In Part 1 of this series, we discussed Row Level Security in Power BI, that it is different from RLS in SQL Server 2016, and then went on to demonstrate two simple scenarios where RLS can be used to filter data in a model based on Role assignment utilizing some DAX filter expressions. We introduced the USERNAME() DAX function and demonstrated its usefulness. In this second article, we’ll be diving a little deeper into RLS.
Row Level Security in Power BI Scenario 3:
Those first two scenarios from Part 1 were not that bad to implement. One line of a DAX expression and you’ve got a simple filter covered. But life seldom ever hands us a scenario that is so cut and dry. For this scenario, we’re going to add some requirements that might get handed down by the business users such as:
- Ability to grant access to a group of countries, like continents, or regions.
- Ability to have any one country be in multiple geo-political groups.
- Ability to grant and revoke access to an entire group of people at once rather than needing to address each person individually
- Ability to grant access on any one country to any one individual
For this we’re going to again go back to the database and create a few tables to help us. We already have the Country table so no need to do anything there. But we’ll add four more:
- Users
- Groups
- User Group Membership
- Country Group Membership
The T-SQL code for this can be found in the attachment, and a simple database drawing is shown here:
Image may be NSFW.
Clik here to view.
We’ll also add some rows to the respective tables via basic INSERT statements:
For those of you who, like me, are full-blooded SQL nerds, I have included the CREATE TABLE, INSERT INTO, and CREATE VIEW scripts that can be executed in your database. But we don’t need to import all four additional tables into your model, we simply need the distinct list of Users and the Countries to which each has access. This is represented in the [dbo].[SecurityQuery] view (Script #05). If you look at the view definition, note the DISTINCT key word in the SELECT clause, the optional fields to show the User’s Full Name and the Country Name, the absence of any fields from the [Group] or[ GroupUser] tables, and the WHERE clause at the bottom that filters for only Active records.
First, we’ll need to remove the Continent table from the model, then import the SecurityQuery dataset. Since this is not an exercise in how to import data, I’ll leave it up to you to get that done. And while you’re adding it, you might as well add the [SecurityReference] view as well and we’ll cover its usage at the end of the article.
Once you have the [SecurityQuery] in your model, it needs to be joined to the Country table, on Country code:
Image may be NSFW.
Clik here to view.
Hint: If you have sufficient rows in the Security Query dataset, when you set up the relationship between [Country Code] in the Country table and the [Country Code] in the Security Query table, Power BI will recognize that this is a one-to-many relationship with the Country table being at the “one” side.
As before, we will use the Manage Roles button on the Modelling tab to configure the DAX Filter for the Roles. We’ll start with what we learned from before: a filter on the [UserName] field and making use of the USERNAME() DAX function as before:
Image may be NSFW.
Clik here to view.
But if we were to browse the model now for any one member in our set, we’ll see that he or she is limited in the rows returned by the Security Query, but NOT limited by rows returned in the Country table. Why not? It worked before. It is because the table on which the USERNAME() filter is applied is not at the top of the hierarchy, it’s at the bottom. To make things more complicated, there are two branches from the top of that hierarchy, one branch to sales, through the country table, and one branch directly to the security table. If only there was a way to filter the countries in that table based on those listed in the Security Query table. We need something more.
Enter the CONTAINS() DAX function. The description from MDSN is as follows:
Syntax
CONTAINS(<table>, <columnName>, <value>[, <columnName>, <value>]…)
Return Value
A value of TRUE if each specified value can be found in the corresponding columnName, or are contained, in those columns; otherwise, the function returns FALSE.
If you’re like me, you’ll read that and think, “say WHAT?” Let’s implement it first, then we’ll sort out and explain how the parameters are used. Back under the Manage Roles, add a Filter to the Country table as the following (hint: it doesn’t matter which column you select as you will be replacing it with the entire text below):
CONTAINS (
‘SecurityQuery’,
‘SecurityQuery'[UserName],
USERNAME(),
‘SecurityQuery'[CountryCode],
‘Country'[CountryCode]
)
Note: The line breaks are not required in this, but are added for clarity of reading.
Now here’s the layman’s description of how the DAX statement above works with the five parameters, in order of their appearance:
Go to the ‘SecurityQuery’ table (1st parameter), and in the column ‘SecurityColumn’[UserName] (2nd parameter), look for any rows that match the value returned by the function USERNAME() (3rd parameter). Then for those rows, take the values in the column ‘SecurityQuery’[CountryCode] (4th), and see if those values exist in the column ‘Country’[CountryCode] (5th). If that [CountryCode] value is found return TRUE for that row, and allow it to be viewed in this context.
There are a couple of things to watch for when implementing this approach. First, you need to have a relationship defined between the Country table and the SecurityQuery table. Second, the <value> parameter (3rd parameter) of the CONTAINS() function can be just that, a single value, not a list or table. But the USERNAME() function fits this bill nicely. Third, you still need the [UserName] = USERNAME() security filter on the Security Query dataset.
If you followed the INSERT scripts included, you will recall that there were only two people involved: Fred and Bob. And you may recall that we granted Fred access to a Group called “Pacific Rim” (which included countries such as Japan, Hong Kong, and the Philippines) but that [GroupMembership] row was NOT flagged with [IsActive] = 0. Fred’s mapped data looks like this:
Image may be NSFW.
Clik here to view.
Now here’s where the [SecurityRefernce] dataset comes in. After adding it as part of the model, you need to make sure that A) it has NO relationships to any other tables and that B) it also has a [UserName] = USERNAME() DAX Filter expression applied.
I added a report page to my Power BI model and a simple Matrix visualization based on this table and configured it as follows:
- Rows = [Country Name]
- Columns = [GroupName]
- Values = [IsActive], with a SUM aggregation
- For the [IsActive], I set up Conditional Formatting with the following properties
Image may be NSFW.
Clik here to view.
The resulting visualization is a nice way to see exactly HOW a person has gotten access to any one particular country (via membership in which Group), and what countries are in any Group they have are also a member of, even if that membership is not active.
Image may be NSFW.
Clik here to view.
It is true that there is a lot introduced here that may not specifically be “Row Level Security” stuff, but rather T-SQL overhead. After all, all you really need for this 3rd scenario to work is the distinct list of Users and Countries that are allowed access. But given the 173 countries in the world and say, two dozen people to which access control is required, that’s potentially up to 4,000 rows of data controlling who has access to what. Breaking it into Users and Groups and Memberships is a way to manage the mess.
About Todd: Todd Chittenden started his programming and reporting career with industrial maintenance applications in the late 1990’s. When SQL Server 2005 was introduced, he quickly became certified in Microsoft’s latest RDBMS technology and has added certifications over the years. He currently holds an MCSE in Business Intelligence . He has applied his knowledge of relational databases, data warehouses, business intelligence and analytics to a variety of projects for BlumShapiro since 2011.
Image may be NSFW.
Clik here to view.