Clear All Tables Within a Schema

August 11, 2015 Leave a comment

Often as part of an ETL process I want to simply clear out all the table within a schema, e.g. ‘landing’ or ‘etl’.

However, as new tables are added to the schema, the ‘clear out’ code needs to be maintained. In addition, if any referential integrity exists, the tables will need to be cleared in a certain order. Finally, the Truncate Table statement is significantly faster than a simple Delete, but there are restrictions on when it can be used.
All of which led me to create the following stored procedure so that I don’t have to think these concerns again, allowing me time to focus on the core of the ETL work.

Feel free to use/copy/amend as appropriate, but please let me have your comments/updates.


Create Procedure etl.ClearSchemaTables
/******************************************************************************
Description:
    Clears down any tables in the target schema.
    Tables are cleared down in referential order using truncate where possible.
In:
    @target_schema_name
        The name of the schema whose tables should be cleared, e.g. 'landing'.
History:
    03-Aug-2015 S.Wilber
        Created
******************************************************************************/
    @target_schema_name sysname
As

Set         NoCount On

-- declare variables
Declare     @object_id int
           ,@table_name sysname
           ,@schema_name sysname
           ,@is_referenced bit
           ,@success bit
           ,@sql nvarchar(max)

If Object_ID('tempdb..#table_dependencies') Is Not Null Drop Table #table_dependencies
Create Table #table_dependencies
(
    source_object_id int
   ,dependancy_object_id int
)

-- declare tables
If Object_ID('tempdb..#tables') Is Not Null Drop Table #tables
Create Table #tables
(
    object_id int
   ,schema_name sysname
   ,table_name sysname
   ,is_referenced bit default(0)
   ,is_processed bit default(0)
)

-- find all tables in the schema - look for fk references that are not from self
Insert      #tables (object_id, schema_name, table_name, is_referenced)
Select      t.object_id
           ,s.name
           ,t.name
           ,Case When fk.object_id Is Not Null Then 1 Else 0 End
From        sys.tables t
Inner Join  sys.schemas s On (s.schema_id = t.schema_id)
Left Join   sys.foreign_keys fk On (fk.referenced_object_id = t.object_id and fk.parent_object_id != t.object_id)
Where       s.name = @target_schema_name

-- retrieve all interdependancies between batch managed tables.
Insert      #table_dependencies (source_object_id, dependancy_object_id)
Select      fk.parent_object_id
           ,fk.referenced_object_id
From        sys.foreign_keys fk
Inner Join  #tables s On (s.object_id = fk.parent_object_id)
Inner Join  #tables t On (t.object_id = fk.referenced_object_id)

-- loop through all the tables delete the content, either via a truncate or delete depending on foreign keys
While Exists (Select 1 From #tables Where is_processed = 0)
  Begin
  
    -- find the first table where all of the dependant tables have been processed
    Select      @object_id = object_id
               ,@schema_name = schema_name
               ,@table_name = table_name
               ,@is_referenced = is_referenced
    From        #tables sbt
    Where       is_processed = 0
    And         Not Exists
                (
                    Select      1
                    From        #table_dependencies td
                    Inner Join  #tables tbt On (tbt.object_id = td.source_object_id)
                    Where       td.dependancy_object_id = sbt.object_id
                    And         tbt.is_processed = 0
                )

    -- if none found, then raise an error as it seems that we have a circular dependancy that we can't resolve
    If @object_id Is Null
      Begin
        RaisError ('A table where all dependant tables have been processed cannot be found. This implies that circular dependencies exist. Batch rollback cannot continue.', 11, 0)
      End

    Begin Try
        -- if referenced then we need to perform a delete otherwise we can perform a truncate
        If @is_referenced = 1 
          Begin
            Set         @sql = 'Delete From [' + @schema_name + '].[' + @table_name + ']'
          End
        Else
          Begin
            Set         @sql = 'Truncate Table [' + @schema_name + '].[' + @table_name + ']'
          End
          
        -- exeute the sql
        Exec        sp_executesql @stmt = @sql          
    
    End Try
    Begin Catch
        -- flag as unsuccessful
        Set         @success = 0

        -- declare variables
        Declare         @error_description varchar(2000)
                       ,@error_severity int
                       ,@error_state int

        -- fetch the error details
        Select         @error_description = ERROR_MESSAGE(), @error_severity = ERROR_SEVERITY(), @error_state = ERROR_STATE()

        -- throw the error
        RaisError (@error_description, @error_severity, @error_state) 
            
    End Catch
    
    -- update the table to state that that it has been processed
    Update      #tables
    Set         is_processed = 1
    Where       object_id = @object_id
    
  End  

Temporarily Disabling Referential Integrity Constraints on a SQL Table

I recently came across a situation where part of the data processing would put the data temporarily into a non-referential state. In order to accommodate this, without scripting Drop/Create statements which could get out of sync with other development work, I put together the following stored procedure which enables/disables check constraints.

Feel free to use/copy/amend as appropriate, but please let me have your comments/updates.

Create Procedure [etl].[SetCheckContraintsState]
/******************************************************************************
Description:
    Enables/Disables any check contraints that reference the target table.
    This is useful for when referenced data is going to be temporarily modified
    in a way that will invalidate any check constraints.
History:
    02-Jun-2015 S.Wilber
        Created
******************************************************************************/
    @target_schema_name sysname
   ,@target_table_name sysname
   ,@do_set_enabled bit
As

-- declare variables
Declare     @constraint_name sysname
           ,@referencing_schema_name sysname
           ,@referencing_table_name sysname
           ,@sql nvarchar(max)

Declare @constraints table
(
    ConstraintName sysname
   ,ReferencingSchemaName sysname
   ,ReferencingTableName sysname
)

Insert      @constraints (ConstraintName, ReferencingSchemaName, ReferencingTableName)
Select      fk.name As ConstraintName
           ,sch.name As ReferencingSchemaName
           ,t.name As ReferencingTableName
From        sys.foreign_keys fk
Inner Join  sys.tables t On (t.object_id = fk.parent_object_id)
Inner Join  sys.schemas sch On (sch.schema_id = t.schema_id)
Where       fk.referenced_object_id = Object_ID(QuoteName(@target_schema_name) + '.' + QuoteName(@target_table_name))

Declare     constraint_cursor Cursor Fast_Forward For
Select      ConstraintName
           ,ReferencingSchemaName
           ,ReferencingTableName
From        @constraints

Open constraint_cursor
Fetch From constraint_cursor Into @constraint_name, @referencing_schema_name, @referencing_table_name

While @@Fetch_Status = 0
  Begin
    If @do_set_enabled = 1
        Set @sql = 'Alter Table ' + QuoteName(@referencing_schema_name) + '.' + QuoteName(@referencing_table_name) + ' With Check Check Constraint ' + QuoteName(@constraint_name)
    Else
        Set @sql = 'Alter Table ' + QuoteName(@referencing_schema_name) + '.' + QuoteName(@referencing_table_name) + ' NoCheck Constraint ' + QuoteName(@constraint_name)

    Exec sp_executesql @sql

    -- get the next constraint
    Fetch From constraint_cursor Into @constraint_name, @referencing_schema_name, @referencing_table_name
End

Close constraint_cursor
Deallocate constraint_cursor

Code Reuse in SSIS – Passing Values To and From Child Packages

August 15, 2014 2 comments

I recently found myself having to replicate the same piece of script code in a SSIS package multiple times, which of course makes any developer feel dirty as copied code tends to drift divergently and not be maintained correctly, especially where the number of copies is beyond a manageable few.

Given restrictions of various clients that I’ve worked with I’ve had to avoid creating SSIS extensions or DLLs that would have to be deployed along with the package. This has meant that code reuse is always a little tricky. The tricky aspect is that in order to reuse a script in any meaningful way parameter values need to be passed in and return values received back. Thus, with plain SSIS this is difficult.

Or so I thought. I’ve now found that it is possible, and in fact easy to pass parameters down to child packages, via the well-known ‘Parent package variable’ configuration, but also to pass values back up to the parent package. Below I’ll walk through a simple example showing how this is done.

Firstly, create a parent package as shown below. We’ll fill in the details shortly.

Contents of Parent Package

 

The parent package will execute a child package two times in parallel to show that there are no issues with parallel running.

The Sequence Container labelled ‘Start’ is purely a placeholder and contains no sub-tasks.

The two parallel Sequence Containers each contain a call to the child package and a script task to message box out the result. Message boxes clearly are not the best thing in an SSIS package, but they suit our purposes for this demo.

Against each Sequence Container I set up two variables, one as a parameter value to pass to the child package and another to store the result coming back. To show that the parameter value passed down is consumed correctly each container is given a different value as shown in the screenshots below.

Sequence 1 Variables

Sequence 2 Variables

 

The code to show the message box containing the resultant value is shown below. Don’t forget to add ‘User::Result’ as a read only variable for the script task.

public void Main()
{
    // display result in a message box
    MessageBox.Show(Dts.Variables["User::Result"].Value.ToString());

    // return success as the task result
    Dts.TaskResult = (int)ScriptResults.Success;
}

 

Before we can set up the task to execute the child package we first need to create the child package.

The child package simply consists of two script tasks. The first to create a resultant value from the parent parameter value and the second to sleep for 5 seconds to show parallel running.

Child Package

 

In order to access the parent parameter value we need to set up package configuration on the child package. To do this, firstly create a variable on the child package to hold the parent value.

Child Package Variables

 

Secondly, right click anywhere on the background of the child package and select ‘Package Configurations…’.

Tick the check box labelled ‘Enable package configurations’.

Click ‘Add…’ to add a new configuration item which will pop up the Package Configuration Wizard as per the screenshot below.

Package Configuration Wizard Step 1

 

Select ‘Parent package variable’ as the configuration type.

For ‘Parent variable’ simply enter the name and not the full path of the parent variable, ‘ParameterValue’. By not entering the path, the variable is searched for initially in the scope of the task calling the child package and if not found, then upwards from there. Since we want to reuse the child task it is important that a full path is not specified so that we can pick up the value from wherever the child task is called from.

Click ‘Next >’.

Now we need to assign the parent value to a property within the child package. Select the ‘Value’ property of the variable we created earlier called ‘ParentParameterValue’ as per the screenshot below.

Package Configuration Wizard Step 2

 

Click ‘Next >’.

Finally, give the configuration a name as per the screenshot below. Again, I’ve chosen ‘ParentParameterValue’.

Package Configuration Wizard Step 3

 

Finally, click ‘Finish’ and then ‘Close’.

So now, when the parent package is executed, the variable values set in the parent package can be picked up in the child package. Let’s now see how we can pass result values back. It is surprisingly simple.

Edit the script task in the child package called ‘Set Result Value’. In the initial dialogue box we need to set up the variables that the script can access. Set them up as per the screenshot below, i.e. adding ‘User::ParentParameterValue’ as a read only variable and ‘User::Result’ as a read/write variable.

Child Script Task Configuration

 

But wait! What is this ‘User::Result’ variable. We’ve not set it up in the child package. The script task won’t complain. What this in fact will refer to, once the package is run, is the nearest variable of that name in scope, starting from the child package, but then bubbling up the search through to the parent package starting at the point where the child package is called and then onwards and upwards. One way to think about it is to imagine that a copy of the contents of the child package is in fact inserted into the parent package in place of the ‘Execute Child Package’ task.

Once you think of it like that, the next logical question is whether it is possible to read parent package variables without having to add in the ‘Parent Package Variable’ configuration, and it is. So long as no explicit paths are given a variable reference will bubble up from the scope of the child package through to the parent calling package just as if the child were in fact part of the parent.

So why do we need ‘Parent Package Variable’ configuration at all? It is needed because validation of the child package for nearly all task types will fail if the variable references don’t point to variables that the validation process can see, i.e. are in scope in the current package as packages are validated in their own right and not in the context of any calling package. For Script Tasks this variable validation is not carried out. Variables within the script itself are available through the Dts.Variables object which is an aggregation of all variables in scope at the time of execution going up from the child package up to and including the parent package. Dts.Variables also has a ‘Contains’ method to allow you to check for the existence of a variable before use.

So, back to the demo. The script itself is shown below and simply shows that the parent value is accessible and also adds in a message from the child.

public void Main()
{
    // set the result value to be the parent value plus some changes
    Dts.Variables["User::Result"].Value = "Result: '" + Dts.Variables["User::ParentParameterValue"].Value + "'. Child says hello.";

    // return success as the task result
    Dts.TaskResult = (int)ScriptResults.Success;
}

 

The script for the ‘Sleep for 5 Seconds’ task is as below. You will need to add a ‘using’ to ‘System.Threading’ at the top of the script.

public void Main()
{
    // set the result value to be the parent value plus some changes
    Thread.Sleep(5000);

    // return success as the task result
    Dts.TaskResult = (int)ScriptResults.Success;
}

 

Finally, we can return to the parent package and complete the details required in the ‘Execute Child Package’ tasks, as per the screenshot below.

Execute Child Task

 

Executing the parent package, following the 5 second sleep, results in the following two message boxes popping up simultaneously.

Result 1Result 2

 

From this we can see that both child tasks executed in parallel, both could access variables in the parent package and could set values in the parent package and that the parent package itself could see those changes.

Going back to the initial code reuse problem this means that script tasks can be moved into child packages and can pick up in scope variables values from the executing parent and return values back for further consumption within the parent package. Brilliant.

 

Categories: SSIS Tags: , , ,

Show Expression Logic in SSIS Task Links (Constraints)

June 16, 2014 1 comment

As part of my campaign to make logic discovery easier in SSIS for my future self I discovered by chance today a great little feature.
Often key logic can be build into the constraints (the red/green arrows) between tasks in the form of expressions. The logic is not visible without showing the properties of the constraint or by double clicking on it.

 

A simple SSIS package including some constraint expressions where the logic is not clear.

Often the logic for a package can be hidden in constraint expressions.

 

However in the properties for a constraint there is an item called “ShowAnnotation”. Set this to “ConstraintOptions” and the logic of the constraint is shown as an annotation next to the constraint. Brilliant!

There are a few other options, such as “ConstraintDescription” which can be useful to summarise the logic if it grows to long. An example is shown below.

 

Properties window in SSIS highlighting the 'ShowAnnotation' property.

Properties window in SSIS highlighting the ‘ShowAnnotation’ property.

 

Rather sadly the layout of the constraint annotation is not brilliant, so sometimes it overlaps other items when ‘Auto Layout’ is used. However, in allowing an understanding of process flow logic at a glance, it’s great.

 

A simple SSIS package including some constraint expressions where the logic is made visible via annotations.

The logic of the constraint expressions shown by setting ‘ShowAnnotation’ to ‘ConstraintOptions’.

 

 

Fast Row Count for All Tables

November 12, 2013 Leave a comment

 

Using Count(*) to obtain a row count can be very slow on large tables. This is the only truly reliable way to obtain an accurate row count but, post SQL Server 2000, the mess that was sys.indexes, etc. was largely cleaned up and nowadays the row count can be obtained using the following code. Microsoft still states that the row count in sys.partitions is an “Approximate number of rows in this partition”, but it appears to be stable and accurate with appropriate caveats with regards to uncommitted transactions.

 

Select      s.Name
           ,t.Name
           ,Sum(p.rows) As Rows
From        sys.tables t
Inner Join  sys.schemas s On (s.schema_id = t.schema_id)
Inner Join  sys.indexes i On (i.object_id = t.object_id and i.index_id < 2)
Inner Join  sys.partitions p On (p.object_id = t.object_id)
Where       p.index_id In (0, 1)
Group By    s.Name
           ,t.Name
Order By    s.name
           ,t.name
Categories: SQL Server, T-SQL Tags: , , ,

Mark Items as Default in Reference Data … and Ensure There is Only One

November 6, 2013 Leave a comment

 

Often it is convenient to be able to default values in where they are not specified, or to not require items to be specified to make life easier.
For maintainability it is best to mark default values in the data itself, rather than in some code somewhere, and where better than in the source reference table.
By adding a bit column ‘IsDefault’ it is simple to mark a particular record as the default, however it is also easy to mark several items as default by mistake.
But with careful use of a filtered unique index it is possible to apply a constraint such that there can be only one default record as below.

 

Create Unique NonClustered Index IX_UNIQUE_IS_DEFAULT On MyReferenceTable
(
	IsDefault Asc
)
Where (IsDefault=(1))

 

By filtering, we don’t index the ‘false’ values so that more than one record can be marked as “not the default”, and by making it unique we ensure that we only have a single record marked as default.

Generate Strong Password in T-SQL

October 30, 2013 Leave a comment

 

Recently I needed to create a stored procedure to automate the creation of user logins within SQL Server and required a way to generate strong passwords. The following is the resultant code. It is fairly simplistic, but flexible and convenient.
It uses a recursive common table expression to create a number list and then uses the Crypt_Gen_Random function to generate a random number in the binary variable which is then used to dip into the character string to pull out the characters to build up the password. Due to the use of “For XML Path” characters which would be HTML encoded have been excluded.

 

Declare     @length int = 10
Declare     @chars char(92)
           ,@bytes binary(128)
           ,@password varchar(128)

Set         @chars = 'abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789!"£$%^*()_+-={[}]:;@~#,.?/'
Set         @bytes = Crypt_Gen_Random(128)
  
;With cte_numbers (number) As
(
    Select      Cast(1 As int)
    Union All
    Select      number + 1
    From        cte_numbers
    Where       number < @length
)
Select      @password = 
            (
                Select      SubString(@chars, (SubString(@bytes, n.Number, 1) % Len(@chars)) + 1, 1)
                From        cte_numbers n
                For         XML Path ('')
            )
            
-- display the resultant password
Select      @password
Categories: SQL Server, T-SQL
Follow

Get every new post delivered to your Inbox.