I have had success testing with another method that can also be used to mimic RLS on a shared dataset. In this use case for mimicking RLS, it's important that the users don't have build access to the dataset and can only export summarized data from any visuals, as they'd otherwise be able to access everything.
Being a live connection, all of the measures I'm mentioning below are report-level measures. Here are the steps:
- Create a measure like ACTIVE_USER = "stringliteral@sampleemail.com", typing in the user's email that you'd like to test. This is the same as testing RLS for "other role" in the "View as" Modeling menu.
- Using a measure you'd like to test, create a new measure such as:
SAMPLE_MEASURE =
VAR _user = [ACTIVE_USER]
RETURN
CALCULATE( [TARGET_MEASURE], USER_TABLE[EMAIL] = _user )
This isn't practical for testing across a wide range of measures, but it does allow for testing different users across a number of measures, by turning [ACTIVE_USER] into variable that can be set in one place (within that measure).
This does need to be set as a variable, which is just a manipulation of the calculation order in DAX, computing [ACTIVE_USER] before [SAMPLE_MEASURE] and then making that availalbe for the CALCULATE() function.
In the scenario where this is used to mimic RLS filtering, you would simply replace the string literal in the [ACTIVE_USER] measure, as: ACTIVE_USER = USERPRINCIPALNAME().
