VARBINARY and semantic model

[BastiaanK]

In my fact table in the Fabric data warehouse, I use a key column with the data type varbinary(32). What strikes me is that this column is not visible in my Power BI service semantic model.

Does anyone know if this is due to the fact that varbinary still has status preview in Fabric?

[DataBard]

Hi @BastiaanK 

Last I've seen in documentation, binary and varbinary have limited support in Power BI.
https://learn.microsoft.com/en-us/power-bi/connect-data/desktop-data-types#determine-and-specify-a-c...

And as you asserted, varbinary is still in preview for Fabric Warehouse:

https://learn.microsoft.com/en-us/fabric/data-warehouse/data-types#data-types-in-warehouse

Knowing that each 'product' within Fabric has its own development teams, I'd expect we have to look at the warehouse preview as exclusive - Just because it is in preview for Warehouse doesn't mean that Power BI's support of varbinary has changed. I would expect that it's not visible in your semantic model because the semantic model itself doesn't support it.

[BastiaanK]

Hi,

I understand what you say but I don't see the Power BI service separate from Warehouse as the semantic model functionality of the Power BI / Fabric service is part of Warehouse.

The moment you would create a semantic model in Power BI Desktop based on Warehouse and then be limited in terms of binary data type usage I would understand.

[DataBard]

I understand that having both solutions closely tied within Fabric makes it complicated in situations like this.

Think of how Microsoft is supporting the Fabric product as a whole. The many features of Fabric are not developed by a centralized team, but by different teams supporting different features. While creating a warehouse does generate a semantic model by default, it is still two separate technologies within Fabric that are developed by two separate product teams with two different backlogs to prioritize their work.

In the architecture you are looking to use which employs both technologies, varbinary has to be supported by both technologies for the full solution to work. Based on available documentation, it is not, and resolving it involves both product teams enhancing both the warehouse and semantic model further.

It can be frustrating to see the current state of the product in circumstances like this. The best we can do is raise up this issue in the community. Taking a look at the Ideas page, I found a similar request to better support binary identifiers: https://community.fabric.microsoft.com/t5/Fabric-Ideas/Support-for-uniqueidentifier-keys-as-16-byte-.... I'm not sure if this would specifically support your use case, but if we can get enough votes on ideas like this, we can get this fixed.

Hopefully this helps. If it does, be sure to accept an answer so others in the community can learn from your question.

[BastiaanK · Accepted solution]

Hi @DataBard,

My preliminary conclusion is that Fabric is a nice product with a lot of potential. Unfortunately, I notice when you start deploying it for real customer solutions that the various components like the pipelines in the Azure DF variant and both lakehouse and warehouse fall short.

Functionally, I was now looking for a solution to hash key columns in my dimensional model in Warehouse so that I can refresh dimension and fact tables in isolation. In Databricks and Snowflake, this works without any problems and so I expected that given that in Fabric Delta and Spark is the standard, this would also work.

Thanks for your thoughts and tips