Ust Oldfield's Blog

Process a Tabular model from a Stored Procedure

The Challenge

Recently, at a client, I was challenged to create a stored procedure that would process a tabular model. This stored procedure would then be executed from a web application. The process behind it being: a user enters data into a web application, which gets written to a database. That data then needs to be immediately surfaced up into reports, with additional calculations and measures along the way. Therefore the tabular model, which does all the additional calculation and measures, needs to be processed by a user from the web application.


That’s the challenge – process a tabular model quickly, but should be processed by users on-demand.

The Solution

Part one: Method of process

There’s quite a few methods to process a tabular model: use an SSIS package, PowerShell, SQL Agent Job and others. I opted for SQL Agent Job because it was the most simple method of execution without having to reconfigure the server or introduce technologies and languages that weren’t already in use.

First things first, create a SQL Agent Job, I called mine ProcessTabular. Then create a Step. The Type should be SQL Server Analysis Services Command, input the server address and input the refresh script. I’m using SQL Server 2016, so using Tabular Model Scripting Language (TMSL) for my command. XMLA commands also work for older versions. A full list of commands for processing a tabular database can be found here.


Part two: Start the Agent Job

Now that we have a SQL Agent job, we’ll need to start that job using SQL. Luckily, there’s a system stored procedure that can start agent jobs: msdb.dbo.sp_start_job

Method for calling it is

EXEC msdb.dbo.sp_start_job 'ProcessTabular'

producing the following successful message


Part three: The Stored Procedure

sp_start_job works, but it doesn’t accommodate for providing a completion message, or informing a user that a process is in progress.

Introducing the code for the stored procedure:

CREATE PROCEDURE [dbo].[ProcessTabularDatabase]
DECLARE @JobName nvarchar(50) = 'ProcessTabular',
@ResultCheck INT
         SELECT 1
         FROM msdb.dbo.sysjobs_view job 
         JOIN msdb.dbo.sysjobactivity activity ON job.job_id = activity.job_id
             activity.run_Requested_date IS NOT NULL 
         AND activity.stop_execution_date IS NULL 
         AND = @JobName
     PRINT 'Starting job ''' + @JobName + '''';
     EXEC msdb.dbo.sp_start_job @JobName;
  WHILE (1 = 1)
   SELECT @ResultCheck =  IIF(stop_execution_date IS NULL,1,0)
      FROM msdb.dbo.sysjobactivity AS sja
      JOIN msdb.dbo.sysjobs AS sj ON sja.job_id = sj.job_id
      WHERE = @JobName
   IF @ResultCheck = 0 BREAK;
  PRINT 'Successfully Processed Tabular Database'
     PRINT 'Job ''' + @JobName + ''' is already started ';


This stored procedure can be executed by the web application, enabling users to process a tabular database on-demand and get feedback as to the success of the task.

SQL 2016 CTP3 R Setup and Review

The release of SQL Server 2016 CTP3 is pretty big as it’s the first CTP which incorporates in-database support for R. This means that you can call R scripts and functions within a SQL query without any workarounds as previously demonstrated.

To get yourself up and running on CTP3 you’ll need to follow these steps:

1)      Download, install and configure CTP3

a.       Make sure that Advanced Analytics Extension is installed

b.       verify that the service for Advanced Analytics Extensions starts automatically using the MSSQLLaunchpad account.

c.        On the Database Engine Configuration page, for Authentication mode, select Mixed Mode (SQL Server authentication and Windows authentication).

                                                               i.      For CTP3, a SQL login is required when using the RevoScaleR functions from a remote workstation to run R scripts in the context of the server.

2)      Download, install and configure Revolution R Open

3)      Download, install and configure Revolution R Enterprise

4)      Enable external scripts

exec sp_configure 'external scripts enabled',1;


5)      Run the post-installation script(s)

a.       Open an instance of command prompt with administrator privileges and run the following script

b.       C:\Program Files\RRO\RRO-3.2.2-for-RRE-7.5.0\R-3.2.2\library\RevoScaleR\rxLibs\x64\RegisterRExt.exe /install

                                                               i.      More info can be found here:

Now you should be up and running and ready to fire up a new instance of SQL Server Management Studio to begin working with R. For the rest of the demo I will be using the latest version of Adventure Works for SQL 2016 CTP3, which can be found here: Please restore the backed up databases to continue.

To start off we’re just going to do something really simple, which is return the mean of the UnitPrice from Sales.SalesOrderDetail in a result set, just to demonstrate functionality and ease in which you can write and execute R scripts within SQL.

The script I’m going to use is:

USE AdventureWorks2016CTP3;


EXECUTE sp_execute_external_script

@language = N'R'

,@script = N'OutputDataSet <-data.frame(mean(InputDataSet[,1]))'

,@input_data_1 = N'SELECT UnitPrice FROM [Sales].[SalesOrderDetail]'

WITH RESULT SETS ((col int not null));


What it does is call a stored procedure with three input parameters: @language; @script; and @input_data_1

The only valid input for @language is 'R’.

@script is the external language script specified as a literal or variable input. Datatype of @script is NVARCHAR(MAX).

@input_data_1 specifies the input data for the external script in the form of a T-SQL query. Datatype of @input_data_1 is NVARCHAR(MAX).

              With the specification of _1 it would appear that in future one might be able to pass in more than one lot of data.

The syntax for sp_execute_external_script can be much more complex than demonstrated and more information can be found here:

So, we’ve run our script. SQL has initiated the R script, which is taking a mean of the first column and converting it to a data frame. A data frame is similar to a table and, at the moment, the output data has to be in a data frame structure in order for it to be passed back into SQL. The end result is that the mean of the UnitPrice has been returned.

Query and result should look something like this:


R is a really powerful tool as it is and by incorporating in-database support for R within SQL Server 2016 allows for a convenient, usable and even more powerful toolkit to process and analyse data. The possibilities of application to your data are vast but can range from simple data exploration, as demonstrated above, to predictive modelling.

Be on the lookout for more demonstrations of R functionality within SQL Server 2016.