Circular Dependency in the Desktop but not in the Fabric Service after March 2024 Update

[BI4Kids]

Hello 

After the March 2024 update a Circular Dependency Detected error began to appear in a relationship that seemed to be fine since forever. In fact, the Fabric Service does not issue this error. What's worse, any changes I try to make to the dataset regardless of where the changes are made; this error pops up and completely ignores applying the changes I want to make. Is there a "bug" in the latest Desktop update? Why did this error not exist prior to this? I need help, I can't seem to shake off this issue and I have several updates I need to make to this dataset which cannot be made in the Fabric service.

[BI4Kids]

I just discovered that I could revert back to a previous version of the Power BI Desktop from Power BI updates archive. So I installed the February 2024 update and sure enough the error is gone. I guess the March 2024 is "buggy"...

If anyone else is experiencing similar issues, you can test if its from the most recent update by reverting back to the previous version of Power BI desktop from here:

https://learn.microsoft.com/en-us/power-bi/fundamentals/desktop-latest-update-archive?tabs=powerbi-d...

So will someone from Microsoft Power BI let me know once this bug is corrected??

[v-yetao1-msft]

Hi @BI4Kids 

Please check if there is an error associated with the blank value, and if so, change it to 0. 

Best Regards,
Community Support Team _ Ailsa Tao

[BI4Kids]

@v-yetao1-msft There does not seem to be. I initially thought I did, but to no avail was it that. It seems there is a flaw with the update. I dont get the error in the service and the February update does not give the error.

[NarWind]

I've got the exact same issue with a client. Relationships that worked for months suddenly broke. I'm using a calculated table that is related back to another dimension table. 

[sgroth001]

I have the same thing.

[ebeery]

We're seeing this error too, with a calculated table that has a 1:N relationship with a dimension table.  This table and relationship have existed in this state for > 2 years with no issues until the March 2024 version of PBI Desktop was released.

In our case, the issue appears to be related to the internal VALUES() function that's contained within SUMMARIZECOLUMNS, as described in the articles linked below.  

Replacing the SUMMARIZECOLUMNS function with an equivalent expression using FILTER, ADDCOLUMNS, CROSSJOIN, and DISTINCT seemed to "solve" the issue.

What's not at all clear to me is why this code worked for the last 2+ years, but is only causing circular dependency errors since the March 2024 release.  

Avoiding circular dependency errors in DAX - SQLBI

SUMMARIZECOLUMNS – DAX Guide

[BI4Kids]

@ebeery thanks for the addiional context to the issue. But I still believe SUMMARIZECOLUMNS creates DISTINCT records innately. This calculated table worked all this time up until the March Desktop update. In addition, the SERVICE does not have any issues with the relationship.

I tried changing this:

to this:

but now the query takes astronomically longer; in fact what used to take less than 9 seconds ran for more  than 5 minutes before I canceled it. I believe there is a flaw with the DAX error handling in this latest update. 

[ebeery]

@BI4Kids I agree, something strange is going on.

Double-check that DAX though - did you intend for the DATESBETWEEN to be part of the FILTER function, rather than part of the CROSSJOIN?

[BI4Kids]

@ebeery I did originally, but the process ran unusually long and never finished. So I tried to get the date filter executed a bit deeper in the query to see if it made a performance difference; which it did not. At this point, I am just going to continue to develop this dataset in February 2024 update and wait on MS to resolve the obvious issue with the March update. Hopefully soon, because this is annoying...

[v-yetao1-msft]

Hi all ,

Power BI Desktop has been updated to April version , you can test it on a new version and see if the problem persists. Looking forward to the results of your tests.

Best Regards,
Community Support Team _ Ailsa Tao