY constraint. For each object there are specific rules that apply, therefore for more details check the documentation.

  To explore the functionality here are the definitions of the above objects and the further scripts to test them:

-- creating the test table
CREATE TABLE dbo.TestTable(City nvarchar(50)
, PostalCode nvarchar(50))

-- creating the test view 
CREATE VIEW dbo.TestView 
AS
SELECT 'Test' as Result

-- creating the test stored procedure  
CREATE PROCEDURE dbo.TestProcedure
AS
BEGIN
SELECT 'Test' as Result
END

-- creating the test function
CREATE FUNCTION dbo.TestFunction()
RETURNS nvarchar(50)
BEGIN
    RETURN 'Test'
END

-- creating the test index
CREATE NONCLUSTERED INDEX [IX_TestTable] ON [dbo].[TestTable]
(
 [PostalCode] ASC
)


--testing the table
SELECT *
FROM dbo.TestTable

-- testing the view 
SELECT *
FROM dbo.TestView 

-- testing the procedure 
EXEC dbo.TestProcedure

-- testing the function
SELECT dbo.TestFunction() as Result

   Moreover, the IF EXISTS can be used when dropping the constraint or column of a table:

-- adding a new column 
ALTER TABLE dbo.TestTable
ADD DateFrom datetime2(0)

-- adding a constraint on it
ALTER TABLE dbo.TestTable
ADD CONSTRAINT [DF_DateFrom_Default]  DEFAULT (GetDate()) FOR [DateFrom]

-- inserting test data
INSERT INTO dbo.TestTable(City, PostalCode)
VALUES ('New York', 'XZY')

--testing the changes
SELECT *
FROM dbo.TestTable

-- dropping the constraint
ALTER TABLE dbo.TestTable
DROP CONSTRAINT IF EXISTS DF_DateFrom_Default

-- dropping a column from a table
ALTER TABLE dbo.TestTable
DROP COLUMN IF EXISTS DateFrom

--testing the changes
SELECT *
FROM dbo.TestTable

    If a constraint exists on the column first must be dropped the constraint and after that the column, like in the above example.  

Happy coding!

💎🏭SQL Reloaded: Trimming Strings (Before and After)

One of the annoying things when writing queries is the repetitive lengthy expressions that obfuscate in general the queries making them more difficult to read, understand and troubleshoot, and sometimes such expressions come with a performance penalty as well.    Loading data from Excel, text files and other sources involving poorly formatted data often requires trimming (all) the text values. In the early versions of SQL Server, the equivalent of a Trim function was obtained by using the combined LTrim and RTrim functions. This resumed in writing code like this (based on AdventureWorks 2014 database):

-- trimming via LTrim, RTrim 
SELECT LTrim(RTrim(AddressLine1)) AddressLine1
, LTrim(RTrim(AddressLine2)) AddressLine2
, LTrim(RTrim(City)) City
, LTrim(RTrim(PostalCode)) PostalCode
FROM Person.Address

This might not look much though imagine you have to deal with 30-50 text attributes, that the code is not written in a readable format (e.g. the way is stored in database), that some attributes require further processing (e.g. removal of special characters, splitting, concatenating).

 
 Often developers preferred encapsulating the call to the two functions within a user-defined function:

-- Trim user-defiend function
CREATE FUNCTION dbo.Trim(
@string nvarchar(max))
RETURNS nvarchar(max)
BEGIN
    RETURN LTrim(RTrim(@string))
END

With it the code is somehow simplified, but not by much and includes the costs of calling a user-defined function:

-- trimming via dbo.Trim
SELECT dbo.Trim(AddressLine1) AddressLine1
, dbo.Trim(AddressLine2) AddressLine2
, dbo.Trim(City) City
, dbo.Trim(PostalCode) PostalCode
FROM Person.Address

In SQL Server 2017 was introduced the Trim function which not only replaces the combined use of LTrim and RTrim functions, but it also allows to replace other specified characters (including CR, LF, Tab) from the start or end of a string.

By default the function removes the space from both sides of a string:

-- trimming via Trim
SELECT Trim(AddressLine1) AddressLine1
, Trim(AddressLine2) AddressLine2
, Trim(City) City
, Trim(PostalCode) PostalCode
FROM Person.Address

When a set of characters is provided the function removes the specified characters:

SELECT Trim ('#' FROM '# 843984') Example1
, Trim ('[]' FROM '[843984]') Example2
, Trim ('+' FROM '+49127298000') Example3
, Trim ('+-' FROM '+ 49-12729-8000 ') + ';' Example4
, Trim ('+ ' FROM '+ 49-12729-8000 ') + ';' Example5
, ASCII(Left(Trim (char(13) FROM char(13) + '49127298000'), 1)) Example6

Output:
Example1   Example2     Example3        Example4            Example5            Example6
--------          --------          ------------           -----------------       -----------------        -----------
  843984      843984        49127298000   49-12729-8000 ;  49-12729-8000;    52

As can be seen, when is needed to remove other characters together with the space then is needed to include the space in the list of characters.

Notes:
1) The dbo.Trim function can be created in SQL Server 2017 environments as well.
2) The collation of the database will affect the behavior of Trim function, therefore the results might look different when a case sensitive collection is used.

3)  The queries work also in SQL databases in Microsoft Fabric.

Happy coding!

💎SQL Reloaded: Deterministic Functions (GetDate SQL Server 2000/2005)

In an answer on LinkedIn, one of the users made me attentive that in SQL Server 2005 and later versions the GetDate() and other non-deterministic UDFs (user-defined functions) can be used inside functions and views, thing which was not possible under SQL Server 2000. This lack of functionality involved considerable overhead when was needed to use current date in table-valued functions. The solution was to send the date as parameter to the respective objects.     

For exemplification the following piece of code would run successfully on SQL Server 2005 but fail on SQL Server 2000, returning the below error message. 

-- creating the test function
CREATE FUNCTION dbo.fGetCurrentDate() 
RETURNS smalldatetime AS 
BEGIN 
    RETURN GetDate()
END

-- testing the function
SELECT dbo.fGetCurrentDate() 

Error message:
Server: Msg 443, Level 16, State 1, Procedure GetCurrentDate, Line 6 Invalid use of 'getdate' within a function.

A similar error message resulted when using the GetDate() function inside of a view:

-- creating the test view
CREATE VIEW dbo.vGetCurrentDate
AS
SELECT GetDate() AS CurrentDate

-- testing the view
SELECT * 
FROM dbo.vGetCurrentDate

The alternative was to use an UDF instead, which functions like a parameterized view, something like:

-- creating the test UDF
CREATE FUNCTION dbo.udfFilterByDate(
 @date datetime)
 RETURNS TABLE
 AS 
 RETURN(
 SELECT * 
 FROM [AdventureWorks2014].Production.Product
 WHERE [SellStartDate]>=@Date)

 -- testing the UDF
 SELECT *
 FROM dbo.udfFilterByDate(GetDate())

 -- testing the UDF (alternative)
 DECLARE @date as datetime
 SET @date = GetDate()
 SELECT *
 FROM dbo.udfFilterByDate(@date)

Using an UDF can still be necessary when is needed to pass dates and other types of parameters used inside of a correlated query or within the logic.